i965_drv_video/shaders/mpeg2/Makefile
i965_drv_video/shaders/mpeg2/vld/Makefile
i965_drv_video/shaders/render/Makefile
+ i965_drv_video/shaders/post_processing/Makefile
test/Makefile
test/basic/Makefile
test/decode/Makefile
-SUBDIRS = h264 mpeg2 render
+SUBDIRS = h264 mpeg2 render post_processing
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: AYUV_Load_16x8.asm
+//----------------------------------------------------------------
+
+
+#include "AYUV_Load_16x8.inc"
+
+// In order to load 64x8 AYUV data (16x8 pixels), we need to divide the data
+// into two regions and load them separately.
+//
+// 32 byte 32 byte
+//|----------------|----------------|
+//| | |
+//| A | B |8
+//| | |
+//| | |
+//|----------------|----------------|
+
+// Load the first 32x8 data block
+// Packed data block should be loaded as 32x8 pixel block
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Block origin
+ shl (1) rMSGSRC.0<1>:d acc0:w 2:w { NoDDClr } // H. block origin need to be four times larger
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_YUV:ud { NoDDChk } // Block width and height (32x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_YUV(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_YUV+nBI_CURRENT_SRC_YUV:ud
+
+//Load the second 32x8 data block
+// Offset the origin X - move to next 32 colomns
+ add (1) rMSGSRC.0<1>:d rMSGSRC.0<0;1,0>:d 32:w // Increase X origin by 8
+
+// Size stays the same - 32x8
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud // Copy message description to message header
+ send (8) udSRC_YUV(8)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_YUV+nBI_CURRENT_SRC_YUV:ud
+
+// Give AYUV region addresses to address register
+ mov (1) SRC_YUV_OFFSET<1>:ud 0x00400038*32:ud //Address registers contain starting addresses of two halves
+
+//Directly move the data to destination
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) uwDEST_Y(%1)<1> r[SRC_YUV_OFFSET,%1*32+2]<8,4>:ub
+ mov (16) uwDEST_U(%1)<1> r[SRC_YUV_OFFSET,%1*32+1]<8,4>:ub
+ mov (16) uwDEST_V(%1)<1> r[SRC_YUV_OFFSET,%1*32+0]<8,4>:ub
+ }
+
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: AYUV_Load_16x8.inc
+//
+// AYUV data are first loaded to bottom I/O REGION_2, then unpacked to planar data
+// and stored in top I/O REGION_1
+
+#undef nY_NUM_OF_ROWS
+
+#define nY_NUM_OF_ROWS 8 // Number of Y rows per block
+
+#define nDPR_BLOCK_SIZE_YUV nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // Y block size 32x8
+#define nDPR_MSG_SIZE_YUV nRESLEN_8 // # of MRF's to hold Y block data (8)
+
+//Temporary storage for unpacked AYUV data
+#define rUNPACK_TEMP REG(r,nTEMP0)
+.declare udUNPACK_TEMP Base=rUNPACK_TEMP ElementSize=4 SrcRegion=<8;8,1> Type=ud //1 GRF
+.declare ubUNPACK_TEMP Base=rUNPACK_TEMP ElementSize=1 SrcRegion=<32;32,1> Type=ub //1 GRF
+
+.declare ubBOT_Y_IO Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(32,1) Type=ub
+
+
+#define udSRC_YUV udBOT_Y_IO
+#define ubSRC_YUV ubBOT_Y_IO
+#define nSRC_YUV_REG nBOT_Y
+
+#define uwDEST_Y uwTOP_Y
+#define uwDEST_U uwTOP_U
+#define uwDEST_V uwTOP_V
+
+#define SRC_YUV_OFFSET a0.0
+
+#define nSRC_REGION nREGION_1 // REGION_1 will be the source region for first kernel
+
+// End of AYUV_Load_16x8.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: Expansion.inc
+// Number of U/V rows per block definition
+#undef nUV_NUM_OF_ROWS
+#ifdef EXPAND_9x5
+ #define nUV_NUM_OF_ROWS 6
+#else
+ #define nUV_NUM_OF_ROWS 8
+#endif
+
+// Source/destination region definitions
+#undef uwDEST_U
+#undef uwDEST_V
+#if (nSRC_REGION==nREGION_1)
+ #define uwDEST_U uwTOP_U
+ #define uwDEST_V uwTOP_V
+#elif (nSRC_REGION==nREGION_2)
+ #define uwDEST_U uwBOT_U
+ #define uwDEST_V uwBOT_V
+#endif
+
+// End of Expansion.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: IMC3_Load_8x4.asm
+//
+//----------------------------------------------------------------
+
+#define IMC3_LOAD_8x4
+#include "PL3_Load.inc"
+
+// Load 16x8 planar Y ----------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 8x4 planar U and V -----------------------------------------------------
+ asr (2) rMSGSRC.0<1>:d rMSGSRC.0<2;2,1>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (8x4)
+
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_U:ud
+ mov (8) mMSGHDRV<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_V(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_V:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ mov (16) uwDEST_U(0, %1*16)<1> ubSRC_U(0, %1*16)
+ mov (16) uwDEST_V(0, %1*16)<1> ubSRC_V(0, %1*16)
+ }
+
+// End of IMC3_Load_8x4
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: IMC3_Load_8x5.asm
+//
+//----------------------------------------------------------------
+
+#define IMC3_LOAD_8x5
+#include "PL3_Load.inc"
+
+// Load 16x8 planar Y ----------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 8x5 planar U and V -----------------------------------------------------
+ asr (2) rMSGSRC.0<1>:d rMSGSRC.0<2;2,1>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (8x5)
+
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_U:ud
+ mov (8) mMSGHDRV<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_V(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_V:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ mov (16) uwDEST_U(0, %1*16)<1> ubSRC_U(0, %1*16)
+ mov (16) uwDEST_V(0, %1*16)<1> ubSRC_V(0, %1*16)
+ }
+
+// End of IMC3_Load_8x5
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: IMC3_Load_9x5.asm
+//
+//----------------------------------------------------------------
+// This module loads 16x8 Y, 9x5 U and 9x5 V planar data blocks for CSC module
+// and stores it in byte-aligned format.
+//----------------------------------------------------------------
+
+#define IMC3_LOAD_9x5
+#include "PL3_Load.inc"
+
+// Load 16x8 planar Y ----------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 9x5 planar U and V -----------------------------------------------------
+ asr (2) rMSGSRC.0<1>:d rMSGSRC.0<2;2,1>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (12x5)
+
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_U:ud
+ mov (8) mMSGHDRV<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_V(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_V:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for(nUV_NUM_OF_ROWS-2; >-1; -1) {
+ mov (16) uwDEST_U(0, %1*16)<1> ubSRC_U(0, %1*16)
+ mov (16) uwDEST_V(0, %1*16)<1> ubSRC_V(0, %1*16)
+ }
+
+// End of IMC3_Load_9x5
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#ifdef GT // to remove error messages of un-initialized GRF
+ .declare udGRF_space Base=r0.0 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+
+ $for (7; <80; 1) {
+ mov (8) udGRF_space(%1)<1> 0:ud
+ }
+#else
+#endif
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+/////////////////////////////////////////////////////////////////////////////////
+// Multiple_Loop.asm
+
+
+// This lable is for satisfying component kernel build.
+// DL will remove this label and reference the real one in Multiple_Loop_Head.asm.
+#if defined(COMPONENT)
+VIDEO_PROCESSING_LOOP:
+#endif
+
+
+//===== Possible build flags for component kernels
+// 1) INC_SCALING
+// 2) INC_BLENDING
+// 3) INC_BLENDING and INC_SCALING
+// 4) (no flags)
+
+
+#define MxN_MULTIPLE_BLOCKS
+
+//------------------------------------------------------------------------------
+#if defined(MxN_MULTIPLE_BLOCKS)
+// Do Multiple Block Processing ------------------------------------------------
+
+ // The 1st block has been processed before entering the loop
+
+ // Processed all blocks?
+ add.z.f0.0 (1) wNUM_BLKS:w wNUM_BLKS:w -1:w
+
+ // Reached multi-block width?
+ add (1) wORIX:w wORIX:w 16:w
+ cmp.l.f0.1 (1) null:w acc0.0:w wFRAME_ENDX:w // acc0.0 has wORIX
+
+ #if defined(INC_SCALING)
+ // Update SRC_VID_H_ORI for scaling
+ mul (1) REG(r,nTEMP0):f fVIDEO_STEP_X:f 16.0:f
+ add (1) fSRC_VID_H_ORI:f REG(r,nTEMP0):f fSRC_VID_H_ORI:f
+ #endif
+
+ #if defined(INC_BLENDING)
+ // Update SRC_ALPHA_H_ORI for blending
+ mul (1) REG(r,nTEMP0):f fALPHA_STEP_X:f 16.0:f
+ add (1) fSRC_ALPHA_H_ORI:f REG(r,nTEMP0):f fSRC_ALPHA_H_ORI:f
+ #endif
+
+ (f0.0)jmpi (1) END_VIDEO_PROCESSING // All blocks are done - Exit loop
+
+ (f0.1)jmpi (1) VIDEO_PROCESSING_LOOP // If not the end of row, goto the beginning of the loop
+
+ //If end of row, restart Horizontal offset and calculate Vertical offsets next row.
+ mov (1) wORIX:w wCOPY_ORIX:w
+ add (1) wORIY:w wORIY:w 8:w
+
+ #if defined(INC_SCALING)
+ // Update SRC_VID_H_ORI and SRC_VID_V_ORI for scaling
+ mov (1) fSRC_VID_H_ORI:f fFRAME_VID_ORIX:f // Reset normalised X origin to 0 for video and alpha
+ mul (1) REG(r,nTEMP0):f fVIDEO_STEP_Y:f 8.0:f
+ add (1) fSRC_VID_V_ORI:f REG(r,nTEMP0):f fSRC_VID_V_ORI:f
+ #endif
+
+ #if defined(INC_BLENDING)
+ // Update SRC_ALPHA_H_ORI and SRC_ALPHA_V_ORI for blending
+ mov (1) fSRC_ALPHA_H_ORI:f fFRAME_ALPHA_ORIX:f // Reset normalised X origin to 0 for video and alpha
+ mul (1) REG(r,nTEMP0):f fALPHA_STEP_Y:f 8.0:f
+ add (1) fSRC_ALPHA_V_ORI:f REG(r,nTEMP0):f fSRC_ALPHA_V_ORI:f
+ #endif
+
+ jmpi (1) VIDEO_PROCESSING_LOOP // Continue Loop
+
+END_VIDEO_PROCESSING:
+ nop
+
+#endif
+END_THREAD // End of Thread
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//////////////////////////////////////////////////////////////////////////////////
+// Multiple_Loop_Head.asm
+// This code sets up the loop control for multiple blocks per thread
+
+ mul (1) wFRAME_ENDX:w ubBLK_CNT_X:ub 16:uw { NoDDClr } // Build multi-block loop counters
+ mov (1) wNUM_BLKS:w ubNUM_BLKS:ub { NoDDClr, NoDDChk } // Copy num blocks to word variable
+ mov (1) wCOPY_ORIX:w wORIX:w { NoDDChk } // Copy multi-block origin in pixel
+ mov (2) fFRAME_VID_ORIX<1>:f fSRC_VID_H_ORI<4;2,2>:f // Copy src video origin for scaling, and alpha origin for blending
+ add (1) wFRAME_ENDX:w wFRAME_ENDX:w wORIX:w // Continue building multi-block loop counters
+
+VIDEO_PROCESSING_LOOP: // Loop back entry point as the biginning of the loop for multiple blocks
+
+// Beginning of the loop
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: NV11_Load_4x8.asm
+//----------------------------------------------------------------
+
+#define NV11_LOAD_4x8
+#include "PL2_Load.inc"
+
+// Load 16x8 NV11 Y ------------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Y block width and height (16x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 8x8 NV11 UV ----------------------------------------------------------
+ asr (1) rMSGSRC.0<1>:d rMSGSRC.0<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (8x8)
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_UV:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (nUV_NUM_OF_ROWS/4-1; >-1; -1) {
+ mov (16) uwDEST_U(0,%1*16)<1> ubSRC_U(0,%1*32)<32;16,2>
+ mov (16) uwDEST_V(0,%1*16)<1> ubSRC_U(0,%1*32+1)<32;16,2>
+ }
+
+// End of NV11_Load_4x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: NV11_Load_5x8.asm
+//----------------------------------------------------------------
+
+#define NV11_LOAD_5x8
+#include "PL2_Load.inc"
+
+// Load 16x8 NV11 Y ------------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Y block width and height (16x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 12x8 NV11 UV ---------------------------------------------------------
+ asr (1) rMSGSRC.0<1>:d rMSGSRC.0<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (12x8)
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_UV:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ mov (16) uwDEST_U(0,%1*16)<1> ubSRC_U(0,%1*32)<16;8,2>
+ mov (16) uwDEST_V(0,%1*16)<1> ubSRC_U(0,%1*32+1)<16;8,2>
+ }
+
+// End of NV11_Load_5x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: NV12_Load_8x4.asm
+//----------------------------------------------------------------
+
+#define NV12_LOAD_8x4
+#include "PL2_Load.inc"
+
+// Load 16x8 planar Y ----------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Y block width and height (16x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 8x4 planar U and V -----------------------------------------------------
+ asr (1) rMSGSRC.1<1>:d rMSGSRC.1<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (16x4)
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_UV:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ mov (16) uwDEST_U(0,%1*16)<1> ubSRC_U(0,%1*32)<32;16,2>
+ mov (16) uwDEST_V(0,%1*16)<1> ubSRC_U(0,%1*32+1)<32;16,2>
+ }
+
+// End of NV12_Load_8x4
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: NV12_Load_8x5.asm
+//----------------------------------------------------------------
+
+#define NV12_LOAD_8x5
+#include "PL2_Load.inc"
+
+// Load 16x8 planar Y ----------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Y block width and height (16x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 8x5 planar U and V -----------------------------------------------------
+ asr (1) rMSGSRC.1<1>:d rMSGSRC.1<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (16x5)
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_UV:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ mov (16) uwDEST_U(0,%1*16)<1> ubSRC_U(0,%1*32)<16;8,2>
+ mov (16) uwDEST_V(0,%1*16)<1> ubSRC_U(0,%1*32+1)<16;8,2>
+ }
+
+// End of NV12_Load_8x5
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: NV12_Load_9x5.asm
+//----------------------------------------------------------------
+
+#define NV12_LOAD_9x5
+#include "PL2_Load.inc"
+
+// Load 16x8 planar Y ----------------------------------------------------------
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Y block width and height (16x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+// Load 9x5 planar U and V -----------------------------------------------------
+ asr (1) rMSGSRC.1<1>:d rMSGSRC.1<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (20x5)
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_UV:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (nY_NUM_OF_ROWS-1; >-1; -1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for(nUV_NUM_OF_ROWS-2; >-1; -1) {
+ mov (16) uwDEST_U(0,%1*16)<1> ubSRC_U(0,%1*32)<16;8,2>
+ mov (16) uwDEST_V(0,%1*16)<1> ubSRC_U(0,%1*32+1)<16;8,2>
+ }
+
+// End of NV12_Load_9x5
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: P208_Load_8x8.asm
+//----------------------------------------------------------------
+
+#define P208_LOAD_8x8
+#include "PL2_Load.inc"
+
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Y Block width and height (16x8) (U/V block size is the same)
+
+// Load 16x8 P208 Y ------------------------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+ // Load 16x8 planar UV -----------------------------------------------------
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_UV:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) uwDEST_Y(0,%1*16) ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_U(0,%1*16) ubSRC_U(0,%1*32)<32;16,2>
+ mov (16) uwDEST_V(0,%1*16) ubSRC_U(0,%1*32+1)<32;16,2>
+ }
+
+// End of P208_Load_8x8.asm
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: P208_Load_9x8.asm
+//----------------------------------------------------------------
+
+#define P208_LOAD_9x8
+#include "PL2_Load.inc"
+
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+
+// Load 16x8 P208 Y ------------------------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_Y:ud // Y block width and height (16x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_CURRENT_SRC_Y:ud
+#endif
+
+ // Load 16x8 planar UV -----------------------------------------------------
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (20x8)
+ mov (8) mMSGHDRU<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDRU udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_CURRENT_SRC_UV:ud
+
+// Convert to word-aligned format ----------------------------------------------
+#if !defined(LOAD_UV_ONLY)
+ $for (0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) uwDEST_Y(0,%1*16) ubSRC_Y(0,%1*16)
+ }
+#endif
+ $for (0; <nUV_NUM_OF_ROWS; 1) {
+ mov (16) uwDEST_U(0,%1*16) ubSRC_U(0,%1*32)<32;16,2>
+ mov (16) uwDEST_V(0,%1*16) ubSRC_U(0,%1*32+1)<32;16,2>
+ }
+
+// End of P208_Load_9x8.asm
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PA_Load.inc
+//
+// YUV422 data are first loaded to bottom I/O REGION_2, then unpacked to planar data
+// and stored in top I/O REGION_1
+
+#undef nY_NUM_OF_ROWS
+#undef nUV_NUM_OF_ROWS
+
+#define nY_NUM_OF_ROWS 8 // Number of Y rows per block
+#define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+
+#if defined(PA_LOAD_8x8)
+ #define nDPR_BLOCK_SIZE_YUV nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // Y block size 32x8
+ #define nDPR_MSG_SIZE_YUV nRESLEN_8 // # of MRF's to hold Y block data (8)
+#endif
+#if defined(PA_LOAD_9x8)
+ #define nDPR_BLOCK_SIZE_YUV_MAIN nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // Main YUV block size 32x8
+ #define nDPR_MSG_SIZE_YUV_MAIN nRESLEN_8 // # of MRF's to hold Y block data (8)
+ #define nDPR_BLOCK_SIZE_YUV_ADDITION nBLOCK_WIDTH_4+nBLOCK_HEIGHT_8 // Additional YUV block size 4x8
+ #define nDPR_MSG_SIZE_YUV_ADDITION nRESLEN_1 // # of MRF's to hold Y block data (8)
+#endif
+
+#define udSRC_YUV udBOT_Y_IO
+#define nSRC_YUV_REG nBOT_Y
+
+#define uwDEST_Y uwTOP_Y
+#define uwDEST_U uwTOP_U
+#define uwDEST_V uwTOP_V
+
+#define nSRC_REGION nREGION_1 // REGION_1 will be the source region for first kernel
+
+// End of PA_Load.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PA_Load_8x8.asm
+//----------------------------------------------------------------
+
+#define PA_LOAD_8x8
+#include "PA_Load.inc"
+
+// Load 16x8 packed data block
+// Packed data block should be loaded as 32x8 pixel block
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Block origin
+ shl (1) rMSGSRC.0<1>:d acc0:w 1:w // H. block origin need to be doubled
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_YUV:ud // Block width and height (32x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_YUV(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_YUV+nBI_CURRENT_SRC_YUV:ud
+
+// Unpack to "planar" YUV422 format in word-aligned bytes
+ add (4) pCF_Y_OFFSET<1>:uw ubSRC_CF_OFFSET<4;4,1>:ub nSRC_YUV_REG*nGRFWIB:w // Initial Y,U,V offset in YUV422 block
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) uwDEST_Y(0, %1*16)<1> r[pCF_Y_OFFSET, %1*nGRFWIB]REGION(16,2)
+ mov (8) uwDEST_U(0, %1*8)<1> r[pCF_U_OFFSET, %1*nGRFWIB]REGION(8,4)
+ mov (8) uwDEST_V(0, %1*8)<1> r[pCF_V_OFFSET, %1*nGRFWIB]REGION(8,4)
+ }
+
+// End of PA_Load_8x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PA_Load_9x8.asm
+//----------------------------------------------------------------
+// This module loads 16x8 Y, 9x8 U and 9x8 V planar data blocks for CSC module
+// and stores it in word-aligned format.
+//----------------------------------------------------------------
+
+#define PA_LOAD_9x8
+#include "PA_Load.inc"
+
+// Load 18x8 packed data block
+// Packed data block should be loaded as 36x8 pixel block
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Block origin
+ shl (1) rMSGSRC.0<1>:d acc0:w 1:w // H. block origin need to be doubled
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_YUV_MAIN:ud // Block width and height (32x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_YUV(0)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_YUV_MAIN+nBI_CURRENT_SRC_YUV:ud
+
+ add (1) rMSGSRC.0<1>:d rMSGSRC.0:d 32:w //the last 4 pixels are read again for optimization
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_YUV_ADDITION:ud // Block width and height (4x8)
+ mov (8) mMSGHDRY<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_YUV(8)<1> mMSGHDRY udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_YUV_ADDITION+nBI_CURRENT_SRC_YUV:ud
+
+// Unpack to "planar" YUV422 format in word-aligned bytes
+ add (4) pCF_Y_OFFSET<1>:uw ubSRC_CF_OFFSET<4;4,1>:ub nSRC_YUV_REG*nGRFWIB:w // Initial Y,U,V offset in YUV422 block
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) uwDEST_Y(0, %1*16)<1> r[pCF_Y_OFFSET, %1*nGRFWIB]REGION(16,2)
+ mov (8) uwDEST_U(0, %1*16)<1> r[pCF_U_OFFSET, %1*nGRFWIB]REGION(8,4)
+ mov (8) uwDEST_V(0, %1*16)<1> r[pCF_V_OFFSET, %1*nGRFWIB]REGION(8,4)
+ }
+
+ $for(0; <nUV_NUM_OF_ROWS; 1) {
+ mov (1) uwDEST_U(0, %1*16+8)<1> r[pCF_U_OFFSET, %1*4+256]REGION(1,0)
+ mov (1) uwDEST_V(0, %1*16+8)<1> r[pCF_V_OFFSET, %1*4+256]REGION(1,0)
+ }
+ //UV expansion done in PL9x8_PL16x8.asm module
+
+// End of PA_Load_9x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL16x8_PL8x4.asm
+//----------------------------------------------------------------
+
+#include "common.inc"
+
+#ifndef DEST_U //DEST_U, DEST_V not defined
+
+ #if (nSRC_REGION==nREGION_1)
+ #define DEST_Y uwTOP_Y
+ #define DEST_U uwTOP_U
+ #define DEST_V uwTOP_V
+ #elif (nSRC_REGION==nREGION_2)
+ #define DEST_Y uwBOT_Y
+ #define DEST_U uwBOT_U
+ #define DEST_V uwBOT_V
+ #endif
+
+#endif
+
+//Convert 444 from sampler to 422
+$for (0, 0; <8; 2, 1) {
+ mov (8) DEST_U(0,%2*8)<1> DEST_U(%1)<16;8,2>
+ mov (8) DEST_V(0,%2*8)<1> DEST_V(%1)<16;8,2>
+}
+
+// Re-define new number of lines
+#undef nUV_NUM_OF_ROWS
+#define nUV_NUM_OF_ROWS 4
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL16x8_PL8x8.asm
+//----------------------------------------------------------------
+
+#include "common.inc"
+
+#ifndef DEST_U
+
+ //DEST_U, DEST_V not defined
+ #if (nSRC_REGION==nREGION_1)
+ #define DEST_Y uwTOP_Y
+ #define DEST_U uwTOP_U
+ #define DEST_V uwTOP_V
+ #elif (nSRC_REGION==nREGION_2)
+ #define DEST_Y uwBOT_Y
+ #define DEST_U uwBOT_U
+ #define DEST_V uwBOT_V
+ #endif
+
+#endif
+
+
+//Convert 444 from sampler to 422
+$for (0, 0; <8; 2, 1) {
+ mov DEST_U(%2)<1> DEST_U(%1)<16;8,2>
+ mov DEST_V(%2)<1> DEST_V(%1)<16;8,2>
+}
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL2_Load.inc
+
+#undef nY_NUM_OF_ROWS
+#undef nUV_NUM_OF_ROWS
+
+#define nY_NUM_OF_ROWS 8 // Number of Y rows per block
+#define nDPR_BLOCK_SIZE_Y nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // Y block size 16x8
+#define nDPR_MSG_SIZE_Y nRESLEN_4 // # of MRF's to hold Y block data (4)
+
+
+#if defined(NV11_LOAD_4x8)
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_8+nBLOCK_HEIGHT_8 // U/V block size 8x8
+ #define nDPR_MSG_SIZE_UV nRESLEN_2 // # of MRF's to hold U/V block data (2)
+#endif
+
+#if defined(NV11_LOAD_5x8)
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_12+nBLOCK_HEIGHT_8 // U/V block size 12x8
+ #define nDPR_MSG_SIZE_UV nRESLEN_4 // # of MRF's to hold U/V block data (4)
+#endif
+#if defined(NV12_LOAD_8x4)
+ #define nUV_NUM_OF_ROWS 4 // Number of U/V rows per block
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_16+nBLOCK_HEIGHT_4 // U/V block size 16x4
+ #define nDPR_MSG_SIZE_UV nRESLEN_2 // # of MRF's to hold U/V block data (2)
+#endif
+#if defined(NV12_LOAD_8x5)
+ #define nUV_NUM_OF_ROWS 6 // Number of U/V rows per block (Rounded Up to Nearest Even Number)
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_16+nBLOCK_HEIGHT_5 // U/V block size 16x5
+ #define nDPR_MSG_SIZE_UV nRESLEN_3 // # of MRF's to hold U/V block data (3)
+#endif
+#if defined(NV12_LOAD_9x5)
+ #define nUV_NUM_OF_ROWS 6 // Number of U/V rows per block (Rounded Up to Nearest Even Number)
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_20+nBLOCK_HEIGHT_5 // U/V block size 20x5
+ #define nDPR_MSG_SIZE_UV nRESLEN_5 // # of MRF's to hold U/V block data (5)
+#endif
+#if defined(P208_LOAD_8x8)
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // U/V block size 16x8
+ #define nDPR_MSG_SIZE_UV nRESLEN_4 // # of MRF's to hold U/V block data (4)
+#endif
+#if defined(P208_LOAD_9x8)
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_20+nBLOCK_HEIGHT_8 // U/V block size 20x8
+ #define nDPR_MSG_SIZE_UV nRESLEN_8 // # of MRF's to hold U/V block data (8)
+#endif
+
+// Source/destination region definitions
+#if !defined(udSRC_Y)
+ #define udSRC_Y udBOT_Y_IO // Default Y source region is top Y region
+#endif
+
+#if !defined(udSRC_U)
+ #define udSRC_U udBOT_U_IO // Default U source region is top U region
+#endif
+
+#define ubSRC_Y ubBOT_Y
+#define nSRC_Y_REG nBOT_Y
+#define ubSRC_U ubBOT_U
+#define nSRC_U_REG nBOT_U
+
+#define uwDEST_Y uwTOP_Y // However they can be transferred to word-aligned byte if desired
+#define uwDEST_U uwTOP_U
+#define uwDEST_V uwTOP_V
+
+#define nSRC_REGION nREGION_1 // REGION_1 will be the source region for first kernel
+
+// End of PL2_Load.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL3_Load.inc
+
+#undef nY_NUM_OF_ROWS
+#undef nUV_NUM_OF_ROWS
+
+#define nY_NUM_OF_ROWS 8 // Number of Y rows per block
+#define nDPR_BLOCK_SIZE_Y nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // Y block size 16x8
+#define nDPR_MSG_SIZE_Y nRESLEN_4 // # of MRF's to hold Y block data (4)
+
+#if defined(IMC3_LOAD_8x4)
+ #define nUV_NUM_OF_ROWS 4 // Number of U/V rows per block
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_8+nBLOCK_HEIGHT_4 // U/V block size 8x4
+ #define nDPR_MSG_SIZE_UV nRESLEN_1 // # of MRF's to hold U/V block data (1)
+#endif
+#if defined(IMC3_LOAD_8x5)
+ #define nUV_NUM_OF_ROWS 6 // Number of U/V rows per block (Rounded Up to Nearest Even Number)
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_8+nBLOCK_HEIGHT_5 // U/V block size 8x5
+ #define nDPR_MSG_SIZE_UV nRESLEN_2 // # of MRF's to hold U/V block data (2)
+#endif
+#if defined(IMC3_LOAD_9x5)
+ #define nUV_NUM_OF_ROWS 6 // Number of U/V rows per block (Rounded Up to Nearest Even Number)
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_12+nBLOCK_HEIGHT_5 // U/V block size 12x5
+ #define nDPR_MSG_SIZE_UV nRESLEN_3 // # of MRF's to hold U/V block data (3)
+#endif
+
+// Source/destination region definitions
+#if !defined(udSRC_Y)
+ #define udSRC_Y udBOT_Y_IO // Default Y source region is top Y region
+#endif
+
+#if !defined(udSRC_U)
+ #define udSRC_U udBOT_U_IO // Default U source region is top U region
+#endif
+
+#if !defined(udSRC_V)
+ #define udSRC_V udBOT_V_IO // Default V source region is top V region
+#endif
+
+#define ubSRC_Y ubBOT_Y // Loading data are always in byte type
+#define ubSRC_U ubBOT_U
+#define ubSRC_V ubBOT_V
+
+#define uwDEST_Y uwTOP_Y // However they can be transferred to word-aligned byte if desired
+#define uwDEST_U uwTOP_U
+#define uwDEST_V uwTOP_V
+
+#define nSRC_REGION nREGION_1 // REGION_1 will be the source region for first kernel
+
+// End of PL3_Load.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+
+#include "PL4x8_Save_NV11.inc"
+
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+#if !defined(SAVE_UV_ONLY)
+// Save current planar frame Y block data (16x8) -------------------------------
+
+ mov (2) mMSGHDR.0<1>:d wORIX<2;2,1>:w // Block origin
+ mov (1) mMSGHDR.2<1>:ud nDPW_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+
+///* Yoni - masking is not relevant for ILK?!?
+//
+// //Use the mask to determine which pixels shouldn't be over-written
+// cmp.ge.f0.0 (1) NULLREG BLOCK_MASK_D:ud 0x00FFFFFF:ud //Check if all pixels in the block need to be modified
+// (f0.0) jmpi WritePlanarToDataPort
+//
+// //If mask is not all 1's, then load the entire 16x8 block
+// //so that only those bytes may be modified that need to be (using the mask)
+// send (8) SRC_YD(0)<1> MSGHDR MSGSRC<8;8,1>:ud DWBRMSGDSC+0x00040000+BI_DEST_Y:ud //16x8
+//
+// asr (2) MSGSRC.0<1>:ud ORIX<2;2,1>:w 1:w // U/V block origin should be half of Y's
+// mov (1) MSGSRC.2<1>:ud 0x00030007:ud // Block width and height (8x4)
+// send (8) SRC_UD(0)<1> MSGHDR MSGSRC<8;8,1>:ud DWBRMSGDSC+0x00010000+BI_DEST_U:ud
+// send (8) SRC_VD(0)<1> MSGHDR MSGSRC<8;8,1>:ud DWBRMSGDSC+0x00010000+BI_DEST_V:ud
+//
+// //Restore the origin information
+// mov (2) MSGSRC.0<1>:ud ORIX<2;2,1>:w // Block origin
+// mov (1) MSGSRC.2<1>:ud 0x0007000F:ud // Block width and height (16x8)
+//
+// //expand U and V to be aligned on word boundary
+// mov (16) SRC_UW(1)<1> SRC_U(0,16)
+// mov (16) SRC_UW(0)<1> SRC_U(0, 0)
+// mov (16) SRC_VW(1)<1> SRC_V(0,16)
+// mov (16) SRC_VW(0)<1> SRC_V(0, 0)
+//
+// //Merge the data
+// mov (1) f0.1:uw BLOCK_MASK_V:uw //Load the mask on flag reg
+// (f0.1) mov (8) TEMP0<1>:uw BLOCK_MASK_H:uw
+// (-f0.1) mov (8) TEMP0<1>:uw 0:uw
+//
+// // Destination is Word aligned
+// $for(0; <Y_ROW_SIZE; 2) {
+// mov (1) f0.1:uw TEMP(0,%1)<0;1,0>
+// (-f0.1) mov (16) DEST_Y(0, %1*32)<2> SRC_Y(0, %1*16)
+// (-f0.1) mov (16) DEST_U(0, %1*8)<1> SRC_U(0, %1*8) //only works for Word aligned Byte data
+// (-f0.1) mov (16) DEST_V(0, %1*8)<1> SRC_V(0, %1*8) //only works for Word aligned Byte data
+//
+// mov (1) f0.1:uw TEMP(0,1+%1)<0;1,0>
+// (-f0.1) mov (16) DEST_Y(0, 1+%1*32)<2> SRC_Y(0, 1+%1*16)
+//
+// }
+//
+//*/ Yoni - masking is not relevant for ILK?!?
+
+WritePlanarToDataPort:
+ $for(0,0; <nY_NUM_OF_ROWS; 2,1) {
+ mov (16) mubMSGPAYLOAD(%2,0)<1> ub2DEST_Y(%1)REGION(16,2)
+ mov (16) mubMSGPAYLOAD(%2,16)<1> ub2DEST_Y(%1+1)REGION(16,2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_Y+nBI_DESTINATION_Y:ud
+#endif
+
+// Save U/V data block in planar format (4x8) ----------------------------------
+ mov (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w // Block origin
+ asr (1) rMSGSRC.0<1>:d rMSGSRC.0<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ $for(0,0; <nY_NUM_OF_ROWS;4,1) {
+ mov (16) mubMSGPAYLOAD(%2,0)<2> ub2DEST_U(%2)REGION(16,2)
+ mov (16) mubMSGPAYLOAD(%2,1)<2> ub2DEST_V(%2)REGION(16,2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_UV+nBI_DESTINATION_UV:ud
+
+// End of PL4x8_Save_NV11
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//Module name: PL8x4_Save_NV11.inc
+//
+// Setup for storing planar data
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+#define nDPW_BLOCK_SIZE_Y nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // Y block size 16x8
+#define nDPW_MSG_SIZE_Y nMSGLEN_4 // # of MRF's to hold Y block data (4)
+#define nDPW_BLOCK_SIZE_UV nBLOCK_WIDTH_8+nBLOCK_HEIGHT_8 // U/V interleaved block width and height (8x8)
+#define nDPW_MSG_SIZE_UV nMSGLEN_2 // # of MRF's to hold U/V block data (2)
+
+#if (nSRC_REGION==nREGION_1)
+ #define udSRC_Y udBOT_Y_IO
+ #define udSRC_U udBOT_U_IO
+ #define udSRC_V udBOT_V_IO
+ #define ubSRC_Y ubBOT_Y
+ #define ubSRC_U ubBOT_U
+ #define ubSRC_V ubBOT_V
+
+ #define uwSRC_U uwBOT_U //For masking operation
+ #define uwSRC_V uwBOT_V
+
+ #define ub2DEST_Y ub2TOP_Y
+ #define ub2DEST_U ub2TOP_U
+ #define ub2DEST_V ub2TOP_V
+
+#elif (nSRC_REGION==nREGION_2)
+ #define udSRC_Y udTOP_Y_IO
+ #define udSRC_U udTOP_U_IO
+ #define udSRC_V udTOP_V_IO
+ #define ubSRC_Y ubTOP_Y
+ #define ubSRC_U ubTOP_U
+ #define ubSRC_V ubTOP_V
+
+ #define uwSRC_U uwTOP_U //For masking operation
+ #define uwSRC_V uwTOP_V
+
+ #define ub2DEST_Y ub2BOT_Y
+ #define ub2DEST_U ub2BOT_U
+ #define ub2DEST_V ub2BOT_V
+
+#endif
+
+///* Yoni - masking is not relevant for ILK?!?
+//#define TEMP0 REG(r,54)
+//.declare TEMP Base=TEMP0 ElementSize=2 SrcRegion=<8;8,1> Type=uw
+///* Yoni - masking is not relevant for ILK?!?
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL5x8_PL16x8.asm
+
+#include "Expansion.inc"
+
+//------------------------------ Horizontal Upconversion -----------------------------
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ avg.sat (16) uwDEST_U(0, %1*32+16) uwDEST_U(0, %1*16+7)<1;2,0> uwDEST_U(0, %1*16+7)<1;2,1>
+ avg.sat (16) uwDEST_V(0, %1*32+16) uwDEST_V(0, %1*16+7)<1;2,0> uwDEST_V(0, %1*16+7)<1;2,1>
+ avg.sat (16) uwDEST_U(0, %1*32) uwDEST_U(0, %1*16)<1;2,0> uwDEST_U(0, %1*16)<1;2,1>
+ avg.sat (16) uwDEST_V(0, %1*32) uwDEST_V(0, %1*16)<1;2,0> uwDEST_V(0, %1*16)<1;2,1>
+ }
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ avg.sat (16) uwDEST_U(0, %1*32+16) uwDEST_U(0, %1*32+18)<1;2,0> uwDEST_U(0, %1*32+18)<1;2,1>
+ avg.sat (16) uwDEST_V(0, %1*32+16) uwDEST_V(0, %1*32+18)<1;2,0> uwDEST_V(0, %1*32+18)<1;2,1>
+ avg.sat (16) uwDEST_U(0, %1*32) uwDEST_U(0, %1*32)<1;2,0> uwDEST_U(0, %1*32)<1;2,1>
+ avg.sat (16) uwDEST_V(0, %1*32) uwDEST_V(0, %1*32)<1;2,0> uwDEST_V(0, %1*32)<1;2,1>
+ }
+
+// End of PL5x8_PL16x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL5x8_PL8x8.asm
+
+#include "Expansion.inc"
+
+//------------------------------ Horizontal Upconversion -----------------------------
+ $for (0; <nUV_NUM_OF_ROWS; 1) {
+ avg.sat (8) uwDEST_U(0, %1*8) uwDEST_U(0, %1*8)<1;2,0> uwDEST_U(0, %1*8)<1;2,1>
+ avg.sat (8) uwDEST_V(0, %1*8) uwDEST_V(0, %1*8)<1;2,0> uwDEST_V(0, %1*8)<1;2,1>
+ }
+
+// End of PL5x8_PL8x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL8x4_Save_IMC3.asm
+//
+// Save planar YUV420 frame data block of size 16x8
+
+#include "PL8x4_Save_IMC3.inc"
+
+//Use the mask to determine which pixels shouldn't be over-written
+ and (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0x00FFFFFF:ud
+ cmp.ge.f0.0(1) dNULLREG acc0.0<0;1,0>:ud 0x00FFFFFF:ud //Check if all pixels in the block need to be modified
+ (f0.0) jmpi WritePlanarToDataPort
+
+ //If mask is not all 1's, then load the entire 16x8 block
+ //so that only those bytes may be modified that need to be (using the mask)
+
+ // Load 16x8 planar Y ----------------------------------------------------------
+ mov (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w // Block origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_Y(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_DESTINATION_Y:ud
+ // Load 8x4 planar U and V -----------------------------------------------------
+ asr (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_UV:ud // Block width and height (8x4)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_U(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_DESTINATION_U:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ send (8) udSRC_V(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_DESTINATION_V:ud
+
+ //expand U and V to be aligned on word boundary - Y remains in bytes
+ $for (nUV_NUM_OF_ROWS/2-1; >-1; -1) {
+ mov (16) uwSRC_U(0, %1*16)<1> ubSRC_U(0, %1*16)
+ mov (16) uwSRC_V(0, %1*16)<1> ubSRC_V(0, %1*16)
+ }
+
+ //Merge the data
+ mov (1) f0.0:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.0) mov (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw
+ (-f0.0) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+ // Destination is Word aligned
+ $for(0; <nY_NUM_OF_ROWS; 2) {
+ mov (1) f0.1:uw uwMASK_TEMP(0,%1)<0;1,0>
+ (-f0.1) mov (16) ub2DEST_Y(0, %1*32)<2> ubSRC_Y(0, %1*16)
+ (-f0.1) mov (16) ub2DEST_U(0, %1*8)<1> ubSRC_U(0, %1*8) //only works for Word aligned Byte data
+ (-f0.1) mov (16) ub2DEST_V(0, %1*8)<1> ubSRC_V(0, %1*8) //only works for Word aligned Byte data
+
+ mov (1) f0.1:uw uwMASK_TEMP(0,1+%1)<0;1,0>
+ (-f0.1) mov (16) ub2DEST_Y(0, 1+%1*32)<2> ubSRC_Y(0, 1+%1*16)
+ }
+
+WritePlanarToDataPort:
+#if !defined(SAVE_UV_ONLY)
+// Save current planar frame Y block data (16x8) -------------------------------
+ mov (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w // Block origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ $for(0,0; <nY_NUM_OF_ROWS; 2,1) {
+ mov(16) mubMSGPAYLOAD(%2,0)<1> ub2DEST_Y(%1)REGION(16,2)
+ mov(16) mubMSGPAYLOAD(%2,16)<1> ub2DEST_Y(%1+1)REGION(16,2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_Y+nBI_DESTINATION_Y:ud
+#endif
+// Save U/V data block in planar format (8x4) ----------------------------------
+ asr (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_UV:ud // Block width and height (8x4)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+// Save U picture data ---------------------------------------------------------
+ mov (16) mubMSGPAYLOAD(0,0)<1> ub2DEST_U(0)REGION(16,2) // U rows 0,1
+ mov (16) mubMSGPAYLOAD(0,16)<1> ub2DEST_U(1)REGION(16,2) // U rows 2,3
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_UV+nBI_DESTINATION_U:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+// Save V picture data ---------------------------------------------------------
+ mov (16) mubMSGPAYLOAD(0,0)<1> ub2DEST_V(0)REGION(16,2) // V rows 0,1
+ mov (16) mubMSGPAYLOAD(0,16)<1> ub2DEST_V(1)REGION(16,2) // V rows 2,3
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_UV+nBI_DESTINATION_V:ud
+
+// End of PL8x4_Save_IMC3
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL8x4_Save_IMC3.inc
+//
+// Setup for storing planar data
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+// For saving
+#define nDPW_BLOCK_SIZE_Y nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // Y block size 16x8
+#define nDPW_MSG_SIZE_Y nMSGLEN_4 // # of MRF's to hold Y block data (4)
+#define nDPW_BLOCK_SIZE_UV nBLOCK_WIDTH_8+nBLOCK_HEIGHT_4 // U/V block size 8x4
+#define nDPW_MSG_SIZE_UV nMSGLEN_1 // # of MRF's to hold U/V block data (1)
+
+// For masking
+#undef nDPR_MSG_SIZE_Y
+#define nDPR_MSG_SIZE_Y nRESLEN_4 // # of MRF's to hold Y block data (4)
+#undef nDPR_MSG_SIZE_UV
+#define nDPR_MSG_SIZE_UV nRESLEN_1 // # of MRF's to hold U/V block data (1)
+#define rMASK_TEMP REG(r,nTEMP0)
+.declare uwMASK_TEMP Base=rMASK_TEMP ElementSize=2 SrcRegion=<8;8,1> Type=uw //1 GRF
+
+#if (nSRC_REGION==nREGION_1)
+ // For saving
+ #define ub2DEST_Y ub2TOP_Y
+ #define ub2DEST_U ub2TOP_U
+ #define ub2DEST_V ub2TOP_V
+ //For masking operation
+ #define udSRC_Y udBOT_Y_IO
+ #define udSRC_U udBOT_U_IO
+ #define udSRC_V udBOT_V_IO
+ #define ubSRC_Y ubBOT_Y
+ #define ubSRC_U ubBOT_U
+ #define ubSRC_V ubBOT_V
+ #define uwSRC_U uwBOT_U
+ #define uwSRC_V uwBOT_V
+
+#elif (nSRC_REGION==nREGION_2)
+ // For saving
+ #define ub2DEST_Y ub2BOT_Y
+ #define ub2DEST_U ub2BOT_U
+ #define ub2DEST_V ub2BOT_V
+ //For masking operation
+ #define udSRC_Y udTOP_Y_IO
+ #define udSRC_U udTOP_U_IO
+ #define udSRC_V udTOP_V_IO
+ #define ubSRC_Y ubTOP_Y
+ #define ubSRC_U ubTOP_U
+ #define ubSRC_V ubTOP_V
+ #define uwSRC_U uwTOP_U
+ #define uwSRC_V uwTOP_V
+
+#endif
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+
+// Module name: PL8x4_Save_NV12.asm
+//
+// Save entire current planar frame data block of size 16x8
+//---------------------------------------------------------------
+// Symbols needed to be defined before including this module
+//
+// DWORD_ALIGNED_DEST: only if DEST_Y, DEST_U, DEST_V data are DWord aligned
+// ORIX:
+//---------------------------------------------------------------
+
+#include "PL8x4_Save_NV12.inc"
+
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+#if !defined(SAVE_UV_ONLY)
+// Save current planar frame Y block data (16x8) -------------------------------
+
+ mov (2) mMSGHDR.0<1>:d wORIX<2;2,1>:w // Block origin
+ mov (1) mMSGHDR.2<1>:ud nDPW_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+#endif
+
+//Use the mask to determine which pixels shouldn't be over-written
+ and (1) acc0<1>:ud udBLOCK_MASK<0;1,0>:ud 0x00FFFFFF:ud
+ cmp.ge.f0.0 (1) dNULLREG acc0<0;1,0>:ud 0x00FFFFFF:ud //Check if all pixels in the block need to be modified
+ (f0.0) jmpi WritePlanarToDataPort
+
+//If mask is not all 1's, then load the entire 16x8 block
+//so that only those bytes may be modified that need to be (using the mask)
+ send (8) udSRC_Y(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_Y+nBI_DESTINATION_Y:ud //16x8
+
+ asr (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w 1:w { NoDDClr } // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_UV:ud { NoDDChk } // Block width and height (16x4)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud //move message desrcptor to the message header
+ send (8) udSRC_U(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_UV+nBI_DESTINATION_UV:ud
+
+//Restore the origin information
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // Block origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud //move message desrcptor to the message header
+
+//Merge the data
+ mov (1) f0.1:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.1) mov (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw
+ (-f0.1) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+//convert the mask from 16bits to 8bits by selecting every other bit
+ mov (1) udMASK_TEMP1(0,0)<1> 0x00040001:ud
+ mov (1) udMASK_TEMP1(0,1)<1> 0x00400010:ud
+ mov (1) udMASK_TEMP1(0,2)<1> 0x04000100:ud
+ mov (1) udMASK_TEMP1(0,3)<1> 0x40001000:ud
+
+//merge the loaded block with the current block
+ $for(0,0; <nY_NUM_OF_ROWS; 2,1) {
+ mov (1) f0.1:uw uwMASK_TEMP(0, %1)<0;1,0>
+ (-f0.1) mov (16) ubDEST_Y(0,%1*32)<2> ubSRC_Y(0,%1*16)
+
+ and.nz.f0.1 (8) wNULLREG uwMASK_TEMP(0,%1)<0;1,0> uwMASK_TEMP1(0,0) //change the mask by selecting every other bit
+ (-f0.1) mov (8) ubDEST_U(0, %2*16)<2> ub2SRC_U(0, %1*8)<16;8,2>
+ (-f0.1) mov (8) ubDEST_V(0, %2*16)<2> ub2SRC_U(0, %1*8+1)<16;8,2>
+
+ mov (1) f0.1:uw uwMASK_TEMP(0,1+%1)<0;1,0>
+ (-f0.1) mov (16) ubDEST_Y(0, (1+%1)*32)<2> ubSRC_Y(0, (1+%1)*16)
+
+ }
+
+WritePlanarToDataPort:
+#if !defined(SAVE_UV_ONLY)
+ $for(0,0; <nY_NUM_OF_ROWS; 2,1) {
+ mov (16) mubMSGPAYLOAD(%2,0)<1> ub2DEST_Y(%1)REGION(16,2)
+ mov (16) mubMSGPAYLOAD(%2,16)<1> ub2DEST_Y(%1+1)REGION(16,2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_Y+nBI_DESTINATION_Y:ud
+#endif
+
+//** Save 8x4 packed U and V -----------------------------------------------------
+// we could write directly wORIX to mMSGHDR and then execute asr on it, that way we could
+// avoid using rMSGSRC as a buffer and have one command less in code, but it is unknown whether
+//it is possible to do asr on mMSGHDR so we use rMSGSRC.
