[Pixman] [PATCH] ARM: added 'neon_composite_out_reverse_8_0565' fast path

[Siarhei Siamashka]

From: Siarhei Siamashka <siarhei.siamashka at nokia.com>

== before ==

    outrev_8_0565 =  L1:  22.91  L2:  22.40  M: 18.75 ( 10.47%)
                     HT: 12.62   VT: 12.22   R: 11.32  RT:  5.30 (  58Kops/s)

== after ==

    outrev_8_0565 =  L1: 176.27  L2: 151.70  M:108.79 ( 60.81%)
                     HT: 50.43   VT: 37.16   R: 32.26  RT:  9.62 (  97Kops/s)
---
 pixman/pixman-arm-neon-asm.S |   49 ++++++++++++++++++++++++++++++++++++++++++
 pixman/pixman-arm-neon.c     |    4 +++
 2 files changed, 53 insertions(+), 0 deletions(-)

diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index cbcc720..09f0268 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -1924,3 +1924,52 @@ generate_composite_function \
     10, /* dst_r_basereg */ \
     8,  /* src_basereg   */ \
     15  /* mask_basereg  */
+
+/******************************************************************************/
+
+.macro pixman_composite_out_reverse_8_0565_process_pixblock_head
+    /* mask is in d15 */
+    convert_0565_to_x888 q5, d6, d5, d4
+    /* destination pixel data is in {d4, d5, d6, xx} */
+    vmvn.8      d24, d15 /* get inverted alpha */
+    /* now do alpha blending */
+    vmull.u8    q8, d24, d4
+    vmull.u8    q9, d24, d5
+    vmull.u8    q10, d24, d6
+.endm
+
+.macro pixman_composite_out_reverse_8_0565_process_pixblock_tail
+    vrshr.u16   q14, q8, #8
+    vrshr.u16   q15, q9, #8
+    vrshr.u16   q12, q10, #8
+    vraddhn.u16 d0, q14, q8
+    vraddhn.u16 d1, q15, q9
+    vraddhn.u16 d2, q12, q10
+    /* 32bpp result is in {d0, d1, d2, XX} */
+    convert_8888_to_0565 d2, d1, d0, q14, q15, q3
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_out_reverse_8_0565_process_pixblock_tail_head
+    vld1.8     {d15}, [SRC]!
+    pixman_composite_out_reverse_8_0565_process_pixblock_tail
+    vld1.16    {d10, d11}, [DST_R, :128]!
+    cache_preload 8, 8
+    pixman_composite_out_reverse_8_0565_process_pixblock_head
+    vst1.16    {d28, d29}, [DST_W, :128]!
+.endm
+
+generate_composite_function \
+    pixman_composite_out_reverse_8_0565_asm_neon, 8, 0, 16, \
+    FLAG_DST_READWRITE, \
+    8, /* number of pixels, processed in a single block */ \
+    5, /* prefetch distance */ \
+    default_init_need_all_regs, \
+    default_cleanup_need_all_regs, \
+    pixman_composite_out_reverse_8_0565_process_pixblock_head, \
+    pixman_composite_out_reverse_8_0565_process_pixblock_tail, \
+    pixman_composite_out_reverse_8_0565_process_pixblock_tail_head, \
+    28, /* dst_w_basereg */ \
+    10, /* dst_r_basereg */ \
+    15, /* src_basereg   */ \
+    0   /* mask_basereg  */
diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index 705631e..fa3b1ae 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -60,6 +60,8 @@ PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, over_8888_0565,
                                    uint32_t, 1, uint16_t, 1)
 PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, over_8888_8888,
                                    uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, out_reverse_8_0565,
+                                   uint8_t, 1, uint16_t, 1)
 
 PIXMAN_ARM_BIND_FAST_PATH_N_DST (neon, over_n_0565,
                                  uint16_t, 1)
@@ -266,6 +268,8 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
     PIXMAN_STD_FAST_PATH (ADD,  a8b8g8r8, null,     a8b8g8r8, neon_composite_add_8888_8888),
     PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8r8g8b8, neon_composite_over_reverse_n_8888),
     PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8b8g8r8, neon_composite_over_reverse_n_8888),
+    PIXMAN_STD_FAST_PATH (OUT_REVERSE,  a8,    null, r5g6b5,   neon_composite_out_reverse_8_0565),
+    PIXMAN_STD_FAST_PATH (OUT_REVERSE,  a8,    null, b5g6r5,   neon_composite_out_reverse_8_0565),
 
     { PIXMAN_OP_NONE },
 };
-- 
1.7.2.2