add (1) tmp_reg0.0<1>:d mbb_result.0<0,1,0>:d mbc_result.0<0,1,0>:d {align1};
cmp.z.f0.0 (1) null:d tmp_reg0.0<0,1,0>:d 0:d {align1};
(-f0.0) jmpi (1) mb_median_start;
-cmp.nz.f0.0 (1) null:d mba_result.0<0,1,0>:d 1:d {align1};
+cmp.nz.f0.0 (1) null:d mba_result.0<0,1,0>:d 0:d {align1};
(f0.0) mov (1) mbb_result.4<1>:ud mba_result.4<0,1,0>:ud {align1};
(f0.0) mov (1) mbc_result.4<1>:ud mba_result.4<0,1,0>:ud {align1};
(f0.0) mov (1) mbb_result.20<1>:uw mba_result.20<0,1,0>:uw {align1};
__mb_hwdep_end:
+mov (2) mv_cc_ref.0<1>:w mba_result.4<2,2,1>:w {align1};
+
/* Calibrate the ref window for MPEG2 */
mov (1) vme_m0.0<1>:W -16:W {align1};
mov (1) vme_m0.2<1>:W -12:W {align1};
/* Setup the Cost center */
/* currently four 8x8 share the same cost center */
-mov (4) vme_m3.0<2>:ud 0x0:ud {align1};
-mov (4) vme_m3.4<2>:ud 0x0:ud {align1};
+mov (4) vme_m3.0<2>:ud mv_cc_ref.0<0,1,0>:ud {align1};
+mov (4) vme_m3.4<2>:ud mv_cc_ref.0<0,1,0>:ud {align1};
mov (8) vme_msg_3<1>:UD vme_m3.0<8,8,1>:UD {align1};
mov (8) vme_msg_2<1>:UD vme_m2.0<8,8,1>:UD {align1};
rlen vme_wb_length
{align1};
+and.z.f0.0 (1) null:uw mb_hwdep<0,1,0>:uw 0x04:uw {align1};
+(-f0.0) jmpi (1) vme_run_again;
+nop;
+vme_mv_output:
+
add (1) obw_m0.8<1>:UD obw_m0.8<0,1,0>:UD 0x02:UD {align1};
mov (8) msg_reg0.0<1>:UD obw_m0<8,8,1>:UD {align1};
/* write FME info */
nop;
nop;
+vme_run_again:
+
+asr (2) mb_ref_win.0<1>:w mb_mvp_ref.0<2,2,1>:w 2:w {align1};
+mov (2) tmp_reg0.0<1>:w mb_ref_win.0<2,2,1>:w {align1};
+add (2) mb_ref_win.8<1>:w mb_ref_win.0<2,2,1>:w 3:w {align1};
+and (2) mb_ref_win.16<1>:uw mb_ref_win.8<2,2,1>:uw 0xFFFC:uw {align1};
+
+cmp.l.f0.0 (1) null:w tmp_reg0.0<0,1,0>:w 0:w {align1};
+(f0.0) mul (1) tmp_reg0.0<1>:w tmp_reg0.0<0,1,0>:w -1:w {align1};
+cmp.l.f0.0 (1) null:w tmp_reg0.2<0,1,0>:w 0:w {align1};
+(f0.0) mul (1) tmp_reg0.2<1>:w tmp_reg0.2<0,1,0>:w -1:w {align1};
+
+cmp.ge.f0.0 (1) null:w tmp_reg0.0<0,1,0>:w 4:w {align1};
+(f0.0) jmpi (1) vme_start;
+cmp.ge.f0.0 (1) null:w tmp_reg0.2<0,1,0>:w 4:w {align1};
+(f0.0) jmpi (1) vme_start;
+
+jmpi (1) vme_done;
+
+vme_start:
+ mov (8) tmp_vme_wb0.0<1>:ud vme_wb0.0<8,8,1>:ud {align1};
+ mov (8) tmp_vme_wb1.0<1>:ud vme_wb1.0<8,8,1>:ud {align1};
+
+/* Calibrate the ref window for MPEG2 */
+mov (1) vme_m0.0<1>:W -16:W {align1};
+mov (1) vme_m0.2<1>:W -12:W {align1};
+mov (1) INPUT_ARG0.8<1>:ud vme_m0.8<0,1,0>:ud {align1};
+add (2) INPUT_ARG0.0<1>:w vme_m0.0<2,2,1>:w mb_ref_win.16<2,2,1>:w {align1};
+mov (8) INPUT_ARG1.0<1>:ud pic_ref.0<8,8,1>:ud {align1};
+
+SAVE_RET {align1};
+jmpi (1) ref_boundary_check;
+mov (2) vme_m0.0<1>:w RET_ARG<2,2,1>:w {align1};
+
+/* IME search */
+mov (1) vme_m0.12<1>:UD SEARCH_CTRL_SINGLE + INTER_PART_MASK + INTER_SAD_HAAR:UD {align1}; /* 16x16 Source, harr */
+mov (1) vme_m0.22<1>:UW REF_REGION_SIZE {align1}; /* Reference Width&Height, 48x40 */
+
+mov (1) vme_m0.4<1>:UD vme_m0.0<0,1,0>:UD {align1};
+
+mov (8) vme_msg_0.0<1>:UD vme_m0.0<8,8,1>:UD {align1};
+
+mov (8) vme_m1.0<1>:ud 0x0:UD {align1};
+
+mov (1) vme_m1.0<1>:UD ADAPTIVE_SEARCH_ENABLE:ud {align1} ;