+ mov (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w // Block origin
+ asr (1) rMSGSRC.1<1>:d rMSGSRC.1<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_UV:ud // U/V block width and height (16x4)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ $for(0,0; <nY_NUM_OF_ROWS;4,1) {
+ mov (16) mubMSGPAYLOAD(%2,0)<2> ub2DEST_U(%2)REGION(16,2)
+ mov (16) mubMSGPAYLOAD(%2,1)<2> ub2DEST_V(%2)REGION(16,2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_UV+nBI_DESTINATION_UV:ud
+
+// End of PL8x4_Save_NV12
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//Module name: PL8x4_Save_NV12.inc
+//
+// Setup for storing planar data
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+#undef nDPW_BLOCK_SIZE_Y
+#undef nDPW_MSG_SIZE_Y
+#undef nDPW_BLOCK_SIZE_UV
+#undef nDPW_MSG_SIZE_UV
+
+#define nDPW_BLOCK_SIZE_Y nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // Y block size 16x8
+#define nDPW_MSG_SIZE_Y nMSGLEN_4 // # of MRF's to hold Y block data (4)
+#define nDPW_BLOCK_SIZE_UV nBLOCK_WIDTH_16+nBLOCK_HEIGHT_4 // U/V interleaved block width and height (16x4)
+#define nDPW_MSG_SIZE_UV nMSGLEN_2 // # of MRF's to hold U/V block data (2)
+
+// For masking
+#undef nDPR_MSG_SIZE_Y
+#define nDPR_MSG_SIZE_Y nRESLEN_4 // # of MRF's to hold Y block data (4)
+#undef nDPR_MSG_SIZE_UV
+#define nDPR_MSG_SIZE_UV nRESLEN_2
+#define rMASK_TEMP REG(r,nTEMP0)
+.declare uwMASK_TEMP Base=rMASK_TEMP ElementSize=2 SrcRegion=<8;8,1> Type=uw //1 GRF
+#define rMASK_TEMP1 REG(r,nTEMP1)
+.declare udMASK_TEMP1 Base=rMASK_TEMP1 ElementSize=4 SrcRegion=<4;4,1> Type=ud //1 GRF
+.declare uwMASK_TEMP1 Base=rMASK_TEMP1 ElementSize=2 SrcRegion=<8;8,1> Type=uw //1 GRF
+
+
+#if (nSRC_REGION==nREGION_1)
+ #define udSRC_Y udBOT_Y_IO
+ #define udSRC_U udBOT_U_IO
+ #define udSRC_V udBOT_V_IO
+ #define ubSRC_Y ubBOT_Y
+ #define ubSRC_U ubBOT_U
+ #define ubSRC_V ubBOT_V
+
+ #define uwSRC_U uwBOT_U //For masking operation
+ #define uwSRC_V uwBOT_V
+
+ #define ub2DEST_Y ub2TOP_Y
+ #define ub2DEST_U ub2TOP_U
+ #define ub2DEST_V ub2TOP_V
+
+ #define ubDEST_Y ubTOP_Y
+ #define ubDEST_U ubTOP_U
+ #define ubDEST_V ubTOP_V
+
+ #define ub2SRC_U ub2BOT_U
+#elif (nSRC_REGION==nREGION_2)
+ #define udSRC_Y udTOP_Y_IO
+ #define udSRC_U udTOP_U_IO
+ #define udSRC_V udTOP_V_IO
+ #define ubSRC_Y ubTOP_Y
+ #define ubSRC_U ubTOP_U
+ #define ubSRC_V ubTOP_V
+
+ #define uwSRC_U uwTOP_U //For masking operation
+ #define uwSRC_V uwTOP_V
+
+ #define ub2DEST_Y ub2BOT_Y
+ #define ub2DEST_U ub2BOT_U
+ #define ub2DEST_V ub2BOT_V
+
+ #define ubDEST_Y ubBOT_Y
+ #define ubDEST_U ubBOT_U
+ #define ubDEST_V ubBOT_V
+
+ #define ub2SRC_U ub2TOP_U
+#endif
+
+///* Yoni - masking is not relevant for ILK?!?
+//#define TEMP0 REG(r,54)
+//.declare TEMP Base=TEMP0 ElementSize=2 SrcRegion=<8;8,1> Type=uw
+///* Yoni - masking is not relevant for ILK?!?
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL8x5_PL8x8.asm
+
+#include "Expansion.inc"
+
+//------------------------------- Vertical Upconversion ------------------------------
+ avg.sat (8) uwDEST_U(0, 3*16+8)<1> uwDEST_U(0, 3*8) uwDEST_U(0, (1+3)*8) // Optimization
+ avg.sat (8) uwDEST_V(0, 3*16+8)<1> uwDEST_V(0, 3*8) uwDEST_V(0, (1+3)*8) // Optimization
+
+ $for(nUV_NUM_OF_ROWS/2-2; >-1; -1) {
+ mov (8) uwDEST_U(0, (1+%1)*16)<1> uwDEST_U(0, (1+%1)*8)
+ avg.sat (8) uwDEST_U(0, %1*16+8)<1> uwDEST_U(0, %1*8) uwDEST_U(0, (1+%1)*8)
+
+ mov (8) uwDEST_V(0, (1+%1)*16)<1> uwDEST_V(0, (1+%1)*8)
+ avg.sat (8) uwDEST_V(0, %1*16+8)<1> uwDEST_V(0, %1*8) uwDEST_V(0, (1+%1)*8)
+ }
+
+// End of PL8x5_PL8x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL8x8_PL8x4.asm
+//
+// Convert PL 8x8 to PL8x4 in GRF
+//---------------------------------------------------------------
+// Symbols needed to be defined before including this module
+//
+// DWORD_ALIGNED_DEST: only if DEST_Y, DEST_U, DEST_V data are DWord aligned
+// ORIX:
+//---------------------------------------------------------------
+
+#include "PL8x8_PL8x4.inc"
+
+// Convert PL8x8 to PL8x4 ---------------------------------------------------------
+
+ mov (8) ubDEST_U(0,16)<2> ubDEST_U(1)<16;8,2> //selecting U every other row
+ mov (16) ubDEST_U(0,32)<2> ubDEST_U(2)<32;8,2> //selecting U every other row
+ mov (8) ubDEST_V(0,16)<2> ubDEST_V(1)<16;8,2> //selecting V every other row
+ mov (16) ubDEST_V(0,32)<2> ubDEST_V(2)<32;8,2> //selecting V every other row
+
+// End of PL8x8_PL8x4.asm -------------------------------------------------------
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL8x8_PL8x4.inc
+//
+// Setup module for convert PL8x8 to PL8x4
+//
+//
+
+// Source/destination region definitions
+//
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+#if (nSRC_REGION==nREGION_1)
+
+ //REGION_1 selected
+ #define ubDEST_Y ubTOP_Y
+ #define ubDEST_U ubTOP_U
+ #define ubDEST_V ubTOP_V
+
+#elif (nSRC_REGION==nREGION_2)
+
+ //REGION_2 selected
+ #define ubDEST_Y ubBOT_Y
+ #define ubDEST_U ubBOT_U
+ #define ubDEST_V ubBOT_V
+
+
+#endif
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+
+// Module name: PL8x8_Save_P208.asm
+//
+// Save entire current planar frame data block of size 16x8
+//---------------------------------------------------------------
+// Symbols needed to be defined before including this module
+//
+// DWORD_ALIGNED_DEST: only if DEST_Y, DEST_U, DEST_V data are DWord aligned
+// ORIX:
+//---------------------------------------------------------------
+
+#include "PL8x8_Save_P208.inc"
+
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+#if !defined(SAVE_UV_ONLY)
+// Save current planar frame Y block data (16x8) -------------------------------
+
+ mov (2) mMSGHDR.0<1>:d wORIX<2;2,1>:w // Block origin
+ mov (1) mMSGHDR.2<1>:ud nDPW_BLOCK_SIZE_Y:ud // Block width and height (16x8)
+
+WritePlanarToDataPort:
+ $for(0,0; <nY_NUM_OF_ROWS; 2,1) {
+ mov (16) mubMSGPAYLOAD(%2,0)<1> ub2DEST_Y(%1)REGION(16,2)
+ mov (16) mubMSGPAYLOAD(%2,16)<1> ub2DEST_Y(%1+1)REGION(16,2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_Y+nBI_DESTINATION_Y:ud
+#endif
+
+//** Save 8x8 packed U and V -----------------------------------------------------
+// we could write directly wORIX to mMSGHDR and then execute asr on it, that way we could
+// avoid using rMSGSRC as a buffer and have one command less in code, but it is unknown whether
+//it is possible to do asr on mMSGHDR so we use rMSGSRC.
+ mov (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w // Block origin
+
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_UV:ud // U/V block width and height (16x4)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ $for(0,0; <nY_NUM_OF_ROWS;2,1) {
+ mov (16) mubMSGPAYLOAD(%2,0)<2> ub2DEST_U(%2)REGION(16,2)
+ mov (16) mubMSGPAYLOAD(%2,1)<2> ub2DEST_V(%2)REGION(16,2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_UV+nBI_DESTINATION_UV:ud
+
+//End of PL8x8_Save_P208.asm
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+
+//Module name: PL8x8_Save_P208.inc
+//
+// Setup for storing planar data
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+#define nDPW_BLOCK_SIZE_Y nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // Y block size 16x8
+#define nDPW_MSG_SIZE_Y nMSGLEN_4 // # of MRF's to hold Y block data (4)
+#define nDPW_BLOCK_SIZE_UV nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // U/V interleaved block width and height (16x8)
+#define nDPW_MSG_SIZE_UV nMSGLEN_4 // # of MRF's to hold U/V block data (4)
+
+#if (nSRC_REGION==nREGION_1)
+ #define udSRC_Y udBOT_Y_IO
+ #define udSRC_U udBOT_U_IO
+ #define udSRC_V udBOT_V_IO
+ #define ubSRC_Y ubBOT_Y
+ #define ubSRC_U ubBOT_U
+ #define ubSRC_V ubBOT_V
+
+ #define uwSRC_U uwBOT_U //For masking operation
+ #define uwSRC_V uwBOT_V
+
+ #define ub2DEST_Y ub2TOP_Y
+ #define ub2DEST_U ub2TOP_U
+ #define ub2DEST_V ub2TOP_V
+
+#elif (nSRC_REGION==nREGION_2)
+ #define udSRC_Y udTOP_Y_IO
+ #define udSRC_U udTOP_U_IO
+ #define udSRC_V udTOP_V_IO
+ #define ubSRC_Y ubTOP_Y
+ #define ubSRC_U ubTOP_U
+ #define ubSRC_V ubTOP_V
+
+ #define uwSRC_U uwTOP_U //For masking operation
+ #define uwSRC_V uwTOP_V
+
+ #define ub2DEST_Y ub2BOT_Y
+ #define ub2DEST_U ub2BOT_U
+ #define ub2DEST_V ub2BOT_V
+
+#endif
+
+///* Yoni - masking is not relevant for ILK?!?
+//#define TEMP0 REG(r,54)
+//.declare TEMP Base=TEMP0 ElementSize=2 SrcRegion=<8;8,1> Type=uw
+///* Yoni - masking is not relevant for ILK?!?
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL8x8_Save_PA.asm
+//
+// Save planar YUV422 to packed YUV422 format data
+//
+// Note: SRC_* must reference to regions with data type "BYTE"
+// in order to save to byte-aligned byte location
+
+#include "PL8x8_Save_PA.inc"
+
+ add (4) pCF_Y_OFFSET<1>:uw ubDEST_CF_OFFSET<4;4,1>:ub nDEST_YUV_REG*nGRFWIB:w // Initial Y,U,V offset in YUV422 block
+
+ // Pack Y
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) r[pCF_Y_OFFSET, %1*nGRFWIB]<2> ubSRC_Y(0,%1*32)
+ }
+
+ // Pack U/V
+ $for(0; <nUV_NUM_OF_ROWS; 1) {
+ mov (8) r[pCF_U_OFFSET, %1*nGRFWIB]<4> ubSRC_U(0, %1*16)
+ mov (8) r[pCF_V_OFFSET, %1*nGRFWIB]<4> ubSRC_V(0, %1*16)
+ }
+
+ shl (1) rMSGSRC.0<1>:d wORIX<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+ mov (1) rMSGSRC.1<1>:d wORIY<0;1,0>:w { NoDDClr, NoDDChk } // Block origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_YUV:ud { NoDDChk } // Block width and height (32x8)
+
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+//Use the mask to determine which pixels shouldn't be over-written
+ and (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0x00FFFFFF:ud
+ cmp.ge.f0.0(1) dNULLREG acc0.0<0;1,0>:ud 0x00FFFFFF:ud //Check if all pixels in the block need to be modified
+ (f0.0) jmpi WritePackedToDataPort
+
+ //If mask is not all 1's, then load the entire 32x8 block
+ //so that only those bytes may be modified that need to be (using the mask)
+
+ // Load 32x8 packed YUV 422 ----------------------------------------------------
+ send (8) udSRC_YUV(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_YUV+nBI_DESTINATION_YUV:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ //Merge the data
+ mov (1) f0.0:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.0) mov (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw
+ (-f0.0) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+ // Destination is Byte aligned
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (1) f0.1:uw uwMASK_TEMP(0,%1)<0;1,0>
+ (-f0.1) mov (16) uwDEST_YUV(%1)<1> uwSRC_YUV(%1) //check the UV merge - vK
+ }
+
+WritePackedToDataPort:
+ // Packed YUV data are stored in one of the I/O regions before moving to MRF
+ // Note: This is necessary since indirect addressing is not supported for MRF.
+ // Packed data block should be saved as 32x8 pixel block
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_YUV(%1)REGION(8,1)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_YUV+nBI_DESTINATION_YUV:ud
+
+// End of PL8x8_Save_PA
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL8x8_Save_PA.inc
+//
+// Setup for storing packed data
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+// For saving
+#define nDPW_BLOCK_SIZE_YUV nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // YUV block size 32x8
+#define nDPW_MSG_SIZE_YUV nMSGLEN_8 // # of MRF's to hold YUV block data (8)
+
+// For masking
+#undef nDPR_MSG_SIZE_YUV
+#define nDPR_MSG_SIZE_YUV nRESLEN_8 // # of MRF's to hold YUV block data (8)
+#define rMASK_TEMP REG(r,nTEMP0)
+.declare uwMASK_TEMP Base=rMASK_TEMP ElementSize=2 SrcRegion=<8;8,1> Type=uw //1 GRF
+
+#if (nSRC_REGION==nREGION_1)
+ // For saving
+ #define udSRC_YUV udTOP_Y_IO
+ #define udDEST_YUV udBOT_Y_IO
+ #define nDEST_YUV_REG nBOT_Y
+ //For masking operation
+ #define ubSRC_Y ub2TOP_Y
+ #define ubSRC_U ub2TOP_U
+ #define ubSRC_V ub2TOP_V
+ #define uwSRC_YUV uwTOP_Y
+ #define uwDEST_YUV uwBOT_Y
+
+#elif (nSRC_REGION==nREGION_2)
+ // For saving
+ #define udSRC_YUV udBOT_Y_IO
+ #define udDEST_YUV udTOP_Y_IO
+ #define nDEST_YUV_REG nTOP_Y
+ //For masking operation
+ #define ubSRC_Y ub2BOT_Y
+ #define ubSRC_U ub2BOT_U
+ #define ubSRC_V ub2BOT_V
+ #define uwSRC_YUV uwBOT_Y
+ #define uwDEST_YUV uwTOP_Y
+
+#endif
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL9x5_PL16x8.asm
+
+#define EXPAND_9x5
+#include "Expansion.inc"
+
+//------------------------------ Horizontal Upconversion -----------------------------
+ $for (nUV_NUM_OF_ROWS-2; >-1; -1) {
+ avg.sat (16) uwDEST_U(0, %1*16)<1> uwDEST_U(0, %1*16)<1;2,0> uwDEST_U(0, %1*16)<1;2,1>
+ avg.sat (16) uwDEST_V(0, %1*16)<1> uwDEST_V(0, %1*16)<1;2,0> uwDEST_V(0, %1*16)<1;2,1>
+ }
+
+#undef nUV_NUM_OF_ROWS
+#define nUV_NUM_OF_ROWS 8 //use packed version of all post-processing kernels
+
+//------------------------------- Vertical Upconversion ------------------------------
+ avg.sat (16) uwDEST_U(0, 3*32+16)<1> uwDEST_U(0, 3*16) uwDEST_U(0, (1+3)*16)
+ avg.sat (16) uwDEST_V(0, 3*32+16)<1> uwDEST_V(0, 3*16) uwDEST_V(0, (1+3)*16)
+
+ $for(nUV_NUM_OF_ROWS/2-2; >-1; -1) {
+ mov (16) uwDEST_U(0, (1+%1)*32)<1> uwDEST_U(0, (1+%1)*16)
+ avg.sat (16) uwDEST_U(0, %1*32+16)<1> uwDEST_U(0, %1*16) uwDEST_U(0, (1+%1)*16)
+
+ mov (16) uwDEST_V(0, (1+%1)*32)<1> uwDEST_V(0, (1+%1)*16)
+ avg.sat (16) uwDEST_V(0, %1*32+16)<1> uwDEST_V(0, %1*16) uwDEST_V(0, (1+%1)*16)
+ }
+
+// End of PL9x5_PL16x8
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: PL9x5_PL16x8.asm
+
+#include "Expansion.inc"
+
+//------------------------------ Horizontal Upconversion -----------------------------
+ $for (0; <nUV_NUM_OF_ROWS; 1) {
+ avg.sat (16) uwDEST_U(0, %1*16)<1> uwDEST_U(0, %1*16)<1;2,0> uwDEST_U(0, %1*16)<1;2,1>
+ avg.sat (16) uwDEST_V(0, %1*16)<1> uwDEST_V(0, %1*16)<1;2,0> uwDEST_V(0, %1*16)<1;2,1>
+ }
+
+// End of PL9x5_PL16x8
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: RGB16x8_Save_RGB.asm
+//
+// Save packed ARGB 444 frame data block of size 16x8
+//
+// To save 16x8 block (64x8 byte layout for ARGB8888) we need 2 send instructions
+// ---------
+// | 1 | 2 |
+// ---------
+
+#include "RGB16x8_Save_RGB.inc"
+
+ shl (1) rMSGSRC.0<1>:d wORIX<0;1,0>:w 2:w { NoDDClr } // H. block origin need to be quadrupled
+ mov (1) rMSGSRC.1<1>:d wORIY<0;1,0>:w { NoDDClr, NoDDChk } // Block origin (1st quadrant)
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_ARGB:ud { NoDDChk } // Block width and height (32x8)
+
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+//Use the mask to determine which pixels shouldn't be over-written
+ and (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0x00FFFFFF:ud
+ cmp.ge.f0.0(1) dNULLREG acc0.0<0;1,0>:ud 0x00FFFFFF:ud //Check if all pixels in the block need to be modified
+ (f0.0) jmpi WriteARGBToDataPort
+
+ //If mask is not all 1's, then load the entire 64x8 block
+ //so that only those bytes may be modified that need to be (using the mask)
+
+ // Load first block 16x8 packed ARGB 444 ---------------------------------------
+ or (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0xFF00FF00:ud //Check first block
+ cmp.e.f0.0 (1) dNULLREG acc0.0<0;1,0>:ud 0xFFFFFFFF:ud
+ (f0.0) jmpi SkipFirstBlockMerge //If full mask then skip this block
+
+ send (8) udSRC_ARGB(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ //Merge the data
+ mov (1) f0.0:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.0) mov (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw //use sel instruction - vK
+ (-f0.0) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+ $for(0, 0; <nY_NUM_OF_ROWS; 1, 2) { //take care of the lines in the block, they are different in the src and dest
+ mov (1) f0.1:uw uwMASK_TEMP(0,%1)<0;1,0>
+ (-f0.1) mov (8) udDEST_ARGB(%2)<1> udSRC_ARGB(%1)
+ }
+
+SkipFirstBlockMerge:
+ // Load second block 16x8 packed ARGB 444 ---------------------------------------
+ or (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0xFF0000FF:ud //Check second block
+ cmp.e.f0.0 (1) dNULLREG acc0.0<0;1,0>:ud 0xFFFFFFFF:ud
+ (f0.0) jmpi WriteARGBToDataPort //If full mask then skip this block
+
+ add (1) mMSGHDR.0<1>:d rMSGSRC.0<0;1,0>:d 32:d // Point to 2nd part
+ send (8) udSRC_ARGB(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud // Point to 1st part again
+
+ //Merge the data
+ mov (1) f0.0:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.0) shr (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw 8:uw //load the mask for second block
+ (-f0.0) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+ $for(0, 1; <nY_NUM_OF_ROWS; 1, 2) { //take care of the lines in the block, they are different in the src and dest
+ mov (1) f0.1:uw uwMASK_TEMP(0,%1)<0;1,0>
+ (-f0.1) mov (8) udDEST_ARGB(%2)<1> udSRC_ARGB(%1)
+ }
+
+WriteARGBToDataPort:
+ // Move packed data to MRF and output
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_ARGB(%1*2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ add (1) mMSGHDR.0<1>:d rMSGSRC.0<0;1,0>:d 32:d // Point to 2nd part
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_ARGB(%1*2+1)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+
+// End of RGB16x8_Save_RGB
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: RGB16x8_Save_RGB.inc
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+// For saving
+#define nDPW_BLOCK_SIZE_ARGB nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // ARGB block size 32x8
+#define nDPW_MSG_SIZE_ARGB nMSGLEN_8 // # of MRF's to hold ARGB block data (8)
+
+// For masking
+#undef nDPR_MSG_SIZE_ARGB
+#define nDPR_MSG_SIZE_ARGB nRESLEN_8 // # of MRF's to hold ARGB block data (8)
+#define rMASK_TEMP REG(r,nTEMP0)
+.declare uwMASK_TEMP Base=rMASK_TEMP ElementSize=2 SrcRegion=<8;8,1> Type=uw //1 GRF
+
+#if (nSRC_REGION==nREGION_1)
+ // For saving
+ #define udDEST_ARGB udTOP_Y_IO //The output of previous stage is stored here; This is modified and is written to render cache
+ //For masking operation
+ #define udSRC_ARGB udBOT_Y_IO //To hold the destination data that shouldn't be modified
+
+#elif (nSRC_REGION==nREGION_2)
+ // For saving
+ #define udDEST_ARGB udBOT_Y_IO //The output of previous stage is stored here; This is modified and is written to render cache
+ //For masking operation
+ #define udSRC_ARGB udTOP_Y_IO //To hold the destination data that shouldn't be modified
+
+#endif
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: RGB16x8_Save_RGB16.asm
+//
+// Save packed RGB565 frame data block of size 16x8
+//
+// To save 16x8 block (32x8 byte layout for RGB565) we need 1 send instruction
+// -----
+// | 1 |
+// -----
+
+#include "RGB16x8_Save_RGB16.inc"
+
+//convert 32 bit RGB to 16 bit RGB
+ // Truncate A8R8G8B8 to A6R5G6B5 within byte.
+ // That is keeping 5 MSB of R and B, and 6 MSB of G.
+
+ $for (0, 0; <nY_NUM_OF_ROWS; 1, 2) {
+ shr uwCSC_TEMP(%1,0)<1> ubDEST_ARGB(%2,0)<32;8,4> 3:w // B >> 3
+
+ shl (16) uwTEMP_RGB16(0)<1> uwDEST_ARGB(%2,1)<16;8,2> 8:w // R << 8
+ and (16) uwTEMP_RGB16(0)<1> uwTEMP_RGB16(0) 0xF800:uw
+ or (16) uwCSC_TEMP(%1,0)<1> uwCSC_TEMP(%1,0)<16;16,1> uwTEMP_RGB16(0)
+
+ shr (16) uwTEMP_RGB16(0)<1> uwDEST_ARGB(%2,0)<16;8,2> 5:w // G >> 5
+ and (16) uwTEMP_RGB16(0)<1> uwTEMP_RGB16(0) 0x07E0:uw
+ or (16) uwCSC_TEMP(%1,0)<1> uwCSC_TEMP(%1,0)<16;16,1> uwTEMP_RGB16(0)
+ }
+
+ mov (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w // Block origin (1st quadrant)
+ shl (1) rMSGSRC.0<1>:d wORIX<0;1,0>:w 1:w // H. block origin need to be doubled for byte offset
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_RGB16:ud // Block width and height (32x8)
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+//Use the mask to determine which pixels shouldn't be over-written
+ and (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0x00FFFFFF:ud
+ cmp.ge.f0.0(1) dNULLREG acc0.0<0;1,0>:ud 0x00FFFFFF:ud //Check if all pixels in the block need to be modified
+ (f0.0) jmpi WriteRGB16ToDataPort
+
+ //If mask is not all 1's, then load the entire 32x8 block
+ //so that only those bytes may be modified that need to be (using the mask)
+
+ // Load 32x8 packed RGB565 -----------------------------------------------------
+ send (8) udSRC_RGB16(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_RGB16+nBI_DESTINATION_RGB:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ //Merge the data
+ mov (1) f0.0:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.0) mov (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw //use sel instruction - vK
+ (-f0.0) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (1) f0.1:uw uwMASK_TEMP(0,%1)<0;1,0>
+ (-f0.1) mov (16) uwCSC_TEMP(%1)<1> uwSRC_RGB16(%1)
+ }
+
+WriteRGB16ToDataPort:
+ // Move packed data to MRF and output
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udCSC_TEMP(%1)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_RGB16+nBI_DESTINATION_RGB:ud
+
+// End of RGB16x8_Save_RGB16
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: RGB16x8_Save_RGB16.inc
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+// For saving
+#define nDPW_BLOCK_SIZE_RGB16 nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // RGB16 block size 32x8
+#define nDPW_MSG_SIZE_RGB16 nMSGLEN_8 // # of MRF's to hold RGB16 block data (8)
+
+// For conversion to 16bit
+.declare uwTEMP_RGB16 Base=REG(r,nTEMP1) ElementSize=2 SrcRegion=<16;16,1> Type=uw //1 GRF
+
+// For masking
+#undef nDPR_MSG_SIZE_RGB16
+#define nDPR_MSG_SIZE_RGB16 nRESLEN_8 // # of MRF's to hold ARGB block data (8)
+#define rMASK_TEMP REG(r,nTEMP0)
+.declare uwMASK_TEMP Base=rMASK_TEMP ElementSize=2 SrcRegion=<8;8,1> Type=uw //1 GRF
+
+#if (nSRC_REGION==nREGION_1)
+ // For saving
+ #define ubDEST_ARGB ubTOP_Y //Data from previous module
+ #define uwDEST_ARGB uwTOP_Y //Data from previous module
+ #define udCSC_TEMP udBOT_Y_IO //Data Converted to 16 bits
+ #define uwCSC_TEMP uwBOT_Y
+ //For masking operation
+ #define udSRC_RGB16 udTOP_Y_IO //To hold the destination data that shouldn't be modified
+ #define uwSRC_RGB16 uwTOP_Y
+
+#elif (nSRC_REGION==nREGION_2)
+ // For saving
+ #define ubDEST_ARGB ubBOT_Y //Data from previous module
+ #define uwDEST_ARGB uwBOT_Y //Data from previous module
+ #define udCSC_TEMP udTOP_Y_IO //Data Converted to 16 bits
+ #define uwCSC_TEMP uwTOP_Y
+ //For masking operation
+ #define udSRC_RGB16 udBOT_Y_IO //To hold the destination data that shouldn't be modified
+ #define uwSRC_RGB16 uwBOT_Y
+
+#endif
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: RGB16x8_Save_Y416.asm
+//
+// Save packed ARGB 444 frame data block of size 16x8
+//
+// To save 16x8 block (128x8 byte layout for ARGB 16bit per component) we need 4 send instructions
+// -----------------
+// | 1 | 2 | 3 | 4 |
+// -----------------
+
+#include "RGB16x8_Save_RGB.inc"
+
+ shl (1) rMSGSRC.0<1>:d wORIX<0;1,0>:w 3:w { NoDDClr } // H. block origin need to become 8 times
+ mov (1) rMSGSRC.1<1>:d wORIY<0;1,0>:w { NoDDClr, NoDDChk } // Block origin (1st quadrant)
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_ARGB:ud { NoDDChk } // Block width and height (32x8)
+
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+/* Not needed for validation kernels for now -vK
+//Use the mask to determine which pixels shouldn't be over-written
+ and (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0x00FFFFFF:ud
+ cmp.ge.f0.0(1) dNULLREG acc0.0<0;1,0>:ud 0x00FFFFFF:ud //Check if all pixels in the block need to be modified
+ (f0.0) jmpi WriteARGBToDataPort
+
+ //If mask is not all 1's, then load the entire 64x8 block
+ //so that only those bytes may be modified that need to be (using the mask)
+
+ // Load first block 16x8 packed ARGB 444 ---------------------------------------
+ or (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0xFF00FF00:ud //Check first block
+ cmp.e.f0.0 (1) dNULLREG acc0.0<0;1,0>:ud 0xFFFFFFFF:ud
+ (f0.0) jmpi SkipFirstBlockMerge //If full mask then skip this block
+
+ send (8) udSRC_ARGB(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ //Merge the data
+ mov (1) f0.0:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.0) mov (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw //use sel instruction - vK
+ (-f0.0) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+ $for(0, 0; <nY_NUM_OF_ROWS; 1, 2) { //take care of the lines in the block, they are different in the src and dest
+ mov (1) f0.1:uw uwMASK_TEMP(0,%1)<0;1,0>
+ (-f0.1) mov (8) udDEST_ARGB(%2)<1> udSRC_ARGB(%1)
+ }
+
+SkipFirstBlockMerge:
+ // Load second block 16x8 packed ARGB 444 ---------------------------------------
+ or (1) acc0.0<1>:ud udBLOCK_MASK<0;1,0>:ud 0xFF0000FF:ud //Check second block
+ cmp.e.f0.0 (1) dNULLREG acc0.0<0;1,0>:ud 0xFFFFFFFF:ud
+ (f0.0) jmpi WriteARGBToDataPort //If full mask then skip this block
+
+ add (1) mMSGHDR.0<1>:d rMSGSRC.0<0;1,0>:d 32:d // Point to 2nd part
+ send (8) udSRC_ARGB(0)<1> mMSGHDR udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nDPR_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud // Point to 1st part again
+
+ //Merge the data
+ mov (1) f0.0:uw ubBLOCK_MASK_V:ub //Load the mask on flag reg
+ (f0.0) shr (8) rMASK_TEMP<1>:uw uwBLOCK_MASK_H:uw 8:uw //load the mask for second block
+ (-f0.0) mov (8) rMASK_TEMP<1>:uw 0:uw
+
+ $for(0, 1; <nY_NUM_OF_ROWS; 1, 2) { //take care of the lines in the block, they are different in the src and dest
+ mov (1) f0.1:uw uwMASK_TEMP(0,%1)<0;1,0>
+ (-f0.1) mov (8) udDEST_ARGB(%2)<1> udSRC_ARGB(%1)
+ }
+*/
+WriteARGBToDataPort:
+ // Move packed data to MRF and output
+
+ //Write 1st 4X8 pixels
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_ARGB(%1*4)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+
+ //Write 2nd 4X8 pixels
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ add (1) mMSGHDR.0<1>:d rMSGSRC.0<0;1,0>:d 32:d // Point to 2nd part
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_ARGB(%1*4+1)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+
+ //Write 3rd 4X8 pixels
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ add (1) mMSGHDR.0<1>:d rMSGSRC.0<0;1,0>:d 64:d // Point to 2nd part
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_ARGB(%1*4+2)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+
+ //Write 4th 4X8 pixels
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ add (1) mMSGHDR.0<1>:d rMSGSRC.0<0;1,0>:d 96:d // Point to 2nd part
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_ARGB(%1*4+3)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_ARGB+nBI_DESTINATION_RGB:ud
+
+// End of RGB16x8_Save_Y416
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: RGB16x8_Save_Y416.inc
+//
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+// For saving
+#define nDPW_BLOCK_SIZE_ARGB nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // ARGB block size 32x8
+#define nDPW_MSG_SIZE_ARGB nMSGLEN_8 // # of MRF's to hold ARGB block data (8)
+
+// For masking
+#undef nDPR_MSG_SIZE_ARGB
+#define nDPR_MSG_SIZE_ARGB nRESLEN_8 // # of MRF's to hold ARGB block data (8)
+#define rMASK_TEMP REG(r,nTEMP0)
+.declare uwMASK_TEMP Base=rMASK_TEMP ElementSize=2 SrcRegion=<8;8,1> Type=uw //1 GRF
+
+#if (nSRC_REGION==nREGION_1)
+ // For saving
+ #define udDEST_ARGB udTOP_Y_IO //The output of previous stage is stored here; This is modified and is written to render cache
+ //For masking operation
+ #define udSRC_ARGB udBOT_Y_IO //To hold the destination data that shouldn't be modified
+
+#elif (nSRC_REGION==nREGION_2)
+ // For saving
+ #define udDEST_ARGB udBOT_Y_IO //The output of previous stage is stored here; This is modified and is written to render cache
+ //For masking operation
+ #define udSRC_ARGB udTOP_Y_IO //To hold the destination data that shouldn't be modified
+
+#endif
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+
+.declare SRC_B Base=REG(r,10) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare SRC_G Base=REG(r,18) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare SRC_R Base=REG(r,26) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare SRC_A Base=REG(r,34) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+
+#define DEST_ARGB ubBOT_ARGB
+
+#undef nSRC_REGION
+#define nSRC_REGION nREGION_2
+
+
+//Pack directly to mrf as optimization - vK
+
+$for(0, 0; <8; 1, 2) {
+// mov (16) DEST_ARGB(%2,0)<4> SRC_B(%1) { Compr, NoDDClr } // 16 B
+// mov (16) DEST_ARGB(%2,1)<4> SRC_G(%1) { Compr, NoDDClr, NoDDChk } // 16 G
+// mov (16) DEST_ARGB(%2,2)<4> SRC_R(%1) { Compr, NoDDClr, NoDDChk } // 16 R //these 2 inst can be merged - vK
+// mov (16) DEST_ARGB(%2,3)<4> SRC_A(%1) { Compr, NoDDChk } //DEST_RGB_FORMAT<0;1,0>:ub { Compr, NoDDChk } // 16 A
+
+ mov (8) DEST_ARGB(%2, 0)<4> SRC_B(%1) { NoDDClr } // 8 B
+ mov (8) DEST_ARGB(%2, 1)<4> SRC_G(%1) { NoDDClr, NoDDChk } // 8 G
+ mov (8) DEST_ARGB(%2, 2)<4> SRC_R(%1) { NoDDClr, NoDDChk } // 8 R
+ mov (8) DEST_ARGB(%2, 3)<4> SRC_A(%1) { NoDDChk } // 8 A
+
+ mov (8) DEST_ARGB(%2+1,0)<4> SRC_B(%1,8) { NoDDClr } // 8 B
+ mov (8) DEST_ARGB(%2+1,1)<4> SRC_G(%1,8) { NoDDClr, NoDDChk } // 8 G
+ mov (8) DEST_ARGB(%2+1,2)<4> SRC_R(%1,8) { NoDDClr, NoDDChk } // 8 R
+ mov (8) DEST_ARGB(%2+1,3)<4> SRC_A(%1,8) { NoDDChk } // 8 A
+}
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Modual name: SetupVPKernel.asm
+//
+// Initial setup for running video-processing kernels
+//
+
+#include "common.inc"
+
+//
+// Now, begin source code....
