ARM: added 'neon_composite_add_n_8_8888' fast path
authorSiarhei Siamashka <siarhei.siamashka@nokia.com>
Mon, 29 Nov 2010 00:38:52 +0000 (02:38 +0200)
committerSiarhei Siamashka <siarhei.siamashka@nokia.com>
Fri, 3 Dec 2010 13:37:49 +0000 (15:37 +0200)
pixman/pixman-arm-neon-asm.S
pixman/pixman-arm-neon.c

index 829ef84..dd6f2c5 100644 (file)
@@ -1628,6 +1628,35 @@ generate_composite_function \
 
 /******************************************************************************/
 
+.macro pixman_composite_add_n_8_8888_init
+    add         DUMMY, sp, #ARGS_STACK_OFFSET
+    vld1.32     {d3[0]}, [DUMMY]
+    vdup.8      d0, d3[0]
+    vdup.8      d1, d3[1]
+    vdup.8      d2, d3[2]
+    vdup.8      d3, d3[3]
+.endm
+
+.macro pixman_composite_add_n_8_8888_cleanup
+.endm
+
+generate_composite_function \
+    pixman_composite_add_n_8_8888_asm_neon, 0, 8, 32, \
+    FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+    8, /* number of pixels, processed in a single block */ \
+    5, /* prefetch distance */ \
+    pixman_composite_add_n_8_8888_init, \
+    pixman_composite_add_n_8_8888_cleanup, \
+    pixman_composite_add_8888_8888_8888_process_pixblock_head, \
+    pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
+    pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \
+    28, /* dst_w_basereg */ \
+    4,  /* dst_r_basereg */ \
+    0,  /* src_basereg   */ \
+    27  /* mask_basereg  */
+
+/******************************************************************************/
+
 .macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
     /* expecting source data in {d0, d1, d2, d3} */
     /* destination data in {d4, d5, d6, d7} */
index eaf9787..5ad58bd 100644 (file)
@@ -80,6 +80,8 @@ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_8,
                                       uint8_t, 1, uint8_t, 1)
 PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8,
                                       uint8_t, 1, uint8_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8888,
+                                      uint8_t, 1, uint32_t, 1)
 
 PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_8888,
                                      uint32_t, 1, uint32_t, 1)
@@ -281,6 +283,8 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
     PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, null,     a8r8g8b8, neon_composite_src_x888_8888),
     PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, null,     a8b8g8r8, neon_composite_src_x888_8888),
     PIXMAN_STD_FAST_PATH (ADD,  solid,    a8,       a8,       neon_composite_add_n_8_8),
+    PIXMAN_STD_FAST_PATH (ADD,  solid,    a8,       a8r8g8b8, neon_composite_add_n_8_8888),
+    PIXMAN_STD_FAST_PATH (ADD,  solid,    a8,       a8b8g8r8, neon_composite_add_n_8_8888),
     PIXMAN_STD_FAST_PATH (ADD,  a8,       a8,       a8,       neon_composite_add_8_8_8),
     PIXMAN_STD_FAST_PATH (ADD,  r5g6b5,   a8,       r5g6b5,   neon_composite_add_0565_8_0565),
     PIXMAN_STD_FAST_PATH (ADD,  b5g6r5,   a8,       b5g6r5,   neon_composite_add_0565_8_0565),