+/* the Max MV number is passed by constant buffer */
+mov (1) vme_m1.4<1>:UB r4.28<0,1,0>:UB {align1};
+mov (1) vme_m1.8<1>:UD START_CENTER + SEARCH_PATH_LEN:UD {align1};
+mov (8) vme_msg_1.0<1>:UD vme_m1.0<8,8,1>:UD {align1};
+
+/* Setup the Cost center */
+/* currently four 8x8 share the same cost center */
+mov (4) vme_m3.0<2>:ud mv_cc_ref.0<0,1,0>:ud {align1};
+mov (4) vme_m3.4<2>:ud mv_cc_ref.0<0,1,0>:ud {align1};
+
+mov (8) vme_msg_3<1>:UD vme_m3.0<8,8,1>:UD {align1};
+mov (8) vme_msg_2<1>:UD vme_m2.0<8,8,1>:UD {align1};
+
+/* M4/M5 search path */
+mov (1) vme_msg_4.0<1>:UD 0x01010101:UD {align1};
+mov (1) vme_msg_4.4<1>:UD 0x10010101:UD {align1};
+mov (1) vme_msg_4.8<1>:UD 0x0F0F0F0F:UD {align1};
+mov (1) vme_msg_4.12<1>:UD 0x100F0F0F:UD {align1};
+mov (1) vme_msg_4.16<1>:UD 0x01010101:UD {align1};
+mov (1) vme_msg_4.20<1>:UD 0x10010101:UD {align1};
+mov (1) vme_msg_4.24<1>:UD 0x0F0F0F0F:UD {align1};
+mov (1) vme_msg_4.28<1>:UD 0x100F0F0F:UD {align1};
+
+mov (1) vme_msg_5.0<1>:UD 0x01010101:UD {align1};
+mov (1) vme_msg_5.4<1>:UD 0x10010101:UD {align1};
+mov (1) vme_msg_5.8<1>:UD 0x0F0F0F0F:UD {align1};
+mov (1) vme_msg_5.12<1>:UD 0x000F0F0F:UD {align1};
+
+mov (4) vme_msg_5.16<1>:UD 0x0:UD {align1};
+
+send (8)
+ vme_msg_ind
+ vme_wb<1>:UD
+ null
+ vme(
+ BIND_IDX_VME,
+ 0,
+ 0,
+ VME_IME_MESSAGE_TYPE
+ )
+ mlen ime_vme_msg_length
+ rlen vme_wb_length {align1};
+
+/* Set Macroblock-shape/mode for FBR */
+
+mov (1) vme_m2.20<1>:UD 0x0:UD {align1};
+mov (1) vme_m2.21<1>:UB vme_wb.25<0,1,0>:UB {align1};
+mov (1) vme_m2.22<1>:UB vme_wb.26<0,1,0>:UB {align1};
+
+and (1) tmp_reg0.0<1>:UW vme_wb.0<0,1,0>:UW 0x03:UW {align1};
+mov (1) vme_m2.20<1>:UB tmp_reg0.0<0,1,0>:UB {align1};
+
+/* Send FBR message into CRE */
+
+mov (8) vme_msg_4.0<1>:UD vme_wb1.0<8,8,1>:UD {align1};
+mov (8) vme_msg_5.0<1>:ud vme_wb2.0<8,8,1>:ud {align1};
+mov (8) vme_msg_6.0<1>:ud vme_wb3.0<8,8,1>:ud {align1};
+mov (8) vme_msg_7.0<1>:ud vme_wb4.0<8,8,1>:ud {align1};
+
+mov (1) vme_m0.12<1>:UD INTER_SAD_HAAR + SUB_PEL_MODE_HALF + FBR_BME_DISABLE:UD {align1}; /* 16x16 Source, 1/2 pixel, harr, BME disable */
+/* Bilinear filter */
+mov (1) tmp_reg0.0<1>:uw 0x04:uw {align1};
+add (1) vme_m1.30<1>:ub vme_m1.30<0,1,0>:ub tmp_reg0.0<0,1,0>:ub {align1};
+
+mov (8) vme_msg_0.0<1>:UD vme_m0.0<8,8,1>:UD {align1};
+mov (8) vme_msg_1.0<1>:UD vme_m1.0<8,8,1>:UD {align1};
+
+mov (8) vme_msg_2.0<1>:UD vme_m2.0<8,8,1>:UD {align1};
+mov (8) vme_msg_3.0<1>:UD vme_m3.0<8,8,1>:UD {align1};
+
+/* after verification it will be passed by using payload */
+send (8)
+ vme_msg_ind
+ vme_wb<1>:UD
+ null
+ cre(
+ BIND_IDX_VME,
+ VME_FBR_MESSAGE_TYPE
+ )
+ mlen fbr_vme_msg_length
+ rlen vme_wb_length
+ {align1};
+
+cmp.l.f0.0 (1) null:uw vme_wb0.8<0,1,0>:uw tmp_vme_wb0.8<0,1,0>:uw {align1};
+(f0.0) jmpi (1) vme_done;
+mov (8) vme_wb0.0<1>:ud tmp_vme_wb0.0<8,8,1>:ud {align1};
+mov (8) vme_wb1.0<1>:ud tmp_vme_wb1.0<8,8,1>:ud {align1};
+
+vme_done:
+ jmpi (1) vme_mv_output;
+nop;
+nop;
+nop;
+
projects / platform / upstream / libva-intel-driver.git / blobdiff