+//
+.code
+
+#include "Init_All_Regs.asm"
+
+mov (8) rMSGSRC.0<1>:ud r0.0<8;8,1>:ud // Initialize message payload header with R0
+#if defined (INC_BLENDING)
+ mul (1) fALPHA_STEP_X:f fSCALING_STEP_RATIO:f fVIDEO_STEP_X:f //StepX_ratio = AlphaStepX / VideoStepX
+#endif
+
+// End of SetupVPKernel
+
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#ifndef COMMON_INC
+#define COMMON_INC
+
+// Module name: common.inc
+//
+// Common header file for all Video-Processing kernels
+//
+
+.default_execution_size (16)
+.default_register_type :ub
+
+.reg_count_total 80
+.reg_count_payload 4
+
+
+//========== Common constants ==========
+
+// Bit position constants
+#define BIT0 0x01
+#define BIT1 0x02
+#define BIT2 0x04
+#define BIT3 0x08
+#define BIT4 0x10
+#define BIT5 0x20
+#define BIT6 0x40
+#define BIT7 0x80
+#define BIT8 0x0100
+#define BIT9 0x0200
+#define BIT10 0x0400
+#define BIT11 0x0800
+#define BIT12 0x1000
+#define BIT13 0x2000
+#define BIT14 0x4000
+#define BIT15 0x8000
+#define BIT16 0x00010000
+#define BIT17 0x00020000
+#define BIT18 0x00040000
+#define BIT19 0x00080000
+#define BIT20 0x00100000
+#define BIT21 0x00200000
+#define BIT22 0x00400000
+#define BIT23 0x00800000
+#define BIT24 0x01000000
+#define BIT25 0x02000000
+#define BIT26 0x04000000
+#define BIT27 0x08000000
+#define BIT28 0x10000000
+#define BIT29 0x20000000
+#define BIT30 0x40000000
+#define BIT31 0x80000000
+
+#define nGRFWIB 32 // GRF register width in byte
+#define nGRFWIW 16 // GRF register width in word
+#define nGRFWID 8 // GRF register width in dword
+
+#define nTOP_FIELD 0
+#define nBOTTOM_FIELD 1
+
+#define nPREVIOUS_FRAME 0 // Previous frame
+#define nCURRENT_FRAME 1 // Current frame
+#define nNEXT_FRAME 2 // Next frame
+
+#ifdef GT
+// GT DI Kernel
+#else // ILK
+// ILK DI Kernel
+#endif
+
+//===================================
+
+//========== Macros ==========
+#define REGION(Width,HStride) <Width*HStride;Width,HStride> // Region definition when ExecSize = Width
+
+#define RegFile(a) a
+#define REG(r,n) _REG(RegFile(r),n)
+#define _REG(r,n) __REG(r,n)
+#define __REG(r,n) r##n.0
+#define REG2(r,n,s) _REG2(RegFile(r),n,s)
+#define _REG2(r,n,s) __REG2(r,n,s)
+#define __REG2(r,n,s) r##n.##s
+
+#define dNULLREG null<1>:d
+#define wNULLREG null<1>:w
+
+#define KERNEL_ID(kernel_ID) mov NULLREG kernel_ID:ud
+
+
+#define NODDCLR
+#define NODDCLR_NODDCHK
+#define NODDCHK
+
+//#define NODDCLR { NoDDClr }
+//#define NODDCLR_NODDCHK { NoDDClr, NoDDChk }
+//#define NODDCHK { NoDDChk }
+
+
+//========== Defines ====================
+
+
+//========== GRF partition ==========
+// r0 header : r0 (1 GRF)
+// Static parameters : r1 - r5 (5 GRFS)
+// Inline parameters : r6 - r7 (2 GRFs)
+// MSGSRC : r9 (1 GRF)
+// Top IO region : r10 - r33 (24 GRFS 8 for each component Y,U,V 16X8:w)
+// Free space : r34 - r55 (22 GRFS)
+// Bottom IO region : r56 - r79 (24 GRFS 8 for each component Y,U,V 16X8:w)
+//===================================
+
+
+//========== Static Parameters ==========
+// r1
+#define fPROCAMP_C0 r1.0 // DWORD 0, Procamp constant C0 in :f
+#define wPROCAMP_C0 r1.0 // DWORD 0, Procamp constant C0 in :w
+#define NUMBER_0002 r1.1 // DWORD 0, 0x0002 used in procamp for GT
+#define udCP_MessageFormat r1.0 // DWORD 0, bits 2:3 of DWORD. (CE)
+#define udCP_StatePointer r1.0 // DWORD 0, bits 31:5 of DWORD.(CE)
+
+#define ubSRC_CF_OFFSET r1.4 // DWORD 1, byte 0-2. SRC packed color format YUV offset in :ub
+
+#define ubDEST_RGB_FORMAT r1.8 // DWORD 2, byte 0. Dest RGB color format (0:ARGB FF:XRGB)
+#define ubDEST_CF_OFFSET r1.8 // DWORD 2, byte 0-2. Dest packed color format YUV offset in :ub
+
+#define fPROCAMP_C1 r1.3 // DWORD 3, Procamp constant C1 in :f
+#define wPROCAMP_C1 r1.6 // DWORD 3, Procamp constant C1 in :w
+#define NUMBER_0100 r1.7 // DWORD 3, 0x0100 used in procamp for GT
+
+#define fPROCAMP_C2 r1.4 // DWORD 4, Procamp constant C2 in :f
+#define wPROCAMP_C2 r1.8 // DWORD 4, Procamp constant C2 in :w
+
+#define uwSPITCH_DIV2 r1.10 // DWORD 5, byte 0-1. statistics surface pitch divided by 2
+
+#define fVIDEO_STEP_Y r1.6 // DWORD 6, :f, AVS normalized reciprocal of Y Scaling factor
+#define ubSTMM_SHIFT r1.24 // DWORD 6, byte 0. Amount of right shift for the DI blending equation
+#define ubSTMM_MIN r1.25 // DWORD 6, byte 1. Min STMM for DI blending equation
+#define ubSTMM_MAX r1.26 // DWORD 6, byte 2. Max STMM for DI blending equation
+#define ubTFLD_FIRST r1.27 // DWORD 6, byte 3. Field parity order
+
+#define fPROCAMP_C5 r1.7 // DWORD 7, Procamp constant C3 in :f
+#define wPROCAMP_C5 r1.14 // DWORD 7, Procamp constant C3 in :w
+
+// r2
+#define fPROCAMP_C3 r2.0 // DWORD 0, Procamp constant C4 in :f
+#define wPROCAMP_C3 r2.0 // DWORD 0, Procamp constant C4 in :w
+
+#define fCSC_C5 r2.2 // DWORD 2. WG+CSC constant C5
+#define wCSC_C5 r2.4 // DWORD 2. WG+CSC constant C5
+
+#define fPROCAMP_C4 r2.3 // DWORD 3, Procamp constant C5 in :f
+#define wPROCAMP_C4 r2.6 // DWORD 3, Procamp constant C5 in :w
+
+#define fCSC_C8 r2.4 // DWORD 4. WG+CSC constant C8
+#define wCSC_C8 r2.8 // DWORD 4. WG+CSC constant C8
+#define fCSC_C9 r2.7 // DWORD 7. WG+CSC constant C9
+#define wCSC_C9 r2.14 // DWORD 7. WG+CSC constant C9
+
+// r3
+#define fCSC_C0 r3.0 // DWORD 0. WG+CSC constant C0
+#define wCSC_C0 r3.0 // DWORD 0. WG+CSC constant C0
+
+#define fSCALING_STEP_RATIO r3.1 // DWORD 1, = Alpha_X_Scaling_Step / Video_X_scaling_Step :f (blending)
+#define fALPHA_STEP_X r3.1 // DWORD 1, = 1/Scale X, 0.5 = 2x, in :f (blending)
+
+#define fALPHA_STEP_Y r3.2 // DWORD 2, = 1/Scale Y, in :f
+
+#define fCSC_C4 r3.3 // DWORD 3. WG+CSC constant C4
+#define wCSC_C4 r3.6 // DWORD 3. WG+CSC constant C4
+#define fCSC_C1 r3.4 // DWORD 4. WG+CSC constant C1
+#define wCSC_C1 r3.8 // DWORD 4. WG+CSC constant C1
+
+#define wSRC_H_ORI_OFFSET r3.10 // DWORD 5, bytes 0,1 :w
+#define wSRC_V_ORI_OFFSET r3.11 // DWORD 5, bytes 2,3 :w
+
+#define dCOLOR_PIXEL r3.6 // DWORD 6. Color pixel for Colorfill
+
+#define fCSC_C2 r3.6 // DWORD 6. WG+CSC constant C2
+#define wCSC_C2 r3.12 // DWORD 6. WG+CSC constant C2
+#define fCSC_C3 r3.7 // DWORD 7. WG+CSC constant C3
+#define wCSC_C3 r3.14 // DWORD 7. WG+CSC constant C3
+
+// r4
+#define fCSC_C6 r4.0 // DWORD 0. WG+CSC constant C6
+#define wCSC_C6 r4.0 // DWORD 0. WG+CSC constant C6
+
+#define wFRAME_ENDX r4.2 // DWORD 1, word 0. Horizontal end = Origin+Width (in pixels)(for multiple blocks)
+#define wNUM_BLKS r4.3 // DWORD 1, word 1. Number of blocks to process (for multiple blocks)
+
+#define wCOPY_ORIX r4.5 // DWORD 2, word 1. A copy of X origin (for multiple blocks)
+#define uwNLAS_ENABLE r4.4 // DWORD 2, bit 15, NLAS enble bit
+
+#define fCSC_C7 r4.3 // DWORD 3. WG+CSC constant C7
+#define wCSC_C7 r4.6 // DWORD 3. WG+CSC constant C7
+#define fCSC_C10 r4.4 // DWORD 4. WG+CSC constant C10
+#define wCSC_C10 r4.8 // DWORD 4. WG+CSC constant C10
+
+#define fFRAME_VID_ORIX r4.5 // DWORD 5, Frame horizontal origin normalized for scale kernel
+
+#define fFRAME_ALPHA_ORIX r4.6 // DWORD 6. Normalized alpha horiz origin for the frame
+
+#define fCSC_C11 r4.7 // DWORD 7. WG+CSC constant C11
+#define wCSC_C11 r4.14 // DWORD 7. WG+CSC constant C11
+
+//========================================
+
+//========== Inline parameters ===========
+// r5
+#define wORIX r5.0 // DWORD 0, byte 0-1. :w, Destination Block Horizontal Origin in pel
+#define wORIY r5.1 // DWORD 0, byte 2-3. :w, Destination Block Vertical Origin in pel
+
+#define fSRC_VID_H_ORI r5.1 // DWORD 1, :f, SRC Y horizontal origin normalized for scale kernel
+
+#define fSRC_VID_V_ORI r5.2 // DWORD 2, :f, SRC Y vertical origin normalized for scale kernel
+
+#define fSRC_ALPHA_H_ORI r5.3 // DWORD 3, :f, Normalized alpha horizontal origin
+
+#define fSRC_ALPHA_V_ORI r5.4 // DWORD 4, :f, Normalized alpha vertical origin
+
+#define uwALPHA_MASK_X r5.10 // DWORD 5, byte 0-1 :w, H. alpha mask
+#define ubALPHA_MASK_Y r5.22 // DWORD 5, byte 2. :ub,V. alpha mask
+#define ubBLK_CNT_X r5.23 // DWORD 5, byte 3, :ub, Horizontal Block Count per thread
+
+#define udBLOCK_MASK r5.6 // DWORD 6
+#define uwBLOCK_MASK_H r5.12 // DWORD 6, byte 0-1 :uw, Block horizontal mask used in non-DWord aligned kernels
+#define ubBLOCK_MASK_V r5.26 // DWORD 6, byte 2 :ub, Block vertical mask used in non-DWord aligned kernels
+#define ubNUM_BLKS r5.27 // DWORD 6, byte 3, :ub, Total Block Count per thread
+
+#define fVIDEO_STEP_X r5.7 // DWORD 7. :f, AVS normalized reciprocal of X Scaling factor
+
+// r6
+#define fVIDEO_STEP_DELTA r6.0 // DWORD 0. :f, AVS normalized delta between 2 adjacent scaling steps (used for non-linear scaling)
+
+
+//====================== Binding table =========================================
+
+#if defined(DNDI)
+ // DNDI Surface Binding Table
+ //#define nBI_SRC_CURR 0 // Current input frame surface
+ //#define nBI_SRC_PRIV 1 // Denoised previous input frame surface
+ //#define nBI_SRC_STAT 2 // Statistics input surface (STMM / Noise motion history)
+ //#define nBI_DEST_1ST 3 // 1st deinterlaced output frame surface
+// #define nBI_DEST_YUV 3 // Dest frame YUV (for DN only)
+ //#define nBI_DEST_Y 3 // Dest frame Y (for DN only)
+ //#define nBI_DEST_2ND 4 // 2nd deinterlaced output frame surface
+ //#define nBI_DEST_DN_CURR 6 // Denoised current output frame surface
+ //#define nBI_DEST_STAT 7 // Statistics output surface (STMM / Noise motion history)
+// #define nBI_DEST_U 8 // Dest frame U (for DN only)
+// #define nBI_DEST_V 9 // Dest frame V (for DN only)
+// #define nBI_SRC_U 10 // Src frame U (for DN only)
+// #define nBI_SRC_V 11 // Src frame V (for DN only)
+// #define nBI_SRC_UV 10 // Current src frame for UV
+
+#endif
+
+#if defined(INPUT_PL3)
+ // PL3 Surface Binding Table
+// #define nBI_SRC_ALPHA 0 // Alpha
+// #define nBI_SRC_Y 1 // Current src frame
+// #define nBI_SRC_U 2 // Current src frame
+// #define nBI_SRC_V 3 // Current src frame
+// #define nBI_DEST_Y 10 // Dest frame
+// #define nBI_DEST_U 11 // Dest frame
+// #define nBI_DEST_V 12 // Dest frame
+// #define nBI_DEST_YUV 7 // Dest frame
+// #define nBI_DEST_RGB 7 // same num as BI_DEST_YUV, never used at the same time
+#endif
+
+#if defined(INPUT_PL2)
+ // PL2 Surface Binding Table
+// #define nBI_SRC_ALPHA 0 // Alpha
+// #define nBI_SRC_Y 1 // Current src frame for Y + offseted UV
+// #define nBI_SRC_YUV 1 // Current src frame for YUV in case of NV12_AVS
+// #define nBI_SRC_UV 2 // Current src frame for UV
+// #define nBI_DEST_YUV 7 // Current dest frame for Y + offseted UV
+// #define nBI_DEST_RGB 7 // same num as BI_DEST_YUV, never used at the same time
+// #define nBI_DEST_Y 10 // Dest frame
+// #define nBI_DEST_U 11 // Dest frame
+// #define nBI_DEST_V 12 // Dest frame
+#endif
+
+#if defined(INPUT_PA) || defined(COLORFILL)
+ // Packed Surface Binding Table
+// #define nBI_SRC_ALPHA 0 // Alpha
+// #define nBI_SRC_YUV 1 // Current src frame
+// #define nBI_DEST_YUV 3 // Dest frame
+// #define nBI_DEST_RGB 3 // same num as BI_DEST_YUV, never used at the same time
+#endif
+
+
+//supper binding table
+#define nBI_ALPHA_SRC 0
+#define nBI_CURRENT_SRC_YUV 1
+#define nBI_FIELD_COPY_SRC_1_YUV 1
+#define nBI_CURRENT_SRC_Y 1
+#define nBI_FIELD_COPY_SRC_1_Y 1
+#define nBI_CURRENT_SRC_RGB 1
+#define nBI_CURRENT_SRC_UV 2
+#define nBI_FIELD_COPY_SRC_1_UV 2
+#define nBI_CURRENT_SRC_U 2
+#define nBI_FIELD_COPY_SRC_1_U 2
+#define nBI_CURRENT_SRC_V 3
+#define nBI_FIELD_COPY_SRC_1_V 3
+#define nBI_TEMPORAL_REFERENCE_YUV 4
+#define nBI_FIELD_COPY_SRC_2_YUV 4
+#define nBI_TEMPORAL_REFERENCE_Y 4
+#define nBI_FIELD_COPY_SRC_2_Y 4
+#define nBI_CURRENT_SRC_YUV_HW_DI 4
+#define nBI_TEMPORAL_REFERENCE_UV 5
+#define nBI_FIELD_COPY_SRC_2_UV 5
+#define nBI_TEMPORAL_REFERENCE_U 5
+#define nBI_FIELD_COPY_SRC_2_U 5
+#define nBI_DENOISED_PREV_HW_DI 5
+#define nBI_TEMPORAL_REFERENCE_V 6
+#define nBI_FIELD_COPY_SRC_2_V 6
+#define nBI_STMM_HISTORY 6
+#define nBI_DESTINATION_YUV 7
+#define nBI_DESTINATION_RGB 7
+#define nBI_DESTINATION_Y 7
+#define nBI_DESTINATION_UV 8
+#define nBI_DESTINATION_U 8
+#define nBI_DESTINATION_V 9
+#define nBI_DESTINATION_1_YUV 10
+#define nBI_DESTINATION_1_Y 10
+#define nBI_DESTINATION_1_UV 11
+#define nBI_DESTINATION_1_U 11
+#define nBI_DESTINATION_1_V 12
+#define nBI_DESTINATION_2_YUV 13
+#define nBI_DESTINATION_2_Y 13
+#define nBI_DESTINATION_2_UV 14
+#define nBI_DESTINATION_2_U 14
+#define nBI_DESTINATION_2_V 15
+#define nBI_STMM_HISTORY_OUTPUT 20
+#define nBI_TEMPORAL_REFERENCE_YUV_PDI 21
+#define nBI_TEMPORAL_REFERENCE_Y_PDI 21
+#define nBI_TEMPORAL_REFERENCE_UV_PDI 22
+#define nBI_TEMPORAL_REFERENCE_U_PDI 22
+#define nBI_TEMPORAL_REFERENCE_V_PDI 23
+#define nBI_SUBVIDEO_YUV 26
+#define nBI_SUBVIDEO_Y 26
+#define nBI_SUBVIDEO_UV 27
+#define nBI_SUBVIDEO_U 27
+#define nBI_SUBVIDEO_V 28
+#define nBI_SUBPICTURE_YUV 29
+#define nBI_SUBPICTURE_P8 29
+#define nBI_SUBPICTURE_A8 30
+#define nBI_GRAPHIC_YUV 31
+#define nBI_GRAPHIC_P8 31
+#define nBI_GRAPHIC_A8 32
+
+
+
+//========== Planar Sampler State Table Index ==========
+#define nSI_SRC_ALPHA 0x000 // Sampler State for Alpha
+
+//Sampler Index for AVS/IEF messages
+#define nSI_SRC_Y 0x400 // Sampler State for Y
+#define nSI_SRC_U 0x800 // Sampler State for U
+#define nSI_SRC_V 0xC00 // Sampler State for V
+#define nSI_SRC_UV 0x800 // For NV12 surfaces
+#define nSI_SRC_YUV 0x400 // For Packed surfaces
+#define nSI_SRC_RGB 0x400 // For ARGB surfaces
+
+//Sampler Index for SIMD16 sampler messages
+#define nSI_SRC_SIMD16_Y 0x100 // Sampler State for Y
+#define nSI_SRC_SIMD16_U 0x200 // Sampler State for U
+#define nSI_SRC_SIMD16_V 0x300 // Sampler State for V
+#define nSI_SRC_SIMD16_UV 0x200 // For NV12 surfaces
+#define nSI_SRC_SIMD16_YUV 0x100 // For Packed surfaces
+#define nSI_SRC_SIMD16_RGB 0x100 // For ARGB surfaces
+
+
+
+// Common Registers
+#define pCF_Y_OFFSET a0.4 // Address register holding Y offset
+#define pCF_U_OFFSET a0.5 // Address register holding U offset
+#define pCF_V_OFFSET a0.6 // Address register holding V offset
+
+// #define YUV_ORI ORIX // Used by writing packed data to dport
+
+
+//================= Message Payload Header fields ==============================
+#define IDP r0.2:ud // Interface Descriptor Pointer
+
+//================= Common Message Descriptor TBD add common load and save =====
+// Message descriptor for dataport media write
+#ifdef GT
+ // Message Descriptors
+ // = 000 0001 (min message len 1 - add later) 00000 (resp len 0)
+ // 1 (header present 1) 0 0 1010 (media block write) 00000
+ // 00000000 (binding table index - set later)
+ // = 0x02094000
+ #define nDPMW_MSGDSC 0x02094000
+ #define nDPMR_MSGDSC 0x02098000 // Data Port Media Block Read Message Descriptor
+ // TBD
+#else // ILK
+ // Message Descriptors
+ // = 000 0001 (min message len 1 - add later) 00000 (resp len 0)
+ // 1 (header present 1) 000 0 010 (media block write) 0000
+ // 00000000 (binding table index - set later)
+ // = 0x02082000
+ #define nDPMW_MSGDSC 0x02082000 // Data Port Media Block Write Message Descriptor
+ #define nDPMR_MSGDSC 0x0208A000 // Data Port Media Block Read Message Descriptor
+#endif
+
+// Message Length defines
+#define nMSGLEN_1 0x02000000 // Message Length of 1 GRF for Send
+#define nMSGLEN_2 0x04000000 // Message Length of 2 GRF for Send
+#define nMSGLEN_4 0x08000000 // Message Length of 4 GRF for Send
+#define nMSGLEN_8 0x10000000 // Message Length of 8 GRF for Send
+
+// Response Length defines
+#define nRESLEN_1 0x00100000 // Message Response Length of 1 GRF from Send
+#define nRESLEN_2 0x00200000 // Message Response Length of 2 GRF from Send
+#define nRESLEN_3 0x00300000 // Message Response Length of 3 GRF from Send
+#define nRESLEN_4 0x00400000 // Message Response Length of 4 GRF from Send
+#define nRESLEN_5 0x00500000 // Message Response Length of 5 GRF from Send
+#define nRESLEN_8 0x00800000 // Message Response Length of 8 GRF from Send
+#define nRESLEN_9 0x00900000 // Message Response Length of 9 GRF from Send
+#define nRESLEN_11 0x00B00000 // Message Response Length of 11 GRF from Send
+#define nRESLEN_12 0x00C00000 // Message Response Length of 12 GRF from Send
+#define nRESLEN_16 0x01000000 // Message Response Length of 16 GRF from Send
+
+// Block Width and Height Size defines
+#define nBLOCK_WIDTH_4 0x00000003 // Block Width 4
+#define nBLOCK_WIDTH_5 0x00000004 // Block Width 5
+#define nBLOCK_WIDTH_8 0x00000007 // Block Width 8
+#define nBLOCK_WIDTH_9 0x00000008 // Block Width 9
+#define nBLOCK_WIDTH_12 0x0000000B // Block Width 12
+#define nBLOCK_WIDTH_16 0x0000000F // Block Width 16
+#define nBLOCK_WIDTH_20 0x00000013 // Block Width 20
+#define nBLOCK_WIDTH_32 0x0000001F // Block Width 32
+#define nBLOCK_HEIGHT_1 0x00000000 // Block Height 1
+#define nBLOCK_HEIGHT_2 0x00010000 // Block Height 2
+#define nBLOCK_HEIGHT_4 0x00030000 // Block Height 4
+#define nBLOCK_HEIGHT_5 0x00040000 // Block Height 5
+#define nBLOCK_HEIGHT_8 0x00070000 // Block Height 8
+
+// Extended Message Descriptors
+#define nEXTENDED_MATH 0x1
+#define nSMPL_ENGINE 0x2
+#define nMESSAGE_GATEWAY 0x3
+#define nDATAPORT_READ 0x4
+#define nDATAPORT_WRITE 0x5
+#define nURB 0x6
+#define nTS_EOT 0x27 // with End-Of-Thread bit ON
+
+// Common message descriptors:
+#ifdef GT
+ #define nEOT_MSGDSC 0x02000010 // End of Thread Message Descriptor
+ #define IF_NULL null:uw null:uw null:uw //for different if instructions on ILK and Gen6
+#else //ILK
+ #define nEOT_MSGDSC 0x02000000 // End of Thread Message Descriptor
+ #define IF_NULL
+#endif
+
+
+//===================== Math Function Control ===================================
+#define mfcINV 0x1 // reciprocal
+#define mfcLOG 0x2 // log
+#define mfcEXP 0x3 // exponent
+#define mfcSQRT 0x4 // square root
+#define mfcRSQ 0x5 // reciprocal square root
+#define mfcSIN 0x6 // sine (in radians)
+#define mfcCOS 0x7 // cosine (in radians)
+#define mfcSINCOS 0x8 // dst0 = sin of src0, dst1 = cosine of src0 (in radians) - GT+ ONLY
+#define mfcPOW 0xA // abs(src0) raised to the src1 power
+#define mfcINT_DIV_QR 0xB // return quotient and remainder
+#define mfcINT_DIV_Q 0xC // return quotient
+#define mfcINT_DIV_R 0xD // return remainder
+
+
+//=================== Message related registers =================================
+
+#ifdef GT
+ #define udDUMMY_NULL
+#else // _ILK
+ #define udDUMMY_NULL null:ud // Used in send inst as src0
+#endif
+
+
+//----------- Message Registers ------------
+#define mMSGHDR m1 // Message Payload Header
+#define mMSGHDRY m1 // Message Payload Header register for Y data
+#define mMSGHDRU m2 // Message Payload Header register for U data
+#define mMSGHDRV m3 // Message Payload Header register for V data
+#define mMSGHDRYA m4 // Second Message Payload Header register for Y data
+#define mMSGHDRH m5 // Message Payload Header register for motion history
+#define mMSGHDRY1 m1 // Message Payload Header register for first Y data
+#define mMSGHDRY2 m2 // Message Payload Header register for second Y data
+#define mMSGHDRY3 m3 // Message Payload Header register for third Y data
+#define mMSGHDRY4 m4 // Message Payload Header register for fourth Y data
+#define mMSGHDRY5 m5 // Message Payload Header register for fifth Y data
+#define mMSGHDRY6 m6 // Message Payload Header register for sixth Y data
+#define mMSGHDR_EOT m15 // Dummy Message Register for EOT
+
+#define rMSGSRC r8 // Message source register
+#define pMSGDSC a0.0:ud // Message Descriptor register (type DWORD)
+
+#define udMH_ORI rMSGSRC.0 // Data Port Media Block R/W message header block offset
+#define udMH_ORIX rMSGSRC.0 // Data Port Media Block R/W message header X offset
+#define udMH_ORIY rMSGSRC.1 // Data Port Media Block R/W message header Y offset
+#define udMH_SIZE rMSGSRC.2 // Data Port Media Block R/W message header block width & height
+
+// M2 - M9 for message data payload
+.declare mubMSGPAYLOAD Base=m2 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare muwMSGPAYLOAD Base=m2 ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+.declare mudMSGPAYLOAD Base=m2 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare mfMSGPAYLOAD Base=m2 ElementSize=4 SrcRegion=REGION(8,1) Type=f
+
+//=================== End of thread instruction ===========================
+#ifdef GT
+ #define END_THREAD mov (8) mMSGHDR_EOT<1>:ud r0.0<8;8,1>:ud \n\
+ send (1) null<1>:d mMSGHDR_EOT nTS_EOT nEOT_MSGDSC
+#else // ILK This should be changed to 1 instruction; I have tested it and it works - vK
+ #define END_THREAD mov (8) mMSGHDR_EOT<1>:ud r0.0<8;8,1>:ud \n\
+ send (1) dNULLREG mMSGHDR_EOT udDUMMY_NULL nTS_EOT nEOT_MSGDSC:ud
+#endif
+
+
+//=======================================================================
+// Region declarations for SRC and DEST as TOP and BOT
+
+// Common I/O regions
+#define nREGION_1 1
+#define nREGION_2 2
+
+//*** These region base GRFs are fixed regardless planar/packed, and data alignment.
+//*** Each kernel is responsible to select the correct region declaration below.
+//*** YUV regions are not necessarily next to each other.
+#define nTOP_Y 10 // r10 - r17 (8 GRFs)
+#define nTOP_U 18 // r18 - r25 (8 GRFs)
+#define nTOP_V 26 // r26 - r33 (8 GRFs)
+
+#define nBOT_Y 56 // r56 - r63 (8 GRFs)
+#define nBOT_U 64 // r64 - r71 (8 GRFs)
+#define nBOT_V 72 // r72 - r79 (8 GRFs)
+
+// Define temp space for any usages
+#define nTEMP0 34
+#define nTEMP1 35
+#define nTEMP2 36
+#define nTEMP3 37
+#define nTEMP4 38
+#define nTEMP5 39
+#define nTEMP6 40
+#define nTEMP7 41
+#define nTEMP8 42
+#define nTEMP10 44
+#define nTEMP12 46
+#define nTEMP14 48
+#define nTEMP16 50
+#define nTEMP17 51
+#define nTEMP18 52
+
+#define nTEMP24 58
+
+// Common region 1
+.declare ubTOP_Y Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(16,1) DstRegion=<1> Type=ub
+.declare ubTOP_U Base=REG(r,nTOP_U) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub
+.declare ubTOP_V Base=REG(r,nTOP_V) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub
+
+.declare uwTOP_Y Base=REG(r,nTOP_Y) ElementSize=2 SrcRegion=REGION(16,1) DstRegion=<1> Type=uw
+.declare uwTOP_U Base=REG(r,nTOP_U) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare uwTOP_V Base=REG(r,nTOP_V) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare ub2TOP_Y Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(16,2) DstRegion=<1> Type=ub
+.declare ub2TOP_U Base=REG(r,nTOP_U) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub
+.declare ub2TOP_V Base=REG(r,nTOP_V) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub
+
+.declare ub4TOP_Y Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+.declare ub4TOP_U Base=REG(r,nTOP_U) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+.declare ub4TOP_V Base=REG(r,nTOP_V) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+
+.declare ubTOP_ARGB Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+
+// Used by "send" instruction
+.declare udTOP_Y_IO Base=REG(r,nTOP_Y) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare udTOP_U_IO Base=REG(r,nTOP_U) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare udTOP_V_IO Base=REG(r,nTOP_V) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+
+// Common region 2
+.declare ubBOT_Y Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(16,1) DstRegion=<1> Type=ub
+.declare ubBOT_U Base=REG(r,nBOT_U) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub
+.declare ubBOT_V Base=REG(r,nBOT_V) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub
+
+.declare uwBOT_Y Base=REG(r,nBOT_Y) ElementSize=2 SrcRegion=REGION(16,1) DstRegion=<1> Type=uw
+.declare uwBOT_U Base=REG(r,nBOT_U) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare uwBOT_V Base=REG(r,nBOT_V) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare ub2BOT_Y Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(16,2) DstRegion=<1> Type=ub
+.declare ub2BOT_U Base=REG(r,nBOT_U) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub
+.declare ub2BOT_V Base=REG(r,nBOT_V) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub
+
+.declare ubBOT_ARGB Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+
+// Used by "send" instruction
+.declare udBOT_Y_IO Base=REG(r,nBOT_Y) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare udBOT_U_IO Base=REG(r,nBOT_U) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare udBOT_V_IO Base=REG(r,nBOT_V) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+
+// End of common.inc
+
+#endif // COMMON_INC
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: readSampler16x1.asm
+//
+// Read one row of pix through sampler
+//
+
+
+
+//#define SAMPLER_MSG_DSC 0x166A0000 // ILK Sampler Message Descriptor
+
+
+
+// Send Message [DevILK] Message Descriptor
+// MBZ MsgL=5 MsgR=8 H MBZ SIMD MsgType SmplrIndx BindTab
+// 000 0 101 0 1000 1 0 10 0000 0000 00000000
+// 0 A 8 A 0 0 0 0
+
+// MsgL=1+2*2(u,v)=5 MsgR=8
+
+#define SAMPLER_MSG_DSC 0x0A8A0000 // ILK Sampler Message Descriptor
+
+
+
+
+
+
+
+ // Assume MSGSRC is set already in the caller
+ //mov (8) rMSGSRC.0<1>:ud 0:ud // Unused fileds
+
+
+
+ // Read 16 sampled pixels and stored them in float32 in 8 GRFs
+ // 422 data is expanded to 444, return 8 GRF in the order of RGB- (UYV-).
+ // 420 data has three surfaces, return 8 GRF. Valid is always in the 1st GRF when in R8. Make sure no overwrite the following 3 GRFs.
+ // alpha data is expanded to 4444, return 8 GRF in the order of RGBA (UYVA).
+
+ mov(16) mMSGHDR<1>:uw rMSGSRC<16;16,1>:uw
+ send (16) DATABUF(0)<1> mMSGHDR udDUMMY_NULL 0x2 SAMPLER_MSG_DSC+SAMPLER_IDX+BINDING_IDX:ud
+
+
+
+
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Modual name: undefall.inc
+//
+// undefine all global symbol for new process
+//
+
+//Source definitions
+#undef ubSRC_Y
+#undef ubSRC_U
+#undef ubSRC_V
+
+#undef ub2SRC_Y
+#undef ub2SRC_U
+#undef ub2SRC_V
+
+#undef ub4SRC_Y
+#undef ub4SRC_U
+#undef ub4SRC_V
+
+#undef uwSRC_Y
+#undef uwSRC_U
+#undef uwSRC_V
+
+#undef udSRC_Y
+#undef udSRC_U
+#undef udSRC_V
+
+#undef udSRC_YUV
+#undef nSRC_YUV_REG
+
+//Destination definitions
+#undef ubDEST_Y
+#undef ubDEST_U
+#undef ubDEST_V
+
+#undef ub2DEST_Y
+#undef ub2DEST_U
+#undef ub2DEST_V
+
+#undef ub4DEST_Y
+#undef ub4DEST_U
+#undef ub4DEST_V
+
+#undef uwDEST_Y
+#undef uwDEST_U
+#undef uwDEST_V
+
+#undef udDEST_Y
+#undef udDEST_U
+#undef udDEST_V
+
+#undef udDEST_YUV
+#undef nDEST_YUV_REG
+#undef ubDEST_ARGB
+
+// End of undefall.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: AVS_IEF.inc
+
+#ifndef _AVS_INF_INC_
+#define _AVS_INF_INC_
+
+#include "undefall.inc" //Undefine the SRC and DEST sysmbols
+
+ // Message Header
+ // m0.7 31:0 Debug
+ // m0.6 31:0 Debug
+ // m0.5 31:0 Ignored
+ // m0.4 31:0 Ignored
+ // m0.3 31:0 Ignored
+ // m0.2 31:16 Ignored
+ // 15 Alpha Write Channel Mask enable=0, disable=1
+ // 14 Blue Write Channel Mask (V)
+ // 13 Green Write Channel Mask (Y)
+ // 12 Red Write Channel Mask (U)
+ // 11:0 Ignored
+ // m0.1 Ignored
+ // m0.0 Ignored
+
+#define mAVS_8x8_HDR m0 // Message Header
+#define mAVS_PAYLOAD m1 // Message Payload Header
+
+#define mAVS_8x8_HDR_2 m2 // Message Header
+#define mAVS_PAYLOAD_2 m3 // Message Payload Header
+
+#define mAVS_8x8_HDR_UV m2 // Message Header
+#define mAVS_PAYLOAD_UV m3 // Message Payload Header
+
+#define rAVS_8x8_HDR rMSGSRC // Mirror of Message Header
+#define rAVS_PAYLOAD r9 // Mirror of Message Payload Header
+
+ // AVS payload
+ // m1.7 Ignored
+ // m1.6 Pixel 0 V Address ---> ORIY (Y0)
+ // m1.5 Delta V ---> Step Y
+ // m1.4 Ignored
+ // m1.3 Ignored
+ // m1.2 Pixel 0 U Address ---> ORIX (X0)
+ // m1.1 U 2nd Derivative ---> NLAS dx
+ // m1.0 Delta U ---> Step X
+
+ // Sampler Message Descriptor
+ // 31:29 Reserved 000
+ // 28:25 Message length 0010
+ // 24:20 Response length xxxxx ---> 4GRFs for each enabled channel
+ // 19 Header Present 1
+ // 18 MBZ 0
+ // 17:16 SIMD Mode 11 ---> SIMD64
+ // 15:12 Message Type 0011 ---> sample_8x8
+ // 11:8 Sampler Index xxxx
+ // 7:0 Binding Table Index xxxxxxxx
+#define nAVS_MSG_DSC_1CH 0x044BB000
+#define nAVS_MSG_DSC_2CH 0x048BB000
+#define nAVS_MSG_DSC_3CH 0x04CBB000
+#define nAVS_MSG_DSC_4CH 0x050BB000
+
+#define nAVS_RED_CHANNEL_ONLY 0x0000E000 // Enable Red channel only
+#define nAVS_GREEN_CHANNEL_ONLY 0x0000D000 // Enable Green channel only
+#define nAVS_RED_BLUE_CHANNELS 0x0000A000 // Enable Red and Blue channels
+#define nAVS_RGB_CHANNELS 0x00008000 // Enable RGB(YUV) channels
+#define nAVS_ALL_CHANNELS 0x00000000 // Enable all channels (ARGB\AYUV)
+
+
+
+.declare ubAVS_RESPONSE Base=REG(r,nTEMP8) ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare uwAVS_RESPONSE Base=REG(r,nTEMP8) ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+
+.declare ubAVS_RESPONSE_2 Base=REG(r,nTEMP24) ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare uwAVS_RESPONSE_2 Base=REG(r,nTEMP24) ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+
+
+#if (nSRC_REGION==nREGION_2)
+ #define uwDEST_Y uwBOT_Y
+ #define uwDEST_U uwBOT_U
+ #define uwDEST_V uwBOT_V
+
+ #define ubDEST_Y ubBOT_Y
+
+ #undef nSRC_REGION
+ #define nSRC_REGION nREGION_2
+
+#else //(nSRC_REGION==nREGION_1)
+ #define uwDEST_Y uwTOP_Y
+ #define uwDEST_U uwTOP_U
+ #define uwDEST_V uwTOP_V
+
+ #define ubDEST_Y ubTOP_Y
+
+ #undef nSRC_REGION
+ #define nSRC_REGION nREGION_1
+
+#endif
+
+
+#endif //_AVS_INF_INC_
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//------------------------------------------------------------------------------
+// AVS_SetupFirstBlock.asm
+//------------------------------------------------------------------------------
+
+ // Setup Message Header
+// mov (8) mAVS_8x8_HDR<1>:ud rMSGSRC<8;8,1>:ud
+
+ // Check NLAS Enable bit
+ and.z.f0.0 (1) wNULLREG uwNLAS_ENABLE:uw BIT15:uw
+ (f0.0)mov (1) fVIDEO_STEP_DELTA:f 0.0:f
+
+ // Setup Message Payload Header for 1st block of Media Sampler 8x8
+ mov (1) rAVS_PAYLOAD.0:f fVIDEO_STEP_DELTA:f //NLAS dx
+ mov (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_X:f //Step X
+ mov (1) rAVS_PAYLOAD.5:f fVIDEO_STEP_Y:f //Step Y
+ mov (2) rAVS_PAYLOAD.2<4>:f fSRC_VID_H_ORI<2;2,1>:f //Orig X and Y
+
+
+
+
+
+
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//------------------------------------------------------------------------------
+// AVS_SetupSecondBlock.asm
+//------------------------------------------------------------------------------
+
+ //NLAS calculations for 2nd block of Media Sampler 8x8:
+ // X(i) = X0 + dx*i + ddx*i*(i-1)/2 ==> X(8) = X0 + dx*8 +ddx*28
+ // dx(i)= dx(0) + ddx*i ==> dx(8)= dx + ddx*8
+
+ // Calculating X(8)
+ mov (1) acc0.2<1>:f fSRC_VID_H_ORI:f
+ mac (1) acc0.2<1>:f fVIDEO_STEP_X:f 8.0:f
+ mac (1) rAVS_PAYLOAD.2:f fVIDEO_STEP_DELTA:f 28.0:f
+
+ // Calculating dx(8)
+ mov (1) acc0.1<1>:f fVIDEO_STEP_X:f
+ mac (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_DELTA:f 8.0:f
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: DI.inc
+
+#ifdef GT
+// GT DI Kernel
+#else // ILK
+// ILK DI Kernel
+#endif
+
+
+//---------------------------------------------------------------------------
+// Binding table indices
+//---------------------------------------------------------------------------
+#define nBIDX_DI_PRV 10 // Previous DI-ed frame
+#define nBIDX_DI_CUR 13 // Current DI-ed frame
+#define nBIDX_DN 7 // Denoised frame
+#define nBIDX_STAT 20 // Statistics
+#define nBIDX_DI_Source 4 // Source Surface
+
+
+//---------------------------------------------------------------------------
+// Message descriptors
+//---------------------------------------------------------------------------
+// Extended message descriptor
+#define nSMPL_ENGINE 0x2
+#define nDATAPORT_WRITE 0x5
+#define nTS_EOT 0x27 // with End-Of-Thread bit ON
+
+ // Message descriptor for end-of-thread
+ // = 000 0001 (message len) 00000 (resp len)
+ // 0 (header present 0) 00000000000000 0 (URB dereferenced) 0000
+#define nEOT_MSGDSC 0x02000000
+
+ // Message descriptor for sampler read
+ // = 000 0010 (message len 2) 00000 (resp len - set later, 12 or 5 or 11)
+ // 1 (header present 1) 0 11 (SIMD32/64 mode)
+ // 1000 (message type) 0000 (DI state index)
+ // 00000000 (binding table index - set later)
+ // = 0x040b8000
+
+// comment begin
+// The following is commented out because of walker feature
+// It corresponds to the #ifdef GT #else and #endif
+//#define nSMPL_MSGDSC 0x040b8000
+//#define nSMPL_RESP_LEN_DI 0x00c00000 // 12
+//#define nSMPL_RESP_LEN_NODI_PL 0x00500000 // 5
+//#define nSMPL_RESP_LEN_NODI_PA 0x00900000 // 9
+//#define nSMPL_RESP_LEN_NODN 0x00900000 // 9
+//#define nSMPL_RESP_LEN_PDI 0x00b00000 // 11
+// comment end
+
+#ifdef GT
+
+#define nSMPL_MSGDSC 0x040b8000
+#define nSMPL_RESP_LEN_DI 0x00c00000 // 12
+#define nSMPL_RESP_LEN_NODI_PL 0x00500000 // 5 //DI disable, the XY stored in 5th GRF, no impact to return length
+#define nSMPL_RESP_LEN_NODI_PA 0x00900000 // 9 //DI disable, the XY stored in 5th GRF, no impact to return length
+#define nSMPL_RESP_LEN_NODN 0x00a00000 // 10 //NO DN, originally use 9, now we need use 10 to store the XY with walker
+#define nSMPL_RESP_LEN_PDI 0x00b00000 // 11
+
+#else
+
+#define nSMPL_MSGDSC 0x040b8000
+#define nSMPL_RESP_LEN_DI 0x00c00000 // 12
+#define nSMPL_RESP_LEN_NODI_PL 0x00500000 // 5
+#define nSMPL_RESP_LEN_NODI_PA 0x00900000 // 9
+#define nSMPL_RESP_LEN_NODN 0x00900000 // 9
+#define nSMPL_RESP_LEN_PDI 0x00b00000 // 11
+
+#endif
+
+ // Message descriptor for dataport media write
+#ifdef GT
+ // = 000 0000 (message len - set later) 00000 (resp len 0)
+ // 1 (header present 1) 0 0 1010 (media block write) 00000
+ // 00000000 (binding table index - set later)
+ // = 0x00094000
+#define nDPMW_MSGDSC 0x00094000
+#else // ILK
+ // = 000 0000 (message len - set later) 00000 (resp len 0)
+ // 1 (header present 1) 000 0 010 (media block write) 0000
+ // 00000000 (binding table index - set later)
+ // = 0x00082000
+#define nDPMW_MSGDSC 0x00082000
+#endif
+#define nDPMW_MSG_LEN_STMM 0x04000000 // 2 - STMM
+#define nDPMW_MSG_LEN_DH 0x04000000 // 2 - Denoise history
+#define nDPMW_MSG_LEN_PA_DN 0x0a000000 // 5 - Denoised output
+#define nDPMW_MSG_LEN_PA_NODI 0x12000000 // 9 - Denoised output - denoise only - DI disabled
+#define nDPMW_MSG_LEN_PL_DN 0x06000000 // 3 - Denoised output
+#define nDPMW_MSG_LEN_PL_NODI 0x0a000000 // 5 - Denoised output - denoise only - DI disabled
+#define nDPMW_MSG_LEN_DI 0x0a000000 // 5 - DI output
+
+
+//---------------------------------------------------------------------------
+// Static and inline parameters
+//---------------------------------------------------------------------------
+// Static parameters
+.declare ubTFLD_FIRST Base=r1.27 ElementSize=1 Type=ub // top field first
+.declare ubSRCYUVOFFSET Base=r1.4 ElementSize=1 Type=ub // source packed format
+.declare ubDSTYUVOFFSET Base=r1.8 ElementSize=1 Type=ub // destination packed format
+.declare uwSPITCH_DIV2 Base=r1.10 ElementSize=2 Type=uw // statistics surface pitch divided by 2
+
+// Inline parameters
+.declare uwXORIGIN Base=r5.0 ElementSize=2 Type=uw // X and Y origin
+.declare uwYORIGIN Base=r5.1 ElementSize=2 Type=uw
+
+
+//---------------------------------------------------------------------------
+// Kernel GRF variables
+//---------------------------------------------------------------------------
+// Message response (Denoised & DI-ed pixels & statistics)
+.declare dRESP Base=r8 ElementSize=4 Type=d // Response message (12 or 5 or 11)
+.declare ubRESP Base=r8 ElementSize=1 Type=ub
+
+.declare dSTMM Base=r16 ElementSize=4 Type=d // STMM
+.declare ubDN_HIST_NODI Base=r12 ElementSize=1 Type=ub // Denoise history data (DI disabled)
+.declare ubDN_HIST_DI Base=r17 ElementSize=1 Type=ub // Denoise history data (DI enabled)
+.declare uwRETURNED_POSITION_DI Base=r17 ElementSize=2 Type=uw // XY_Return_Data (DI enabled)
+.declare uwRETURNED_POSITION_DN Base=r12 ElementSize=2 Type=uw // XY_Return_Data (DI disabled)
+
+.declare ub1ST_FLD_DN Base=r12 ElementSize=1 Type=ub // 1st field Denoised data (DI enabled)
+.declare d1ST_FLD_DN Base=r12 ElementSize=4 Type=d
+.declare ub2ND_FLD_DN Base=r18 ElementSize=1 Type=ub // 2nd field Denoised data (DI enabled)
+.declare d2ND_FLD_DN Base=r18 ElementSize=4 Type=d
+.declare ubPRV_DI Base=r8 ElementSize=1 Type=ub // Previous frame DI (DI enabled)
+.declare ubCUR_DI Base=r12 ElementSize=1 Type=ub // Previous frame DI (DI enabled)
+
+// Packed denoised output
+.declare ubDN_YUV Base=r22 ElementSize=1 Type=ub // Denoised YUV422
+.declare dDN_YUV Base=r22 ElementSize=4 Type=d
+#define npDN_YUV 704 // = 22*32 = 0x280
+
+// Packed DI output
+.declare dDI_YUV_PRV Base=r32 ElementSize=4 Type=d // Previous frame DI output
+.declare dDI_YUV_CUR Base=r36 ElementSize=4 Type=d // Current frame DI output
+#define npDI_YUV 1024 // = 32*32 = 0x
+
+// For packed output
+#define p422_YOFFSET a0.2
+#define p422_UOFFSET a0.3
+#define p422_VOFFSET a0.4
+#define pDN_TFLDSRC a0.6
+#define pDN_BFLDSRC a0.7
+#define npRESP 192 // = 6*32
+
+// Message source
+.declare udMSGSRC Base=r70 ElementSize=4 Type=ud
+.declare uwMSGSRC Base=r70 ElementSize=2 Type=uw
+.declare dMSGSRC Base=r70 ElementSize=4 Type=d
+
+
+//---------------------------------------------------------------------------
+// Kernel MRF variables
+//---------------------------------------------------------------------------
+#define mMSGHDR_SMPL m1 // Sampler response: m1~m2
+.declare mudMSGHDR_SMPL Base=m1 ElementSize=4 Type=ud
+.declare muwMSGHDR_SMPL Base=m1 ElementSize=2 Type=uw
+#define mMSGHDR_DN m3 // Denoise output: m3~m7 for PA, m3~m5 for PL
+.declare mdMSGHDR_DN Base=m3 ElementSize=4 Type=d
+#define mMSGHDR_STAT m8 // Statistics output: m8~m9
+.declare mdMSGHDR_STAT Base=m8 ElementSize=4 Type=d
+.declare mubMSGHDR_STAT Base=m8 ElementSize=1 Type=ub
+#define mMSGHDR_DI m10 // DI output: m10~m14
+.declare mdMSGHDR_DI Base=m10 ElementSize=4 Type=d
+#define mMSGHDR_EOT m15 // EOT
+
+#ifdef GT
+#define MSGSRC
+#else
+#define MSGSRC null:ud
+#endif
+
+
+//---------------------------------------------------------------------------
+// End of thread instruction
+//---------------------------------------------------------------------------
+#ifdef GT
+#define END_THREAD send (8) null<1>:d mMSGHDR_EOT nTS_EOT nEOT_MSGDSC
+#else // ILK
+#define END_THREAD send (8) null<1>:d mMSGHDR_EOT null:ud nTS_EOT nEOT_MSGDSC
+#endif
+
+
+// end of DI.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+
+// Write denoise history to memory
+shr (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w 2:w NODDCLR // X,Y origin / 4
+add (1) rMSGSRC.0<1>:ud rMSGSRC.0<0;1,0>:ud uwSPITCH_DIV2<0;1,0>:uw NODDCLR_NODDCHK // Add pitch to X origin
+mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_HIST:ud NODDCHK // block width and height (4x2)
+
+mov (8) mMSGHDR_HIST<1>:ud rMSGSRC.0<8;8,1>:ud // message header
+mov (1) mudMSGHDR_HIST(1)<1> udRESP(nDI_HIST_OFFSET,0)<0;1,0> // Move denoise history to MRF
+
+send (8) dNULLREG mMSGHDR_HIST udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_HIST+nBI_STMM_HISTORY_OUTPUT:ud
+
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+ shl (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w NODDCLR // H. block origin need to be doubled
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w NODDCLR_NODDCHK // Block origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DI:ud NODDCHK // Block width and height (32x8)
+
+
+ add (4) pCF_Y_OFFSET<1>:uw ubDEST_CF_OFFSET<4;4,1>:ub nDEST_YUV_REG*nGRFWIB:w // Initial Y,U,V offset in YUV422 block
+
+ // Pack 2nd field Y
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) r[pCF_Y_OFFSET, %1*nGRFWIB]<2> ubRESP(nDI_PREV_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // Pack 1st field Y
+ $for(0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) r[pCF_Y_OFFSET, %1+4*nGRFWIB]<2> ubRESP(nDI_CURR_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // Pack 2nd field U
+ $for(0; <nUV_NUM_OF_ROWS; 1) {
+ mov (8) r[pCF_U_OFFSET, %1*nGRFWIB]<4> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ }
+ // Pack 1st field U
+ $for(0; <nUV_NUM_OF_ROWS; 1) {
+ mov (8) r[pCF_U_OFFSET, %1+4*nGRFWIB]<4> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ }
