ARM: added 'neon_composite_src_0565_8888' fast path

diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index d35704d..bffc676 100644 (file)
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -387,6 +387,41 @@ generate_composite_function \
 
 /******************************************************************************/
 
+.macro pixman_composite_src_0565_8888_process_pixblock_head
+    vshrn.u16   d30, q0, #8
+    vshrn.u16   d29, q0, #3
+    vsli.u16    q0, q0, #5
+    vmov.u8     d31, #255
+    vsri.u8     d30, d30, #5
+    vsri.u8     d29, d29, #6
+    vshrn.u16   d28, q0, #2
+.endm
+
+.macro pixman_composite_src_0565_8888_process_pixblock_tail
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_src_0565_8888_process_pixblock_tail_head
+    pixman_composite_src_0565_8888_process_pixblock_tail
+    vst4.8     {d28, d29, d30, d31}, [DST_W, :128]!
+    vld1.16    {d0, d1}, [SRC]!
+    pixman_composite_src_0565_8888_process_pixblock_head
+    cache_preload 8, 8
+.endm
+
+generate_composite_function \
+    pixman_composite_src_0565_8888_asm_neon, 16, 0, 32, \
+    FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+    8, /* number of pixels, processed in a single block */ \
+    10, /* prefetch distance */ \
+    default_init, \
+    default_cleanup, \
+    pixman_composite_src_0565_8888_process_pixblock_head, \
+    pixman_composite_src_0565_8888_process_pixblock_tail, \
+    pixman_composite_src_0565_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
 .macro pixman_composite_add_8000_8000_process_pixblock_head
     vqadd.u8    q14, q0, q2
     vqadd.u8    q15, q1, q3

diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index c7f0870..4894285 100644 (file)
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -214,6 +214,7 @@ BIND_SRC_NULL_DST(src_8888_8888, uint32_t, 1, uint32_t, 1)
 BIND_SRC_NULL_DST(src_0565_0565, uint16_t, 1, uint16_t, 1)
 BIND_SRC_NULL_DST(src_0888_0888, uint8_t, 3, uint8_t, 3)
 BIND_SRC_NULL_DST(src_8888_0565, uint32_t, 1, uint16_t, 1)
+BIND_SRC_NULL_DST(src_0565_8888, uint16_t, 1, uint32_t, 1)
 BIND_SRC_NULL_DST(add_8000_8000, uint8_t, 1, uint8_t, 1)
 BIND_SRC_NULL_DST(add_8888_8888, uint32_t, 1, uint32_t, 1)
 
@@ -344,6 +345,10 @@ static const pixman_fast_path_t arm_neon_fast_path_array[] =
     { PIXMAN_OP_SRC,  PIXMAN_x8r8g8b8, PIXMAN_null,     PIXMAN_r5g6b5,   neon_composite_src_8888_0565    },
     { PIXMAN_OP_SRC,  PIXMAN_a8b8g8r8, PIXMAN_null,     PIXMAN_b5g6r5,   neon_composite_src_8888_0565    },
     { PIXMAN_OP_SRC,  PIXMAN_x8b8g8r8, PIXMAN_null,     PIXMAN_b5g6r5,   neon_composite_src_8888_0565    },
+    { PIXMAN_OP_SRC,  PIXMAN_r5g6b5,   PIXMAN_null,     PIXMAN_a8r8g8b8, neon_composite_src_0565_8888    },
+    { PIXMAN_OP_SRC,  PIXMAN_r5g6b5,   PIXMAN_null,     PIXMAN_x8r8g8b8, neon_composite_src_0565_8888    },
+    { PIXMAN_OP_SRC,  PIXMAN_b5g6r5,   PIXMAN_null,     PIXMAN_a8b8g8r8, neon_composite_src_0565_8888    },
+    { PIXMAN_OP_SRC,  PIXMAN_b5g6r5,   PIXMAN_null,     PIXMAN_x8b8g8r8, neon_composite_src_0565_8888    },
     { PIXMAN_OP_SRC,  PIXMAN_a8r8g8b8, PIXMAN_null,     PIXMAN_x8r8g8b8, neon_composite_src_8888_8888    },
     { PIXMAN_OP_SRC,  PIXMAN_x8r8g8b8, PIXMAN_null,     PIXMAN_x8r8g8b8, neon_composite_src_8888_8888    },
     { PIXMAN_OP_SRC,  PIXMAN_a8b8g8r8, PIXMAN_null,     PIXMAN_x8b8g8r8, neon_composite_src_8888_8888    },