ARM: NEON: Standard fast path src_n_8_8

Performance numbers of before/after on cortex-a8 @ 1GHz

- before
L1:  28.05  L2:  28.26  M: 26.97 (  4.48%)  HT: 19.79  VT: 19.14  R: 17.61  RT:  9.88 ( 101Kops/s)

- after
L1:1430.28  L2:1252.10  M:421.93 ( 75.48%)  HT:170.16  VT:138.03  R:145.86  RT: 35.51 ( 255Kops/s)

2 files changed, 69 insertions, 0 deletions

diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index 1db02db..da8f054 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -1292,6 +1292,72 @@ generate_composite_function \
 
 /******************************************************************************/
 
+.macro pixman_composite_src_n_8_8_process_pixblock_head
+    vmull.u8    q0, d24, d16
+    vmull.u8    q1, d25, d16
+    vmull.u8    q2, d26, d16
+    vmull.u8    q3, d27, d16
+    vrsra.u16   q0, q0,  #8
+    vrsra.u16   q1, q1,  #8
+    vrsra.u16   q2, q2,  #8
+    vrsra.u16   q3, q3,  #8
+.endm
+
+.macro pixman_composite_src_n_8_8_process_pixblock_tail
+    vrshrn.u16  d28, q0, #8
+    vrshrn.u16  d29, q1, #8
+    vrshrn.u16  d30, q2, #8
+    vrshrn.u16  d31, q3, #8
+.endm
+
+.macro pixman_composite_src_n_8_8_process_pixblock_tail_head
+    fetch_mask_pixblock
+                                    PF add PF_X, PF_X, #8
+        vrshrn.u16  d28, q0, #8
+                                    PF tst PF_CTL, #0x0F
+        vrshrn.u16  d29, q1, #8
+                                    PF addne PF_X, PF_X, #8
+        vrshrn.u16  d30, q2, #8
+                                    PF subne PF_CTL, PF_CTL, #1
+        vrshrn.u16  d31, q3, #8
+                                    PF cmp PF_X, ORIG_W
+    vmull.u8    q0,  d24, d16
+                                    PF pld, [PF_MASK, PF_X, lsl #mask_bpp_shift]
+    vmull.u8    q1,  d25, d16
+                                    PF subge PF_X, PF_X, ORIG_W
+    vmull.u8    q2,  d26, d16
+                                    PF subges PF_CTL, PF_CTL, #0x10
+    vmull.u8    q3,  d27, d16
+                                    PF ldrgeb DUMMY, [PF_MASK, MASK_STRIDE, lsl #mask_bpp_shift]!
+        vst1.8      {d28, d29, d30, d31}, [DST_W, :128]!
+    vrsra.u16   q0, q0,  #8
+    vrsra.u16   q1, q1,  #8
+    vrsra.u16   q2, q2,  #8
+    vrsra.u16   q3, q3,  #8
+.endm
+
+.macro pixman_composite_src_n_8_8_init
+    add         DUMMY, sp, #ARGS_STACK_OFFSET
+    vld1.32     {d16[0]}, [DUMMY]
+    vdup.8      d16, d16[3]
+.endm
+
+.macro pixman_composite_src_n_8_8_cleanup
+.endm
+
+generate_composite_function \
+    pixman_composite_src_n_8_8_asm_neon, 0, 8, 8, \
+    FLAG_DST_WRITEONLY, \
+    32, /* number of pixels, processed in a single block */ \
+    5, /* prefetch distance */ \
+    pixman_composite_src_n_8_8_init, \
+    pixman_composite_src_n_8_8_cleanup, \
+    pixman_composite_src_n_8_8_process_pixblock_head, \
+    pixman_composite_src_n_8_8_process_pixblock_tail, \
+    pixman_composite_src_n_8_8_process_pixblock_tail_head
+
+/******************************************************************************/
+
 .macro pixman_composite_over_n_8_8888_process_pixblock_head
     /* expecting deinterleaved source data in {d8, d9, d10, d11} */
     /* d8 - blue, d9 - green, d10 - red, d11 - alpha */
diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index 3db9adf..ca139de 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -92,6 +92,8 @@ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, add_n_8_8888,
                                       uint8_t, 1, uint32_t, 1)
 PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (0, neon, src_n_8_8888,
                                       uint8_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (0, neon, src_n_8_8,
+                                      uint8_t, 1, uint8_t, 1)
 
 PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_8888_n_8888,
                                      uint32_t, 1, uint32_t, 1)
@@ -295,6 +297,7 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
     PIXMAN_STD_FAST_PATH (SRC,  solid,    a8,       x8r8g8b8, neon_composite_src_n_8_8888),
     PIXMAN_STD_FAST_PATH (SRC,  solid,    a8,       a8b8g8r8, neon_composite_src_n_8_8888),
     PIXMAN_STD_FAST_PATH (SRC,  solid,    a8,       x8b8g8r8, neon_composite_src_n_8_8888),
+    PIXMAN_STD_FAST_PATH (SRC,  solid,    a8,       a8,       neon_composite_src_n_8_8),
 
     PIXMAN_STD_FAST_PATH (OVER, solid,    a8,       a8,       neon_composite_over_n_8_8),
     PIXMAN_STD_FAST_PATH (OVER, solid,    a8,       r5g6b5,   neon_composite_over_n_8_0565),