+ // Pack 2nd field V
+ $for(0; <nUV_NUM_OF_ROWS; 1) {
+ mov (8) r[pCF_V_OFFSET, %1*nGRFWIB]<4> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //Vpixels
+ }
+ // Packs1st field V
+ $for(0; <nUV_NUM_OF_ROWS; 1) {
+ mov (8) r[pCF_V_OFFSET, %1+4*nGRFWIB]<4> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //Vpixels
+ }
+
+ //save the previous frame
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ $for(0; <4; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_YUV(%1)REGION(8,1)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_DI+nBI_DESTINATION_1_YUV:ud
+
+ //save the current frame
+ mov (8) mMSGHDR<1>:ud rMSGSRC<8;8,1>:ud
+ $for(0; <4; 1) {
+ mov (8) mudMSGPAYLOAD(%1)<1> udDEST_YUV(%1+4)REGION(8,1)
+ }
+ send (8) dNULLREG mMSGHDR udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPW_MSG_SIZE_DI+nBI_DESTINATION_2_YUV:ud
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: DI.inc
+
+#ifdef GT
+// GT DI Kernel
+#else // ILK
+// ILK DI Kernel
+#endif
+
+#include "undefall.inc"
+
+//---------------------------------------------------------------------------
+// Message descriptors
+//---------------------------------------------------------------------------
+// Extended message descriptor
+ // Message descriptor for sampler read
+// // = 000 0010 (message len 2) 00000 (resp len - set later, 12 or 5 or 11)
+// // 1 (header present 1) 0 11 (SIMD32/64 mode)
+// // 1000 (message type) 0000 (DI state index)
+// // 00000000 (binding table index - set later)
+// // = 0x040b8000
+#define nSMPL_DI_MSGDSC 0x040b8000
+
+#define nSMPL_RESP_LEN_DNDI nRESLEN_12 // 12 - for DN + DI Alg
+#define nSMPL_RESP_LEN_DN_PL nRESLEN_5 // 5 - for DN Planar Alg
+#define nSMPL_RESP_LEN_DN_PA nRESLEN_9 // 9 - for DN Packed Alg
+#define nSMPL_RESP_LEN_DI nRESLEN_9 // 9 - for DI Only Alg
+#define nSMPL_RESP_LEN_PDI nRESLEN_11 // 11 - for Partial DI Alg
+
+// Attention: The Message Length is The Number of GRFs with Data Only, without the Header
+#define nDPMW_MSG_LEN_STMM nMSGLEN_1 // 1 - For STMM Save
+#define nDPMW_MSG_LEN_HIST nMSGLEN_1 // 1 - For Denoise History Save
+#define nDPMW_MSG_LEN_PA_DN_DI nMSGLEN_4 // 4 - For DN Curr Save
+#define nDPMW_MSG_LEN_PA_DN_NODI nMSGLEN_8 // 8 - For DN Curr Save (denoise only - DI disabled)
+#define nDPMW_MSG_LEN_PL_DN_DI nMSGLEN_2 // 2 - For DN Curr Save
+#define nDPMW_MSG_LEN_PL_DN_NODI nMSGLEN_4 // 4 - For DN Curr Save (denoise only - DI disabled)
+
+#define nDPW_BLOCK_SIZE_STMM nBLOCK_WIDTH_8+nBLOCK_HEIGHT_4 // Y block size 8x4
+
+#undef nDPW_BLOCK_SIZE_DI
+#undef nDPW_MSG_SIZE_DI
+#define nDPW_BLOCK_SIZE_DI nBLOCK_WIDTH_32+nBLOCK_HEIGHT_4
+#define nDPW_MSG_SIZE_DI nMSGLEN_4
+
+
+//---------------------------------------------------------------------------
+// Kernel GRF variables
+//---------------------------------------------------------------------------
+// Defines for DI enabled
+#define nDI_PREV_FRAME_LUMA_OFFSET 0
+#define nDI_PREV_FRAME_CHROMA_OFFSET 2
+#define nDI_CURR_FRAME_LUMA_OFFSET 4
+#define nDI_CURR_FRAME_CHROMA_OFFSET 6
+#define nDI_STMM_OFFSET 8
+#define nDI_HIST_OFFSET 9
+#define nDI_CURR_2ND_FIELD_LUMA_OFFSET 10
+#define nDI_CURR_2ND_FIELD_CHROMA_OFFSET 11
+
+// Defines for DI disabled
+#define nNODI_LUMA_OFFSET 0
+#define nNODI_HIST_OFFSET 4
+#define nNODI_CHROMA_OFFSET 5
+
+#ifdef DI_ENABLE
+ #define nHIST_OFFSET nDI_HIST_OFFSET
+ #undef nY_NUM_OF_ROWS
+ #define nY_NUM_OF_ROWS 8 // Number of Y rows per block (4 rows for each frame)
+ #undef nUV_NUM_OF_ROWS
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+
+#endif
+
+#ifdef DI_DISABLE
+ #define nHIST_OFFSET nNODI_HIST_OFFSET
+#endif
+
+#if (nSRC_REGION==nREGION_2)
+ #define ub2SRC_Y ub2BOT_Y
+ #define ub2SRC_U ub2BOT_U
+ #define ub2SRC_V ub2BOT_V
+ #define uwDEST_Y uwBOT_Y
+ #define uwDEST_U uwBOT_U
+ #define uwDEST_V uwBOT_V
+ #define nDEST_YUV_REG nTOP_Y
+ #define udDEST_YUV udTOP_Y_IO
+
+ #define nRESP nTEMP0 // DI return message requires 12 GRFs
+ #define nDN_YUV nTOP_Y // Space for Packing DN for next run requires 8 GRFs
+
+ #undef nSRC_REGION
+ #define nSRC_REGION nREGION_2
+
+#else
+ #define ub2SRC_Y ub2TOP_Y
+ #define ub2SRC_U ub2TOP_U
+ #define ub2SRC_V ub2TOP_V
+ #define uwDEST_Y uwTOP_Y
+ #define uwDEST_U uwTOP_U
+ #define uwDEST_V uwTOP_V
+ #define nDEST_YUV_REG nBOT_Y
+ #define udDEST_YUV udBOT_Y_IO
+ #define nRESP nTEMP0 // DI return message requires 12 GRFs
+ #define nDN_YUV nBOT_Y // Space for Packing DN for next run requires 8 GRFs
+
+ #undef nSRC_REGION
+ #define nSRC_REGION nREGION_1 // REGION_1 will be the source region for first kernel
+
+#endif
+
+
+
+
+
+
+
+
+
+// Message response (Denoised & DI-ed pixels & statistics)
+.declare udRESP Base=REG(r,nRESP) ElementSize=4 SrcRegion=REGION(8,1) DstRegion=<1> Type=ud
+.declare ubRESP Base=REG(r,nRESP) ElementSize=1 SrcRegion=REGION(16,1) DstRegion=<1> Type=ub
+
+// For Denoised Curr Output (Used as Priv in Next Run)
+.declare ubDN_YUV Base=REG(r,nDN_YUV) ElementSize=1 Type=ub
+.declare udDN_YUV Base=REG(r,nDN_YUV) ElementSize=4 Type=ud
+#define npDN_YUV nDN_YUV*nGRFWIB
+
+// For DI Process Output (1st and 2nd Frames Output)
+//.declare udDI_YUV_PRIV Base=REG(r,nTEMP0) ElementSize=4 Type=ud // Previous frame DI output
+//.declare udDI_YUV_CURR Base=REG(r,nTEMP0) ElementSize=4 Type=ud // Current frame DI output
+//#define npDI_YUV nTEMP0*nGRFWIB
+
+//---------------------------------------------------------------------------
+// Kernel MRF variables
+//---------------------------------------------------------------------------
+#define mMSG_SMPL m1 // Sampler Command is in: m1~m2
+.declare mudMSG_SMPL Base=mMSG_SMPL ElementSize=4 Type=ud
+.declare muwMSG_SMPL Base=mMSG_SMPL ElementSize=2 Type=uw
+
+#define mMSGHDR_DN m1 // Denoise Output: m1~m9 for PA, m3~m5 for PL
+.declare mudMSGHDR_DN Base=mMSGHDR_DN ElementSize=4 Type=ud
+.declare mubMSGHDR_DN Base=mMSGHDR_DN ElementSize=1 Type=ub
+
+#define mMSGHDR_STMM m11 // STMM Output: m11~m12
+.declare mudMSGHDR_STMM Base=mMSGHDR_STMM ElementSize=4 Type=ud
+#define mMSGHDR_HIST m13 // HIST Output: m13~m14
+.declare mudMSGHDR_HIST Base=mMSGHDR_HIST ElementSize=1 Type=ud
+
+#define mMSGHDR_DI_1ST m1 // DI output: m1~m5
+.declare mudMSGHDR_DI_1ST Base=mMSGHDR_DI_1ST ElementSize=4 Type=ud
+#define mMSGHDR_DI_2ND m6 // DI output: m6~m10
+.declare mudMSGHDR_DI_2ND Base=mMSGHDR_DI_2ND ElementSize=4 Type=ud
+
+// end of DNDI.inc
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Activate the DNDI send command
+mov (8) mudMSG_SMPL(0)<1> rMSGSRC.0<8;8,1>:ud NODDCLR // message header
+mov (1) muwMSG_SMPL(1,4)<1> wORIX<0;1,0>:w NODDCLR_NODDCHK// horizontal origin
+mov (1) muwMSG_SMPL(1,12)<1> wORIY<0;1,0>:w NODDCLR_NODDCHK // vertical origin
+//mov (2) muwMSG_SMPL(1,4)<2> wORIX<2;2,1>:w NODDCHK// problem during compile !! when using this line
+
+send (8) udRESP(0)<1> mMSG_SMPL udDUMMY_NULL nSMPL_ENGINE nSMPL_DI_MSGDSC+nSMPL_RESP_LEN+nBI_CURRENT_SRC_YUV_HW_DI:ud
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+
+// Write denoise history to memory
+shr (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w 2:w NODDCLR // X,Y origin / 4
+add (1) rMSGSRC.0<1>:ud rMSGSRC.0<0;1,0>:ud uwSPITCH_DIV2<0;1,0>:uw NODDCLR_NODDCHK// Add pitch to X origin
+mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_HIST:ud NODDCHK // block width and height (4x2)
+
+mov (8) mMSGHDR_HIST<1>:ud rMSGSRC.0<8;8,1>:ud // message header
+mov (2) mudMSGHDR_HIST(1)<1> udRESP(nNODI_HIST_OFFSET,0)<2;2,1> // Move denoise history to MRF
+
+send (8) dNULLREG mMSGHDR_HIST udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_HIST+nBI_STMM_HISTORY_OUTPUT:ud
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_AVS_IEF_16x8.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 YUV packed
+//------------------------------------------------------------------------------
+#include "PA_AVS_IEF_Sample.asm"
+
+//------------------------------------------------------------------------------
+// Unpacking sampler reads to 4:4:4 internal planar
+//------------------------------------------------------------------------------
+#include "PA_AVS_IEF_Unpack_16x8.asm"
+
+//------------------------------------------------------------------------------
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_AVS_IEF_8x4.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 YUV packed
+//------------------------------------------------------------------------------
+#include "PA_AVS_IEF_Sample.asm"
+
+//------------------------------------------------------------------------------
+// Unpacking sampler data to 4:2:0 internal planar
+//------------------------------------------------------------------------------
+#include "PA_AVS_IEF_Unpack_8x4.asm"
+
+//------------------------------------------------------------------------------
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_AVS_IEF_8x8.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 YUV packed
+//------------------------------------------------------------------------------
+#include "PA_AVS_IEF_Sample.asm"
+
+//------------------------------------------------------------------------------
+// Unpacking sampler data to 4:2:2 internal planar
+//------------------------------------------------------------------------------
+#include "PA_AVS_IEF_Unpack_8x8.asm"
+
+//------------------------------------------------------------------------------
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_AVS_IEF_Sample.asm ----------
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 YUV packed
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ // Enable RGB(YUV) channels
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RGB_CHANNELS:ud
+
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_3CH+nSI_SRC_YUV+nBI_CURRENT_SRC_YUV
+ // Return YUV in 12 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+
+ mov (16) mAVS_8x8_HDR_2.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(0)<1> mAVS_8x8_HDR_2 udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_3CH+nSI_SRC_YUV+nBI_CURRENT_SRC_YUV
+ // Return YUV in 12 GRFs
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_AVS_IEF_Unpack_16x8.asm ----------
+
+#ifdef AVS_OUTPUT_16_BIT //Output is packed in AVYU format
+// Move first 8x8 words of Y to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,1)<4> uwAVS_RESPONSE(2,0)<4;4,1>
+ mov (4) uwDEST_Y(1,1)<4> uwAVS_RESPONSE(2,8)<4;4,1>
+ mov (4) uwDEST_Y(4,1)<4> uwAVS_RESPONSE(2,4)<4;4,1>
+ mov (4) uwDEST_Y(5,1)<4> uwAVS_RESPONSE(2,12)<4;4,1>
+ mov (4) uwDEST_Y(8,1)<4> uwAVS_RESPONSE(3,0)<4;4,1>
+ mov (4) uwDEST_Y(9,1)<4> uwAVS_RESPONSE(3,8)<4;4,1>
+ mov (4) uwDEST_Y(12,1)<4> uwAVS_RESPONSE(3,4)<4;4,1>
+ mov (4) uwDEST_Y(13,1)<4> uwAVS_RESPONSE(3,12)<4;4,1>
+ mov (4) uwDEST_Y(16,1)<4> uwAVS_RESPONSE(8,0)<4;4,1>
+ mov (4) uwDEST_Y(17,1)<4> uwAVS_RESPONSE(8,8)<4;4,1>
+ mov (4) uwDEST_Y(20,1)<4> uwAVS_RESPONSE(8,4)<4;4,1>
+ mov (4) uwDEST_Y(21,1)<4> uwAVS_RESPONSE(8,12)<4;4,1>
+ mov (4) uwDEST_Y(24,1)<4> uwAVS_RESPONSE(9,0)<4;4,1>
+ mov (4) uwDEST_Y(25,1)<4> uwAVS_RESPONSE(9,8)<4;4,1>
+ mov (4) uwDEST_Y(28,1)<4> uwAVS_RESPONSE(9,4)<4;4,1>
+ mov (4) uwDEST_Y(29,1)<4> uwAVS_RESPONSE(9,12)<4;4,1>
+
+// Move first 8x8 words of U to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,0)<4> uwAVS_RESPONSE(4,0)<4;4,1>
+ mov (4) uwDEST_Y(1,0)<4> uwAVS_RESPONSE(4,8)<4;4,1>
+ mov (4) uwDEST_Y(4,0)<4> uwAVS_RESPONSE(4,4)<4;4,1>
+ mov (4) uwDEST_Y(5,0)<4> uwAVS_RESPONSE(4,12)<4;4,1>
+ mov (4) uwDEST_Y(8,0)<4> uwAVS_RESPONSE(5,0)<4;4,1>
+ mov (4) uwDEST_Y(9,0)<4> uwAVS_RESPONSE(5,8)<4;4,1>
+ mov (4) uwDEST_Y(12,0)<4> uwAVS_RESPONSE(5,4)<4;4,1>
+ mov (4) uwDEST_Y(13,0)<4> uwAVS_RESPONSE(5,12)<4;4,1>
+ mov (4) uwDEST_Y(16,0)<4> uwAVS_RESPONSE(10,0)<4;4,1>
+ mov (4) uwDEST_Y(17,0)<4> uwAVS_RESPONSE(10,8)<4;4,1>
+ mov (4) uwDEST_Y(20,0)<4> uwAVS_RESPONSE(10,4)<4;4,1>
+ mov (4) uwDEST_Y(21,0)<4> uwAVS_RESPONSE(10,12)<4;4,1>
+ mov (4) uwDEST_Y(24,0)<4> uwAVS_RESPONSE(11,0)<4;4,1>
+ mov (4) uwDEST_Y(25,0)<4> uwAVS_RESPONSE(11,8)<4;4,1>
+ mov (4) uwDEST_Y(28,0)<4> uwAVS_RESPONSE(11,4)<4;4,1>
+ mov (4) uwDEST_Y(29,0)<4> uwAVS_RESPONSE(11,12)<4;4,1>
+
+// Move first 8x8 words of V to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,2)<4> uwAVS_RESPONSE(0,0)<4;4,1>
+ mov (4) uwDEST_Y(1,2)<4> uwAVS_RESPONSE(0,8)<4;4,1>
+ mov (4) uwDEST_Y(4,2)<4> uwAVS_RESPONSE(0,4)<4;4,1>
+ mov (4) uwDEST_Y(5,2)<4> uwAVS_RESPONSE(0,12)<4;4,1>
+ mov (4) uwDEST_Y(8,2)<4> uwAVS_RESPONSE(1,0)<4;4,1>
+ mov (4) uwDEST_Y(9,2)<4> uwAVS_RESPONSE(1,8)<4;4,1>
+ mov (4) uwDEST_Y(12,2)<4> uwAVS_RESPONSE(1,4)<4;4,1>
+ mov (4) uwDEST_Y(13,2)<4> uwAVS_RESPONSE(1,12)<4;4,1>
+ mov (4) uwDEST_Y(16,2)<4> uwAVS_RESPONSE(6,0)<4;4,1>
+ mov (4) uwDEST_Y(17,2)<4> uwAVS_RESPONSE(6,8)<4;4,1>
+ mov (4) uwDEST_Y(20,2)<4> uwAVS_RESPONSE(6,4)<4;4,1>
+ mov (4) uwDEST_Y(21,2)<4> uwAVS_RESPONSE(6,12)<4;4,1>
+ mov (4) uwDEST_Y(24,2)<4> uwAVS_RESPONSE(7,0)<4;4,1>
+ mov (4) uwDEST_Y(25,2)<4> uwAVS_RESPONSE(7,8)<4;4,1>
+ mov (4) uwDEST_Y(28,2)<4> uwAVS_RESPONSE(7,4)<4;4,1>
+ mov (4) uwDEST_Y(29,2)<4> uwAVS_RESPONSE(7,12)<4;4,1>
+
+// Move first 8x8 words of A to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,3)<4> 0:uw
+ mov (4) uwDEST_Y(1,3)<4> 0:uw
+ mov (4) uwDEST_Y(4,3)<4> 0:uw
+ mov (4) uwDEST_Y(5,3)<4> 0:uw
+ mov (4) uwDEST_Y(8,3)<4> 0:uw
+ mov (4) uwDEST_Y(9,3)<4> 0:uw
+ mov (4) uwDEST_Y(12,3)<4> 0:uw
+ mov (4) uwDEST_Y(13,3)<4> 0:uw
+ mov (4) uwDEST_Y(16,3)<4> 0:uw
+ mov (4) uwDEST_Y(17,3)<4> 0:uw
+ mov (4) uwDEST_Y(20,3)<4> 0:uw
+ mov (4) uwDEST_Y(21,3)<4> 0:uw
+ mov (4) uwDEST_Y(24,3)<4> 0:uw
+ mov (4) uwDEST_Y(25,3)<4> 0:uw
+ mov (4) uwDEST_Y(28,3)<4> 0:uw
+ mov (4) uwDEST_Y(29,3)<4> 0:uw
+
+// Move second 8x8 words of Y to dest GRF
+ mov (4) uwDEST_Y(2,1)<4> uwAVS_RESPONSE_2(2,0)<4;4,1>
+ mov (4) uwDEST_Y(3,1)<4> uwAVS_RESPONSE_2(2,8)<4;4,1>
+ mov (4) uwDEST_Y(6,1)<4> uwAVS_RESPONSE_2(2,4)<4;4,1>
+ mov (4) uwDEST_Y(7,1)<4> uwAVS_RESPONSE_2(2,12)<4;4,1>
+ mov (4) uwDEST_Y(10,1)<4> uwAVS_RESPONSE_2(3,0)<4;4,1>
+ mov (4) uwDEST_Y(11,1)<4> uwAVS_RESPONSE_2(3,8)<4;4,1>
+ mov (4) uwDEST_Y(14,1)<4> uwAVS_RESPONSE_2(3,4)<4;4,1>
+ mov (4) uwDEST_Y(15,1)<4> uwAVS_RESPONSE_2(3,12)<4;4,1>
+ mov (4) uwDEST_Y(18,1)<4> uwAVS_RESPONSE_2(8,0)<4;4,1>
+ mov (4) uwDEST_Y(19,1)<4> uwAVS_RESPONSE_2(8,8)<4;4,1>
+ mov (4) uwDEST_Y(22,1)<4> uwAVS_RESPONSE_2(8,4)<4;4,1>
+ mov (4) uwDEST_Y(23,1)<4> uwAVS_RESPONSE_2(8,12)<4;4,1>
+ mov (4) uwDEST_Y(26,1)<4> uwAVS_RESPONSE_2(9,0)<4;4,1>
+ mov (4) uwDEST_Y(27,1)<4> uwAVS_RESPONSE_2(9,8)<4;4,1>
+ mov (4) uwDEST_Y(30,1)<4> uwAVS_RESPONSE_2(9,4)<4;4,1>
+ mov (4) uwDEST_Y(31,1)<4> uwAVS_RESPONSE_2(9,12)<4;4,1>
+
+// Move second 8x8 words of U to dest GRF
+ mov (4) uwDEST_Y(2,0)<4> uwAVS_RESPONSE_2(4,0)<4;4,1>
+ mov (4) uwDEST_Y(3,0)<4> uwAVS_RESPONSE_2(4,8)<4;4,1>
+ mov (4) uwDEST_Y(6,0)<4> uwAVS_RESPONSE_2(4,4)<4;4,1>
+ mov (4) uwDEST_Y(7,0)<4> uwAVS_RESPONSE_2(4,12)<4;4,1>
+ mov (4) uwDEST_Y(10,0)<4> uwAVS_RESPONSE_2(5,0)<4;4,1>
+ mov (4) uwDEST_Y(11,0)<4> uwAVS_RESPONSE_2(5,8)<4;4,1>
+ mov (4) uwDEST_Y(14,0)<4> uwAVS_RESPONSE_2(5,4)<4;4,1>
+ mov (4) uwDEST_Y(15,0)<4> uwAVS_RESPONSE_2(5,12)<4;4,1>
+ mov (4) uwDEST_Y(18,0)<4> uwAVS_RESPONSE_2(10,0)<4;4,1>
+ mov (4) uwDEST_Y(19,0)<4> uwAVS_RESPONSE_2(10,8)<4;4,1>
+ mov (4) uwDEST_Y(22,0)<4> uwAVS_RESPONSE_2(10,4)<4;4,1>
+ mov (4) uwDEST_Y(23,0)<4> uwAVS_RESPONSE_2(10,12)<4;4,1>
+ mov (4) uwDEST_Y(26,0)<4> uwAVS_RESPONSE_2(11,0)<4;4,1>
+ mov (4) uwDEST_Y(27,0)<4> uwAVS_RESPONSE_2(11,8)<4;4,1>
+ mov (4) uwDEST_Y(30,0)<4> uwAVS_RESPONSE_2(11,4)<4;4,1>
+ mov (4) uwDEST_Y(31,0)<4> uwAVS_RESPONSE_2(11,12)<4;4,1>
+
+// Move second 8x8 words of V to dest GRF
+ mov (4) uwDEST_Y(2,2)<4> uwAVS_RESPONSE_2(0,0)<4;4,1>
+ mov (4) uwDEST_Y(3,2)<4> uwAVS_RESPONSE_2(0,8)<4;4,1>
+ mov (4) uwDEST_Y(6,2)<4> uwAVS_RESPONSE_2(0,4)<4;4,1>
+ mov (4) uwDEST_Y(7,2)<4> uwAVS_RESPONSE_2(0,12)<4;4,1>
+ mov (4) uwDEST_Y(10,2)<4> uwAVS_RESPONSE_2(1,0)<4;4,1>
+ mov (4) uwDEST_Y(11,2)<4> uwAVS_RESPONSE_2(1,8)<4;4,1>
+ mov (4) uwDEST_Y(14,2)<4> uwAVS_RESPONSE_2(1,4)<4;4,1>
+ mov (4) uwDEST_Y(15,2)<4> uwAVS_RESPONSE_2(1,12)<4;4,1>
+ mov (4) uwDEST_Y(18,2)<4> uwAVS_RESPONSE_2(6,0)<4;4,1>
+ mov (4) uwDEST_Y(19,2)<4> uwAVS_RESPONSE_2(6,8)<4;4,1>
+ mov (4) uwDEST_Y(22,2)<4> uwAVS_RESPONSE_2(6,4)<4;4,1>
+ mov (4) uwDEST_Y(23,2)<4> uwAVS_RESPONSE_2(6,12)<4;4,1>
+ mov (4) uwDEST_Y(26,2)<4> uwAVS_RESPONSE_2(7,0)<4;4,1>
+ mov (4) uwDEST_Y(27,2)<4> uwAVS_RESPONSE_2(7,8)<4;4,1>
+ mov (4) uwDEST_Y(30,2)<4> uwAVS_RESPONSE_2(7,4)<4;4,1>
+ mov (4) uwDEST_Y(31,2)<4> uwAVS_RESPONSE_2(7,12)<4;4,1>
+
+// Move second 8x8 words of A to dest GRF
+ mov (4) uwDEST_Y(2,3)<4> 0:uw
+ mov (4) uwDEST_Y(3,3)<4> 0:uw
+ mov (4) uwDEST_Y(6,3)<4> 0:uw
+ mov (4) uwDEST_Y(7,3)<4> 0:uw
+ mov (4) uwDEST_Y(10,3)<4> 0:uw
+ mov (4) uwDEST_Y(11,3)<4> 0:uw
+ mov (4) uwDEST_Y(14,3)<4> 0:uw
+ mov (4) uwDEST_Y(15,3)<4> 0:uw
+ mov (4) uwDEST_Y(18,3)<4> 0:uw
+ mov (4) uwDEST_Y(19,3)<4> 0:uw
+ mov (4) uwDEST_Y(22,3)<4> 0:uw
+ mov (4) uwDEST_Y(23,3)<4> 0:uw
+ mov (4) uwDEST_Y(26,3)<4> 0:uw
+ mov (4) uwDEST_Y(27,3)<4> 0:uw
+ mov (4) uwDEST_Y(30,3)<4> 0:uw
+ mov (4) uwDEST_Y(31,3)<4> 0:uw
+
+/* This section will be used if 16-bit output is needed in planar format -vK
+ // Move first 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0)<1> uwAVS_RESPONSE(2,0)<8;4,1>
+ mov (8) uwDEST_Y(1)<1> uwAVS_RESPONSE(2,8)<8;4,1>
+ mov (8) uwDEST_Y(2)<1> uwAVS_RESPONSE(3,0)<8;4,1>
+ mov (8) uwDEST_Y(3)<1> uwAVS_RESPONSE(3,8)<8;4,1>
+ mov (8) uwDEST_Y(4)<1> uwAVS_RESPONSE(8,0)<8;4,1>
+ mov (8) uwDEST_Y(5)<1> uwAVS_RESPONSE(8,8)<8;4,1>
+ mov (8) uwDEST_Y(6)<1> uwAVS_RESPONSE(9,0)<8;4,1>
+ mov (8) uwDEST_Y(7)<1> uwAVS_RESPONSE(9,8)<8;4,1>
+
+ // Move first 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0)<1> ubAVS_RESPONSE(0,0)<8;4,1>
+ mov (8) uwDEST_V(1)<1> ubAVS_RESPONSE(0,8)<8;4,1>
+ mov (8) uwDEST_V(2)<1> ubAVS_RESPONSE(1,0)<8;4,1>
+ mov (8) uwDEST_V(3)<1> ubAVS_RESPONSE(1,8)<8;4,1>
+ mov (8) uwDEST_V(4)<1> ubAVS_RESPONSE(6,0)<8;4,1>
+ mov (8) uwDEST_V(5)<1> ubAVS_RESPONSE(6,8)<8;4,1>
+ mov (8) uwDEST_V(6)<1> ubAVS_RESPONSE(7,0)<8;4,1>
+ mov (8) uwDEST_V(7)<1> ubAVS_RESPONSE(7,8)<8;4,1>
+
+ // Move first 8x8 words of U to dest GRF
+ mov (8) uwDEST_U(0)<1> ubAVS_RESPONSE(4,0)<8;4,1>
+ mov (8) uwDEST_U(1)<1> ubAVS_RESPONSE(4,8)<8;4,1>
+ mov (8) uwDEST_U(2)<1> ubAVS_RESPONSE(5,0)<8;4,1>
+ mov (8) uwDEST_U(3)<1> ubAVS_RESPONSE(5,8)<8;4,1>
+ mov (8) uwDEST_U(4)<1> ubAVS_RESPONSE(10,0)<8;4,1>
+ mov (8) uwDEST_U(5)<1> ubAVS_RESPONSE(10,8)<8;4,1>
+ mov (8) uwDEST_U(6)<1> ubAVS_RESPONSE(11,0)<8;4,1>
+ mov (8) uwDEST_U(7)<1> ubAVS_RESPONSE(11,8)<8;4,1>
+
+ // Move second 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0,8)<1> uwAVS_RESPONSE_2(2,0)<8;4,1>
+ mov (8) uwDEST_Y(1,8)<1> uwAVS_RESPONSE_2(2,8)<8;4,1>
+ mov (8) uwDEST_Y(2,8)<1> uwAVS_RESPONSE_2(3,0)<8;4,1>
+ mov (8) uwDEST_Y(3,8)<1> uwAVS_RESPONSE_2(3,8)<8;4,1>
+ mov (8) uwDEST_Y(4,8)<1> uwAVS_RESPONSE_2(8,0)<8;4,1>
+ mov (8) uwDEST_Y(5,8)<1> uwAVS_RESPONSE_2(8,8)<8;4,1>
+ mov (8) uwDEST_Y(6,8)<1> uwAVS_RESPONSE_2(9,0)<8;4,1>
+ mov (8) uwDEST_Y(7,8)<1> uwAVS_RESPONSE_2(9,8)<8;4,1>
+
+ // Move second 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0,8)<1> ubAVS_RESPONSE_2(0,0)<8;4,1>
+ mov (8) uwDEST_V(1,8)<1> ubAVS_RESPONSE_2(0,8)<8;4,1>
+ mov (8) uwDEST_V(2,8)<1> ubAVS_RESPONSE_2(1,0)<8;4,1>
+ mov (8) uwDEST_V(3,8)<1> ubAVS_RESPONSE_2(1,8)<8;4,1>
+ mov (8) uwDEST_V(4,8)<1> ubAVS_RESPONSE_2(6,0)<8;4,1>
+ mov (8) uwDEST_V(5,8)<1> ubAVS_RESPONSE_2(6,8)<8;4,1>
+ mov (8) uwDEST_V(6,8)<1> ubAVS_RESPONSE_2(7,0)<8;4,1>
+ mov (8) uwDEST_V(7,8)<1> ubAVS_RESPONSE_2(7,8)<8;4,1>
+
+ // Move second 8x8 words of U to dest GRF
+ mov (8) uwDEST_U(0,8)<1> ubAVS_RESPONSE_2(4,0)<8;4,1>
+ mov (8) uwDEST_U(1,8)<1> ubAVS_RESPONSE_2(4,8)<8;4,1>
+ mov (8) uwDEST_U(2,8)<1> ubAVS_RESPONSE_2(5,0)<8;4,1>
+ mov (8) uwDEST_U(3,8)<1> ubAVS_RESPONSE_2(5,8)<8;4,1>
+ mov (8) uwDEST_U(4,8)<1> ubAVS_RESPONSE_2(10,0)<8;4,1>
+ mov (8) uwDEST_U(5,8)<1> ubAVS_RESPONSE_2(10,8)<8;4,1>
+ mov (8) uwDEST_U(6,8)<1> ubAVS_RESPONSE_2(11,0)<8;4,1>
+ mov (8) uwDEST_U(7,8)<1> ubAVS_RESPONSE_2(11,8)<8;4,1>
+*/
+#else /* OUTPUT_8_BIT */
+ // Move first 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0)<1> ubAVS_RESPONSE(2,1)<16;4,2>
+ mov (8) uwDEST_Y(1)<1> ubAVS_RESPONSE(2,8+1)<16;4,2>
+ mov (8) uwDEST_Y(2)<1> ubAVS_RESPONSE(3,1)<16;4,2>
+ mov (8) uwDEST_Y(3)<1> ubAVS_RESPONSE(3,8+1)<16;4,2>
+ mov (8) uwDEST_Y(4)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) uwDEST_Y(5)<1> ubAVS_RESPONSE(8,8+1)<16;4,2>
+ mov (8) uwDEST_Y(6)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) uwDEST_Y(7)<1> ubAVS_RESPONSE(9,8+1)<16;4,2>
+
+ // Move first 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0)<1> ubAVS_RESPONSE(0,1)<16;4,2>
+ mov (8) uwDEST_V(1)<1> ubAVS_RESPONSE(0,8+1)<16;4,2>
+ mov (8) uwDEST_V(2)<1> ubAVS_RESPONSE(1,1)<16;4,2>
+ mov (8) uwDEST_V(3)<1> ubAVS_RESPONSE(1,8+1)<16;4,2>
+ mov (8) uwDEST_V(4)<1> ubAVS_RESPONSE(6,1)<16;4,2>
+ mov (8) uwDEST_V(5)<1> ubAVS_RESPONSE(6,8+1)<16;4,2>
+ mov (8) uwDEST_V(6)<1> ubAVS_RESPONSE(7,1)<16;4,2>
+ mov (8) uwDEST_V(7)<1> ubAVS_RESPONSE(7,8+1)<16;4,2>
+
+ // Move first 8x8 words of U to dest GRF
+ mov (8) uwDEST_U(0)<1> ubAVS_RESPONSE(4,1)<16;4,2>
+ mov (8) uwDEST_U(1)<1> ubAVS_RESPONSE(4,8+1)<16;4,2>
+ mov (8) uwDEST_U(2)<1> ubAVS_RESPONSE(5,1)<16;4,2>
+ mov (8) uwDEST_U(3)<1> ubAVS_RESPONSE(5,8+1)<16;4,2>
+ mov (8) uwDEST_U(4)<1> ubAVS_RESPONSE(10,1)<16;4,2>
+ mov (8) uwDEST_U(5)<1> ubAVS_RESPONSE(10,8+1)<16;4,2>
+ mov (8) uwDEST_U(6)<1> ubAVS_RESPONSE(11,1)<16;4,2>
+ mov (8) uwDEST_U(7)<1> ubAVS_RESPONSE(11,8+1)<16;4,2>
+
+ // Move second 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0,8)<1> ubAVS_RESPONSE_2(2,1)<16;4,2>
+ mov (8) uwDEST_Y(1,8)<1> ubAVS_RESPONSE_2(2,8+1)<16;4,2>
+ mov (8) uwDEST_Y(2,8)<1> ubAVS_RESPONSE_2(3,1)<16;4,2>
+ mov (8) uwDEST_Y(3,8)<1> ubAVS_RESPONSE_2(3,8+1)<16;4,2>
+ mov (8) uwDEST_Y(4,8)<1> ubAVS_RESPONSE_2(8,1)<16;4,2>
+ mov (8) uwDEST_Y(5,8)<1> ubAVS_RESPONSE_2(8,8+1)<16;4,2>
+ mov (8) uwDEST_Y(6,8)<1> ubAVS_RESPONSE_2(9,1)<16;4,2>
+ mov (8) uwDEST_Y(7,8)<1> ubAVS_RESPONSE_2(9,8+1)<16;4,2>
+
+ // Move second 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0,8)<1> ubAVS_RESPONSE_2(0,1)<16;4,2>
+ mov (8) uwDEST_V(1,8)<1> ubAVS_RESPONSE_2(0,8+1)<16;4,2>
+ mov (8) uwDEST_V(2,8)<1> ubAVS_RESPONSE_2(1,1)<16;4,2>
+ mov (8) uwDEST_V(3,8)<1> ubAVS_RESPONSE_2(1,8+1)<16;4,2>
+ mov (8) uwDEST_V(4,8)<1> ubAVS_RESPONSE_2(6,1)<16;4,2>
+ mov (8) uwDEST_V(5,8)<1> ubAVS_RESPONSE_2(6,8+1)<16;4,2>
+ mov (8) uwDEST_V(6,8)<1> ubAVS_RESPONSE_2(7,1)<16;4,2>
+ mov (8) uwDEST_V(7,8)<1> ubAVS_RESPONSE_2(7,8+1)<16;4,2>
+
+ // Move second 8x8 words of U to dest GRF
+ mov (8) uwDEST_U(0,8)<1> ubAVS_RESPONSE_2(4,1)<16;4,2>
+ mov (8) uwDEST_U(1,8)<1> ubAVS_RESPONSE_2(4,8+1)<16;4,2>
+ mov (8) uwDEST_U(2,8)<1> ubAVS_RESPONSE_2(5,1)<16;4,2>
+ mov (8) uwDEST_U(3,8)<1> ubAVS_RESPONSE_2(5,8+1)<16;4,2>
+ mov (8) uwDEST_U(4,8)<1> ubAVS_RESPONSE_2(10,1)<16;4,2>
+ mov (8) uwDEST_U(5,8)<1> ubAVS_RESPONSE_2(10,8+1)<16;4,2>
+ mov (8) uwDEST_U(6,8)<1> ubAVS_RESPONSE_2(11,1)<16;4,2>
+ mov (8) uwDEST_U(7,8)<1> ubAVS_RESPONSE_2(11,8+1)<16;4,2>
+#endif
+//------------------------------------------------------------------------------
+
+ // Re-define new number of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_AVS_IEF_Unpack_8x8.asm ----------
+
+// Yoni: In order to optimize unpacking, 3 methods are being checked:
+// 1. AVS_ORIGINAL
+// 2. AVS_ROUND_TO_8_BITS
+// 3. AVS_INDIRECT_ACCESS
+//
+// Only 1 method should stay in the code
+
+
+//#define AVS_ROUND_TO_8_BITS
+//#define AVS_INDIRECT_ACCESS
+
+
+ // Move first 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0)<1> ubAVS_RESPONSE(2,1)<16;4,2>
+ mov (8) uwDEST_Y(1)<1> ubAVS_RESPONSE(2,8+1)<16;4,2>
+ mov (8) uwDEST_Y(2)<1> ubAVS_RESPONSE(3,1)<16;4,2>
+ mov (8) uwDEST_Y(3)<1> ubAVS_RESPONSE(3,8+1)<16;4,2>
+ mov (8) uwDEST_Y(4)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) uwDEST_Y(5)<1> ubAVS_RESPONSE(8,8+1)<16;4,2>
+ mov (8) uwDEST_Y(6)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) uwDEST_Y(7)<1> ubAVS_RESPONSE(9,8+1)<16;4,2>
+
+ // Move first 4x8 words of V to dest GRF
+ mov (4) uwDEST_V(0)<1> ubAVS_RESPONSE(0,1)<16;2,4>
+ mov (4) uwDEST_V(0,8)<1> ubAVS_RESPONSE(1,1)<16;2,4>
+ mov (4) uwDEST_V(1)<1> ubAVS_RESPONSE(6,1)<16;2,4>
+ mov (4) uwDEST_V(1,8)<1> ubAVS_RESPONSE(7,1)<16;2,4>
+
+ // Move first 4x8 words of U to dest GRF
+ mov (4) uwDEST_U(0)<1> ubAVS_RESPONSE(4,1)<16;2,4>
+ mov (4) uwDEST_U(0,8)<1> ubAVS_RESPONSE(5,1)<16;2,4>
+ mov (4) uwDEST_U(1)<1> ubAVS_RESPONSE(10,1)<16;2,4>
+ mov (4) uwDEST_U(1,8)<1> ubAVS_RESPONSE(11,1)<16;2,4>
+
+ // Move second 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0,8)<1> ubAVS_RESPONSE_2(2,1)<16;4,2>
+ mov (8) uwDEST_Y(1,8)<1> ubAVS_RESPONSE_2(2,8+1)<16;4,2>
+ mov (8) uwDEST_Y(2,8)<1> ubAVS_RESPONSE_2(3,1)<16;4,2>
+ mov (8) uwDEST_Y(3,8)<1> ubAVS_RESPONSE_2(3,8+1)<16;4,2>
+ mov (8) uwDEST_Y(4,8)<1> ubAVS_RESPONSE_2(8,1)<16;4,2>
+ mov (8) uwDEST_Y(5,8)<1> ubAVS_RESPONSE_2(8,8+1)<16;4,2>
+ mov (8) uwDEST_Y(6,8)<1> ubAVS_RESPONSE_2(9,1)<16;4,2>
+ mov (8) uwDEST_Y(7,8)<1> ubAVS_RESPONSE_2(9,8+1)<16;4,2>
+
+ // Move second 4x8 words of V to dest GRF
+ mov (4) uwDEST_V(0,4)<1> ubAVS_RESPONSE_2(0,1)<16;2,4>
+ mov (4) uwDEST_V(0,12)<1> ubAVS_RESPONSE_2(1,1)<16;2,4>
+ mov (4) uwDEST_V(1,4)<1> ubAVS_RESPONSE_2(6,1)<16;2,4>
+ mov (4) uwDEST_V(1,12)<1> ubAVS_RESPONSE_2(7,1)<16;2,4>
+
+ // Move second 4x8 words of U to dest GRF
+ mov (4) uwDEST_U(0,4)<1> ubAVS_RESPONSE_2(4,1)<16;2,4>
+ mov (4) uwDEST_U(0,12)<1> ubAVS_RESPONSE_2(5,1)<16;2,4>
+ mov (4) uwDEST_U(1,4)<1> ubAVS_RESPONSE_2(10,1)<16;2,4>
+ mov (4) uwDEST_U(1,12)<1> ubAVS_RESPONSE_2(11,1)<16;2,4>
+
+//------------------------------------------------------------------------------
+
+ // Re-define new number of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_AVS_IEF_Unpack_8x8.asm ----------
+
+// Yoni: In order to optimize unpacking, 3 methods are being checked:
+// 1. AVS_ORIGINAL
+// 2. AVS_ROUND_TO_8_BITS
+// 3. AVS_INDIRECT_ACCESS
+//
+// Only 1 method should stay in the code
+
+
+//#define AVS_ROUND_TO_8_BITS
+//#define AVS_INDIRECT_ACCESS
+
+
+ // Move first 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0)<1> ubAVS_RESPONSE(2,1)<16;4,2>
+ mov (8) uwDEST_Y(1)<1> ubAVS_RESPONSE(2,8+1)<16;4,2>
+ mov (8) uwDEST_Y(2)<1> ubAVS_RESPONSE(3,1)<16;4,2>
+ mov (8) uwDEST_Y(3)<1> ubAVS_RESPONSE(3,8+1)<16;4,2>
+ mov (8) uwDEST_Y(4)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) uwDEST_Y(5)<1> ubAVS_RESPONSE(8,8+1)<16;4,2>
+ mov (8) uwDEST_Y(6)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) uwDEST_Y(7)<1> ubAVS_RESPONSE(9,8+1)<16;4,2>
+
+ // Move first 4x8 words of V to dest GRF
+ mov (4) uwDEST_V(0)<1> ubAVS_RESPONSE(0,1)<16;2,4>
+ mov (4) uwDEST_V(0,8)<1> ubAVS_RESPONSE(0,8+1)<16;2,4>
+ mov (4) uwDEST_V(1)<1> ubAVS_RESPONSE(1,1)<16;2,4>
+ mov (4) uwDEST_V(1,8)<1> ubAVS_RESPONSE(1,8+1)<16;2,4>
+ mov (4) uwDEST_V(2)<1> ubAVS_RESPONSE(6,1)<16;2,4>
+ mov (4) uwDEST_V(2,8)<1> ubAVS_RESPONSE(6,8+1)<16;2,4>
+ mov (4) uwDEST_V(3)<1> ubAVS_RESPONSE(7,1)<16;2,4>
+ mov (4) uwDEST_V(3,8)<1> ubAVS_RESPONSE(7,8+1)<16;2,4>
+
+ // Move first 4x8 words of U to dest GRF
+ mov (4) uwDEST_U(0)<1> ubAVS_RESPONSE(4,1)<16;2,4>
+ mov (4) uwDEST_U(0,8)<1> ubAVS_RESPONSE(4,8+1)<16;2,4>
+ mov (4) uwDEST_U(1)<1> ubAVS_RESPONSE(5,1)<16;2,4>
+ mov (4) uwDEST_U(1,8)<1> ubAVS_RESPONSE(5,8+1)<16;2,4>
+ mov (4) uwDEST_U(2)<1> ubAVS_RESPONSE(10,1)<16;2,4>
+ mov (4) uwDEST_U(2,8)<1> ubAVS_RESPONSE(10,8+1)<16;2,4>
+ mov (4) uwDEST_U(3)<1> ubAVS_RESPONSE(11,1)<16;2,4>
+ mov (4) uwDEST_U(3,8)<1> ubAVS_RESPONSE(11,8+1)<16;2,4>
+
+ // Move second 8x8 words of Y to dest GRF
+ mov (8) uwDEST_Y(0,8)<1> ubAVS_RESPONSE_2(2,1)<16;4,2>
+ mov (8) uwDEST_Y(1,8)<1> ubAVS_RESPONSE_2(2,8+1)<16;4,2>
+ mov (8) uwDEST_Y(2,8)<1> ubAVS_RESPONSE_2(3,1)<16;4,2>
+ mov (8) uwDEST_Y(3,8)<1> ubAVS_RESPONSE_2(3,8+1)<16;4,2>
+ mov (8) uwDEST_Y(4,8)<1> ubAVS_RESPONSE_2(8,1)<16;4,2>
+ mov (8) uwDEST_Y(5,8)<1> ubAVS_RESPONSE_2(8,8+1)<16;4,2>
+ mov (8) uwDEST_Y(6,8)<1> ubAVS_RESPONSE_2(9,1)<16;4,2>
+ mov (8) uwDEST_Y(7,8)<1> ubAVS_RESPONSE_2(9,8+1)<16;4,2>
+
+ // Move second 4x8 words of V to dest GRF
+ mov (4) uwDEST_V(0,4)<1> ubAVS_RESPONSE_2(0,1)<16;2,4>
+ mov (4) uwDEST_V(0,12)<1> ubAVS_RESPONSE_2(0,8+1)<16;2,4>
+ mov (4) uwDEST_V(1,4)<1> ubAVS_RESPONSE_2(1,1)<16;2,4>
+ mov (4) uwDEST_V(1,12)<1> ubAVS_RESPONSE_2(1,8+1)<16;2,4>
+ mov (4) uwDEST_V(2,4)<1> ubAVS_RESPONSE_2(6,1)<16;2,4>
+ mov (4) uwDEST_V(2,12)<1> ubAVS_RESPONSE_2(6,8+1)<16;2,4>
+ mov (4) uwDEST_V(3,4)<1> ubAVS_RESPONSE_2(7,1)<16;2,4>
+ mov (4) uwDEST_V(3,12)<1> ubAVS_RESPONSE_2(7,8+1)<16;2,4>
+
+ // Move second 4x8 words of U to dest GRF
+ mov (4) uwDEST_U(0,4)<1> ubAVS_RESPONSE_2(4,1)<16;2,4>
+ mov (4) uwDEST_U(0,12)<1> ubAVS_RESPONSE_2(4,8+1)<16;2,4>
+ mov (4) uwDEST_U(1,4)<1> ubAVS_RESPONSE_2(5,1)<16;2,4>
+ mov (4) uwDEST_U(1,12)<1> ubAVS_RESPONSE_2(5,8+1)<16;2,4>
+ mov (4) uwDEST_U(2,4)<1> ubAVS_RESPONSE_2(10,1)<16;2,4>
+ mov (4) uwDEST_U(2,12)<1> ubAVS_RESPONSE_2(10,8+1)<16;2,4>
+ mov (4) uwDEST_U(3,4)<1> ubAVS_RESPONSE_2(11,1)<16;2,4>
+ mov (4) uwDEST_U(3,12)<1> ubAVS_RESPONSE_2(11,8+1)<16;2,4>
+
+//------------------------------------------------------------------------------
+
+ // Re-define new number of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_ENABLE
+
+ #include "DNDI.inc"
+
+ #ifdef DI_ONLY
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DI // set the number of GRF
+ #else
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DNDI // set the number of GRF
+ #endif
+
+ #undef nDPW_BLOCK_SIZE_HIST
+ #define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_1 // HIST Block Size for Write is 4x2
+ #undef nDPW_BLOCK_SIZE_DN
+ #define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_32+nBLOCK_HEIGHT_4 // DN Block Size for Write is 32x4
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+ #include "DNDI_Command.asm"
+
+////////////////////////////////////// Rearrange for Internal Planar //////////////////////////////
+ //// move the previous frame Y component to internal planar format
+ //$for (0; <nY_NUM_OF_ROWS/2; 1) {
+ // mov (16) uwDEST_Y(%1,0)<1> ubRESP(nDI_PREV_FRAME_LUMA_OFFSET,%1*16)
+ //}
+ //// move the previous frame U,V components to internal planar format
+ //$for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ // mov (8) uwDEST_U(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ // mov (8) uwDEST_V(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ //}
+ //// move the current frame Y component to internal planar format
+ //$for (0; <nY_NUM_OF_ROWS/2; 1) {
+ // mov (16) uwDEST_Y(%1+4,0)<1> ubRESP(nDI_CURR_FRAME_LUMA_OFFSET,%1*16)
+ //}
+ //// move the current frame U,V components to internal planar format
+ //$for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ // mov (8) uwDEST_U(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ // mov (8) uwDEST_V(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ //}
+
+////////////////////////////////////// Save the STMM Data for Next Run /////////////////////////
+ // Write STMM to memory
+ shr (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w NODDCLR // X origin / 2
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w NODDCLR_NODDCHK // Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_STMM:ud NODDCHK // block width and height (8x4)
+ mov (8) mudMSGHDR_STMM(0)<1> rMSGSRC.0<8;8,1>:ud // message header
+ mov (8) mudMSGHDR_STMM(1)<1> udRESP(nDI_STMM_OFFSET,0) // Move STMM to MRF
+ send (8) dNULLREG mMSGHDR_STMM udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_STMM+nBI_STMM_HISTORY_OUTPUT:ud
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+#ifdef DI_ONLY
+#else
+
+ #include "DI_Hist_Save.asm"
+
+////////////////////////////////////// Pack and Save the DN Curr Frame for Next Run ///////////////
+ // check top/bottom field first
+ cmp.e.f0.0 (1) null<1>:w ubTFLD_FIRST<0;1,0>:ub 1:w
+
+ add (4) pCF_Y_OFFSET<1>:uw ubSRC_CF_OFFSET<4;4,1>:ub npDN_YUV:uw
+ //set the save DN position
+ shl (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w NODDCLR // X origin * 2
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w NODDCLR_NODDCHK // Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud NODDCHK // block width and height (8x4)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+
+
+ (f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ //$for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ // mov (16) r[pCF_Y_OFFSET, %1*32]<2>:ub ubRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*16) // 2nd field luma from current frame (line 0,2)
+ // mov (16) r[pCF_Y_OFFSET, %1+1*32]<2>:ub ubRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,16) // 1st field luma from current frame (line 1,3)
+ // mov (8) r[pCF_U_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16+1)<16;8,2> // 2nd field U from current frame (line 0,2)
+ // mov (8) r[pCF_V_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16)<16;8,2> // 2nd field V from current frame (line 0,2)
+ // mov (8) r[pCF_U_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,16+1)<16;8,2> // 1st field U from current frame (line 1,3)
+ // mov (8) r[pCF_V_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,16)<16;8,2> // 1st field U from current frame (line 1,3)
+ //}
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (16) r[pCF_Y_OFFSET, %1*32]<2>:ub ubRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*16) // 2nd field luma from current frame (line 0,2)
+ mov (16) r[pCF_Y_OFFSET, %1+1*32]<2>:ub ubRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,16) // 1st field luma from current frame (line 1,3)
+ }
+
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (8) r[pCF_U_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16+1)<16;8,2> // 2nd field U from current frame (line 0,2)
+ mov (8) r[pCF_U_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,16+1)<16;8,2> // 1st field U from current frame (line 1,3)
+ }
+
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (8) r[pCF_V_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16)<16;8,2> // 2nd field V from current frame (line 0,2)
+ mov (8) r[pCF_V_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,16)<16;8,2> // 1st field U from current frame (line 1,3)
+ }
+
+ jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ //$for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ // mov (16) r[pCF_Y_OFFSET, %1*32]<2>:ub ubRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,0) // 1st field luma from current frame (line 0,2)
+ // mov (16) r[pCF_Y_OFFSET, %1+1*32]<2>:ub ubRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*16) // 2nd field luma from current frame (line 1,3)
+ // mov (8) r[pCF_U_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,1)<16;8,2> // 1st field U from current frame (line 0,2)
+ // mov (8) r[pCF_V_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,0)<16;8,2> // 1st field V from current frame (line 0,2)
+ // mov (8) r[pCF_U_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16+1)<16;8,2> // 2nd field U from current frame (line 1,3)
+ // mov (8) r[pCF_V_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16)<16;8,2> // 2nd field V from current frame (line 1,3)
+ //}
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (16) r[pCF_Y_OFFSET, %1*32]<2>:ub ubRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,0) // 1st field luma from current frame (line 0,2)
+ mov (16) r[pCF_Y_OFFSET, %1+1*32]<2>:ub ubRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*16) // 2nd field luma from current frame (line 1,3)
+ }
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (8) r[pCF_U_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,1)<16;8,2> // 1st field U from current frame (line 0,2)
+ mov (8) r[pCF_U_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16+1)<16;8,2> // 2nd field U from current frame (line 1,3)
+ }
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (8) r[pCF_V_OFFSET, %1*32]<4>:ub ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET+%2,0)<16;8,2> // 1st field V from current frame (line 0,2)
+ mov (8) r[pCF_V_OFFSET, %1+1*32]<4>:ub ubRESP(nDI_CURR_2ND_FIELD_CHROMA_OFFSET,%2*16)<16;8,2> // 2nd field V from current frame (line 1,3)
+ }
+
+SAVE_DN_CURR:
+ $for(0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (8) mudMSGHDR_DN(%1+1)<1> udDN_YUV(%1)REGION(8,1)
+ }
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PA_DN_DI+nBI_DESTINATION_YUV:ud
+#endif
+
+// Save Processed frames
+#include "DI_Save_PA.asm"
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_DISABLE
+
+#include "DNDI.inc"
+
+#undef nY_NUM_OF_ROWS
+#define nY_NUM_OF_ROWS 8 // Number of Y rows per block
+#undef nUV_NUM_OF_ROWS
+#define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+
+#undef nSMPL_RESP_LEN
+#define nSMPL_RESP_LEN nSMPL_RESP_LEN_DN_PA // Set the Number of GRFs in DNDI response
+#undef nDPW_BLOCK_SIZE_DN
+#define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_32+nBLOCK_HEIGHT_8 // DN Curr Block Size for Write is 32x8
+#undef nDPW_BLOCK_SIZE_HIST
+#define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_2 // HIST Block Size for Write is 4x2
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+#include "DNDI_COMMAND.asm"
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+#include "DNDI_Hist_Save.asm"
+
+////////////////////////////////////// Pack and Save the DN Curr Frame for Next Run ///////////////
+add (4) pCF_Y_OFFSET<1>:uw ubDEST_CF_OFFSET<4;4,1>:ub npDN_YUV:w
+$for (0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) r[pCF_Y_OFFSET, %1*32]<2>:ub ubRESP(nNODI_LUMA_OFFSET,%1*16)<16;16,1> // copy line of Y
+}
+$for (0; <nUV_NUM_OF_ROWS; 1) {
+ mov (8) r[pCF_U_OFFSET, %1*32]<4>:ub ubRESP(nNODI_CHROMA_OFFSET,%1*16+1)<16;8,2> // copy line of U
+ mov (8) r[pCF_V_OFFSET, %1*32]<4>:ub ubRESP(nNODI_CHROMA_OFFSET,%1*16)<16;8,2> // copy line of V
+}
+
+shl (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w // X origin * 2 (422 output)
+mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w // Y origin
+mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud // block width and height (32x8)
+mov (8) mMSGHDR_DN<1>:ud rMSGSRC<8;8,1>:ud // message header
+
+$for(0; <nY_NUM_OF_ROWS; 2) {
+ mov (16) mudMSGHDR_DN(1+%1)<1> udDN_YUV(%1)REGION(8,1) // Move DN Curr to MRF
+}
+send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PA_DN_NODI+nBI_DESTINATION_YUV:ud
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PA_Scaling.asm ----------
+#include "Scaling.inc"
+
+ // Build 16 elements ramp in float32 and normalized it
+// mov (8) SAMPLER_RAMP(0)<1> 0x76543210:v
+// add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+mov (4) SAMPLER_RAMP(0)<1> 0x48403000:vf //3, 2, 1, 0 in float vector
+mov (4) SAMPLER_RAMP(0,4)<1> 0x5C585450:vf //7, 6, 5, 4 in float vector
+add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+
+//Module: PrepareScaleCoord.asm
+
+ // Setup for sampler msg hdr
+ mov (2) rMSGSRC.0<1>:ud 0:ud { NoDDClr } // Unused fields
+ mov (1) rMSGSRC.2<1>:ud 0:ud { NoDDChk } // Write and offset
+
+ // Calculate 16 v based on the step Y and vertical origin
+ mov (16) mfMSGPAYLOAD(2)<1> fSRC_VID_V_ORI<0;1,0>:f
+ mov (16) SCALE_COORD_Y<1>:f fSRC_VID_V_ORI<0;1,0>:f
+
+ // Calculate 16 u based on the step X and hori origin
+// line (16) mfMSGPAYLOAD(0)<1> SCALE_STEP_X<0;1,0>:f SAMPLER_RAMP(0) // Assign to mrf directly
+ mov (16) acc0:f fSRC_VID_H_ORI<0;1,0>:f { Compr }
+ mac (16) mfMSGPAYLOAD(0)<1> fVIDEO_STEP_X<0;1,0>:f SAMPLER_RAMP(0) { Compr }
+
+ //Setup the constants for line instruction
+ mov (1) SCALE_LINE_P255<1>:f 255.0:f { NoDDClr } //{ NoDDClr, NoDDChk }
+ mov (1) SCALE_LINE_P0_5<1>:f 0.5:f { NoDDChk }
+
+//------------------------------------------------------------------------------
+
+$for (0; <nY_NUM_OF_ROWS; 1) {
+
+ // Read 16 sampled pixels and store them in float32 in 8 GRFs in the order of BGRA (VYUA).
