ARM: added 'neon_composite_over_n_8_8' fast path

diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index 91ec27d..a3875ee 100644 (file)
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -1203,6 +1203,74 @@ generate_composite_function \
 
 /******************************************************************************/
 
+.macro pixman_composite_over_n_8_8_process_pixblock_head
+    vmull.u8    q0,  d24, d8
+    vmull.u8    q1,  d25, d8
+    vmull.u8    q6,  d26, d8
+    vmull.u8    q7,  d27, d8
+    vrshr.u16   q10, q0,  #8
+    vrshr.u16   q11, q1,  #8
+    vrshr.u16   q12, q6,  #8
+    vrshr.u16   q13, q7,  #8
+    vraddhn.u16 d0,  q0,  q10
+    vraddhn.u16 d1,  q1,  q11
+    vraddhn.u16 d2,  q6,  q12
+    vraddhn.u16 d3,  q7,  q13
+    vmvn.8      q12, q0
+    vmvn.8      q13, q1
+    vmull.u8    q8,  d24, d4
+    vmull.u8    q9,  d25, d5
+    vmull.u8    q10, d26, d6
+    vmull.u8    q11, d27, d7
+.endm
+
+.macro pixman_composite_over_n_8_8_process_pixblock_tail
+    vrshr.u16   q14, q8,  #8
+    vrshr.u16   q15, q9,  #8
+    vrshr.u16   q12, q10, #8
+    vrshr.u16   q13, q11, #8
+    vraddhn.u16 d28, q14, q8
+    vraddhn.u16 d29, q15, q9
+    vraddhn.u16 d30, q12, q10
+    vraddhn.u16 d31, q13, q11
+    vqadd.u8    q14, q0,  q14
+    vqadd.u8    q15, q1,  q15
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_n_8_8_process_pixblock_tail_head
+    vld1.8      {d4, d5, d6, d7}, [DST_R, :128]!
+    pixman_composite_over_n_8_8_process_pixblock_tail
+    vld1.8      {d24, d25, d26, d27}, [MASK]!
+    cache_preload 32, 32
+    vst1.8      {d28, d29, d30, d31}, [DST_W, :128]!
+    pixman_composite_over_n_8_8_process_pixblock_head
+.endm
+
+.macro pixman_composite_over_n_8_8_init
+    add         DUMMY, sp, #ARGS_STACK_OFFSET
+    vpush       {d8-d15}
+    vld1.32     {d8[0]}, [DUMMY]
+    vdup.8      d8, d8[3]
+.endm
+
+.macro pixman_composite_over_n_8_8_cleanup
+    vpop        {d8-d15}
+.endm
+
+generate_composite_function \
+    pixman_composite_over_n_8_8_asm_neon, 0, 8, 8, \
+    FLAG_DST_READWRITE, \
+    32, /* number of pixels, processed in a single block */ \
+    5, /* prefetch distance */ \
+    pixman_composite_over_n_8_8_init, \
+    pixman_composite_over_n_8_8_cleanup, \
+    pixman_composite_over_n_8_8_process_pixblock_head, \
+    pixman_composite_over_n_8_8_process_pixblock_tail, \
+    pixman_composite_over_n_8_8_process_pixblock_tail_head
+
+/******************************************************************************/
+
 .macro pixman_composite_over_n_8888_8888_ca_process_pixblock_head
     /*
      * 'combine_mask_ca' replacement

diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index 2f82069..72ef75e 100644 (file)
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -76,6 +76,8 @@ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_8888,
                                       uint8_t, 1, uint32_t, 1)
 PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8888_8888_ca,
                                       uint32_t, 1, uint32_t, 1)
+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)
 
@@ -235,6 +237,7 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
     PIXMAN_STD_FAST_PATH (SRC,  b8g8r8,   null,     x8r8g8b8, neon_composite_src_0888_8888_rev),
     PIXMAN_STD_FAST_PATH (SRC,  b8g8r8,   null,     r5g6b5,   neon_composite_src_0888_0565_rev),
     PIXMAN_STD_FAST_PATH (SRC,  pixbuf,   pixbuf,   a8r8g8b8, neon_composite_src_pixbuf_8888),
+    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),
     PIXMAN_STD_FAST_PATH (OVER, solid,    a8,       b5g6r5,   neon_composite_over_n_8_0565),
     PIXMAN_STD_FAST_PATH (OVER, solid,    a8,       a8r8g8b8, neon_composite_over_n_8_8888),