[Pixman] [PATCH 2/5 v2] armv7: Add in_8888_8 fast path

[Ben Avison]

This is tuned for the Cortex-A7 (Raspberry Pi 2).
lowlevel-blt-bench results, compared to the ARMv6 fast path:

    Before          After
    Mean   StdDev   Mean   StdDev  Confidence  Change
L1  146.0  0.7      231.4  1.2     100.0%      +58.5%
L2  143.1  0.9      222.1  1.7     100.0%      +55.3%
M   110.9  0.0      129.0  0.5     100.0%      +16.3%
HT  57.3   0.6      73.0   0.3     100.0%      +27.4%
VT  46.6   0.5      61.6   0.4     100.0%      +32.3%
R   42.3   0.2      51.7   0.2     100.0%      +22.2%
RT  19.1   0.1      21.0   0.1     100.0%      +9.9%
---
 pixman/pixman-arm-neon-asm.S |   35 +++++++++++++++++++++++++++++++++++
 pixman/pixman-arm-neon.c     |    5 +++++
 2 files changed, 40 insertions(+), 0 deletions(-)

diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index 7e949a3..2fecb5b 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -2757,6 +2757,41 @@ generate_composite_function \
 
 /******************************************************************************/
 
+.macro pixman_composite_in_8888_8_process_pixblock_head
+    /* src is in d0-d3 (deinterleaved) */
+    /* destination pixel data is in d4 */
+    vmull.u8    q8, d3, d4
+.endm
+
+.macro pixman_composite_in_8888_8_process_pixblock_tail
+    vrshr.u16   q9, q8, #8
+    vraddhn.u16 d28, q8, q9
+    /* result is in d28 */
+.endm
+
+.macro pixman_composite_in_8888_8_process_pixblock_tail_head
+        vld4.8      {d0-d3}, [SRC]!
+    vrshr.u16   q9, q8, #8
+        vld1.8      {d4}, [DST_R :64]!
+            cache_preload 8, 8
+    vraddhn.u16 d28, q8, q9
+        vmull.u8    q8, d3, d4
+    vst1.8      {d28}, [DST_W :64]!
+.endm
+
+generate_composite_function \
+    pixman_composite_in_8888_8_asm_neon, 32, 0, 8, \
+    FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+    8, /* number of pixels, processed in a single block */ \
+    4, /* prefetch distance */ \
+    default_init, \
+    default_cleanup, \
+    pixman_composite_in_8888_8_process_pixblock_head, \
+    pixman_composite_in_8888_8_process_pixblock_tail, \
+    pixman_composite_in_8888_8_process_pixblock_tail_head
+
+/******************************************************************************/
+
 generate_composite_function_nearest_scanline \
     pixman_scaled_nearest_scanline_8888_8888_OVER_asm_neon, 32, 0, 32, \
     FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index 2f669cb..963d27c 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -66,6 +66,8 @@ PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, out_reverse_8_0565,
                                    uint8_t, 1, uint16_t, 1)
 PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, out_reverse_8_8888,
                                    uint8_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, in_8888_8,
+                                   uint32_t, 1, uint8_t, 1)
 
 PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_0565,
                                  uint16_t, 1)
@@ -372,6 +374,9 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
     PIXMAN_STD_FAST_PATH (ADD,  a8r8g8b8, null,     a8r8g8b8, neon_composite_add_8888_8888),
     PIXMAN_STD_FAST_PATH (ADD,  a8b8g8r8, null,     a8b8g8r8, neon_composite_add_8888_8888),
     PIXMAN_STD_FAST_PATH (IN,   solid,    null,     a8,       neon_composite_in_n_8),
+    PIXMAN_STD_FAST_PATH (IN,   a8r8g8b8, null,     a8,       neon_composite_in_8888_8),
+    PIXMAN_STD_FAST_PATH (IN,   a8b8g8r8, null,     a8,       neon_composite_in_8888_8),
+    PIXMAN_STD_FAST_PATH (IN,   a8r8g8b8_sRGB, null, a8,       neon_composite_in_8888_8),
     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),
-- 
1.7.5.4