+ mov (8) MSGHDR_SCALE.0:ud rMSGSRC.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (16) SCALE_RESPONSE_YW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_YUV+nBI_CURRENT_SRC_YUV
+
+ // Calculate 16 v for next line
+ add (16) mfMSGPAYLOAD(2)<1> SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+ add (16) SCALE_COORD_Y<1>:f SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+
+ // Scale back to [0, 255], convert f to ud
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(0) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(0)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(2) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(2)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(4) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(4)<1> acc0:f { Compr }
+
+ mov (16) DEST_V(%1)<1> SCALE_RESPONSE_YB(0) //possible error due to truncation - vK
+ mov (16) DEST_Y(%1)<1> SCALE_RESPONSE_YB(2) //possible error due to truncation - vK
+ mov (16) DEST_U(%1)<1> SCALE_RESPONSE_YB(4) //possible error due to truncation - vK
+
+}
+
+ #define nSRC_REGION nREGION_1
+
+//------------------------------------------------------------------------------
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL2_AVS_IEF_16x8.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 Y each
+// 2 sampler read for 8x8 U and 8x8 V (NV11\P208 input surface)
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ // Enable green channel only
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud
+
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+ // 8x8 U and V sampling
+ // Enable red and blue channels
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_BLUE_CHANNELS:ud
+
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_2CH+nSI_SRC_UV+nBI_CURRENT_SRC_UV
+ // Return U and V in 8 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+
+ // 2nd 8x8 Y sampling
+ // Enable green channel only
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud
+
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+
+ // 2nd 8x8 U and V sampling
+ // Enable red and blue channels
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_BLUE_CHANNELS:ud
+
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_2CH+nSI_SRC_UV+nBI_CURRENT_SRC_UV
+ // Return U and V in 8 GRFs
+
+//------------------------------------------------------------------------------
+// Unpacking sampler reads to 4:4:4 internal planar
+//------------------------------------------------------------------------------
+ #include "PL2_AVS_IEF_Unpack_16x8.asm"
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL2_AVS_IEF_8x4.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 Y each
+// 1 sampler read for 8x8 U and 8x8 V (NV11\NV12 input surface)
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ // Enable green channel only
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud
+
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+ // 8x8 U and V sampling
+ // Enable red and blue channels
+ //Only 8x4 wil be used
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_BLUE_CHANNELS:ud
+
+ // Calculate Chroma Step Size:
+ // for H direction: 16 Luma samples are covered by 8 Chroma samples. Thus Chroma_Step_X = 2 * Luma_Step_X
+ // for V direction: 8 Luma samples are covered by 8 Chroma samples. Thus Chroma_Step_Y = Luma_Step_Y
+ mul (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_X:f 2.0:f // Step X for chroma
+
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_2CH+nSI_SRC_UV+nBI_CURRENT_SRC_UV
+ // Return U and V in 8 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+
+ // 2nd 8x8 Y sampling
+ // Enable green channel only
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud
+
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+
+//------------------------------------------------------------------------------
+// Unpacking sampler reads to 4:2:0 internal planar
+//------------------------------------------------------------------------------
+ #include "PL2_AVS_IEF_Unpack_8x4.asm"
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL2_AVS_IEF_8x8.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 Y each
+// 1 sampler read for 8x8 U and 8x8 V (NV11\NV12 input surface)
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ // Enable green channel only
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud
+
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+ // 8x8 U and V sampling
+ // Enable red and blue channels
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_BLUE_CHANNELS:ud
+
+ // Calculate Chroma Step Size:
+ // for H direction: 16 Luma samples are covered by 8 Chroma samples. Thus Chroma_Step_X = 2 * Luma_Step_X
+ // for V direction: 8 Luma samples are covered by 8 Chroma samples. Thus Chroma_Step_Y = Luma_Step_Y
+ mul (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_X:f 2.0:f // Step X for chroma
+
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_2CH+nSI_SRC_UV+nBI_CURRENT_SRC_UV
+ // Return U and V in 8 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+
+ // 2nd 8x8 Y sampling
+ // Enable green channel only
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud
+
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+
+//------------------------------------------------------------------------------
+// Unpacking sampler reads to 4:2:2 internal planar
+//------------------------------------------------------------------------------
+ #include "PL2_AVS_IEF_Unpack_8x8.asm"
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL2_AVS_IEF_Unpack_16x8.asm ----------
+
+#ifdef AVS_OUTPUT_16_BIT //Output is packed in AVYU format
+// Move first 8x8 words of Y to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,1)<4> uwAVS_RESPONSE(0,0)<4;4,1>
+ mov (4) uwDEST_Y(1,1)<4> uwAVS_RESPONSE(0,8)<4;4,1>
+ mov (4) uwDEST_Y(4,1)<4> uwAVS_RESPONSE(0,4)<4;4,1>
+ mov (4) uwDEST_Y(5,1)<4> uwAVS_RESPONSE(0,12)<4;4,1>
+ mov (4) uwDEST_Y(8,1)<4> uwAVS_RESPONSE(1,0)<4;4,1>
+ mov (4) uwDEST_Y(9,1)<4> uwAVS_RESPONSE(1,8)<4;4,1>
+ mov (4) uwDEST_Y(12,1)<4> uwAVS_RESPONSE(1,4)<4;4,1>
+ mov (4) uwDEST_Y(13,1)<4> uwAVS_RESPONSE(1,12)<4;4,1>
+ mov (4) uwDEST_Y(16,1)<4> uwAVS_RESPONSE(2,0)<4;4,1>
+ mov (4) uwDEST_Y(17,1)<4> uwAVS_RESPONSE(2,8)<4;4,1>
+ mov (4) uwDEST_Y(20,1)<4> uwAVS_RESPONSE(2,4)<4;4,1>
+ mov (4) uwDEST_Y(21,1)<4> uwAVS_RESPONSE(2,12)<4;4,1>
+ mov (4) uwDEST_Y(24,1)<4> uwAVS_RESPONSE(3,0)<4;4,1>
+ mov (4) uwDEST_Y(25,1)<4> uwAVS_RESPONSE(3,8)<4;4,1>
+ mov (4) uwDEST_Y(28,1)<4> uwAVS_RESPONSE(3,4)<4;4,1>
+ mov (4) uwDEST_Y(29,1)<4> uwAVS_RESPONSE(3,12)<4;4,1>
+
+// Move first 8x8 words of U to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,0)<4> uwAVS_RESPONSE(4,0)<4;4,1>
+ mov (4) uwDEST_Y(1,0)<4> uwAVS_RESPONSE(4,8)<4;4,1>
+ mov (4) uwDEST_Y(4,0)<4> uwAVS_RESPONSE(4,4)<4;4,1>
+ mov (4) uwDEST_Y(5,0)<4> uwAVS_RESPONSE(4,12)<4;4,1>
+ mov (4) uwDEST_Y(8,0)<4> uwAVS_RESPONSE(5,0)<4;4,1>
+ mov (4) uwDEST_Y(9,0)<4> uwAVS_RESPONSE(5,8)<4;4,1>
+ mov (4) uwDEST_Y(12,0)<4> uwAVS_RESPONSE(5,4)<4;4,1>
+ mov (4) uwDEST_Y(13,0)<4> uwAVS_RESPONSE(5,12)<4;4,1>
+ mov (4) uwDEST_Y(16,0)<4> uwAVS_RESPONSE(8,0)<4;4,1>
+ mov (4) uwDEST_Y(17,0)<4> uwAVS_RESPONSE(8,8)<4;4,1>
+ mov (4) uwDEST_Y(20,0)<4> uwAVS_RESPONSE(8,4)<4;4,1>
+ mov (4) uwDEST_Y(21,0)<4> uwAVS_RESPONSE(8,12)<4;4,1>
+ mov (4) uwDEST_Y(24,0)<4> uwAVS_RESPONSE(9,0)<4;4,1>
+ mov (4) uwDEST_Y(25,0)<4> uwAVS_RESPONSE(9,8)<4;4,1>
+ mov (4) uwDEST_Y(28,0)<4> uwAVS_RESPONSE(9,4)<4;4,1>
+ mov (4) uwDEST_Y(29,0)<4> uwAVS_RESPONSE(9,12)<4;4,1>
+
+// Move first 8x8 words of V to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,2)<4> uwAVS_RESPONSE(6,0)<4;4,1>
+ mov (4) uwDEST_Y(1,2)<4> uwAVS_RESPONSE(6,8)<4;4,1>
+ mov (4) uwDEST_Y(4,2)<4> uwAVS_RESPONSE(6,4)<4;4,1>
+ mov (4) uwDEST_Y(5,2)<4> uwAVS_RESPONSE(6,12)<4;4,1>
+ mov (4) uwDEST_Y(8,2)<4> uwAVS_RESPONSE(7,0)<4;4,1>
+ mov (4) uwDEST_Y(9,2)<4> uwAVS_RESPONSE(7,8)<4;4,1>
+ mov (4) uwDEST_Y(12,2)<4> uwAVS_RESPONSE(7,4)<4;4,1>
+ mov (4) uwDEST_Y(13,2)<4> uwAVS_RESPONSE(7,12)<4;4,1>
+ mov (4) uwDEST_Y(16,2)<4> uwAVS_RESPONSE(10,0)<4;4,1>
+ mov (4) uwDEST_Y(17,2)<4> uwAVS_RESPONSE(10,8)<4;4,1>
+ mov (4) uwDEST_Y(20,2)<4> uwAVS_RESPONSE(10,4)<4;4,1>
+ mov (4) uwDEST_Y(21,2)<4> uwAVS_RESPONSE(10,12)<4;4,1>
+ mov (4) uwDEST_Y(24,2)<4> uwAVS_RESPONSE(11,0)<4;4,1>
+ mov (4) uwDEST_Y(25,2)<4> uwAVS_RESPONSE(11,8)<4;4,1>
+ mov (4) uwDEST_Y(28,2)<4> uwAVS_RESPONSE(11,4)<4;4,1>
+ mov (4) uwDEST_Y(29,2)<4> uwAVS_RESPONSE(11,12)<4;4,1>
+
+// Move first 8x8 words of A to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,3)<4> 0:uw
+ mov (4) uwDEST_Y(1,3)<4> 0:uw
+ mov (4) uwDEST_Y(4,3)<4> 0:uw
+ mov (4) uwDEST_Y(5,3)<4> 0:uw
+ mov (4) uwDEST_Y(8,3)<4> 0:uw
+ mov (4) uwDEST_Y(9,3)<4> 0:uw
+ mov (4) uwDEST_Y(12,3)<4> 0:uw
+ mov (4) uwDEST_Y(13,3)<4> 0:uw
+ mov (4) uwDEST_Y(16,3)<4> 0:uw
+ mov (4) uwDEST_Y(17,3)<4> 0:uw
+ mov (4) uwDEST_Y(20,3)<4> 0:uw
+ mov (4) uwDEST_Y(21,3)<4> 0:uw
+ mov (4) uwDEST_Y(24,3)<4> 0:uw
+ mov (4) uwDEST_Y(25,3)<4> 0:uw
+ mov (4) uwDEST_Y(28,3)<4> 0:uw
+ mov (4) uwDEST_Y(29,3)<4> 0:uw
+
+// Move second 8x8 words of Y to dest GRF
+ mov (4) uwDEST_Y(2,1)<4> uwAVS_RESPONSE_2(0,0)<4;4,1>
+ mov (4) uwDEST_Y(3,1)<4> uwAVS_RESPONSE_2(0,8)<4;4,1>
+ mov (4) uwDEST_Y(6,1)<4> uwAVS_RESPONSE_2(0,4)<4;4,1>
+ mov (4) uwDEST_Y(7,1)<4> uwAVS_RESPONSE_2(0,12)<4;4,1>
+ mov (4) uwDEST_Y(10,1)<4> uwAVS_RESPONSE_2(1,0)<4;4,1>
+ mov (4) uwDEST_Y(11,1)<4> uwAVS_RESPONSE_2(1,8)<4;4,1>
+ mov (4) uwDEST_Y(14,1)<4> uwAVS_RESPONSE_2(1,4)<4;4,1>
+ mov (4) uwDEST_Y(15,1)<4> uwAVS_RESPONSE_2(1,12)<4;4,1>
+ mov (4) uwDEST_Y(18,1)<4> uwAVS_RESPONSE_2(2,0)<4;4,1>
+ mov (4) uwDEST_Y(19,1)<4> uwAVS_RESPONSE_2(2,8)<4;4,1>
+ mov (4) uwDEST_Y(22,1)<4> uwAVS_RESPONSE_2(2,4)<4;4,1>
+ mov (4) uwDEST_Y(23,1)<4> uwAVS_RESPONSE_2(2,12)<4;4,1>
+ mov (4) uwDEST_Y(26,1)<4> uwAVS_RESPONSE_2(3,0)<4;4,1>
+ mov (4) uwDEST_Y(27,1)<4> uwAVS_RESPONSE_2(3,8)<4;4,1>
+ mov (4) uwDEST_Y(30,1)<4> uwAVS_RESPONSE_2(3,4)<4;4,1>
+ mov (4) uwDEST_Y(31,1)<4> uwAVS_RESPONSE_2(3,12)<4;4,1>
+
+// Move second 8x8 words of U to dest GRF
+ mov (4) uwDEST_Y(2,0)<4> uwAVS_RESPONSE_2(4,0)<4;4,1>
+ mov (4) uwDEST_Y(3,0)<4> uwAVS_RESPONSE_2(4,8)<4;4,1>
+ mov (4) uwDEST_Y(6,0)<4> uwAVS_RESPONSE_2(4,4)<4;4,1>
+ mov (4) uwDEST_Y(7,0)<4> uwAVS_RESPONSE_2(4,12)<4;4,1>
+ mov (4) uwDEST_Y(10,0)<4> uwAVS_RESPONSE_2(5,0)<4;4,1>
+ mov (4) uwDEST_Y(11,0)<4> uwAVS_RESPONSE_2(5,8)<4;4,1>
+ mov (4) uwDEST_Y(14,0)<4> uwAVS_RESPONSE_2(5,4)<4;4,1>
+ mov (4) uwDEST_Y(15,0)<4> uwAVS_RESPONSE_2(5,12)<4;4,1>
+ mov (4) uwDEST_Y(18,0)<4> uwAVS_RESPONSE_2(8,0)<4;4,1>
+ mov (4) uwDEST_Y(19,0)<4> uwAVS_RESPONSE_2(8,8)<4;4,1>
+ mov (4) uwDEST_Y(22,0)<4> uwAVS_RESPONSE_2(8,4)<4;4,1>
+ mov (4) uwDEST_Y(23,0)<4> uwAVS_RESPONSE_2(8,12)<4;4,1>
+ mov (4) uwDEST_Y(26,0)<4> uwAVS_RESPONSE_2(9,0)<4;4,1>
+ mov (4) uwDEST_Y(27,0)<4> uwAVS_RESPONSE_2(9,8)<4;4,1>
+ mov (4) uwDEST_Y(30,0)<4> uwAVS_RESPONSE_2(9,4)<4;4,1>
+ mov (4) uwDEST_Y(31,0)<4> uwAVS_RESPONSE_2(9,12)<4;4,1>
+
+// Move second 8x8 words of V to dest GRF
+ mov (4) uwDEST_Y(2,2)<4> uwAVS_RESPONSE_2(6,0)<4;4,1>
+ mov (4) uwDEST_Y(3,2)<4> uwAVS_RESPONSE_2(6,8)<4;4,1>
+ mov (4) uwDEST_Y(6,2)<4> uwAVS_RESPONSE_2(6,4)<4;4,1>
+ mov (4) uwDEST_Y(7,2)<4> uwAVS_RESPONSE_2(6,12)<4;4,1>
+ mov (4) uwDEST_Y(10,2)<4> uwAVS_RESPONSE_2(7,0)<4;4,1>
+ mov (4) uwDEST_Y(11,2)<4> uwAVS_RESPONSE_2(7,8)<4;4,1>
+ mov (4) uwDEST_Y(14,2)<4> uwAVS_RESPONSE_2(7,4)<4;4,1>
+ mov (4) uwDEST_Y(15,2)<4> uwAVS_RESPONSE_2(7,12)<4;4,1>
+ mov (4) uwDEST_Y(18,2)<4> uwAVS_RESPONSE_2(10,0)<4;4,1>
+ mov (4) uwDEST_Y(19,2)<4> uwAVS_RESPONSE_2(10,8)<4;4,1>
+ mov (4) uwDEST_Y(22,2)<4> uwAVS_RESPONSE_2(10,4)<4;4,1>
+ mov (4) uwDEST_Y(23,2)<4> uwAVS_RESPONSE_2(10,12)<4;4,1>
+ mov (4) uwDEST_Y(26,2)<4> uwAVS_RESPONSE_2(11,0)<4;4,1>
+ mov (4) uwDEST_Y(27,2)<4> uwAVS_RESPONSE_2(11,8)<4;4,1>
+ mov (4) uwDEST_Y(30,2)<4> uwAVS_RESPONSE_2(11,4)<4;4,1>
+ mov (4) uwDEST_Y(31,2)<4> uwAVS_RESPONSE_2(11,12)<4;4,1>
+
+// Move second 8x8 words of A to dest GRF
+ mov (4) uwDEST_Y(2,3)<4> 0:uw
+ mov (4) uwDEST_Y(3,3)<4> 0:uw
+ mov (4) uwDEST_Y(6,3)<4> 0:uw
+ mov (4) uwDEST_Y(7,3)<4> 0:uw
+ mov (4) uwDEST_Y(10,3)<4> 0:uw
+ mov (4) uwDEST_Y(11,3)<4> 0:uw
+ mov (4) uwDEST_Y(14,3)<4> 0:uw
+ mov (4) uwDEST_Y(15,3)<4> 0:uw
+ mov (4) uwDEST_Y(18,3)<4> 0:uw
+ mov (4) uwDEST_Y(19,3)<4> 0:uw
+ mov (4) uwDEST_Y(22,3)<4> 0:uw
+ mov (4) uwDEST_Y(23,3)<4> 0:uw
+ mov (4) uwDEST_Y(26,3)<4> 0:uw
+ mov (4) uwDEST_Y(27,3)<4> 0:uw
+ mov (4) uwDEST_Y(30,3)<4> 0:uw
+ mov (4) uwDEST_Y(31,3)<4> 0:uw
+
+/* This section will be used if 16-bit output is needed in planar format -vK
+ // Move 1st 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> uwAVS_RESPONSE(%1,0)<8;4,1>
+ mov (8) uwDEST_Y(%1*2+1)<1> uwAVS_RESPONSE(%1,8)<8;4,1>
+ }
+
+ // Move 1st 8x8 words of U to dest GRF (Copy high byte in a word)
+ mov (8) uwDEST_U(0)<1> uwAVS_RESPONSE(4,0)<8;4,1>
+ mov (8) uwDEST_U(1)<1> uwAVS_RESPONSE(4,8)<8;4,1>
+ mov (8) uwDEST_U(2)<1> uwAVS_RESPONSE(5,0)<8;4,1>
+ mov (8) uwDEST_U(3)<1> uwAVS_RESPONSE(5,8)<8;4,1>
+ mov (8) uwDEST_U(4)<1> uwAVS_RESPONSE(8,0)<8;4,1>
+ mov (8) uwDEST_U(5)<1> uwAVS_RESPONSE(8,8)<8;4,1>
+ mov (8) uwDEST_U(6)<1> uwAVS_RESPONSE(9,0)<8;4,1>
+ mov (8) uwDEST_U(7)<1> uwAVS_RESPONSE(9,8)<8;4,1>
+
+ // Move 1st 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0)<1> uwAVS_RESPONSE(6,0)<8;4,1>
+ mov (8) uwDEST_V(1)<1> uwAVS_RESPONSE(6,8)<8;4,1>
+ mov (8) uwDEST_V(2)<1> uwAVS_RESPONSE(7,0)<8;4,1>
+ mov (8) uwDEST_V(3)<1> uwAVS_RESPONSE(7,8)<8;4,1>
+ mov (8) uwDEST_V(4)<1> uwAVS_RESPONSE(10,0)<8;4,1>
+ mov (8) uwDEST_V(5)<1> uwAVS_RESPONSE(10,8)<8;4,1>
+ mov (8) uwDEST_V(6)<1> uwAVS_RESPONSE(11,0)<8;4,1>
+ mov (8) uwDEST_V(7)<1> uwAVS_RESPONSE(11,8)<8;4,1>
+
+ // Move 2nd 8x8 words of Y to dest GRF at higher 8 words of each GRF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> uwAVS_RESPONSE_2(%1,0)<8;4,1>
+ mov (8) uwDEST_Y(%1*2+1,8)<1> uwAVS_RESPONSE_2(%1,8)<8;4,1>
+ }
+
+ // Move 2st 8x8 words of U to dest GRF (Copy high byte in a word)
+ mov (8) uwDEST_U(0,8)<1> uwAVS_RESPONSE_2(4,0)<8;4,1>
+ mov (8) uwDEST_U(1,8)<1> uwAVS_RESPONSE_2(4,8)<8;4,1>
+ mov (8) uwDEST_U(2,8)<1> uwAVS_RESPONSE_2(5,0)<8;4,1>
+ mov (8) uwDEST_U(3,8)<1> uwAVS_RESPONSE_2(5,8)<8;4,1>
+ mov (8) uwDEST_U(4,8)<1> uwAVS_RESPONSE_2(8,0)<8;4,1>
+ mov (8) uwDEST_U(5,8)<1> uwAVS_RESPONSE_2(8,8)<8;4,1>
+ mov (8) uwDEST_U(6,8)<1> uwAVS_RESPONSE_2(9,0)<8;4,1>
+ mov (8) uwDEST_U(7,8)<1> uwAVS_RESPONSE_2(9,8)<8;4,1>
+
+ // Move 2st 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0,8)<1> uwAVS_RESPONSE_2(6,0)<8;4,1>
+ mov (8) uwDEST_V(1,8)<1> uwAVS_RESPONSE_2(6,8)<8;4,1>
+ mov (8) uwDEST_V(2,8)<1> uwAVS_RESPONSE_2(7,0)<8;4,1>
+ mov (8) uwDEST_V(3,8)<1> uwAVS_RESPONSE_2(7,8)<8;4,1>
+ mov (8) uwDEST_V(4,8)<1> uwAVS_RESPONSE_2(10,0)<8;4,1>
+ mov (8) uwDEST_V(5,8)<1> uwAVS_RESPONSE_2(10,8)<8;4,1>
+ mov (8) uwDEST_V(6,8)<1> uwAVS_RESPONSE_2(11,0)<8;4,1>
+ mov (8) uwDEST_V(7,8)<1> uwAVS_RESPONSE_2(11,8)<8;4,1>
+*/
+#else
+ // Move 1st 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> ubAVS_RESPONSE(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1)<1> ubAVS_RESPONSE(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 1st 8x8 words of U to dest GRF (Copy high byte in a word)
+ mov (8) uwDEST_U(0)<1> ubAVS_RESPONSE(4,1)<16;4,2>
+ mov (8) uwDEST_U(1)<1> ubAVS_RESPONSE(4,8+1)<16;4,2>
+ mov (8) uwDEST_U(2)<1> ubAVS_RESPONSE(5,1)<16;4,2>
+ mov (8) uwDEST_U(3)<1> ubAVS_RESPONSE(5,8+1)<16;4,2>
+ mov (8) uwDEST_U(4)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) uwDEST_U(5)<1> ubAVS_RESPONSE(8,8+1)<16;4,2>
+ mov (8) uwDEST_U(6)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) uwDEST_U(7)<1> ubAVS_RESPONSE(9,8+1)<16;4,2>
+
+ // Move 1st 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0)<1> ubAVS_RESPONSE(6,1)<16;4,2>
+ mov (8) uwDEST_V(1)<1> ubAVS_RESPONSE(6,8+1)<16;4,2>
+ mov (8) uwDEST_V(2)<1> ubAVS_RESPONSE(7,1)<16;4,2>
+ mov (8) uwDEST_V(3)<1> ubAVS_RESPONSE(7,8+1)<16;4,2>
+ mov (8) uwDEST_V(4)<1> ubAVS_RESPONSE(10,1)<16;4,2>
+ mov (8) uwDEST_V(5)<1> ubAVS_RESPONSE(10,8+1)<16;4,2>
+ mov (8) uwDEST_V(6)<1> ubAVS_RESPONSE(11,1)<16;4,2>
+ mov (8) uwDEST_V(7)<1> ubAVS_RESPONSE(11,8+1)<16;4,2>
+
+ // Move 2nd 8x8 words of Y to dest GRF at higher 8 words of each GRF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> ubAVS_RESPONSE_2(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1,8)<1> ubAVS_RESPONSE_2(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 2st 8x8 words of U to dest GRF (Copy high byte in a word)
+ mov (8) uwDEST_U(0,8)<1> ubAVS_RESPONSE_2(4,1)<16;4,2>
+ mov (8) uwDEST_U(1,8)<1> ubAVS_RESPONSE_2(4,8+1)<16;4,2>
+ mov (8) uwDEST_U(2,8)<1> ubAVS_RESPONSE_2(5,1)<16;4,2>
+ mov (8) uwDEST_U(3,8)<1> ubAVS_RESPONSE_2(5,8+1)<16;4,2>
+ mov (8) uwDEST_U(4,8)<1> ubAVS_RESPONSE_2(8,1)<16;4,2>
+ mov (8) uwDEST_U(5,8)<1> ubAVS_RESPONSE_2(8,8+1)<16;4,2>
+ mov (8) uwDEST_U(6,8)<1> ubAVS_RESPONSE_2(9,1)<16;4,2>
+ mov (8) uwDEST_U(7,8)<1> ubAVS_RESPONSE_2(9,8+1)<16;4,2>
+
+ // Move 2st 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0,8)<1> ubAVS_RESPONSE_2(6,1)<16;4,2>
+ mov (8) uwDEST_V(1,8)<1> ubAVS_RESPONSE_2(6,8+1)<16;4,2>
+ mov (8) uwDEST_V(2,8)<1> ubAVS_RESPONSE_2(7,1)<16;4,2>
+ mov (8) uwDEST_V(3,8)<1> ubAVS_RESPONSE_2(7,8+1)<16;4,2>
+ mov (8) uwDEST_V(4,8)<1> ubAVS_RESPONSE_2(10,1)<16;4,2>
+ mov (8) uwDEST_V(5,8)<1> ubAVS_RESPONSE_2(10,8+1)<16;4,2>
+ mov (8) uwDEST_V(6,8)<1> ubAVS_RESPONSE_2(11,1)<16;4,2>
+ mov (8) uwDEST_V(7,8)<1> ubAVS_RESPONSE_2(11,8+1)<16;4,2>
+#endif
+
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL2_AVS_IEF_8x4.asm ----------
+
+ // Move first 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> ubAVS_RESPONSE(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1)<1> ubAVS_RESPONSE(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 8x4 words of U to dest GRF (Copy high byte in a word)
+ mov (8) uwDEST_U(0)<1> ubAVS_RESPONSE(4,1)<16;4,2>
+ mov (8) uwDEST_U(0,8)<1> ubAVS_RESPONSE(5,1)<16;4,2>
+ mov (8) uwDEST_U(1)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) uwDEST_U(1,8)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+
+ // Move 8x4 words of V to dest GRF
+ mov (8) uwDEST_V(0)<1> ubAVS_RESPONSE(6,1)<16;4,2>
+ mov (8) uwDEST_V(0,8)<1> ubAVS_RESPONSE(7,1)<16;4,2>
+ mov (8) uwDEST_V(1)<1> ubAVS_RESPONSE(10,1)<16;4,2>
+ mov (8) uwDEST_V(1,8)<1> ubAVS_RESPONSE(11,1)<16;4,2>
+
+ // Move 2nd 8x8 words of Y to dest GRF at higher 8 words of each GRF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> ubAVS_RESPONSE_2(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1,8)<1> ubAVS_RESPONSE_2(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+//------------------------------------------------------------------------------
+
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 4
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL2_AVS_IEF_8x8.asm ----------
+
+ // Move first 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> ubAVS_RESPONSE(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1)<1> ubAVS_RESPONSE(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 8x8 words of U to dest GRF (Copy high byte in a word)
+ mov (8) uwDEST_U(0)<1> ubAVS_RESPONSE(4,1)<16;4,2>
+ mov (8) uwDEST_U(0,8)<1> ubAVS_RESPONSE(4,8+1)<16;4,2>
+ mov (8) uwDEST_U(1)<1> ubAVS_RESPONSE(5,1)<16;4,2>
+ mov (8) uwDEST_U(1,8)<1> ubAVS_RESPONSE(5,8+1)<16;4,2>
+ mov (8) uwDEST_U(2)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) uwDEST_U(2,8)<1> ubAVS_RESPONSE(8,8+1)<16;4,2>
+ mov (8) uwDEST_U(3)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) uwDEST_U(3,8)<1> ubAVS_RESPONSE(9,8+1)<16;4,2>
+
+ // Move 8x8 words of V to dest GRF
+ mov (8) uwDEST_V(0)<1> ubAVS_RESPONSE(6,1)<16;4,2>
+ mov (8) uwDEST_V(0,8)<1> ubAVS_RESPONSE(6,8+1)<16;4,2>
+ mov (8) uwDEST_V(1)<1> ubAVS_RESPONSE(7,1)<16;4,2>
+ mov (8) uwDEST_V(1,8)<1> ubAVS_RESPONSE(7,8+1)<16;4,2>
+ mov (8) uwDEST_V(2)<1> ubAVS_RESPONSE(10,1)<16;4,2>
+ mov (8) uwDEST_V(2,8)<1> ubAVS_RESPONSE(10,8+1)<16;4,2>
+ mov (8) uwDEST_V(3)<1> ubAVS_RESPONSE(11,1)<16;4,2>
+ mov (8) uwDEST_V(3,8)<1> ubAVS_RESPONSE(11,8+1)<16;4,2>
+
+ // Move 2nd 8x8 words of Y to dest GRF at higher 8 words of each GRF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> ubAVS_RESPONSE_2(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1,8)<1> ubAVS_RESPONSE_2(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+//------------------------------------------------------------------------------
+
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL2_Scaling.asm ----------
+#include "Scaling.inc"
+
+ // Build 16 elements ramp in float32 and normalized it
+// mov (8) SAMPLER_RAMP(0)<1> 0x76543210:v
+// add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+mov (4) SAMPLER_RAMP(0)<1> 0x48403000:vf //3, 2, 1, 0 in float vector
+mov (4) SAMPLER_RAMP(0,4)<1> 0x5C585450:vf //7, 6, 5, 4 in float vector
+add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+
+//Module: PrepareScaleCoord.asm
+
+ // Setup for sampler msg hdr
+ mov (2) rMSGSRC.0<1>:ud 0:ud { NoDDClr } // Unused fields
+ mov (1) rMSGSRC.2<1>:ud 0:ud { NoDDChk } // Write and offset
+
+ // Calculate 16 v based on the step Y and vertical origin
+ mov (16) mfMSGPAYLOAD(2)<1> fSRC_VID_V_ORI<0;1,0>:f
+ mov (16) SCALE_COORD_Y<1>:f fSRC_VID_V_ORI<0;1,0>:f
+
+ // Calculate 16 u based on the step X and hori origin
+// line (16) mfMSGPAYLOAD(0)<1> SCALE_STEP_X<0;1,0>:f SAMPLER_RAMP(0) // Assign to mrf directly
+ mov (16) acc0:f fSRC_VID_H_ORI<0;1,0>:f { Compr }
+ mac (16) mfMSGPAYLOAD(0)<1> fVIDEO_STEP_X<0;1,0>:f SAMPLER_RAMP(0) { Compr }
+
+ //Setup the constants for line instruction
+ mov (1) SCALE_LINE_P255<1>:f 255.0:f { NoDDClr } //{ NoDDClr, NoDDChk }
+ mov (1) SCALE_LINE_P0_5<1>:f 0.5:f { NoDDChk }
+
+//------------------------------------------------------------------------------
+
+$for (0; <nY_NUM_OF_ROWS; 1) {
+
+ // Read 16 sampled pixels and store them in float32 in 8 GRFs in the order of BGRA (VYUA).
+ mov (8) MSGHDR_SCALE.0:ud rMSGSRC.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (16) SCALE_RESPONSE_YW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_Y+nBI_CURRENT_SRC_Y
+ send (16) SCALE_RESPONSE_UW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_UV+nBI_CURRENT_SRC_UV
+
+ // Calculate 16 v for next line
+ add (16) mfMSGPAYLOAD(2)<1> SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+ add (16) SCALE_COORD_Y<1>:f SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+
+ // Scale back to [0, 255], convert f to ud
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(0) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(0)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_UF(0) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_UD(0)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_UF(2) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_UD(2)<1> acc0:f { Compr }
+
+ mov (16) DEST_Y(%1)<1> SCALE_RESPONSE_YB(0) //possible error due to truncation - vK
+ mov (16) DEST_U(%1)<1> SCALE_RESPONSE_UB(0) //possible error due to truncation - vK
+ mov (16) DEST_V(%1)<1> SCALE_RESPONSE_UB(2) //possible error due to truncation - vK
+
+}
+
+ #define nSRC_REGION nREGION_1
+
+//------------------------------------------------------------------------------
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL3_AVS_IEF_16x8.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 Y surface
+// 2 sampler read for 8x8 U surface
+// 2 sampler read for 8x8 V surface
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ // 1st 8x8 Y sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud // Enable green channel
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+ // 1st 8x8 U sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Enable red channel
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_U+nBI_CURRENT_SRC_U
+ // Return U in 4 GRFs
+
+ // 1st 8x8 V sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Dummy instruction to avoid back-2-back send instructions
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(8)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_V+nBI_CURRENT_SRC_V
+ // Return V in 4 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+
+ // 2nd 8x8 Y sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud // Enable green channel
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+ // 2nd 8x8 U sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Enable red channel
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_U+nBI_CURRENT_SRC_U
+ // Return U in 4 GRFs
+
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Dummy instruction just in order to avoid back-2-back send instructions!
