ARM: NEON optimization for bilinear scaled 'src_8888_0565'

Benchmark on ARM Cortex-A8 r2p2 @1GHz, 32-bit LPDDR @200MHz:
 Microbenchmark (scaling 2000x2000 image with scale factor close to 1x):
  before: op=1, src=20028888, dst=10020565, speed=6.56 MPix/s
  after:  op=1, src=20028888, dst=10020565, speed=61.65 MPix/s

diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index 52dc444..f0b42ca 100644 (file)
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -2625,3 +2625,6 @@ pixman_asm_function fname
 
 generate_bilinear_scanline_func \
     pixman_scaled_bilinear_scanline_8888_8888_SRC_asm_neon, 8888, 8888, 2, 28
+
+generate_bilinear_scanline_func \
+    pixman_scaled_bilinear_scanline_8888_0565_SRC_asm_neon, 8888, 0565, 2, 28

diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index 98ad5f2..ba6de66 100644 (file)
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -129,6 +129,8 @@ PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_A8_DST (SKIP_ZERO_SRC, neon, 0565_8_0565,
 
 PIXMAN_ARM_BIND_SCALED_BILINEAR_SRC_DST (0, neon, 8888_8888, SRC,
                                          uint32_t, uint32_t)
+PIXMAN_ARM_BIND_SCALED_BILINEAR_SRC_DST (0, neon, 8888_0565, SRC,
+                                         uint32_t, uint16_t)
 
 void
 pixman_composite_src_n_8_asm_neon (int32_t   w,
@@ -350,6 +352,9 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
     SIMPLE_BILINEAR_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8, neon_8888_8888),
     SIMPLE_BILINEAR_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8, neon_8888_8888),
 
+    SIMPLE_BILINEAR_FAST_PATH (SRC, a8r8g8b8, r5g6b5, neon_8888_0565),
+    SIMPLE_BILINEAR_FAST_PATH (SRC, x8r8g8b8, r5g6b5, neon_8888_0565),
+
     { PIXMAN_OP_NONE },
 };