+
+ // 2nd 8x8 V sampling
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(8)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_V+nBI_CURRENT_SRC_V
+ // Return V in 4 GRFs
+
+//------------------------------------------------------------------------------
+// Unpacking sampler reads to 4:4:4 internal planar
+//------------------------------------------------------------------------------
+ #include "PL3_AVS_IEF_Unpack_16x8.asm"
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL3_AVS_IEF_8x4.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 Y surface
+// 1 sampler read for 8x8 U surface
+// 1 sampler read for 8x8 V surface
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ // 1st 8x8 Y sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud // Enable green channel
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+ // 8x8 U sampling ; Only 8x4 will be used
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Enable red channel
+ mul (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_X:f 2.0:f // Calculate Step X for chroma
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_U+nBI_CURRENT_SRC_U
+ // Return U in 4 GRFs
+
+ // 8x8 V sampling ; Only 8x4 will be used
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Dummy instruction just in order to avoid back-2-back send instructions!
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(8)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_V+nBI_CURRENT_SRC_V
+ // Return V in 4 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+
+ // 2nd 8x8 Y sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud // Enable green channel
+ mov (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_X:f // Restore Step X for luma
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(12)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+//------------------------------------------------------------------------------
+// Unpacking sampler reads to 4:2:0 internal planar
+//------------------------------------------------------------------------------
+ #include "PL3_AVS_IEF_Unpack_8x4.asm"
+
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL3_AVS_IEF_8x8.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 Y surface
+// 1 sampler read for 8x8 U surface
+// 1 sampler read for 8x8 V surface
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ // 1st 8x8 Y sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud // Enable green channel
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+ // 8x8 U sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Enable red channel
+ mul (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_X:f 2.0:f // Calculate Step X for chroma
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(4)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_U+nBI_CURRENT_SRC_U
+ // Return U in 4 GRFs
+
+ // 8x8 V sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_RED_CHANNEL_ONLY:ud // Dummy instruction just in order to avoid back-2-back send instructions!
+ mov (16) mAVS_8x8_HDR_UV.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(8)<1> mAVS_8x8_HDR_UV udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_V+nBI_CURRENT_SRC_V
+ // Return V in 4 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+
+ // 2nd 8x8 Y sampling
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_GREEN_CHANNEL_ONLY:ud // Enable green channel
+ mov (1) rAVS_PAYLOAD.1:f fVIDEO_STEP_X:f // Restore Step X for luma
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(12)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_1CH+nSI_SRC_Y+nBI_CURRENT_SRC_Y
+ // Return Y in 4 GRFs
+
+//------------------------------------------------------------------------------
+// Unpacking sampler reads to 4:2:2 internal planar
+//------------------------------------------------------------------------------
+ #include "PL3_AVS_IEF_Unpack_8x8.asm"
+
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL3_AVS_IEF_Unpack_16x8.asm ----------
+
+#ifdef AVS_OUTPUT_16_BIT //Output is packed in AVYU format
+// Move first 8x8 words of Y to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,1)<4> uwAVS_RESPONSE(0,0)<4;4,1>
+ mov (4) uwDEST_Y(1,1)<4> uwAVS_RESPONSE(0,8)<4;4,1>
+ mov (4) uwDEST_Y(4,1)<4> uwAVS_RESPONSE(0,4)<4;4,1>
+ mov (4) uwDEST_Y(5,1)<4> uwAVS_RESPONSE(0,12)<4;4,1>
+ mov (4) uwDEST_Y(8,1)<4> uwAVS_RESPONSE(1,0)<4;4,1>
+ mov (4) uwDEST_Y(9,1)<4> uwAVS_RESPONSE(1,8)<4;4,1>
+ mov (4) uwDEST_Y(12,1)<4> uwAVS_RESPONSE(1,4)<4;4,1>
+ mov (4) uwDEST_Y(13,1)<4> uwAVS_RESPONSE(1,12)<4;4,1>
+ mov (4) uwDEST_Y(16,1)<4> uwAVS_RESPONSE(2,0)<4;4,1>
+ mov (4) uwDEST_Y(17,1)<4> uwAVS_RESPONSE(2,8)<4;4,1>
+ mov (4) uwDEST_Y(20,1)<4> uwAVS_RESPONSE(2,4)<4;4,1>
+ mov (4) uwDEST_Y(21,1)<4> uwAVS_RESPONSE(2,12)<4;4,1>
+ mov (4) uwDEST_Y(24,1)<4> uwAVS_RESPONSE(3,0)<4;4,1>
+ mov (4) uwDEST_Y(25,1)<4> uwAVS_RESPONSE(3,8)<4;4,1>
+ mov (4) uwDEST_Y(28,1)<4> uwAVS_RESPONSE(3,4)<4;4,1>
+ mov (4) uwDEST_Y(29,1)<4> uwAVS_RESPONSE(3,12)<4;4,1>
+
+// Move first 8x8 words of U to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,0)<4> uwAVS_RESPONSE(4,0)<4;4,1>
+ mov (4) uwDEST_Y(1,0)<4> uwAVS_RESPONSE(4,8)<4;4,1>
+ mov (4) uwDEST_Y(4,0)<4> uwAVS_RESPONSE(4,4)<4;4,1>
+ mov (4) uwDEST_Y(5,0)<4> uwAVS_RESPONSE(4,12)<4;4,1>
+ mov (4) uwDEST_Y(8,0)<4> uwAVS_RESPONSE(5,0)<4;4,1>
+ mov (4) uwDEST_Y(9,0)<4> uwAVS_RESPONSE(5,8)<4;4,1>
+ mov (4) uwDEST_Y(12,0)<4> uwAVS_RESPONSE(5,4)<4;4,1>
+ mov (4) uwDEST_Y(13,0)<4> uwAVS_RESPONSE(5,12)<4;4,1>
+ mov (4) uwDEST_Y(16,0)<4> uwAVS_RESPONSE(6,0)<4;4,1>
+ mov (4) uwDEST_Y(17,0)<4> uwAVS_RESPONSE(6,8)<4;4,1>
+ mov (4) uwDEST_Y(20,0)<4> uwAVS_RESPONSE(6,4)<4;4,1>
+ mov (4) uwDEST_Y(21,0)<4> uwAVS_RESPONSE(6,12)<4;4,1>
+ mov (4) uwDEST_Y(24,0)<4> uwAVS_RESPONSE(7,0)<4;4,1>
+ mov (4) uwDEST_Y(25,0)<4> uwAVS_RESPONSE(7,8)<4;4,1>
+ mov (4) uwDEST_Y(28,0)<4> uwAVS_RESPONSE(7,4)<4;4,1>
+ mov (4) uwDEST_Y(29,0)<4> uwAVS_RESPONSE(7,12)<4;4,1>
+
+// Move first 8x8 words of V to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,2)<4> uwAVS_RESPONSE(8,0)<4;4,1>
+ mov (4) uwDEST_Y(1,2)<4> uwAVS_RESPONSE(8,8)<4;4,1>
+ mov (4) uwDEST_Y(4,2)<4> uwAVS_RESPONSE(8,4)<4;4,1>
+ mov (4) uwDEST_Y(5,2)<4> uwAVS_RESPONSE(8,12)<4;4,1>
+ mov (4) uwDEST_Y(8,2)<4> uwAVS_RESPONSE(9,0)<4;4,1>
+ mov (4) uwDEST_Y(9,2)<4> uwAVS_RESPONSE(9,8)<4;4,1>
+ mov (4) uwDEST_Y(12,2)<4> uwAVS_RESPONSE(9,4)<4;4,1>
+ mov (4) uwDEST_Y(13,2)<4> uwAVS_RESPONSE(9,12)<4;4,1>
+ mov (4) uwDEST_Y(16,2)<4> uwAVS_RESPONSE(10,0)<4;4,1>
+ mov (4) uwDEST_Y(17,2)<4> uwAVS_RESPONSE(10,8)<4;4,1>
+ mov (4) uwDEST_Y(20,2)<4> uwAVS_RESPONSE(10,4)<4;4,1>
+ mov (4) uwDEST_Y(21,2)<4> uwAVS_RESPONSE(10,12)<4;4,1>
+ mov (4) uwDEST_Y(24,2)<4> uwAVS_RESPONSE(11,0)<4;4,1>
+ mov (4) uwDEST_Y(25,2)<4> uwAVS_RESPONSE(11,8)<4;4,1>
+ mov (4) uwDEST_Y(28,2)<4> uwAVS_RESPONSE(11,4)<4;4,1>
+ mov (4) uwDEST_Y(29,2)<4> uwAVS_RESPONSE(11,12)<4;4,1>
+
+// Move first 8x8 words of A to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,3)<4> 0:uw
+ mov (4) uwDEST_Y(1,3)<4> 0:uw
+ mov (4) uwDEST_Y(4,3)<4> 0:uw
+ mov (4) uwDEST_Y(5,3)<4> 0:uw
+ mov (4) uwDEST_Y(8,3)<4> 0:uw
+ mov (4) uwDEST_Y(9,3)<4> 0:uw
+ mov (4) uwDEST_Y(12,3)<4> 0:uw
+ mov (4) uwDEST_Y(13,3)<4> 0:uw
+ mov (4) uwDEST_Y(16,3)<4> 0:uw
+ mov (4) uwDEST_Y(17,3)<4> 0:uw
+ mov (4) uwDEST_Y(20,3)<4> 0:uw
+ mov (4) uwDEST_Y(21,3)<4> 0:uw
+ mov (4) uwDEST_Y(24,3)<4> 0:uw
+ mov (4) uwDEST_Y(25,3)<4> 0:uw
+ mov (4) uwDEST_Y(28,3)<4> 0:uw
+ mov (4) uwDEST_Y(29,3)<4> 0:uw
+
+// Move second 8x8 words of Y to dest GRF
+ mov (4) uwDEST_Y(2,1)<4> uwAVS_RESPONSE_2(0,0)<4;4,1>
+ mov (4) uwDEST_Y(3,1)<4> uwAVS_RESPONSE_2(0,8)<4;4,1>
+ mov (4) uwDEST_Y(6,1)<4> uwAVS_RESPONSE_2(0,4)<4;4,1>
+ mov (4) uwDEST_Y(7,1)<4> uwAVS_RESPONSE_2(0,12)<4;4,1>
+ mov (4) uwDEST_Y(10,1)<4> uwAVS_RESPONSE_2(1,0)<4;4,1>
+ mov (4) uwDEST_Y(11,1)<4> uwAVS_RESPONSE_2(1,8)<4;4,1>
+ mov (4) uwDEST_Y(14,1)<4> uwAVS_RESPONSE_2(1,4)<4;4,1>
+ mov (4) uwDEST_Y(15,1)<4> uwAVS_RESPONSE_2(1,12)<4;4,1>
+ mov (4) uwDEST_Y(18,1)<4> uwAVS_RESPONSE_2(2,0)<4;4,1>
+ mov (4) uwDEST_Y(19,1)<4> uwAVS_RESPONSE_2(2,8)<4;4,1>
+ mov (4) uwDEST_Y(22,1)<4> uwAVS_RESPONSE_2(2,4)<4;4,1>
+ mov (4) uwDEST_Y(23,1)<4> uwAVS_RESPONSE_2(2,12)<4;4,1>
+ mov (4) uwDEST_Y(26,1)<4> uwAVS_RESPONSE_2(3,0)<4;4,1>
+ mov (4) uwDEST_Y(27,1)<4> uwAVS_RESPONSE_2(3,8)<4;4,1>
+ mov (4) uwDEST_Y(30,1)<4> uwAVS_RESPONSE_2(3,4)<4;4,1>
+ mov (4) uwDEST_Y(31,1)<4> uwAVS_RESPONSE_2(3,12)<4;4,1>
+
+// Move second 8x8 words of U to dest GRF
+ mov (4) uwDEST_Y(2,0)<4> uwAVS_RESPONSE_2(4,0)<4;4,1>
+ mov (4) uwDEST_Y(3,0)<4> uwAVS_RESPONSE_2(4,8)<4;4,1>
+ mov (4) uwDEST_Y(6,0)<4> uwAVS_RESPONSE_2(4,4)<4;4,1>
+ mov (4) uwDEST_Y(7,0)<4> uwAVS_RESPONSE_2(4,12)<4;4,1>
+ mov (4) uwDEST_Y(10,0)<4> uwAVS_RESPONSE_2(5,0)<4;4,1>
+ mov (4) uwDEST_Y(11,0)<4> uwAVS_RESPONSE_2(5,8)<4;4,1>
+ mov (4) uwDEST_Y(14,0)<4> uwAVS_RESPONSE_2(5,4)<4;4,1>
+ mov (4) uwDEST_Y(15,0)<4> uwAVS_RESPONSE_2(5,12)<4;4,1>
+ mov (4) uwDEST_Y(18,0)<4> uwAVS_RESPONSE_2(6,0)<4;4,1>
+ mov (4) uwDEST_Y(19,0)<4> uwAVS_RESPONSE_2(6,8)<4;4,1>
+ mov (4) uwDEST_Y(22,0)<4> uwAVS_RESPONSE_2(6,4)<4;4,1>
+ mov (4) uwDEST_Y(23,0)<4> uwAVS_RESPONSE_2(6,12)<4;4,1>
+ mov (4) uwDEST_Y(26,0)<4> uwAVS_RESPONSE_2(7,0)<4;4,1>
+ mov (4) uwDEST_Y(27,0)<4> uwAVS_RESPONSE_2(7,8)<4;4,1>
+ mov (4) uwDEST_Y(30,0)<4> uwAVS_RESPONSE_2(7,4)<4;4,1>
+ mov (4) uwDEST_Y(31,0)<4> uwAVS_RESPONSE_2(7,12)<4;4,1>
+
+// Move second 8x8 words of V to dest GRF
+ mov (4) uwDEST_Y(2,2)<4> uwAVS_RESPONSE_2(8,0)<4;4,1>
+ mov (4) uwDEST_Y(3,2)<4> uwAVS_RESPONSE_2(8,8)<4;4,1>
+ mov (4) uwDEST_Y(6,2)<4> uwAVS_RESPONSE_2(8,4)<4;4,1>
+ mov (4) uwDEST_Y(7,2)<4> uwAVS_RESPONSE_2(8,12)<4;4,1>
+ mov (4) uwDEST_Y(10,2)<4> uwAVS_RESPONSE_2(9,0)<4;4,1>
+ mov (4) uwDEST_Y(11,2)<4> uwAVS_RESPONSE_2(9,8)<4;4,1>
+ mov (4) uwDEST_Y(14,2)<4> uwAVS_RESPONSE_2(9,4)<4;4,1>
+ mov (4) uwDEST_Y(15,2)<4> uwAVS_RESPONSE_2(9,12)<4;4,1>
+ mov (4) uwDEST_Y(18,2)<4> uwAVS_RESPONSE_2(10,0)<4;4,1>
+ mov (4) uwDEST_Y(19,2)<4> uwAVS_RESPONSE_2(10,8)<4;4,1>
+ mov (4) uwDEST_Y(22,2)<4> uwAVS_RESPONSE_2(10,4)<4;4,1>
+ mov (4) uwDEST_Y(23,2)<4> uwAVS_RESPONSE_2(10,12)<4;4,1>
+ mov (4) uwDEST_Y(26,2)<4> uwAVS_RESPONSE_2(11,0)<4;4,1>
+ mov (4) uwDEST_Y(27,2)<4> uwAVS_RESPONSE_2(11,8)<4;4,1>
+ mov (4) uwDEST_Y(30,2)<4> uwAVS_RESPONSE_2(11,4)<4;4,1>
+ mov (4) uwDEST_Y(31,2)<4> uwAVS_RESPONSE_2(11,12)<4;4,1>
+
+// Move second 8x8 words of A to dest GRF
+ mov (4) uwDEST_Y(2,3)<4> 0:uw
+ mov (4) uwDEST_Y(3,3)<4> 0:uw
+ mov (4) uwDEST_Y(6,3)<4> 0:uw
+ mov (4) uwDEST_Y(7,3)<4> 0:uw
+ mov (4) uwDEST_Y(10,3)<4> 0:uw
+ mov (4) uwDEST_Y(11,3)<4> 0:uw
+ mov (4) uwDEST_Y(14,3)<4> 0:uw
+ mov (4) uwDEST_Y(15,3)<4> 0:uw
+ mov (4) uwDEST_Y(18,3)<4> 0:uw
+ mov (4) uwDEST_Y(19,3)<4> 0:uw
+ mov (4) uwDEST_Y(22,3)<4> 0:uw
+ mov (4) uwDEST_Y(23,3)<4> 0:uw
+ mov (4) uwDEST_Y(26,3)<4> 0:uw
+ mov (4) uwDEST_Y(27,3)<4> 0:uw
+ mov (4) uwDEST_Y(30,3)<4> 0:uw
+ mov (4) uwDEST_Y(31,3)<4> 0:uw
+
+/* This section will be used if 16-bit output is needed in planar format -vK
+ // Move 1st 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> uwAVS_RESPONSE(%1)<8;4,1>
+ mov (8) uwDEST_Y(%1*2+1)<1> uwAVS_RESPONSE(%1,8)<8;4,1>
+ }
+
+ // Move 8x8 words of U to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_U(%1*2)<1> uwAVS_RESPONSE(%1+4)<8;4,1>
+ mov (8) uwDEST_U(%1*2+1)<1> uwAVS_RESPONSE(%1+4,8)<8;4,1>
+ }
+
+ // Move 8x8 words of V to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_V(%1*2)<1> uwAVS_RESPONSE(%1+8)<8;4,1>
+ mov (8) uwDEST_V(%1*2+1)<1> uwAVS_RESPONSE(%1+8,8)<8;4,1>
+ }
+
+ // Move 2nd 8x8 words of Y to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> uwAVS_RESPONSE_2(%1)<8;4,1>
+ mov (8) uwDEST_Y(%1*2+1,8)<1> uwAVS_RESPONSE_2(%1,8)<8;4,1>
+ }
+
+ // Move 2nd 8x8 words of U to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_U(%1*2,8)<1> uwAVS_RESPONSE_2(%1+4)<8;4,1>
+ mov (8) uwDEST_U(%1*2+1,8)<1> uwAVS_RESPONSE_2(%1+4,8)<8;4,1>
+ }
+
+ // Move 2nd 8x8 words of V to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_V(%1*2,8)<1> uwAVS_RESPONSE_2(%1+8)<8;4,1>
+ mov (8) uwDEST_V(%1*2+1,8)<1> uwAVS_RESPONSE_2(%1+8,8)<8;4,1>
+ }
+*/
+#else /* OUTPUT_8_BIT */
+ // Move 1st 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> ubAVS_RESPONSE(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1)<1> ubAVS_RESPONSE(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 8x8 words of U to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_U(%1*2)<1> ubAVS_RESPONSE(%1+4,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_U(%1*2+1)<1> ubAVS_RESPONSE(%1+4,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 8x8 words of V to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_V(%1*2)<1> ubAVS_RESPONSE(%1+8,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_V(%1*2+1)<1> ubAVS_RESPONSE(%1+8,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 2nd 8x8 words of Y to dest GRF at higher 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> ubAVS_RESPONSE_2(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1,8)<1> ubAVS_RESPONSE_2(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 2nd 8x8 words of U to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_U(%1*2,8)<1> ubAVS_RESPONSE_2(%1+4,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_U(%1*2+1,8)<1> ubAVS_RESPONSE_2(%1+4,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 2nd 8x8 words of V to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_V(%1*2,8)<1> ubAVS_RESPONSE_2(%1+8,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_V(%1*2+1,8)<1> ubAVS_RESPONSE_2(%1+8,8+1)<16;4,2> // Copy high byte in a word
+ }
+#endif
+//------------------------------------------------------------------------------
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL3_AVS_IEF_Unpack_8x4.asm ----------
+
+ // Move 1st 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> ubAVS_RESPONSE(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1)<1> ubAVS_RESPONSE(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 8x4 words of U to dest GRF (Copy high byte in a word)
+ mov (8) uwDEST_U(0)<1> ubAVS_RESPONSE(4,1)<16;4,2>
+ mov (8) uwDEST_U(0,8)<1> ubAVS_RESPONSE(4,9)<16;4,2>
+ mov (8) uwDEST_U(1)<1> ubAVS_RESPONSE(5,1)<16;4,2>
+ mov (8) uwDEST_U(1,8)<1> ubAVS_RESPONSE(5,9)<16;4,2>
+
+ // Move 8x4 words of V to dest GRF
+ mov (8) uwDEST_V(0)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) uwDEST_V(0,8)<1> ubAVS_RESPONSE(8,9)<16;4,2>
+ mov (8) uwDEST_V(1)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) uwDEST_V(1,8)<1> ubAVS_RESPONSE(9,9)<16;4,2>
+
+ // Move 2nd 8x8 words of Y to dest GRF at higher 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> ubAVS_RESPONSE(%1+12,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1,8)<1> ubAVS_RESPONSE(%1+12,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+//------------------------------------------------------------------------------
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 4
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL3_AVS_IEF_Unpack_8x8.asm ----------
+
+ // Move 1st 8x8 words of Y to dest GRF at lower 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2)<1> ubAVS_RESPONSE(%1,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1)<1> ubAVS_RESPONSE(%1,8+1)<16;4,2> // Copy high byte in a word
+ }
+ // Move 8x8 words of U to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_U(%1)<1> ubAVS_RESPONSE(%1+4,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_U(%1,8)<1> ubAVS_RESPONSE(%1+4,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 8x8 words of V to dest GRF
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_V(%1)<1> ubAVS_RESPONSE(%1+8,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_V(%1,8)<1> ubAVS_RESPONSE(%1+8,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+ // Move 2nd 8x8 words of Y to dest GRF at higher 8 words of each RGF.
+ $for(0; <8/2; 1) {
+ mov (8) uwDEST_Y(%1*2,8)<1> ubAVS_RESPONSE(%1+12,1)<16;4,2> // Copy high byte in a word
+ mov (8) uwDEST_Y(%1*2+1,8)<1> ubAVS_RESPONSE(%1+12,8+1)<16;4,2> // Copy high byte in a word
+ }
+
+//------------------------------------------------------------------------------
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- PL3_Scaling.asm ----------
+#include "Scaling.inc"
+
+ // Build 16 elements ramp in float32 and normalized it
+// mov (8) SAMPLER_RAMP(0)<1> 0x76543210:v
+// add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+mov (4) SAMPLER_RAMP(0)<1> 0x48403000:vf { NoDDClr }//3, 2, 1, 0 in float vector
+mov (4) SAMPLER_RAMP(0,4)<1> 0x5C585450:vf { NoDDChk }//7, 6, 5, 4 in float vector
+add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+
+
+//Module: PrepareScaleCoord.asm
+
+ // Setup for sampler msg hdr
+ mov (2) rMSGSRC.0<1>:ud 0:ud { NoDDClr } // Unused fields
+ mov (1) rMSGSRC.2<1>:ud 0:ud { NoDDChk } // Write and offset
+
+ // Calculate 16 v based on the step Y and vertical origin
+ mov (16) mfMSGPAYLOAD(2)<1> fSRC_VID_V_ORI<0;1,0>:f
+ mov (16) SCALE_COORD_Y<1>:f fSRC_VID_V_ORI<0;1,0>:f
+
+ // Calculate 16 u based on the step X and hori origin
+// line (16) mfMSGPAYLOAD(0)<1> SCALE_STEP_X<0;1,0>:f SAMPLER_RAMP(0) // Assign to mrf directly
+ mov (16) acc0:f fSRC_VID_H_ORI<0;1,0>:f { Compr }
+ mac (16) mfMSGPAYLOAD(0)<1> fVIDEO_STEP_X<0;1,0>:f SAMPLER_RAMP(0) { Compr }
+
+ //Setup the constants for line instruction
+ mov (1) SCALE_LINE_P255<1>:f 255.0:f { NoDDClr } //{ NoDDClr, NoDDChk }
+ mov (1) SCALE_LINE_P0_5<1>:f 0.5:f { NoDDChk }
+
+//------------------------------------------------------------------------------
+
+$for (0; <nY_NUM_OF_ROWS; 1) {
+ // Read 16 sampled pixels and store them in float32 in 8 GRFs in the order of BGRA (VYUA).
+ mov (8) MSGHDR_SCALE<1>:ud rMSGSRC<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (16) SCALE_RESPONSE_VW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_V+nBI_CURRENT_SRC_V
+ send (16) SCALE_RESPONSE_YW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_Y+nBI_CURRENT_SRC_Y
+ send (16) SCALE_RESPONSE_UW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_U+nBI_CURRENT_SRC_U
+
+ // Calculate 16 v for next line
+ add (16) mfMSGPAYLOAD(2)<1> SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+ add (16) SCALE_COORD_Y<1>:f SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+
+ // Scale back to [0, 255], convert f to ud
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_VF(0) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_VD(0)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(0) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(0)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_UF(0) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_UD(0)<1> acc0:f { Compr }
+
+ mov (16) DEST_V(%1)<1> SCALE_RESPONSE_VB(0) //possible error due to truncation - vK
+ mov (16) DEST_Y(%1)<1> SCALE_RESPONSE_YB(0) //possible error due to truncation - vK
+ mov (16) DEST_U(%1)<1> SCALE_RESPONSE_UB(0) //possible error due to truncation - vK
+
+}
+
+ #define nSRC_REGION nREGION_1
+
+//------------------------------------------------------------------------------
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_ENABLE
+
+ #include "DNDI.inc"
+
+ #ifdef DI_ONLY
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DI // set the number of GRF
+ #else
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DNDI // set the number of GRF
+ #endif
+
+ #undef nDPW_BLOCK_SIZE_HIST
+ #define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_1 // HIST Block Size for Write is 4x2
+ #undef nDPW_BLOCK_SIZE_DN
+ #define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_16+nBLOCK_HEIGHT_4 // DN Block Size for Write is 16x4
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+ #include "DNDI_Command.asm"
+
+////////////////////////////////////// Rearrange for Internal Planar //////////////////////////////
+
+////////////////////////////////////// Save the STMM Data for Next Run /////////////////////////
+ // Write STMM to memory
+ shr (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w NODDCLR_NODDCHK // X origin / 2
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w NODDCLR_NODDCHK // Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_STMM:ud NODDCHK // block width and height (8x4)
+ mov (8) mudMSGHDR_STMM(0)<1> rMSGSRC.0<8;8,1>:ud // message header
+ mov (8) mudMSGHDR_STMM(1)<1> udRESP(nDI_STMM_OFFSET,0) // Move STMM to MRF
+ send (8) dNULLREG mMSGHDR_STMM udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_STMM+nBI_STMM_HISTORY_OUTPUT:ud
+
+#ifdef DI_ONLY
+#else
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+ #include "DI_Hist_Save.asm"
+
+////////////////////////////////////// Save the DN Curr Frame for Next Run ////////////////////////
+
+ //set the save DN parameters
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w NODDCLR // X origin and Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud NODDCLR_NODDCHK // block width and height (16x4)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+
+ // check top/bottom field first
+ cmp.e.f0.0 (1) null<1>:w ubTFLD_FIRST<0;1,0>:ub 1:w
+ (f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ }
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+
+ jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,0)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ }
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+
+SAVE_DN_CURR:
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PL_DN_DI+nBI_DESTINATION_Y:ud
+#endif
+
+// Save Processed frames
+#include "DI_Save_PA.asm"
+
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_ENABLE
+
+ #include "DNDI.inc"
+
+ #undef nY_NUM_OF_ROWS
+ #define nY_NUM_OF_ROWS 8 // Number of Y rows per block (4 rows for each frame)
+ #undef nUV_NUM_OF_ROWS
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DNDI // set the number of GRF
+ #undef nDPW_BLOCK_SIZE_HIST
+ #define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_1 // HIST Block Size for Write is 4x2
+ #undef nDPW_BLOCK_SIZE_DN
+ #define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_16+nBLOCK_HEIGHT_4 // DN Block Size for Write is 16x4
+ #undef nDPR_BLOCK_SIZE_UV
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_8+nBLOCK_HEIGHT_4 // DN Block Size for UV Write/Read is 8x4
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+ #include "DNDI_Command.asm"
+
+////////////////////////////////////// Rearrange for Internal Planar //////////////////////////////
+ // move the previous frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1,0)<1> ubRESP(nDI_PREV_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the previous frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+ // move the current frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1+4,0)<1> ubRESP(nDI_CURR_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the current frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+
+////////////////////////////////////// Save the STMM Data for Next Run /////////////////////////
+ // Write STMM to memory
+ shr (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w // X origin / 2
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w // Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_STMM:ud // block width and height (8x4)
+ mov (8) mudMSGHDR_STMM(0)<1> rMSGSRC.0<8;8,1>:ud // message header
+ mov (8) mudMSGHDR_STMM(1)<1> udRESP(nDI_STMM_OFFSET,0) // Move STMM to MRF
+ send (8) dNULLREG mMSGHDR_STMM udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_STMM+nBI_STMM_HISTORY_OUTPUT:ud
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+ #include "DI_Hist_Save.asm"
+
+////////////////////////////////////// Save the DN Curr Frame for Next Run ////////////////////////
+ add (4) pCF_Y_OFFSET<1>:uw ubSRC_CF_OFFSET<4;4,1>:ub npDN_YUV:w
+ // check top/bottom field first
+ cmp.e.f0.0 (1) null<1>:w ubTFLD_FIRST<0;1,0>:ub 1:w
+ (f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+ jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,0)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+SAVE_DN_CURR:
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud // block width and height (16x4)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PL_DN_DI+nBI_DESTINATION_Y:ud
+
+
+/////////////////////////////P208 UV Copy 422/////////////////////////////////////////////////////
+ //Read UV through DATAPORT
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+ asr (1) rMSGSRC.0<1>:d rMSGSRC.0<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (16x2)
+ mov (8) mudMSGHDR_DN<1> rMSGSRC<8;8,1>:ud
+ send (8) udBOT_U_IO(0)<1> mMSGHDR_DN udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nRESLEN_1+nBI_CURRENT_SRC_UV:ud
+
+ //Write UV through DATAPORT
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ asr (1) rMSGSRC.0<1>:d rMSGSRC.0<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // block width and height (16x2)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ mov (8) mudMSGHDR_DN(1)<1> udBOT_U_IO(0)<8;8,1>
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nMSGLEN_1+nBI_DESTINATION_UV:ud
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_ENABLE
+
+ #include "DNDI.inc"
+
+ #undef nY_NUM_OF_ROWS
+ #define nY_NUM_OF_ROWS 8 // Number of Y rows per block (4 rows for each frame)
+ #undef nUV_NUM_OF_ROWS
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DNDI // set the number of GRF
+ #undef nDPW_BLOCK_SIZE_HIST
+ #define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_1 // HIST Block Size for Write is 4x2
+ #undef nDPW_BLOCK_SIZE_DN
+ #define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_16+nBLOCK_HEIGHT_4 // DN Block Size for Write is 16x4
+ #undef nDPR_BLOCK_SIZE_UV
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_16+nBLOCK_HEIGHT_2 // DN Block Size for UV Write/Read is 16x2
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+ #include "DNDI_COMMAND.asm"
+
+////////////////////////////////////// Rearrange for Internal Planar //////////////////////////////
+ // move the previous frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1,0)<1> ubRESP(nDI_PREV_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the previous frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+ // move the current frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1+4,0)<1> ubRESP(nDI_CURR_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the current frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+
+////////////////////////////////////// Save the STMM Data for Next Run /////////////////////////
+ // Write STMM to memory
+ shr (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w // X origin / 2
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w // Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_STMM:ud // block width and height (8x4)
+ mov (8) mudMSGHDR_STMM(0)<1> rMSGSRC.0<8;8,1>:ud // message header
+ mov (8) mudMSGHDR_STMM(1)<1> udRESP(nDI_STMM_OFFSET,0) // Move STMM to MRF
+ send (8) dNULLREG mMSGHDR_STMM udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_STMM+nBI_STMM_HISTORY_OUTPUT:ud
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+ #include "DI_Hist_Save.asm"
+
+////////////////////////////////////// Save the DN Curr Frame for Next Run ////////////////////////
+ add (4) pCF_Y_OFFSET<1>:uw ubSRC_CF_OFFSET<4;4,1>:ub npDN_YUV:w
+ // check top/bottom field first
+ cmp.e.f0.0 (1) null<1>:w ubTFLD_FIRST<0;1,0>:ub 1:w
+ (f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+ jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,0)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+SAVE_DN_CURR:
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud // block width and height (16x4)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PL_DN_DI+nBI_DESTINATION_Y:ud
+
+
+/////////////////////////////NV12 UV Copy 422/////////////////////////////////////////////////////
+ //Read UV through DATAPORT
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+ asr (1) rMSGSRC.1<1>:d rMSGSRC.1<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (16x2)
+ mov (8) mudMSGHDR_DN<1> rMSGSRC<8;8,1>:ud
+ send (8) udBOT_U_IO(0)<1> mMSGHDR_DN udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nRESLEN_1+nBI_CURRENT_SRC_UV:ud
+
+ //Write UV through DATAPORT
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ asr (1) rMSGSRC.1<1>:d rMSGSRC.1<0;1,0>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // block width and height (16x2)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ mov (8) mudMSGHDR_DN(1)<1> udBOT_U_IO(0)<8;8,1>
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nMSGLEN_1+nBI_DESTINATION_UV:ud
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_ENABLE
+
+ #include "DNDI.inc"
+
+ #undef nY_NUM_OF_ROWS
+ #define nY_NUM_OF_ROWS 8 // Number of Y rows per block (4 rows for each frame)
+ #undef nUV_NUM_OF_ROWS
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DNDI // set the number of GRF
+ #undef nDPW_BLOCK_SIZE_HIST
+ #define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_1 // HIST Block Size for Write is 4x2
+ #undef nDPW_BLOCK_SIZE_DN
+ #define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_16+nBLOCK_HEIGHT_4 // DN Block Size for Write is 16x4
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+ #include "DNDI_Command.asm"
+
+////////////////////////////////////// Rearrange for Internal Planar //////////////////////////////
+ // move the previous frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1,0)<1> ubRESP(nDI_PREV_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the previous frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+ // move the current frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1+4,0)<1> ubRESP(nDI_CURR_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the current frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+
+////////////////////////////////////// Save the STMM Data for Next Run /////////////////////////
+ // Write STMM to memory
+ shr (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w // X origin / 2
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w // Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_STMM:ud // block width and height (8x4)
+ mov (8) mudMSGHDR_STMM(0)<1> rMSGSRC.0<8;8,1>:ud // message header
+ mov (8) mudMSGHDR_STMM(1)<1> udRESP(nDI_STMM_OFFSET,0) // Move STMM to MRF
+ send (8) dNULLREG mMSGHDR_STMM udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_STMM+nBI_STMM_HISTORY_OUTPUT:ud
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+ #include "DI_Hist_Save.asm"
+
+////////////////////////////////////// Save the DN Curr Frame for Next Run ////////////////////////
+ add (4) pCF_Y_OFFSET<1>:uw ubSRC_CF_OFFSET<4;4,1>:ub npDN_YUV:w
+ // check top/bottom field first
+ cmp.e.f0.0 (1) null<1>:w ubTFLD_FIRST<0;1,0>:ub 1:w
+ (f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+ jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,0)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+SAVE_DN_CURR:
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud // block width and height (16x4)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PL_DN_DI+nBI_DESTINATION_Y:ud
+
+
+/////////////////////////////P208 UV Copy 422/////////////////////////////////////////////////////
+ //Read UV through DATAPORT
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud // Y Block width and height (16x4) (U/V block size is the same)
+ mov (8) mudMSGHDR_DN<1> rMSGSRC<8;8,1>:ud
+ send (8) udBOT_U_IO(0)<1> mMSGHDR_DN udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nRESLEN_2+nBI_CURRENT_SRC_UV:ud
+
+ //Write UV through DATAPORT
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud // block width and height (16x4)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ mov (8) mudMSGHDR_DN(1)<1> udBOT_U_IO(0)<8;8,1>
+ mov (8) mudMSGHDR_DN(2)<1> udBOT_U_IO(1)<8;8,1>
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PL_DN_DI+nBI_DESTINATION_UV:ud
+
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_ENABLE
+
+ #include "DNDI.inc"
+
+ #undef nY_NUM_OF_ROWS
+ #define nY_NUM_OF_ROWS 8 // Number of Y rows per block (4 rows for each frame)
+ #undef nUV_NUM_OF_ROWS
+ #define nUV_NUM_OF_ROWS 8 // Number of U/V rows per block
+
+ #undef nSMPL_RESP_LEN
+ #define nSMPL_RESP_LEN nSMPL_RESP_LEN_DNDI // set the number of GRF
+ #undef nDPW_BLOCK_SIZE_HIST
+ #define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_1 // HIST Block Size for Write is 4x2
+ #undef nDPW_BLOCK_SIZE_DN
+ #define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_16+nBLOCK_HEIGHT_4 // DN Block Size for Write is 16x4
+ #undef nDPR_BLOCK_SIZE_UV
+ #define nDPR_BLOCK_SIZE_UV nBLOCK_WIDTH_8+nBLOCK_HEIGHT_2 // DN Block Size for UV Write/Read is 8x2
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+ #include "DNDI_Command.asm"
+
+////////////////////////////////////// Rearrange for Internal Planar //////////////////////////////
+ // move the previous frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1,0)<1> ubRESP(nDI_PREV_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the previous frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(0,%1*8)<1> ubRESP(nDI_PREV_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+ // move the current frame Y component to internal planar format
+ $for (0; <nY_NUM_OF_ROWS/2; 1) {
+ mov (16) uwDEST_Y(%1+4,0)<1> ubRESP(nDI_CURR_FRAME_LUMA_OFFSET,%1*16)
+ }
+ // move the current frame U,V components to internal planar format
+ $for (0; <nUV_NUM_OF_ROWS/2; 1) {
+ mov (8) uwDEST_U(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16+1)<16;8,2> //U pixels
+ mov (8) uwDEST_V(2,%1*8)<1> ubRESP(nDI_CURR_FRAME_CHROMA_OFFSET,%1*16)<16;8,2> //V pixels
+ }
+
+////////////////////////////////////// Save the STMM Data for Next Run /////////////////////////
+ // Write STMM to memory
+ shr (1) rMSGSRC.0<1>:ud wORIX<0;1,0>:w 1:w // X origin / 2
+ mov (1) rMSGSRC.1<1>:ud wORIY<0;1,0>:w // Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_STMM:ud // block width and height (8x4)
+ mov (8) mudMSGHDR_STMM(0)<1> rMSGSRC.0<8;8,1>:ud // message header
+ mov (8) mudMSGHDR_STMM(1)<1> udRESP(nDI_STMM_OFFSET,0) // Move STMM to MRF
+ send (8) dNULLREG mMSGHDR_STMM udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_STMM+nBI_STMM_HISTORY_OUTPUT:ud
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+ #include "DI_Hist_Save.asm"
+
+////////////////////////////////////// Save the DN Curr Frame for Next Run ////////////////////////
+ add (4) pCF_Y_OFFSET<1>:uw ubSRC_CF_OFFSET<4;4,1>:ub npDN_YUV:w
+ // check top/bottom field first
+ cmp.e.f0.0 (1) null<1>:w ubTFLD_FIRST<0;1,0>:ub 1:w
+ (f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+ jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ $for (0,0; <nY_NUM_OF_ROWS/2; 2,1) {
+ mov (4) mudMSGHDR_DN(1,%1*4)<1> udRESP(nDI_CURR_FRAME_LUMA_OFFSET+%2,0)<4;4,1> // 2nd field luma from current frame (line 0,2)
+ mov (4) mudMSGHDR_DN(1,%1*4+4)<1> udRESP(nDI_CURR_2ND_FIELD_LUMA_OFFSET,%2*4)<4;4,1> // 1st field luma from current frame (line 1,3)
+ }
+SAVE_DN_CURR:
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ mov (1) rMSGSRC.2<1>:ud nDPW_BLOCK_SIZE_DN:ud // block width and height (16x4)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ send (8) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nDPMW_MSG_LEN_PL_DN_DI+nBI_DESTINATION_Y:ud
+
+
+/////////////////////////////IMC3 UV Copy 422/////////////////////////////////////////////////////
+ //Read UV through DATAPORT
+ add (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w wSRC_H_ORI_OFFSET<2;2,1>:w // Source Y Block origin
+ asr (2) rMSGSRC.0<1>:d rMSGSRC.0<2;2,1>:d 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // U/V block width and height (8x2)
+ mov (8) mudMSGHDR_DN<1> rMSGSRC<8;8,1>:ud
+ send (4) udBOT_U_IO(0)<1> mMSGHDR_DN udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nRESLEN_1+nBI_CURRENT_SRC_U:ud
+ send (4) udBOT_V_IO(0)<1> mMSGHDR_DN udDUMMY_NULL nDATAPORT_READ nDPMR_MSGDSC+nRESLEN_1+nBI_CURRENT_SRC_V:ud
+
+ //Write UV through DATAPORT
+ mov (2) rMSGSRC.0<1>:ud wORIX<2;2,1>:w // X origin and Y origin
+ asr (2) rMSGSRC.0<1>:d wORIX<2;2,1>:w 1:w // U/V block origin should be half of Y's
+ mov (1) rMSGSRC.2<1>:ud nDPR_BLOCK_SIZE_UV:ud // block width and height (8x2)
+ mov (8) mudMSGHDR_DN(0)<1> rMSGSRC.0<8;8,1>:ud
+ mov (4) mudMSGHDR_DN(1)<1> udBOT_U_IO(0)<4;4,1>
+ send (4) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nMSGLEN_1+nBI_DESTINATION_U:ud
+ mov (4) mudMSGHDR_DN(1)<1> udBOT_V_IO(0)<4;4,1>
+ send (4) dNULLREG mMSGHDR_DN udDUMMY_NULL nDATAPORT_WRITE nDPMW_MSGDSC+nMSGLEN_1+nBI_DESTINATION_V:ud
\ No newline at end of file
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#define DI_DISABLE
+
+#include "DNDI.inc"
+
+#undef nY_NUM_OF_ROWS
+#define nY_NUM_OF_ROWS 8 // Number of Y rows per block
+
+#undef nSMPL_RESP_LEN
+#define nSMPL_RESP_LEN nSMPL_RESP_LEN_DN_PL // Set the Number of GRFs in DNDI response
+#undef nDPW_BLOCK_SIZE_DN
+#define nDPW_BLOCK_SIZE_DN nBLOCK_WIDTH_16+nBLOCK_HEIGHT_8 // DN Curr Block Size for Write is 16x8
+#undef nDPW_BLOCK_SIZE_HIST
+#define nDPW_BLOCK_SIZE_HIST nBLOCK_WIDTH_4+nBLOCK_HEIGHT_2 // HIST Block Size for Write is 4x2
+
+////////////////////////////////////// Run the DN Algorithm ///////////////////////////////////////
+#include "DNDI_COMMAND.asm"
+
+////////////////////////////////////// Rearrange for Internal Planar //////////////////////////////
+$for (0; <nY_NUM_OF_ROWS; 1) {
+ mov (16) uwDEST_Y(0,%1*16)<1> ubRESP(nNODI_LUMA_OFFSET,%1*16)<16;16,1> // copy line of Y
+}
+
+////////////////////////////////////// Save the History Data for Next Run /////////////////////////
+#include "DNDI_Hist_Save.asm"
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- RGB_AVS_IEF_16x8.asm ----------
+
+#include "AVS_IEF.inc"
+
+//------------------------------------------------------------------------------
+// 2 sampler reads for 8x8 ARGB packed
+//------------------------------------------------------------------------------
+
+ // 1st 8x8 setup
+ #include "AVS_SetupFirstBlock.asm"
+
+ mov (1) rAVS_8x8_HDR.2:ud nAVS_ALL_CHANNELS:ud // Enable ARGB channels
+ mov (16) mAVS_8x8_HDR.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE(0)<1> mAVS_8x8_HDR udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_4CH+nSI_SRC_RGB+nBI_CURRENT_SRC_YUV
+ // Return ARGB in 16 GRFs
+
+ // 2nd 8x8 setup
+ #include "AVS_SetupSecondBlock.asm"
+ mov (16) mAVS_8x8_HDR_2.0:ud rAVS_8x8_HDR.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (1) uwAVS_RESPONSE_2(0)<1> mAVS_8x8_HDR_2 udDUMMY_NULL nSMPL_ENGINE nAVS_MSG_DSC_4CH+nSI_SRC_RGB+nBI_CURRENT_SRC_YUV
+ // Return ARGB in 16 GRFs
+
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- RGB_AVS_IEF_Unpack_16x8.asm ----------
+#include "AVS_IEF.inc"
+
+#ifdef AVS_OUTPUT_16_BIT
+// Move first 8x8 words of B to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,2)<4> uwAVS_RESPONSE(4,0)<4;4,1>
+ mov (4) uwDEST_Y(1,2)<4> uwAVS_RESPONSE(4,8)<4;4,1>
+ mov (4) uwDEST_Y(4,2)<4> uwAVS_RESPONSE(4,4)<4;4,1>
+ mov (4) uwDEST_Y(5,2)<4> uwAVS_RESPONSE(4,12)<4;4,1>
+ mov (4) uwDEST_Y(8,2)<4> uwAVS_RESPONSE(5,0)<4;4,1>
+ mov (4) uwDEST_Y(9,2)<4> uwAVS_RESPONSE(5,8)<4;4,1>
+ mov (4) uwDEST_Y(12,2)<4> uwAVS_RESPONSE(5,4)<4;4,1>
+ mov (4) uwDEST_Y(13,2)<4> uwAVS_RESPONSE(5,12)<4;4,1>
+ mov (4) uwDEST_Y(16,2)<4> uwAVS_RESPONSE(12,0)<4;4,1>
+ mov (4) uwDEST_Y(17,2)<4> uwAVS_RESPONSE(12,8)<4;4,1>
+ mov (4) uwDEST_Y(20,2)<4> uwAVS_RESPONSE(12,4)<4;4,1>
+ mov (4) uwDEST_Y(21,2)<4> uwAVS_RESPONSE(12,12)<4;4,1>
+ mov (4) uwDEST_Y(24,2)<4> uwAVS_RESPONSE(13,0)<4;4,1>
+ mov (4) uwDEST_Y(25,2)<4> uwAVS_RESPONSE(13,8)<4;4,1>
+ mov (4) uwDEST_Y(28,2)<4> uwAVS_RESPONSE(13,4)<4;4,1>
+ mov (4) uwDEST_Y(29,2)<4> uwAVS_RESPONSE(13,12)<4;4,1>
+
+// Move first 8x8 words of G to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,1)<4> uwAVS_RESPONSE(2,0)<4;4,1>
+ mov (4) uwDEST_Y(1,1)<4> uwAVS_RESPONSE(2,8)<4;4,1>
+ mov (4) uwDEST_Y(4,1)<4> uwAVS_RESPONSE(2,4)<4;4,1>
+ mov (4) uwDEST_Y(5,1)<4> uwAVS_RESPONSE(2,12)<4;4,1>
+ mov (4) uwDEST_Y(8,1)<4> uwAVS_RESPONSE(3,0)<4;4,1>
+ mov (4) uwDEST_Y(9,1)<4> uwAVS_RESPONSE(3,8)<4;4,1>
+ mov (4) uwDEST_Y(12,1)<4> uwAVS_RESPONSE(3,4)<4;4,1>
+ mov (4) uwDEST_Y(13,1)<4> uwAVS_RESPONSE(3,12)<4;4,1>
+ mov (4) uwDEST_Y(16,1)<4> uwAVS_RESPONSE(10,0)<4;4,1>
+ mov (4) uwDEST_Y(17,1)<4> uwAVS_RESPONSE(10,8)<4;4,1>
+ mov (4) uwDEST_Y(20,1)<4> uwAVS_RESPONSE(10,4)<4;4,1>
+ mov (4) uwDEST_Y(21,1)<4> uwAVS_RESPONSE(10,12)<4;4,1>
+ mov (4) uwDEST_Y(24,1)<4> uwAVS_RESPONSE(11,0)<4;4,1>
+ mov (4) uwDEST_Y(25,1)<4> uwAVS_RESPONSE(11,8)<4;4,1>
+ mov (4) uwDEST_Y(28,1)<4> uwAVS_RESPONSE(11,4)<4;4,1>
+ mov (4) uwDEST_Y(29,1)<4> uwAVS_RESPONSE(11,12)<4;4,1>
+
+// Move first 8x8 words of R to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,0)<4> uwAVS_RESPONSE(0,0)<4;4,1>
+ mov (4) uwDEST_Y(1,0)<4> uwAVS_RESPONSE(0,8)<4;4,1>
+ mov (4) uwDEST_Y(4,0)<4> uwAVS_RESPONSE(0,4)<4;4,1>
+ mov (4) uwDEST_Y(5,0)<4> uwAVS_RESPONSE(0,12)<4;4,1>
+ mov (4) uwDEST_Y(8,0)<4> uwAVS_RESPONSE(1,0)<4;4,1>
+ mov (4) uwDEST_Y(9,0)<4> uwAVS_RESPONSE(1,8)<4;4,1>
+ mov (4) uwDEST_Y(12,0)<4> uwAVS_RESPONSE(1,4)<4;4,1>
+ mov (4) uwDEST_Y(13,0)<4> uwAVS_RESPONSE(1,12)<4;4,1>
+ mov (4) uwDEST_Y(16,0)<4> uwAVS_RESPONSE(8,0)<4;4,1>
+ mov (4) uwDEST_Y(17,0)<4> uwAVS_RESPONSE(8,8)<4;4,1>
+ mov (4) uwDEST_Y(20,0)<4> uwAVS_RESPONSE(8,4)<4;4,1>
+ mov (4) uwDEST_Y(21,0)<4> uwAVS_RESPONSE(8,12)<4;4,1>
+ mov (4) uwDEST_Y(24,0)<4> uwAVS_RESPONSE(9,0)<4;4,1>
+ mov (4) uwDEST_Y(25,0)<4> uwAVS_RESPONSE(9,8)<4;4,1>
+ mov (4) uwDEST_Y(28,0)<4> uwAVS_RESPONSE(9,4)<4;4,1>
+ mov (4) uwDEST_Y(29,0)<4> uwAVS_RESPONSE(9,12)<4;4,1>
+
+// Move first 8x8 words of A to dest GRF (as packed)
+ mov (4) uwDEST_Y(0,3)<4> uwAVS_RESPONSE(6,0)<4;4,1>
+ mov (4) uwDEST_Y(1,3)<4> uwAVS_RESPONSE(6,8)<4;4,1>
+ mov (4) uwDEST_Y(4,3)<4> uwAVS_RESPONSE(6,4)<4;4,1>
+ mov (4) uwDEST_Y(5,3)<4> uwAVS_RESPONSE(6,12)<4;4,1>
+ mov (4) uwDEST_Y(8,3)<4> uwAVS_RESPONSE(7,0)<4;4,1>
+ mov (4) uwDEST_Y(9,3)<4> uwAVS_RESPONSE(7,8)<4;4,1>
+ mov (4) uwDEST_Y(12,3)<4> uwAVS_RESPONSE(7,4)<4;4,1>
+ mov (4) uwDEST_Y(13,3)<4> uwAVS_RESPONSE(7,12)<4;4,1>
+ mov (4) uwDEST_Y(16,3)<4> uwAVS_RESPONSE(14,0)<4;4,1>
+ mov (4) uwDEST_Y(17,3)<4> uwAVS_RESPONSE(14,8)<4;4,1>
+ mov (4) uwDEST_Y(20,3)<4> uwAVS_RESPONSE(14,4)<4;4,1>
+ mov (4) uwDEST_Y(21,3)<4> uwAVS_RESPONSE(14,12)<4;4,1>
+ mov (4) uwDEST_Y(24,3)<4> uwAVS_RESPONSE(15,0)<4;4,1>
+ mov (4) uwDEST_Y(25,3)<4> uwAVS_RESPONSE(15,8)<4;4,1>
+ mov (4) uwDEST_Y(28,3)<4> uwAVS_RESPONSE(15,4)<4;4,1>
+ mov (4) uwDEST_Y(29,3)<4> uwAVS_RESPONSE(15,12)<4;4,1>
+
+// Move second 8x8 words of B to dest GRF
+ mov (4) uwDEST_Y(2,2)<4> uwAVS_RESPONSE_2(4,0)<4;4,1>
+ mov (4) uwDEST_Y(3,2)<4> uwAVS_RESPONSE_2(4,8)<4;4,1>
+ mov (4) uwDEST_Y(6,2)<4> uwAVS_RESPONSE_2(4,4)<4;4,1>
+ mov (4) uwDEST_Y(7,2)<4> uwAVS_RESPONSE_2(4,12)<4;4,1>
+ mov (4) uwDEST_Y(10,2)<4> uwAVS_RESPONSE_2(5,0)<4;4,1>
+ mov (4) uwDEST_Y(11,2)<4> uwAVS_RESPONSE_2(5,8)<4;4,1>
+ mov (4) uwDEST_Y(14,2)<4> uwAVS_RESPONSE_2(5,4)<4;4,1>
+ mov (4) uwDEST_Y(15,2)<4> uwAVS_RESPONSE_2(5,12)<4;4,1>
+ mov (4) uwDEST_Y(18,2)<4> uwAVS_RESPONSE_2(12,0)<4;4,1>
+ mov (4) uwDEST_Y(19,2)<4> uwAVS_RESPONSE_2(12,8)<4;4,1>
+ mov (4) uwDEST_Y(22,2)<4> uwAVS_RESPONSE_2(12,4)<4;4,1>
+ mov (4) uwDEST_Y(23,2)<4> uwAVS_RESPONSE_2(12,12)<4;4,1>
+ mov (4) uwDEST_Y(26,2)<4> uwAVS_RESPONSE_2(13,0)<4;4,1>
+ mov (4) uwDEST_Y(27,2)<4> uwAVS_RESPONSE_2(13,8)<4;4,1>
+ mov (4) uwDEST_Y(30,2)<4> uwAVS_RESPONSE_2(13,4)<4;4,1>
+ mov (4) uwDEST_Y(31,2)<4> uwAVS_RESPONSE_2(13,12)<4;4,1>
+
+// Move second 8x8 words of G to dest GRF
+ mov (4) uwDEST_Y(2,1)<4> uwAVS_RESPONSE_2(2,0)<4;4,1>
+ mov (4) uwDEST_Y(3,1)<4> uwAVS_RESPONSE_2(2,8)<4;4,1>
+ mov (4) uwDEST_Y(6,1)<4> uwAVS_RESPONSE_2(2,4)<4;4,1>
+ mov (4) uwDEST_Y(7,1)<4> uwAVS_RESPONSE_2(2,12)<4;4,1>
+ mov (4) uwDEST_Y(10,1)<4> uwAVS_RESPONSE_2(3,0)<4;4,1>
+ mov (4) uwDEST_Y(11,1)<4> uwAVS_RESPONSE_2(3,8)<4;4,1>
+ mov (4) uwDEST_Y(14,1)<4> uwAVS_RESPONSE_2(3,4)<4;4,1>
+ mov (4) uwDEST_Y(15,1)<4> uwAVS_RESPONSE_2(3,12)<4;4,1>
+ mov (4) uwDEST_Y(18,1)<4> uwAVS_RESPONSE_2(10,0)<4;4,1>
+ mov (4) uwDEST_Y(19,1)<4> uwAVS_RESPONSE_2(10,8)<4;4,1>
+ mov (4) uwDEST_Y(22,1)<4> uwAVS_RESPONSE_2(10,4)<4;4,1>
+ mov (4) uwDEST_Y(23,1)<4> uwAVS_RESPONSE_2(10,12)<4;4,1>
+ mov (4) uwDEST_Y(26,1)<4> uwAVS_RESPONSE_2(11,0)<4;4,1>
+ mov (4) uwDEST_Y(27,1)<4> uwAVS_RESPONSE_2(11,8)<4;4,1>
+ mov (4) uwDEST_Y(30,1)<4> uwAVS_RESPONSE_2(11,4)<4;4,1>
+ mov (4) uwDEST_Y(31,1)<4> uwAVS_RESPONSE_2(11,12)<4;4,1>
+
+// Move second 8x8 words of R to dest GRF
+ mov (4) uwDEST_Y(2,0)<4> uwAVS_RESPONSE_2(0,0)<4;4,1>
+ mov (4) uwDEST_Y(3,0)<4> uwAVS_RESPONSE_2(0,8)<4;4,1>
+ mov (4) uwDEST_Y(6,0)<4> uwAVS_RESPONSE_2(0,4)<4;4,1>
+ mov (4) uwDEST_Y(7,0)<4> uwAVS_RESPONSE_2(0,12)<4;4,1>
+ mov (4) uwDEST_Y(10,0)<4> uwAVS_RESPONSE_2(1,0)<4;4,1>
+ mov (4) uwDEST_Y(11,0)<4> uwAVS_RESPONSE_2(1,8)<4;4,1>
+ mov (4) uwDEST_Y(14,0)<4> uwAVS_RESPONSE_2(1,4)<4;4,1>
+ mov (4) uwDEST_Y(15,0)<4> uwAVS_RESPONSE_2(1,12)<4;4,1>
+ mov (4) uwDEST_Y(18,0)<4> uwAVS_RESPONSE_2(8,0)<4;4,1>
+ mov (4) uwDEST_Y(19,0)<4> uwAVS_RESPONSE_2(8,8)<4;4,1>
+ mov (4) uwDEST_Y(22,0)<4> uwAVS_RESPONSE_2(8,4)<4;4,1>
+ mov (4) uwDEST_Y(23,0)<4> uwAVS_RESPONSE_2(8,12)<4;4,1>
+ mov (4) uwDEST_Y(26,0)<4> uwAVS_RESPONSE_2(9,0)<4;4,1>
+ mov (4) uwDEST_Y(27,0)<4> uwAVS_RESPONSE_2(9,8)<4;4,1>
+ mov (4) uwDEST_Y(30,0)<4> uwAVS_RESPONSE_2(9,4)<4;4,1>
+ mov (4) uwDEST_Y(31,0)<4> uwAVS_RESPONSE_2(9,12)<4;4,1>
+
+// Move second 8x8 words of A to dest GRF
+ mov (4) uwDEST_Y(2,3)<4> uwAVS_RESPONSE_2(6,0)<4;4,1>
+ mov (4) uwDEST_Y(3,3)<4> uwAVS_RESPONSE_2(6,8)<4;4,1>
+ mov (4) uwDEST_Y(6,3)<4> uwAVS_RESPONSE_2(6,4)<4;4,1>
+ mov (4) uwDEST_Y(7,3)<4> uwAVS_RESPONSE_2(6,12)<4;4,1>
+ mov (4) uwDEST_Y(10,3)<4> uwAVS_RESPONSE_2(7,0)<4;4,1>
+ mov (4) uwDEST_Y(11,3)<4> uwAVS_RESPONSE_2(7,8)<4;4,1>
+ mov (4) uwDEST_Y(14,3)<4> uwAVS_RESPONSE_2(7,4)<4;4,1>
+ mov (4) uwDEST_Y(15,3)<4> uwAVS_RESPONSE_2(7,12)<4;4,1>
+ mov (4) uwDEST_Y(18,3)<4> uwAVS_RESPONSE_2(14,0)<4;4,1>
+ mov (4) uwDEST_Y(19,3)<4> uwAVS_RESPONSE_2(14,8)<4;4,1>
+ mov (4) uwDEST_Y(22,3)<4> uwAVS_RESPONSE_2(14,4)<4;4,1>
+ mov (4) uwDEST_Y(23,3)<4> uwAVS_RESPONSE_2(14,12)<4;4,1>
+ mov (4) uwDEST_Y(26,3)<4> uwAVS_RESPONSE_2(15,0)<4;4,1>
+ mov (4) uwDEST_Y(27,3)<4> uwAVS_RESPONSE_2(15,8)<4;4,1>
+ mov (4) uwDEST_Y(30,3)<4> uwAVS_RESPONSE_2(15,4)<4;4,1>
+ mov (4) uwDEST_Y(31,3)<4> uwAVS_RESPONSE_2(15,12)<4;4,1>
+
+#else /* OUTPUT_8_BIT */
+// Move first 8x8 words of B to dest GRF
+ mov (8) ubDEST_Y(0,2)<4> ubAVS_RESPONSE(4,1)<16;4,2>
+ mov (8) ubDEST_Y(2,2)<4> ubAVS_RESPONSE(4,8+1)<16;4,2>
+ mov (8) ubDEST_Y(4,2)<4> ubAVS_RESPONSE(5,1)<16;4,2>
+ mov (8) ubDEST_Y(6,2)<4> ubAVS_RESPONSE(5,8+1)<16;4,2>
+ mov (8) ubDEST_Y(8,2)<4> ubAVS_RESPONSE(12,1)<16;4,2>
+ mov (8) ubDEST_Y(10,2)<4> ubAVS_RESPONSE(12,8+1)<16;4,2>
+ mov (8) ubDEST_Y(12,2)<4> ubAVS_RESPONSE(13,1)<16;4,2>
+ mov (8) ubDEST_Y(14,2)<4> ubAVS_RESPONSE(13,8+1)<16;4,2>
+
+// Move first 8x8 words of G to dest GRF
+ mov (8) ubDEST_Y(0,1)<4> ubAVS_RESPONSE(2,1)<16;4,2>
+ mov (8) ubDEST_Y(2,1)<4> ubAVS_RESPONSE(2,8+1)<16;4,2>
+ mov (8) ubDEST_Y(4,1)<4> ubAVS_RESPONSE(3,1)<16;4,2>
+ mov (8) ubDEST_Y(6,1)<4> ubAVS_RESPONSE(3,8+1)<16;4,2>
+ mov (8) ubDEST_Y(8,1)<4> ubAVS_RESPONSE(10,1)<16;4,2>
+ mov (8) ubDEST_Y(10,1)<4> ubAVS_RESPONSE(10,8+1)<16;4,2>
+ mov (8) ubDEST_Y(12,1)<4> ubAVS_RESPONSE(11,1)<16;4,2>
+ mov (8) ubDEST_Y(14,1)<4> ubAVS_RESPONSE(11,8+1)<16;4,2>
+
+// Move first 8x8 words of R to dest GRF
+ mov (8) ubDEST_Y(0,0)<4> ubAVS_RESPONSE(0,1)<16;4,2>
+ mov (8) ubDEST_Y(2,0)<4> ubAVS_RESPONSE(0,8+1)<16;4,2>
+ mov (8) ubDEST_Y(4,0)<4> ubAVS_RESPONSE(1,1)<16;4,2>
+ mov (8) ubDEST_Y(6,0)<4> ubAVS_RESPONSE(1,8+1)<16;4,2>
+ mov (8) ubDEST_Y(8,0)<4> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) ubDEST_Y(10,0)<4> ubAVS_RESPONSE(8,8+1)<16;4,2>
+ mov (8) ubDEST_Y(12,0)<4> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) ubDEST_Y(14,0)<4> ubAVS_RESPONSE(9,8+1)<16;4,2>
+
+// Move first 8x8 words of A to dest GRF
+ mov (8) ubDEST_Y(0,3)<4> ubAVS_RESPONSE(6,1)<16;4,2>
+ mov (8) ubDEST_Y(2,3)<4> ubAVS_RESPONSE(6,8+1)<16;4,2>
+ mov (8) ubDEST_Y(4,3)<4> ubAVS_RESPONSE(7,1)<16;4,2>
+ mov (8) ubDEST_Y(6,3)<4> ubAVS_RESPONSE(7,8+1)<16;4,2>
+ mov (8) ubDEST_Y(8,3)<4> ubAVS_RESPONSE(14,1)<16;4,2>
+ mov (8) ubDEST_Y(10,3)<4> ubAVS_RESPONSE(14,8+1)<16;4,2>
+ mov (8) ubDEST_Y(12,3)<4> ubAVS_RESPONSE(15,1)<16;4,2>
+ mov (8) ubDEST_Y(14,3)<4> ubAVS_RESPONSE(15,8+1)<16;4,2>
+
+// Move second 8x8 words of B to dest GRF
+ mov (8) ubDEST_Y(1,2)<4> ubAVS_RESPONSE_2(4,1)<16;4,2>
+ mov (8) ubDEST_Y(3,2)<4> ubAVS_RESPONSE_2(4,8+1)<16;4,2>
+ mov (8) ubDEST_Y(5,2)<4> ubAVS_RESPONSE_2(5,1)<16;4,2>
+ mov (8) ubDEST_Y(7,2)<4> ubAVS_RESPONSE_2(5,8+1)<16;4,2>
+ mov (8) ubDEST_Y(9,2)<4> ubAVS_RESPONSE_2(12,1)<16;4,2>
+ mov (8) ubDEST_Y(11,2)<4> ubAVS_RESPONSE_2(12,8+1)<16;4,2>
+ mov (8) ubDEST_Y(13,2)<4> ubAVS_RESPONSE_2(13,1)<16;4,2>
+ mov (8) ubDEST_Y(15,2)<4> ubAVS_RESPONSE_2(13,8+1)<16;4,2>
+
+// Move second 8x8 words of G to dest GRF
+ mov (8) ubDEST_Y(1,1)<4> ubAVS_RESPONSE_2(2,1)<16;4,2>
+ mov (8) ubDEST_Y(3,1)<4> ubAVS_RESPONSE_2(2,8+1)<16;4,2>
+ mov (8) ubDEST_Y(5,1)<4> ubAVS_RESPONSE_2(3,1)<16;4,2>
+ mov (8) ubDEST_Y(7,1)<4> ubAVS_RESPONSE_2(3,8+1)<16;4,2>
+ mov (8) ubDEST_Y(9,1)<4> ubAVS_RESPONSE_2(10,1)<16;4,2>
+ mov (8) ubDEST_Y(11,1)<4> ubAVS_RESPONSE_2(10,8+1)<16;4,2>
+ mov (8) ubDEST_Y(13,1)<4> ubAVS_RESPONSE_2(11,1)<16;4,2>
+ mov (8) ubDEST_Y(15,1)<4> ubAVS_RESPONSE_2(11,8+1)<16;4,2>
+
+// Move second 8x8 words of R to dest GRF
+ mov (8) ubDEST_Y(1,0)<4> ubAVS_RESPONSE_2(0,1)<16;4,2>
+ mov (8) ubDEST_Y(3,0)<4> ubAVS_RESPONSE_2(0,8+1)<16;4,2>
+ mov (8) ubDEST_Y(5,0)<4> ubAVS_RESPONSE_2(1,1)<16;4,2>
+ mov (8) ubDEST_Y(7,0)<4> ubAVS_RESPONSE_2(1,8+1)<16;4,2>
+ mov (8) ubDEST_Y(9,0)<4> ubAVS_RESPONSE_2(8,1)<16;4,2>
+ mov (8) ubDEST_Y(11,0)<4> ubAVS_RESPONSE_2(8,8+1)<16;4,2>
+ mov (8) ubDEST_Y(13,0)<4> ubAVS_RESPONSE_2(9,1)<16;4,2>
+ mov (8) ubDEST_Y(15,0)<4> ubAVS_RESPONSE_2(9,8+1)<16;4,2>
+
+// Move second 8x8 words of A to dest GRF
+ mov (8) ubDEST_Y(1,3)<4> ubAVS_RESPONSE_2(6,1)<16;4,2>
+ mov (8) ubDEST_Y(3,3)<4> ubAVS_RESPONSE_2(6,8+1)<16;4,2>
+ mov (8) ubDEST_Y(5,3)<4> ubAVS_RESPONSE_2(7,1)<16;4,2>
+ mov (8) ubDEST_Y(7,3)<4> ubAVS_RESPONSE_2(7,8+1)<16;4,2>
+ mov (8) ubDEST_Y(9,3)<4> ubAVS_RESPONSE_2(14,1)<16;4,2>
+ mov (8) ubDEST_Y(11,3)<4> ubAVS_RESPONSE_2(14,8+1)<16;4,2>
+ mov (8) ubDEST_Y(13,3)<4> ubAVS_RESPONSE_2(15,1)<16;4,2>
+ mov (8) ubDEST_Y(15,3)<4> ubAVS_RESPONSE_2(15,8+1)<16;4,2>
+#endif
+//------------------------------------------------------------------------------
+
+ // Set to write bottom region to memory
+ #define SRC_REGION REGION_2
+
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- RGB_AVS_IEF_Unpack_16x8.asm ----------
+#include "AVS_IEF.inc"
+
+.declare DEST_B Base=REG(r,10) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare DEST_G Base=REG(r,18) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare DEST_R Base=REG(r,26) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare DEST_A Base=REG(r,34) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+
+
+#ifdef AVS_OUTPUT_16_BIT
+//This portion will need to be changed if unpacking is required for Y416 kernels (in case of blending etc) - vK
+
+//// Move first 8x8 words of B to dest GRF (as packed)
+// mov (4) uwDEST_Y(0,2)<4> uwAVS_RESPONSE(4,0)<4;4,1>
+// mov (4) uwDEST_Y(1,2)<4> uwAVS_RESPONSE(4,8)<4;4,1>
+// mov (4) uwDEST_Y(4,2)<4> uwAVS_RESPONSE(4,4)<4;4,1>
+// mov (4) uwDEST_Y(5,2)<4> uwAVS_RESPONSE(4,12)<4;4,1>
+// mov (4) uwDEST_Y(8,2)<4> uwAVS_RESPONSE(5,0)<4;4,1>
+// mov (4) uwDEST_Y(9,2)<4> uwAVS_RESPONSE(5,8)<4;4,1>
+// mov (4) uwDEST_Y(12,2)<4> uwAVS_RESPONSE(5,4)<4;4,1>
+// mov (4) uwDEST_Y(13,2)<4> uwAVS_RESPONSE(5,12)<4;4,1>
+// mov (4) uwDEST_Y(16,2)<4> uwAVS_RESPONSE(12,0)<4;4,1>
+// mov (4) uwDEST_Y(17,2)<4> uwAVS_RESPONSE(12,8)<4;4,1>
+// mov (4) uwDEST_Y(20,2)<4> uwAVS_RESPONSE(12,4)<4;4,1>
+// mov (4) uwDEST_Y(21,2)<4> uwAVS_RESPONSE(12,12)<4;4,1>
+// mov (4) uwDEST_Y(24,2)<4> uwAVS_RESPONSE(13,0)<4;4,1>
+// mov (4) uwDEST_Y(25,2)<4> uwAVS_RESPONSE(13,8)<4;4,1>
+// mov (4) uwDEST_Y(28,2)<4> uwAVS_RESPONSE(13,4)<4;4,1>
+// mov (4) uwDEST_Y(29,2)<4> uwAVS_RESPONSE(13,12)<4;4,1>
+//
+//// Move first 8x8 words of G to dest GRF (as packed)
+// mov (4) uwDEST_Y(0,1)<4> uwAVS_RESPONSE(2,0)<4;4,1>
+// mov (4) uwDEST_Y(1,1)<4> uwAVS_RESPONSE(2,8)<4;4,1>
+// mov (4) uwDEST_Y(4,1)<4> uwAVS_RESPONSE(2,4)<4;4,1>
+// mov (4) uwDEST_Y(5,1)<4> uwAVS_RESPONSE(2,12)<4;4,1>
+// mov (4) uwDEST_Y(8,1)<4> uwAVS_RESPONSE(3,0)<4;4,1>
+// mov (4) uwDEST_Y(9,1)<4> uwAVS_RESPONSE(3,8)<4;4,1>
+// mov (4) uwDEST_Y(12,1)<4> uwAVS_RESPONSE(3,4)<4;4,1>
+// mov (4) uwDEST_Y(13,1)<4> uwAVS_RESPONSE(3,12)<4;4,1>
+// mov (4) uwDEST_Y(16,1)<4> uwAVS_RESPONSE(10,0)<4;4,1>
+// mov (4) uwDEST_Y(17,1)<4> uwAVS_RESPONSE(10,8)<4;4,1>
+// mov (4) uwDEST_Y(20,1)<4> uwAVS_RESPONSE(10,4)<4;4,1>
+// mov (4) uwDEST_Y(21,1)<4> uwAVS_RESPONSE(10,12)<4;4,1>
+// mov (4) uwDEST_Y(24,1)<4> uwAVS_RESPONSE(11,0)<4;4,1>
+// mov (4) uwDEST_Y(25,1)<4> uwAVS_RESPONSE(11,8)<4;4,1>
+// mov (4) uwDEST_Y(28,1)<4> uwAVS_RESPONSE(11,4)<4;4,1>
+// mov (4) uwDEST_Y(29,1)<4> uwAVS_RESPONSE(11,12)<4;4,1>
+//
+//// Move first 8x8 words of R to dest GRF (as packed)
+// mov (4) uwDEST_Y(0,0)<4> uwAVS_RESPONSE(0,0)<4;4,1>
+// mov (4) uwDEST_Y(1,0)<4> uwAVS_RESPONSE(0,8)<4;4,1>
+// mov (4) uwDEST_Y(4,0)<4> uwAVS_RESPONSE(0,4)<4;4,1>
+// mov (4) uwDEST_Y(5,0)<4> uwAVS_RESPONSE(0,12)<4;4,1>
+// mov (4) uwDEST_Y(8,0)<4> uwAVS_RESPONSE(1,0)<4;4,1>
+// mov (4) uwDEST_Y(9,0)<4> uwAVS_RESPONSE(1,8)<4;4,1>
+// mov (4) uwDEST_Y(12,0)<4> uwAVS_RESPONSE(1,4)<4;4,1>
+// mov (4) uwDEST_Y(13,0)<4> uwAVS_RESPONSE(1,12)<4;4,1>
+// mov (4) uwDEST_Y(16,0)<4> uwAVS_RESPONSE(8,0)<4;4,1>
+// mov (4) uwDEST_Y(17,0)<4> uwAVS_RESPONSE(8,8)<4;4,1>
+// mov (4) uwDEST_Y(20,0)<4> uwAVS_RESPONSE(8,4)<4;4,1>
+// mov (4) uwDEST_Y(21,0)<4> uwAVS_RESPONSE(8,12)<4;4,1>
+// mov (4) uwDEST_Y(24,0)<4> uwAVS_RESPONSE(9,0)<4;4,1>
+// mov (4) uwDEST_Y(25,0)<4> uwAVS_RESPONSE(9,8)<4;4,1>
+// mov (4) uwDEST_Y(28,0)<4> uwAVS_RESPONSE(9,4)<4;4,1>
+// mov (4) uwDEST_Y(29,0)<4> uwAVS_RESPONSE(9,12)<4;4,1>
+//
+//// Move first 8x8 words of A to dest GRF (as packed)
+// mov (4) uwDEST_Y(0,3)<4> uwAVS_RESPONSE(6,0)<4;4,1>
+// mov (4) uwDEST_Y(1,3)<4> uwAVS_RESPONSE(6,8)<4;4,1>
+// mov (4) uwDEST_Y(4,3)<4> uwAVS_RESPONSE(6,4)<4;4,1>
+// mov (4) uwDEST_Y(5,3)<4> uwAVS_RESPONSE(6,12)<4;4,1>
+// mov (4) uwDEST_Y(8,3)<4> uwAVS_RESPONSE(7,0)<4;4,1>
+// mov (4) uwDEST_Y(9,3)<4> uwAVS_RESPONSE(7,8)<4;4,1>
+// mov (4) uwDEST_Y(12,3)<4> uwAVS_RESPONSE(7,4)<4;4,1>
+// mov (4) uwDEST_Y(13,3)<4> uwAVS_RESPONSE(7,12)<4;4,1>
+// mov (4) uwDEST_Y(16,3)<4> uwAVS_RESPONSE(14,0)<4;4,1>
+// mov (4) uwDEST_Y(17,3)<4> uwAVS_RESPONSE(14,8)<4;4,1>
+// mov (4) uwDEST_Y(20,3)<4> uwAVS_RESPONSE(14,4)<4;4,1>
+// mov (4) uwDEST_Y(21,3)<4> uwAVS_RESPONSE(14,12)<4;4,1>
+// mov (4) uwDEST_Y(24,3)<4> uwAVS_RESPONSE(15,0)<4;4,1>
+// mov (4) uwDEST_Y(25,3)<4> uwAVS_RESPONSE(15,8)<4;4,1>
+// mov (4) uwDEST_Y(28,3)<4> uwAVS_RESPONSE(15,4)<4;4,1>
+// mov (4) uwDEST_Y(29,3)<4> uwAVS_RESPONSE(15,12)<4;4,1>
+//
+//// Move second 8x8 words of B to dest GRF
+// mov (4) uwDEST_Y(2,2)<4> uwAVS_RESPONSE_2(4,0)<4;4,1>
+// mov (4) uwDEST_Y(3,2)<4> uwAVS_RESPONSE_2(4,8)<4;4,1>
+// mov (4) uwDEST_Y(6,2)<4> uwAVS_RESPONSE_2(4,4)<4;4,1>
+// mov (4) uwDEST_Y(7,2)<4> uwAVS_RESPONSE_2(4,12)<4;4,1>
+// mov (4) uwDEST_Y(10,2)<4> uwAVS_RESPONSE_2(5,0)<4;4,1>
+// mov (4) uwDEST_Y(11,2)<4> uwAVS_RESPONSE_2(5,8)<4;4,1>
+// mov (4) uwDEST_Y(14,2)<4> uwAVS_RESPONSE_2(5,4)<4;4,1>
+// mov (4) uwDEST_Y(15,2)<4> uwAVS_RESPONSE_2(5,12)<4;4,1>
+// mov (4) uwDEST_Y(18,2)<4> uwAVS_RESPONSE_2(12,0)<4;4,1>
+// mov (4) uwDEST_Y(19,2)<4> uwAVS_RESPONSE_2(12,8)<4;4,1>
+// mov (4) uwDEST_Y(22,2)<4> uwAVS_RESPONSE_2(12,4)<4;4,1>
+// mov (4) uwDEST_Y(23,2)<4> uwAVS_RESPONSE_2(12,12)<4;4,1>
+// mov (4) uwDEST_Y(26,2)<4> uwAVS_RESPONSE_2(13,0)<4;4,1>
+// mov (4) uwDEST_Y(27,2)<4> uwAVS_RESPONSE_2(13,8)<4;4,1>
+// mov (4) uwDEST_Y(30,2)<4> uwAVS_RESPONSE_2(13,4)<4;4,1>
+// mov (4) uwDEST_Y(31,2)<4> uwAVS_RESPONSE_2(13,12)<4;4,1>
+//
+//// Move second 8x8 words of G to dest GRF
+// mov (4) uwDEST_Y(2,1)<4> uwAVS_RESPONSE_2(2,0)<4;4,1>
+// mov (4) uwDEST_Y(3,1)<4> uwAVS_RESPONSE_2(2,8)<4;4,1>
+// mov (4) uwDEST_Y(6,1)<4> uwAVS_RESPONSE_2(2,4)<4;4,1>
+// mov (4) uwDEST_Y(7,1)<4> uwAVS_RESPONSE_2(2,12)<4;4,1>
+// mov (4) uwDEST_Y(10,1)<4> uwAVS_RESPONSE_2(3,0)<4;4,1>
+// mov (4) uwDEST_Y(11,1)<4> uwAVS_RESPONSE_2(3,8)<4;4,1>
+// mov (4) uwDEST_Y(14,1)<4> uwAVS_RESPONSE_2(3,4)<4;4,1>
+// mov (4) uwDEST_Y(15,1)<4> uwAVS_RESPONSE_2(3,12)<4;4,1>
+// mov (4) uwDEST_Y(18,1)<4> uwAVS_RESPONSE_2(10,0)<4;4,1>
+// mov (4) uwDEST_Y(19,1)<4> uwAVS_RESPONSE_2(10,8)<4;4,1>
+// mov (4) uwDEST_Y(22,1)<4> uwAVS_RESPONSE_2(10,4)<4;4,1>
+// mov (4) uwDEST_Y(23,1)<4> uwAVS_RESPONSE_2(10,12)<4;4,1>
+// mov (4) uwDEST_Y(26,1)<4> uwAVS_RESPONSE_2(11,0)<4;4,1>
+// mov (4) uwDEST_Y(27,1)<4> uwAVS_RESPONSE_2(11,8)<4;4,1>
+// mov (4) uwDEST_Y(30,1)<4> uwAVS_RESPONSE_2(11,4)<4;4,1>
+// mov (4) uwDEST_Y(31,1)<4> uwAVS_RESPONSE_2(11,12)<4;4,1>
+//
+//// Move second 8x8 words of R to dest GRF
+// mov (4) uwDEST_Y(2,0)<4> uwAVS_RESPONSE_2(0,0)<4;4,1>
+// mov (4) uwDEST_Y(3,0)<4> uwAVS_RESPONSE_2(0,8)<4;4,1>
+// mov (4) uwDEST_Y(6,0)<4> uwAVS_RESPONSE_2(0,4)<4;4,1>
+// mov (4) uwDEST_Y(7,0)<4> uwAVS_RESPONSE_2(0,12)<4;4,1>
+// mov (4) uwDEST_Y(10,0)<4> uwAVS_RESPONSE_2(1,0)<4;4,1>
+// mov (4) uwDEST_Y(11,0)<4> uwAVS_RESPONSE_2(1,8)<4;4,1>
+// mov (4) uwDEST_Y(14,0)<4> uwAVS_RESPONSE_2(1,4)<4;4,1>
+// mov (4) uwDEST_Y(15,0)<4> uwAVS_RESPONSE_2(1,12)<4;4,1>
+// mov (4) uwDEST_Y(18,0)<4> uwAVS_RESPONSE_2(8,0)<4;4,1>
+// mov (4) uwDEST_Y(19,0)<4> uwAVS_RESPONSE_2(8,8)<4;4,1>
+// mov (4) uwDEST_Y(22,0)<4> uwAVS_RESPONSE_2(8,4)<4;4,1>
+// mov (4) uwDEST_Y(23,0)<4> uwAVS_RESPONSE_2(8,12)<4;4,1>
+// mov (4) uwDEST_Y(26,0)<4> uwAVS_RESPONSE_2(9,0)<4;4,1>
+// mov (4) uwDEST_Y(27,0)<4> uwAVS_RESPONSE_2(9,8)<4;4,1>
+// mov (4) uwDEST_Y(30,0)<4> uwAVS_RESPONSE_2(9,4)<4;4,1>
+// mov (4) uwDEST_Y(31,0)<4> uwAVS_RESPONSE_2(9,12)<4;4,1>
+//
+//// Move second 8x8 words of A to dest GRF
+// mov (4) uwDEST_Y(2,3)<4> uwAVS_RESPONSE_2(6,0)<4;4,1>
+// mov (4) uwDEST_Y(3,3)<4> uwAVS_RESPONSE_2(6,8)<4;4,1>
+// mov (4) uwDEST_Y(6,3)<4> uwAVS_RESPONSE_2(6,4)<4;4,1>
+// mov (4) uwDEST_Y(7,3)<4> uwAVS_RESPONSE_2(6,12)<4;4,1>
+// mov (4) uwDEST_Y(10,3)<4> uwAVS_RESPONSE_2(7,0)<4;4,1>
+// mov (4) uwDEST_Y(11,3)<4> uwAVS_RESPONSE_2(7,8)<4;4,1>
+// mov (4) uwDEST_Y(14,3)<4> uwAVS_RESPONSE_2(7,4)<4;4,1>
+// mov (4) uwDEST_Y(15,3)<4> uwAVS_RESPONSE_2(7,12)<4;4,1>
+// mov (4) uwDEST_Y(18,3)<4> uwAVS_RESPONSE_2(14,0)<4;4,1>
+// mov (4) uwDEST_Y(19,3)<4> uwAVS_RESPONSE_2(14,8)<4;4,1>
+// mov (4) uwDEST_Y(22,3)<4> uwAVS_RESPONSE_2(14,4)<4;4,1>
+// mov (4) uwDEST_Y(23,3)<4> uwAVS_RESPONSE_2(14,12)<4;4,1>
+// mov (4) uwDEST_Y(26,3)<4> uwAVS_RESPONSE_2(15,0)<4;4,1>
+// mov (4) uwDEST_Y(27,3)<4> uwAVS_RESPONSE_2(15,8)<4;4,1>
+// mov (4) uwDEST_Y(30,3)<4> uwAVS_RESPONSE_2(15,4)<4;4,1>
+// mov (4) uwDEST_Y(31,3)<4> uwAVS_RESPONSE_2(15,12)<4;4,1>
+
+#else /* OUTPUT_8_BIT */
+
+// Move first 8x8 words of B to dest GRF
+ mov (8) DEST_B(0)<1> ubAVS_RESPONSE(4,1)<16;4,2>
+ mov (8) DEST_B(1)<1> ubAVS_RESPONSE(4,8+1)<16;4,2>
+ mov (8) DEST_B(2)<1> ubAVS_RESPONSE(5,1)<16;4,2>
+ mov (8) DEST_B(3)<1> ubAVS_RESPONSE(5,8+1)<16;4,2>
+ mov (8) DEST_B(4)<1> ubAVS_RESPONSE(12,1)<16;4,2>
+ mov (8) DEST_B(5)<1> ubAVS_RESPONSE(12,8+1)<16;4,2>
+ mov (8) DEST_B(6)<1> ubAVS_RESPONSE(13,1)<16;4,2>
+ mov (8) DEST_B(7)<1> ubAVS_RESPONSE(13,8+1)<16;4,2>
+
+// Move first 8x8 words of G to dest GRF
+ mov (8) DEST_G(0)<1> ubAVS_RESPONSE(2,1)<16;4,2>
+ mov (8) DEST_G(1)<1> ubAVS_RESPONSE(2,8+1)<16;4,2>
+ mov (8) DEST_G(2)<1> ubAVS_RESPONSE(3,1)<16;4,2>
+ mov (8) DEST_G(3)<1> ubAVS_RESPONSE(3,8+1)<16;4,2>
+ mov (8) DEST_G(4)<1> ubAVS_RESPONSE(10,1)<16;4,2>
+ mov (8) DEST_G(5)<1> ubAVS_RESPONSE(10,8+1)<16;4,2>
+ mov (8) DEST_G(6)<1> ubAVS_RESPONSE(11,1)<16;4,2>
+ mov (8) DEST_G(7)<1> ubAVS_RESPONSE(11,8+1)<16;4,2>
+
+// Move first 8x8 words of R to dest GRF
+ mov (8) DEST_R(0)<1> ubAVS_RESPONSE(0,1)<16;4,2>
+ mov (8) DEST_R(1)<1> ubAVS_RESPONSE(0,8+1)<16;4,2>
+ mov (8) DEST_R(2)<1> ubAVS_RESPONSE(1,1)<16;4,2>
+ mov (8) DEST_R(3)<1> ubAVS_RESPONSE(1,8+1)<16;4,2>
+ mov (8) DEST_R(4)<1> ubAVS_RESPONSE(8,1)<16;4,2>
+ mov (8) DEST_R(5)<1> ubAVS_RESPONSE(8,8+1)<16;4,2>
+ mov (8) DEST_R(6)<1> ubAVS_RESPONSE(9,1)<16;4,2>
+ mov (8) DEST_R(7)<1> ubAVS_RESPONSE(9,8+1)<16;4,2>
+
+// Move first 8x8 words of A to dest GRF
+ mov (8) DEST_A(0)<1> ubAVS_RESPONSE(6,1)<16;4,2>
+ mov (8) DEST_A(1)<1> ubAVS_RESPONSE(6,8+1)<16;4,2>
+ mov (8) DEST_A(2)<1> ubAVS_RESPONSE(7,1)<16;4,2>
+ mov (8) DEST_A(3)<1> ubAVS_RESPONSE(7,8+1)<16;4,2>
+ mov (8) DEST_A(4)<1> ubAVS_RESPONSE(14,1)<16;4,2>
+ mov (8) DEST_A(5)<1> ubAVS_RESPONSE(14,8+1)<16;4,2>
+ mov (8) DEST_A(6)<1> ubAVS_RESPONSE(15,1)<16;4,2>
+ mov (8) DEST_A(7)<1> ubAVS_RESPONSE(15,8+1)<16;4,2>
+
+// Move second 8x8 words of B to dest GRF
+ mov (8) DEST_B(0,8)<1> ubAVS_RESPONSE_2(4,1)<16;4,2>
+ mov (8) DEST_B(1,8)<1> ubAVS_RESPONSE_2(4,8+1)<16;4,2>
+ mov (8) DEST_B(2,8)<1> ubAVS_RESPONSE_2(5,1)<16;4,2>
+ mov (8) DEST_B(3,8)<1> ubAVS_RESPONSE_2(5,8+1)<16;4,2>
+ mov (8) DEST_B(4,8)<1> ubAVS_RESPONSE_2(12,1)<16;4,2>
+ mov (8) DEST_B(5,8)<1> ubAVS_RESPONSE_2(12,8+1)<16;4,2>
+ mov (8) DEST_B(6,8)<1> ubAVS_RESPONSE_2(13,1)<16;4,2>
+ mov (8) DEST_B(7,8)<1> ubAVS_RESPONSE_2(13,8+1)<16;4,2>
+
+// Move second 8x8 words of G to dest GRF
+ mov (8) DEST_G(0,8)<1> ubAVS_RESPONSE_2(2,1)<16;4,2>
+ mov (8) DEST_G(1,8)<1> ubAVS_RESPONSE_2(2,8+1)<16;4,2>
+ mov (8) DEST_G(2,8)<1> ubAVS_RESPONSE_2(3,1)<16;4,2>
+ mov (8) DEST_G(3,8)<1> ubAVS_RESPONSE_2(3,8+1)<16;4,2>
+ mov (8) DEST_G(4,8)<1> ubAVS_RESPONSE_2(10,1)<16;4,2>
+ mov (8) DEST_G(5,8)<1> ubAVS_RESPONSE_2(10,8+1)<16;4,2>
+ mov (8) DEST_G(6,8)<1> ubAVS_RESPONSE_2(11,1)<16;4,2>
+ mov (8) DEST_G(7,8)<1> ubAVS_RESPONSE_2(11,8+1)<16;4,2>
+
+// Move second 8x8 words of R to dest GRF
+ mov (8) DEST_R(0,8)<1> ubAVS_RESPONSE_2(0,1)<16;4,2>
+ mov (8) DEST_R(1,8)<1> ubAVS_RESPONSE_2(0,8+1)<16;4,2>
+ mov (8) DEST_R(2,8)<1> ubAVS_RESPONSE_2(1,1)<16;4,2>
+ mov (8) DEST_R(3,8)<1> ubAVS_RESPONSE_2(1,8+1)<16;4,2>
+ mov (8) DEST_R(4,8)<1> ubAVS_RESPONSE_2(8,1)<16;4,2>
+ mov (8) DEST_R(5,8)<1> ubAVS_RESPONSE_2(8,8+1)<16;4,2>
+ mov (8) DEST_R(6,8)<1> ubAVS_RESPONSE_2(9,1)<16;4,2>
+ mov (8) DEST_R(7,8)<1> ubAVS_RESPONSE_2(9,8+1)<16;4,2>
+
+// Move second 8x8 words of A to dest GRF
+ mov (8) DEST_A(0,8)<1> ubAVS_RESPONSE_2(6,1)<16;4,2>
+ mov (8) DEST_A(1,8)<1> ubAVS_RESPONSE_2(6,8+1)<16;4,2>
+ mov (8) DEST_A(2,8)<1> ubAVS_RESPONSE_2(7,1)<16;4,2>
+ mov (8) DEST_A(3,8)<1> ubAVS_RESPONSE_2(7,8+1)<16;4,2>
+ mov (8) DEST_A(4,8)<1> ubAVS_RESPONSE_2(14,1)<16;4,2>
+ mov (8) DEST_A(5,8)<1> ubAVS_RESPONSE_2(14,8+1)<16;4,2>
+ mov (8) DEST_A(6,8)<1> ubAVS_RESPONSE_2(15,1)<16;4,2>
+ mov (8) DEST_A(7,8)<1> ubAVS_RESPONSE_2(15,8+1)<16;4,2>
+#endif
+//------------------------------------------------------------------------------
+
+ // Set to write bottom region to memory
+ #define SRC_REGION REGION_1
+
+ // Re-define new # of lines
+ #undef nUV_NUM_OF_ROWS
+ #undef nY_NUM_OF_ROWS
+
+ #define nY_NUM_OF_ROWS 8
+ #define nUV_NUM_OF_ROWS 8
+
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+//---------- RGB_Scaling.asm ----------
+#include "Scaling.inc"
+
+ // Build 16 elements ramp in float32 and normalized it
+// mov (8) SAMPLER_RAMP(0)<1> 0x76543210:v
+// add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+mov (4) SAMPLER_RAMP(0)<1> 0x48403000:vf //3, 2, 1, 0 in float vector
+mov (4) SAMPLER_RAMP(0,4)<1> 0x5C585450:vf //7, 6, 5, 4 in float vector
+add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
+
+//Module: PrepareScaleCoord.asm
+
+ // Setup for sampler msg hdr
+ mov (2) rMSGSRC.0<1>:ud 0:ud { NoDDClr } // Unused fields
+ mov (1) rMSGSRC.2<1>:ud 0:ud { NoDDChk } // Write and offset
+
+ // Calculate 16 v based on the step Y and vertical origin
+ mov (16) mfMSGPAYLOAD(2)<1> fSRC_VID_V_ORI<0;1,0>:f
+ mov (16) SCALE_COORD_Y<1>:f fSRC_VID_V_ORI<0;1,0>:f
+
+ // Calculate 16 u based on the step X and hori origin
+// line (16) mfMSGPAYLOAD(0)<1> SCALE_STEP_X<0;1,0>:f SAMPLER_RAMP(0) // Assign to mrf directly
+ mov (16) acc0:f fSRC_VID_H_ORI<0;1,0>:f { Compr }
+ mac (16) mfMSGPAYLOAD(0)<1> fVIDEO_STEP_X<0;1,0>:f SAMPLER_RAMP(0) { Compr }
+
+ //Setup the constants for line instruction
+ mov (1) SCALE_LINE_P255<1>:f 255.0:f { NoDDClr } //{ NoDDClr, NoDDChk }
+ mov (1) SCALE_LINE_P0_5<1>:f 0.5:f { NoDDChk }
+
+
+//------------------------------------------------------------------------------
+
+$for (0; <nY_NUM_OF_ROWS; 1) {
+
+ // Read 16 sampled pixels and store them in float32 in 8 GRFs in the order of BGRA (VYUA).
+ mov (8) MSGHDR_SCALE.0:ud rMSGSRC.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
+ send (16) SCALE_RESPONSE_YW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_RGB+nBI_CURRENT_SRC_RGB
+
+ // Calculate 16 v for next line
+ add (16) mfMSGPAYLOAD(2)<1> SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+ add (16) SCALE_COORD_Y<1>:f SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
+
+ // Scale back to [0, 255], convert f to ud
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(0) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(0)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(2) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(2)<1> acc0:f { Compr }
+
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(4) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(4)<1> acc0:f { Compr }
+
+//#if defined(SAVE_ARGB) //Only needed if Alpha value is written to the destination
+ line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(6) { Compr } // Process B, V
+ mov (16) SCALE_RESPONSE_YD(6)<1> acc0:f { Compr }
+//#endif
+
+ mov (16) DEST_R(%1)<1> SCALE_RESPONSE_YB(0) //possible error due to truncation - vK
+ mov (16) DEST_G(%1)<1> SCALE_RESPONSE_YB(2) //possible error due to truncation - vK
+ mov (16) DEST_B(%1)<1> SCALE_RESPONSE_YB(4) //possible error due to truncation - vK
+ mov (16) DEST_A(%1)<1> SCALE_RESPONSE_YB(6) //possible error due to truncation - vK
+}
--- /dev/null
+/*
+ * All Video Processing kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// File name: Scaling.inc
+
+#ifndef _SCALING_INC_
+#define _SCALING_INC_
+
+// Local variables----------------------------------------------------------------------------------
+#define MSGHDR_SCALE m1 // Message Payload Header (Uses m2, m3, m4, m5 implicitly)
+
+//--------------------------------------------------------------------------------------------------
+//r10.0 thru r33.0; Primary surface read from sampler (16x8)
+#define DEST_Y uwTOP_Y
+#define DEST_U uwTOP_U
+#define DEST_V uwTOP_V
+
+//r10.0 thru r41.0
+.declare DEST_B Base=REG(r,10) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare DEST_G Base=REG(r,18) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare DEST_R Base=REG(r,26) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+.declare DEST_A Base=REG(r,34) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw
+
+//r56.0 thru r79.0
+.declare SCALE_RESPONSE_YF Base=REG(r,nBOT_Y) ElementSize=4 SrcRegion=REGION(8,1) Type=f
+.declare SCALE_RESPONSE_UF Base=REG(r,nBOT_U) ElementSize=4 SrcRegion=REGION(8,1) Type=f
+.declare SCALE_RESPONSE_VF Base=REG(r,nBOT_V) ElementSize=4 SrcRegion=REGION(8,1) Type=f
+
+.declare SCALE_RESPONSE_YW Base=REG(r,nBOT_Y) ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+.declare SCALE_RESPONSE_UW Base=REG(r,nBOT_U) ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+.declare SCALE_RESPONSE_VW Base=REG(r,nBOT_V) ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+
+.declare SCALE_RESPONSE_YD Base=REG(r,nBOT_Y) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare SCALE_RESPONSE_UD Base=REG(r,nBOT_U) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare SCALE_RESPONSE_VD Base=REG(r,nBOT_V) ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+
+.declare SCALE_RESPONSE_YB Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+.declare SCALE_RESPONSE_UB Base=REG(r,nBOT_U) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+.declare SCALE_RESPONSE_VB Base=REG(r,nBOT_V) ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+
+.declare SAMPLER_RAMP Base=REG(r,42) ElementSize=4 SrcRegion=<8;8,1> Type=f // 2 GRFs, 16 elements
+
+//#define SCALE_STEP_X REG2(r,43,0)
+//#define SCALE_COORD_X REG2(r,43,3)
+
+#define SCALE_LINE_P255 REG2(r,43,4) // = 255.0 Used in 'line' inst to multiply 255, add 0.5, and round to int.
+#define SCALE_LINE_P0_5 REG2(r,43,7) // = 0.5
+
+//r44.0 thru r45.0
+#define SCALE_COORD_Y REG(r,44) //2GRF
+
+
+// Send Message [DevILK] Message Descriptor
+// MBZ MsgL=5 MsgR=8 H MBZ SIMD MsgType SmplrIndx BindTab
+// 000 0 101 0 1000 1 0 10 0000 0000 00000000
+// 0 A 8 A 0 0 0 0
+// MsgL=1+2*2(u,v)=5 MsgR=8
+#define SMPLR_MSG_DSC 0x0A8A0000 // ILK Sampler Message Descriptor
+
+// Re-define new number of lines
+#undef nY_NUM_OF_ROWS
+#undef nUV_NUM_OF_ROWS
+
+#define nY_NUM_OF_ROWS 8
+#define nUV_NUM_OF_ROWS 8
+
+
+#endif //_SCALING_INC_
--- /dev/null
+
+INTEL_G4I =
+
+INTEL_G4A = null.g4a
+
+INTEL_G4B = null.g4b
+
+INTEL_G4B_GEN5 = null.g4b.gen5
+
+EXTRA_DIST = $(INTEL_G4I) \
+ $(INTEL_G4A) \
+ $(INTEL_G4B) \
+ $(INTEL_G4B_GEN5)
+
+if HAVE_GEN4ASM
+
+SUFFIXES = .g4a .g4b
+.g4a.g4b:
+ m4 $*.g4a > $*.g4m && intel-gen4asm -o $@ $*.g4m && intel-gen4asm -g 5 -o $@.gen5 $*.g4m && rm $*.g4m
+
+$(INTEL_G4B): $(INTEL_G4I)
+
+BUILT_SOURCES= $(INTEL_G4B)
+
+clean-local:
+ -rm -f $(INTEL_G4B)
+ -rm -f $(INTEL_G4B_GEN5)
+endif
--- /dev/null
+/* Just for test */
+
+send(16) 0 acc0<1>UW g0<8,8,1>UW thread_spawner(0, 0, 0) mlen 1 rlen 0 {align1 EOT};
--- /dev/null
+ { 0x00800031, 0x24001d28, 0x008d0000, 0x87100000 },
--- /dev/null
+ { 0x00800031, 0x24001d28, 0x748d0000, 0x82000000 },
--- /dev/null
+// Module name: NV12_AVS_NV12
+.kernel NV12_AVS_NV12
+.code
+
+#define INC_SCALING
+
+#include "SetupVPKernel.asm"
+#include "Multiple_Loop_Head.asm"
+#include "PL2_AVS_IEF_16x8.asm"
+#include "PL8x4_Save_NV12.asm"
+#include "Multiple_Loop.asm"
+
+END_THREAD // End of Thread
+
+.end_code
+
+.end_kernel
+
+// end of nv12_avs_nv12.asm
--- /dev/null
+ { 0x00600001, 0x21000021, 0x008d0000, 0x00000000 },
+ { 0x00000441, 0x20842e2d, 0x000000b7, 0x00100010 },
+ { 0x00000c01, 0x2086022d, 0x000000bb, 0x00000000 },
+ { 0x00000801, 0x208a01ad, 0x000000a0, 0x00000000 },
+ { 0x00200001, 0x209403bd, 0x006600a4, 0x00000000 },
+ { 0x00000040, 0x208435ad, 0x00000084, 0x000000a0 },
+ { 0x01000005, 0x20002d2c, 0x00000088, 0x80008000 },
+ { 0x00010001, 0x20c003fd, 0x00000000, 0x00000000 },
+ { 0x00000001, 0x212003bd, 0x000000c0, 0x00000000 },
+ { 0x00000001, 0x212403bd, 0x000000bc, 0x00000000 },
+ { 0x00000001, 0x213403bd, 0x00000038, 0x00000000 },
+ { 0x00200001, 0x612803bd, 0x004500a4, 0x00000000 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0000d000 },
+ { 0x00802001, 0x20000022, 0x008d0100, 0x00000000 },
+ { 0x00000031, 0x25401c09, 0x208d0000, 0x044bb401 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0000a000 },
+ { 0x00802001, 0x20400022, 0x008d0100, 0x00000000 },
+ { 0x02000031, 0x25c01c09, 0x208d0000, 0x048bb802 },
+ { 0x00000001, 0x240803bc, 0x000000a4, 0x00000000 },
+ { 0x00000048, 0x24087fbc, 0x000000bc, 0x41000000 },
+ { 0x00000048, 0x21287fbd, 0x000000c0, 0x41e00000 },
+ { 0x00000001, 0x240403bc, 0x000000bc, 0x00000000 },
+ { 0x00000048, 0x21247fbd, 0x000000c0, 0x41000000 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0000d000 },
+ { 0x00802001, 0x20000022, 0x008d0100, 0x00000000 },
+ { 0x00000031, 0x27401c09, 0x208d0000, 0x044bb401 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0000a000 },
+ { 0x00802001, 0x20400022, 0x008d0100, 0x00000000 },
+ { 0x02000031, 0x27c01c09, 0x208d0000, 0x048bb802 },
+ { 0x00600001, 0x21400229, 0x00aa0541, 0x00000000 },
+ { 0x00600001, 0x21600229, 0x00aa0549, 0x00000000 },
+ { 0x00600001, 0x21800229, 0x00aa0561, 0x00000000 },
+ { 0x00600001, 0x21a00229, 0x00aa0569, 0x00000000 },
+ { 0x00600001, 0x21c00229, 0x00aa0581, 0x00000000 },
+ { 0x00600001, 0x21e00229, 0x00aa0589, 0x00000000 },
+ { 0x00600001, 0x22000229, 0x00aa05a1, 0x00000000 },
+ { 0x00600001, 0x22200229, 0x00aa05a9, 0x00000000 },
+ { 0x00600001, 0x22400229, 0x00aa05c1, 0x00000000 },
+ { 0x00600001, 0x22600229, 0x00aa05c9, 0x00000000 },
+ { 0x00600001, 0x22800229, 0x00aa05e1, 0x00000000 },
+ { 0x00600001, 0x22a00229, 0x00aa05e9, 0x00000000 },
+ { 0x00600001, 0x22c00229, 0x00aa0641, 0x00000000 },
+ { 0x00600001, 0x22e00229, 0x00aa0649, 0x00000000 },
+ { 0x00600001, 0x23000229, 0x00aa0661, 0x00000000 },
+ { 0x00600001, 0x23200229, 0x00aa0669, 0x00000000 },
+ { 0x00600001, 0x23400229, 0x00aa0601, 0x00000000 },
+ { 0x00600001, 0x23600229, 0x00aa0609, 0x00000000 },
+ { 0x00600001, 0x23800229, 0x00aa0621, 0x00000000 },
+ { 0x00600001, 0x23a00229, 0x00aa0629, 0x00000000 },
+ { 0x00600001, 0x23c00229, 0x00aa0681, 0x00000000 },
+ { 0x00600001, 0x23e00229, 0x00aa0689, 0x00000000 },
+ { 0x00600001, 0x24000229, 0x00aa06a1, 0x00000000 },
+ { 0x00600001, 0x24200229, 0x00aa06a9, 0x00000000 },
+ { 0x00600001, 0x21500229, 0x00aa0741, 0x00000000 },
+ { 0x00600001, 0x21700229, 0x00aa0749, 0x00000000 },
+ { 0x00600001, 0x21900229, 0x00aa0761, 0x00000000 },
+ { 0x00600001, 0x21b00229, 0x00aa0769, 0x00000000 },
+ { 0x00600001, 0x21d00229, 0x00aa0781, 0x00000000 },
+ { 0x00600001, 0x21f00229, 0x00aa0789, 0x00000000 },
+ { 0x00600001, 0x22100229, 0x00aa07a1, 0x00000000 },
+ { 0x00600001, 0x22300229, 0x00aa07a9, 0x00000000 },
+ { 0x00600001, 0x22500229, 0x00aa07c1, 0x00000000 },
+ { 0x00600001, 0x22700229, 0x00aa07c9, 0x00000000 },
+ { 0x00600001, 0x22900229, 0x00aa07e1, 0x00000000 },
+ { 0x00600001, 0x22b00229, 0x00aa07e9, 0x00000000 },
+ { 0x00600001, 0x22d00229, 0x00aa0841, 0x00000000 },
+ { 0x00600001, 0x22f00229, 0x00aa0849, 0x00000000 },
+ { 0x00600001, 0x23100229, 0x00aa0861, 0x00000000 },
+ { 0x00600001, 0x23300229, 0x00aa0869, 0x00000000 },
+ { 0x00600001, 0x23500229, 0x00aa0801, 0x00000000 },
+ { 0x00600001, 0x23700229, 0x00aa0809, 0x00000000 },
+ { 0x00600001, 0x23900229, 0x00aa0821, 0x00000000 },
+ { 0x00600001, 0x23b00229, 0x00aa0829, 0x00000000 },
+ { 0x00600001, 0x23d00229, 0x00aa0881, 0x00000000 },
+ { 0x00600001, 0x23f00229, 0x00aa0889, 0x00000000 },
+ { 0x00600001, 0x24100229, 0x00aa08a1, 0x00000000 },
+ { 0x00600001, 0x24300229, 0x00aa08a9, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00200001, 0x202001a6, 0x004500a0, 0x00000000 },
+ { 0x00000001, 0x20280062, 0x00000000, 0x0007000f },
+ { 0x00000005, 0x24000c20, 0x000000b8, 0x00ffffff },
+ { 0x04000010, 0x20000c04, 0x00000400, 0x00ffffff },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x00000056 },
+ { 0x01600031, 0x27000c01, 0x408d0000, 0x0248a007 },
+ { 0x0000040c, 0x21043da1, 0x000000a2, 0x00010001 },
+ { 0x00000801, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01600031, 0x28000c01, 0x408d0000, 0x0228a008 },
+ { 0x00200001, 0x210001a1, 0x004500a0, 0x00000000 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0007000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00000001, 0x26020228, 0x000000ba, 0x00000000 },
+ { 0x00610001, 0x24400129, 0x028d00b8, 0x00000000 },
+ { 0x00710001, 0x24400169, 0x02000000, 0x00000000 },
+ { 0x00000001, 0x24600061, 0x00000000, 0x00040001 },
+ { 0x00000001, 0x24640061, 0x00000000, 0x00400010 },
+ { 0x00000001, 0x24680061, 0x00000000, 0x04000100 },
+ { 0x00000001, 0x246c0061, 0x00000000, 0x40001000 },
+ { 0x00000001, 0x26020128, 0x00000440, 0x00000000 },
+ { 0x00910001, 0x41400231, 0x02b10700, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000440, 0x008d0460 },
+ { 0x00710001, 0x42400231, 0x02ae0800, 0x00000000 },
+ { 0x00710001, 0x43400231, 0x02ae0801, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000442, 0x00000000 },
+ { 0x00910001, 0x41600231, 0x02b10710, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000444, 0x00000000 },
+ { 0x00910001, 0x41800231, 0x02b10720, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000444, 0x008d0460 },
+ { 0x00710001, 0x42500231, 0x02ae0810, 0x00000000 },
+ { 0x00710001, 0x43500231, 0x02ae0811, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000446, 0x00000000 },
+ { 0x00910001, 0x41a00231, 0x02b10730, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000448, 0x00000000 },
+ { 0x00910001, 0x41c00231, 0x02b10740, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000448, 0x008d0460 },
+ { 0x00710001, 0x42600231, 0x02ae0820, 0x00000000 },
+ { 0x00710001, 0x43600231, 0x02ae0821, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044a, 0x00000000 },
+ { 0x00910001, 0x41e00231, 0x02b10750, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044c, 0x00000000 },
+ { 0x00910001, 0x42000231, 0x02b10760, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x0200044c, 0x008d0460 },
+ { 0x00710001, 0x42700231, 0x02ae0830, 0x00000000 },
+ { 0x00710001, 0x43700231, 0x02ae0831, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044e, 0x00000000 },
+ { 0x00910001, 0x42200231, 0x02b10770, 0x00000000 },
+ { 0x00800001, 0x20400232, 0x00d20140, 0x00000000 },
+ { 0x00800001, 0x20500232, 0x00d20160, 0x00000000 },
+ { 0x00800001, 0x20600232, 0x00d20180, 0x00000000 },
+ { 0x00800001, 0x20700232, 0x00d201a0, 0x00000000 },
+ { 0x00800001, 0x20800232, 0x00d201c0, 0x00000000 },
+ { 0x00800001, 0x20900232, 0x00d201e0, 0x00000000 },
+ { 0x00800001, 0x20a00232, 0x00d20200, 0x00000000 },
+ { 0x00800001, 0x20b00232, 0x00d20220, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x0a082007 },
+ { 0x00200001, 0x210001a5, 0x004500a0, 0x00000000 },
+ { 0x0000000c, 0x21043ca5, 0x00000104, 0x00010001 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00800001, 0x40400232, 0x00d20240, 0x00000000 },
+ { 0x00800001, 0x40410232, 0x00d20340, 0x00000000 },
+ { 0x00800001, 0x40600232, 0x00d20260, 0x00000000 },
+ { 0x00800001, 0x40610232, 0x00d20360, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x06082008 },
+ { 0x01000040, 0x20863dad, 0x00000086, 0xffffffff },
+ { 0x00000040, 0x20a03dad, 0x000000a0, 0x00100010 },
+ { 0x05000010, 0x2000358c, 0x02210400, 0x00000084 },
+ { 0x00000041, 0x24407fbd, 0x000000bc, 0x41800000 },
+ { 0x00000040, 0x20a477bd, 0x00000440, 0x000000a4 },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e },
+ { 0x00010220, 0x34001c00, 0x02001400, 0xfffffede },
+ { 0x00000001, 0x20a001ad, 0x0000008a, 0x00000000 },
+ { 0x00000040, 0x20a23dad, 0x000000a2, 0x00080008 },
+ { 0x00000001, 0x20a403bd, 0x00000094, 0x00000000 },
+ { 0x00000041, 0x24407fbd, 0x00000038, 0x41000000 },
+ { 0x00000040, 0x20a877bd, 0x00000440, 0x000000a8 },
+ { 0x00000220, 0x34001c00, 0x00001400, 0xfffffed2 },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
--- /dev/null
+// Module name: NV12_DNDI_NV12
+.kernel NV12_DNDI_NV12
+.code
+
+#define INC_DNDI
+
+#include "SetupVPKernel.asm"
+#include "Multiple_Loop_Head.asm"
+#include "PL_DNDI_ALG_UVCopy_NV12.asm"
+#include "Multiple_Loop.asm"
+
+END_THREAD // End of Thread
+
+.end_code
+
+.end_kernel
+
+// end of nv12_dndi_nv12.asm
--- /dev/null
+ { 0x00600001, 0x21000021, 0x008d0000, 0x00000000 },
+ { 0x00000441, 0x20842e2d, 0x000000b7, 0x00100010 },
+ { 0x00000c01, 0x2086022d, 0x000000bb, 0x00000000 },
+ { 0x00000801, 0x208a01ad, 0x000000a0, 0x00000000 },
+ { 0x00200001, 0x209403bd, 0x006600a4, 0x00000000 },
+ { 0x00000040, 0x208435ad, 0x00000084, 0x000000a0 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00000001, 0x204801aa, 0x000000a0, 0x00000000 },
+ { 0x00000001, 0x205801aa, 0x000000a2, 0x00000000 },
+ { 0x01600031, 0x24400c01, 0x208d0000, 0x04cb8004 },
+ { 0x00800001, 0x21400229, 0x00b10440, 0x00000000 },
+ { 0x00800001, 0x21600229, 0x00b10450, 0x00000000 },
+ { 0x00800001, 0x21800229, 0x00b10460, 0x00000000 },
+ { 0x00800001, 0x21a00229, 0x00b10470, 0x00000000 },
+ { 0x00600001, 0x22400229, 0x00ae0481, 0x00000000 },
+ { 0x00600001, 0x23400229, 0x00ae0480, 0x00000000 },
+ { 0x00600001, 0x22500229, 0x00ae0491, 0x00000000 },
+ { 0x00600001, 0x23500229, 0x00ae0490, 0x00000000 },
+ { 0x00600001, 0x22600229, 0x00ae04a1, 0x00000000 },
+ { 0x00600001, 0x23600229, 0x00ae04a0, 0x00000000 },
+ { 0x00600001, 0x22700229, 0x00ae04b1, 0x00000000 },
+ { 0x00600001, 0x23700229, 0x00ae04b0, 0x00000000 },
+ { 0x00800001, 0x21c00229, 0x00b104c0, 0x00000000 },
+ { 0x00800001, 0x21e00229, 0x00b104d0, 0x00000000 },
+ { 0x00800001, 0x22000229, 0x00b104e0, 0x00000000 },
+ { 0x00800001, 0x22200229, 0x00b104f0, 0x00000000 },
+ { 0x00600001, 0x22800229, 0x00ae0501, 0x00000000 },
+ { 0x00600001, 0x23800229, 0x00ae0500, 0x00000000 },
+ { 0x00600001, 0x22900229, 0x00ae0511, 0x00000000 },
+ { 0x00600001, 0x23900229, 0x00ae0510, 0x00000000 },
+ { 0x00600001, 0x22a00229, 0x00ae0521, 0x00000000 },
+ { 0x00600001, 0x23a00229, 0x00ae0520, 0x00000000 },
+ { 0x00600001, 0x22b00229, 0x00ae0531, 0x00000000 },
+ { 0x00600001, 0x23b00229, 0x00ae0530, 0x00000000 },
+ { 0x00000008, 0x21003da1, 0x000000a0, 0x00010001 },
+ { 0x00000001, 0x210401a1, 0x000000a2, 0x00000000 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x00030007 },
+ { 0x00600001, 0x21600022, 0x008d0100, 0x00000000 },
+ { 0x00600001, 0x21800022, 0x008d0540, 0x00000000 },
+ { 0x0b600031, 0x20000c04, 0x508d0000, 0x04082014 },
+ { 0x00200008, 0x21003da1, 0x004500a0, 0x00020002 },
+ { 0x00000040, 0x21002421, 0x00000100, 0x00000034 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x00000003 },
+ { 0x00600001, 0x21a00022, 0x008d0100, 0x00000000 },
+ { 0x00000001, 0x21c00022, 0x00000560, 0x00000000 },
+ { 0x0d600031, 0x20000c04, 0x508d0000, 0x04082014 },
+ { 0x00400040, 0x22083e28, 0x00690024, 0x07000700 },
+ { 0x01000010, 0x20003e2c, 0x0000003b, 0x00010001 },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x0000000a },
+ { 0x00400001, 0x20400022, 0x00690580, 0x00000000 },
+ { 0x00400001, 0x20500022, 0x006904d0, 0x00000000 },
+ { 0x00400001, 0x20600022, 0x00690590, 0x00000000 },
+ { 0x00400001, 0x20700022, 0x006904f0, 0x00000000 },
+ { 0x00000220, 0x34001c00, 0x00001400, 0x00000008 },
+ { 0x00400001, 0x20400022, 0x006904c0, 0x00000000 },
+ { 0x00400001, 0x20500022, 0x00690580, 0x00000000 },
+ { 0x00400001, 0x20600022, 0x006904e0, 0x00000000 },
+ { 0x00400001, 0x20700022, 0x00690590, 0x00000000 },
+ { 0x00200001, 0x210001a1, 0x004500a0, 0x00000000 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x06082007 },
+ { 0x00200040, 0x210035a5, 0x004500a0, 0x00450074 },
+ { 0x0000000c, 0x21043ca5, 0x00000104, 0x00010001 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0001000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01600031, 0x28000c01, 0x408d0000, 0x0218a002 },
+ { 0x00200001, 0x210001a1, 0x004500a0, 0x00000000 },
+ { 0x0000000c, 0x21043ca5, 0x00000104, 0x00010001 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0001000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00600001, 0x20400022, 0x008d0800, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x04082008 },
+ { 0x01000040, 0x20863dad, 0x00000086, 0xffffffff },
+ { 0x00000040, 0x20a03dad, 0x000000a0, 0x00100010 },
+ { 0x05000010, 0x2000358c, 0x02210400, 0x00000084 },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x00000008 },
+ { 0x00010220, 0x34001c00, 0x02001400, 0xffffff70 },
+ { 0x00000001, 0x20a001ad, 0x0000008a, 0x00000000 },
+ { 0x00000040, 0x20a23dad, 0x000000a2, 0x00080008 },
+ { 0x00000220, 0x34001c00, 0x00001400, 0xffffff6a },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
--- /dev/null
+// Module name: NV12_LOAD_SAVE_NV12
+.kernel NV12_LOAD_SAVE_NV12
+.code
+
+#include "SetupVPKernel.asm"
+#include "Multiple_Loop_Head.asm"
+#include "NV12_Load_8x4.asm"
+#include "PL8x4_Save_NV12.asm"
+#include "Multiple_Loop.asm"
+
+END_THREAD // End of Thread
+
+.end_code
+
+.end_kernel
+
+// end of nv12_load_save_nv12.asm
--- /dev/null
+ { 0x00600001, 0x21000021, 0x008d0000, 0x00000000 },
+ { 0x00000441, 0x20842e2d, 0x000000b7, 0x00100010 },
+ { 0x00000c01, 0x2086022d, 0x000000bb, 0x00000000 },
+ { 0x00000801, 0x208a01ad, 0x000000a0, 0x00000000 },
+ { 0x00200001, 0x209403bd, 0x006600a4, 0x00000000 },
+ { 0x00000040, 0x208435ad, 0x00000084, 0x000000a0 },
+ { 0x00200040, 0x210035a5, 0x004500a0, 0x00450074 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0007000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01600031, 0x27000c01, 0x408d0000, 0x0248a001 },
+ { 0x0000000c, 0x21043ca5, 0x00000104, 0x00010001 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20400022, 0x008d0100, 0x00000000 },
+ { 0x02600031, 0x28000c01, 0x408d0000, 0x0228a002 },
+ { 0x00800001, 0x22200229, 0x00b10770, 0x00000000 },
+ { 0x00800001, 0x22000229, 0x00b10760, 0x00000000 },
+ { 0x00800001, 0x21e00229, 0x00b10750, 0x00000000 },
+ { 0x00800001, 0x21c00229, 0x00b10740, 0x00000000 },
+ { 0x00800001, 0x21a00229, 0x00b10730, 0x00000000 },
+ { 0x00800001, 0x21800229, 0x00b10720, 0x00000000 },
+ { 0x00800001, 0x21600229, 0x00b10710, 0x00000000 },
+ { 0x00800001, 0x21400229, 0x00b10700, 0x00000000 },
+ { 0x00800001, 0x22600229, 0x00d20820, 0x00000000 },
+ { 0x00800001, 0x23600229, 0x00d20821, 0x00000000 },
+ { 0x00800001, 0x22400229, 0x00d20800, 0x00000000 },
+ { 0x00800001, 0x23400229, 0x00d20801, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00200001, 0x202001a6, 0x004500a0, 0x00000000 },
+ { 0x00000001, 0x20280062, 0x00000000, 0x0007000f },
+ { 0x00000005, 0x24000c20, 0x000000b8, 0x00ffffff },
+ { 0x04000010, 0x20000c04, 0x00000400, 0x00ffffff },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x00000056 },
+ { 0x01600031, 0x27000c01, 0x408d0000, 0x0248a007 },
+ { 0x0000040c, 0x21043da1, 0x000000a2, 0x00010001 },
+ { 0x00000801, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01600031, 0x28000c01, 0x408d0000, 0x0228a008 },
+ { 0x00200001, 0x210001a1, 0x004500a0, 0x00000000 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0007000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00000001, 0x26020228, 0x000000ba, 0x00000000 },
+ { 0x00610001, 0x24400129, 0x028d00b8, 0x00000000 },
+ { 0x00710001, 0x24400169, 0x02000000, 0x00000000 },
+ { 0x00000001, 0x24600061, 0x00000000, 0x00040001 },
+ { 0x00000001, 0x24640061, 0x00000000, 0x00400010 },
+ { 0x00000001, 0x24680061, 0x00000000, 0x04000100 },
+ { 0x00000001, 0x246c0061, 0x00000000, 0x40001000 },
+ { 0x00000001, 0x26020128, 0x00000440, 0x00000000 },
+ { 0x00910001, 0x41400231, 0x02b10700, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000440, 0x008d0460 },
+ { 0x00710001, 0x42400231, 0x02ae0800, 0x00000000 },
+ { 0x00710001, 0x43400231, 0x02ae0801, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000442, 0x00000000 },
+ { 0x00910001, 0x41600231, 0x02b10710, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000444, 0x00000000 },
+ { 0x00910001, 0x41800231, 0x02b10720, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000444, 0x008d0460 },
+ { 0x00710001, 0x42500231, 0x02ae0810, 0x00000000 },
+ { 0x00710001, 0x43500231, 0x02ae0811, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000446, 0x00000000 },
+ { 0x00910001, 0x41a00231, 0x02b10730, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000448, 0x00000000 },
+ { 0x00910001, 0x41c00231, 0x02b10740, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000448, 0x008d0460 },
+ { 0x00710001, 0x42600231, 0x02ae0820, 0x00000000 },
+ { 0x00710001, 0x43600231, 0x02ae0821, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044a, 0x00000000 },
+ { 0x00910001, 0x41e00231, 0x02b10750, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044c, 0x00000000 },
+ { 0x00910001, 0x42000231, 0x02b10760, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x0200044c, 0x008d0460 },
+ { 0x00710001, 0x42700231, 0x02ae0830, 0x00000000 },
+ { 0x00710001, 0x43700231, 0x02ae0831, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044e, 0x00000000 },
+ { 0x00910001, 0x42200231, 0x02b10770, 0x00000000 },
+ { 0x00800001, 0x20400232, 0x00d20140, 0x00000000 },
+ { 0x00800001, 0x20500232, 0x00d20160, 0x00000000 },
+ { 0x00800001, 0x20600232, 0x00d20180, 0x00000000 },
+ { 0x00800001, 0x20700232, 0x00d201a0, 0x00000000 },
+ { 0x00800001, 0x20800232, 0x00d201c0, 0x00000000 },
+ { 0x00800001, 0x20900232, 0x00d201e0, 0x00000000 },
+ { 0x00800001, 0x20a00232, 0x00d20200, 0x00000000 },
+ { 0x00800001, 0x20b00232, 0x00d20220, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x0a082007 },
+ { 0x00200001, 0x210001a5, 0x004500a0, 0x00000000 },
+ { 0x0000000c, 0x21043ca5, 0x00000104, 0x00010001 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00800001, 0x40400232, 0x00d20240, 0x00000000 },
+ { 0x00800001, 0x40410232, 0x00d20340, 0x00000000 },
+ { 0x00800001, 0x40600232, 0x00d20260, 0x00000000 },
+ { 0x00800001, 0x40610232, 0x00d20360, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x06082008 },
+ { 0x01000040, 0x20863dad, 0x00000086, 0xffffffff },
+ { 0x00000040, 0x20a03dad, 0x000000a0, 0x00100010 },
+ { 0x05000010, 0x2000358c, 0x02210400, 0x00000084 },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x00000008 },
+ { 0x00010220, 0x34001c00, 0x02001400, 0xffffff48 },
+ { 0x00000001, 0x20a001ad, 0x0000008a, 0x00000000 },
+ { 0x00000040, 0x20a23dad, 0x000000a2, 0x00080008 },
+ { 0x00000220, 0x34001c00, 0x00001400, 0xffffff42 },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
--- /dev/null
+// Module name: NV12_SCALING_NV12
+.kernel NV12_SCALING_NV12
+.code
+
+#define INC_SCALING
+
+#include "SetupVPKernel.asm"
+#include "Multiple_Loop_Head.asm"
+#include "PL2_Scaling.asm"
+#include "PL16x8_PL8x4.asm"
+#include "PL8x4_Save_NV12.asm"
+#include "Multiple_Loop.asm"
+
+END_THREAD // End of Thread
+
+.end_code
+
+.end_kernel
+
+// end of nv12_scaling_nv12.asm
--- /dev/null
+ { 0x00600001, 0x21000021, 0x008d0000, 0x00000000 },
+ { 0x00000441, 0x20842e2d, 0x000000b7, 0x00100010 },
+ { 0x00000c01, 0x2086022d, 0x000000bb, 0x00000000 },
+ { 0x00000801, 0x208a01ad, 0x000000a0, 0x00000000 },
+ { 0x00200001, 0x209403bd, 0x006600a4, 0x00000000 },
+ { 0x00000040, 0x208435ad, 0x00000084, 0x000000a0 },
+ { 0x00400001, 0x254002fd, 0x00000000, 0x48403000 },
+ { 0x00400001, 0x255002fd, 0x00000000, 0x5c585450 },
+ { 0x00600040, 0x25607fbd, 0x008d0540, 0x41000000 },
+ { 0x00200401, 0x21000061, 0x00000000, 0x00000000 },
+ { 0x00000801, 0x21080061, 0x00000000, 0x00000000 },
+ { 0x00802001, 0x208003be, 0x000000a8, 0x00000000 },
+ { 0x00802001, 0x258003bd, 0x000000a8, 0x00000000 },
+ { 0x00802001, 0x240003bc, 0x000000a4, 0x00000000 },
+ { 0x00802048, 0x204077be, 0x000000bc, 0x008d0540 },
+ { 0x00000401, 0x257003fd, 0x00000000, 0x437f0000 },
+ { 0x00000801, 0x257c03fd, 0x00000000, 0x3f000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x21400229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x22400229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x23400229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x21600229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x22600229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x23600229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x21800229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x22800229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x23800229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x21a00229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x22a00229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x23a00229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x21c00229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x22c00229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x23c00229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x21e00229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x22e00229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x23e00229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x22000229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x23000229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x24000229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01800031, 0x27001c09, 0x208d0000, 0x0a8a0101 },
+ { 0x01800031, 0x28001c09, 0x208d0000, 0x0a8a0202 },
+ { 0x00802040, 0x208077be, 0x008d0580, 0x00000038 },
+ { 0x00802040, 0x258077bd, 0x008d0580, 0x00000038 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0700 },
+ { 0x00802001, 0x27000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0800 },
+ { 0x00802001, 0x28000381, 0x00b10400, 0x00000000 },
+ { 0x00802059, 0x240077bc, 0x00000570, 0x008d0840 },
+ { 0x00802001, 0x28400381, 0x00b10400, 0x00000000 },
+ { 0x00800001, 0x22200229, 0x00cf0700, 0x00000000 },
+ { 0x00800001, 0x23200229, 0x00cf0800, 0x00000000 },
+ { 0x00800001, 0x24200229, 0x00cf0840, 0x00000000 },
+ { 0x00600001, 0x22400129, 0x00ae0240, 0x00000000 },
+ { 0x00600001, 0x23400129, 0x00ae0340, 0x00000000 },
+ { 0x00600001, 0x22500129, 0x00ae0280, 0x00000000 },
+ { 0x00600001, 0x23500129, 0x00ae0380, 0x00000000 },
+ { 0x00600001, 0x22600129, 0x00ae02c0, 0x00000000 },
+ { 0x00600001, 0x23600129, 0x00ae03c0, 0x00000000 },
+ { 0x00600001, 0x22700129, 0x00ae0300, 0x00000000 },
+ { 0x00600001, 0x23700129, 0x00ae0400, 0x00000000 },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00200001, 0x202001a6, 0x004500a0, 0x00000000 },
+ { 0x00000001, 0x20280062, 0x00000000, 0x0007000f },
+ { 0x00000005, 0x24000c20, 0x000000b8, 0x00ffffff },
+ { 0x04000010, 0x20000c04, 0x00000400, 0x00ffffff },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x00000056 },
+ { 0x01600031, 0x27000c01, 0x408d0000, 0x0248a007 },
+ { 0x0000040c, 0x21043da1, 0x000000a2, 0x00010001 },
+ { 0x00000801, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x01600031, 0x28000c01, 0x408d0000, 0x0228a008 },
+ { 0x00200001, 0x210001a1, 0x004500a0, 0x00000000 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0007000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00000001, 0x26020228, 0x000000ba, 0x00000000 },
+ { 0x00610001, 0x24400129, 0x028d00b8, 0x00000000 },
+ { 0x00710001, 0x24400169, 0x02000000, 0x00000000 },
+ { 0x00000001, 0x24600061, 0x00000000, 0x00040001 },
+ { 0x00000001, 0x24640061, 0x00000000, 0x00400010 },
+ { 0x00000001, 0x24680061, 0x00000000, 0x04000100 },
+ { 0x00000001, 0x246c0061, 0x00000000, 0x40001000 },
+ { 0x00000001, 0x26020128, 0x00000440, 0x00000000 },
+ { 0x00910001, 0x41400231, 0x02b10700, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000440, 0x008d0460 },
+ { 0x00710001, 0x42400231, 0x02ae0800, 0x00000000 },
+ { 0x00710001, 0x43400231, 0x02ae0801, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000442, 0x00000000 },
+ { 0x00910001, 0x41600231, 0x02b10710, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000444, 0x00000000 },
+ { 0x00910001, 0x41800231, 0x02b10720, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000444, 0x008d0460 },
+ { 0x00710001, 0x42500231, 0x02ae0810, 0x00000000 },
+ { 0x00710001, 0x43500231, 0x02ae0811, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000446, 0x00000000 },
+ { 0x00910001, 0x41a00231, 0x02b10730, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x00000448, 0x00000000 },
+ { 0x00910001, 0x41c00231, 0x02b10740, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x02000448, 0x008d0460 },
+ { 0x00710001, 0x42600231, 0x02ae0820, 0x00000000 },
+ { 0x00710001, 0x43600231, 0x02ae0821, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044a, 0x00000000 },
+ { 0x00910001, 0x41e00231, 0x02b10750, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044c, 0x00000000 },
+ { 0x00910001, 0x42000231, 0x02b10760, 0x00000000 },
+ { 0x02600005, 0x2000252c, 0x0200044c, 0x008d0460 },
+ { 0x00710001, 0x42700231, 0x02ae0830, 0x00000000 },
+ { 0x00710001, 0x43700231, 0x02ae0831, 0x00000000 },
+ { 0x00000001, 0x26020128, 0x0000044e, 0x00000000 },
+ { 0x00910001, 0x42200231, 0x02b10770, 0x00000000 },
+ { 0x00800001, 0x20400232, 0x00d20140, 0x00000000 },
+ { 0x00800001, 0x20500232, 0x00d20160, 0x00000000 },
+ { 0x00800001, 0x20600232, 0x00d20180, 0x00000000 },
+ { 0x00800001, 0x20700232, 0x00d201a0, 0x00000000 },
+ { 0x00800001, 0x20800232, 0x00d201c0, 0x00000000 },
+ { 0x00800001, 0x20900232, 0x00d201e0, 0x00000000 },
+ { 0x00800001, 0x20a00232, 0x00d20200, 0x00000000 },
+ { 0x00800001, 0x20b00232, 0x00d20220, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x0a082007 },
+ { 0x00200001, 0x210001a5, 0x004500a0, 0x00000000 },
+ { 0x0000000c, 0x21043ca5, 0x00000104, 0x00010001 },
+ { 0x00000001, 0x21080061, 0x00000000, 0x0003000f },
+ { 0x00600001, 0x20200022, 0x008d0100, 0x00000000 },
+ { 0x00800001, 0x40400232, 0x00d20240, 0x00000000 },
+ { 0x00800001, 0x40410232, 0x00d20340, 0x00000000 },
+ { 0x00800001, 0x40600232, 0x00d20260, 0x00000000 },
+ { 0x00800001, 0x40610232, 0x00d20360, 0x00000000 },
+ { 0x01600031, 0x20000c04, 0x508d0000, 0x06082008 },
+ { 0x01000040, 0x20863dad, 0x00000086, 0xffffffff },
+ { 0x00000040, 0x20a03dad, 0x000000a0, 0x00100010 },
+ { 0x05000010, 0x2000358c, 0x02210400, 0x00000084 },
+ { 0x00000041, 0x24407fbd, 0x000000bc, 0x41800000 },
+ { 0x00000040, 0x20a477bd, 0x00000440, 0x000000a4 },
+ { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e },
+ { 0x00010220, 0x34001c00, 0x02001400, 0xfffffe66 },
+ { 0x00000001, 0x20a001ad, 0x0000008a, 0x00000000 },
+ { 0x00000040, 0x20a23dad, 0x000000a2, 0x00080008 },
+ { 0x00000001, 0x20a403bd, 0x00000094, 0x00000000 },
+ { 0x00000041, 0x24407fbd, 0x00000038, 0x41000000 },
+ { 0x00000040, 0x20a877bd, 0x00000440, 0x000000a8 },
+ { 0x00000220, 0x34001c00, 0x00001400, 0xfffffe5a },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
+ { 0x00600001, 0x21e00022, 0x008d0000, 0x00000000 },
+ { 0x0f000031, 0x20000c04, 0x708d0000, 0x82000000 },
projects / platform / upstream / libva.git / commitdiff