From d1410558827fce8aac354274a7150fa915881c50 Mon Sep 17 00:00:00 2001 From: Siarhei Siamashka Date: Thu, 10 Dec 2009 00:51:50 +0200 Subject: [PATCH 5/5] ARM: added NEON optimizations for fetch/store r5g6b5 scanline --- pixman/pixman-access.c | 23 ++++++++++++++++++++++- pixman/pixman-arm-neon-asm.S | 20 ++++++++++++++++++++ pixman/pixman-arm-neon.c | 41 +++++++++++++++++++++++++++++++++++++++++ pixman/pixman-private.h | 5 +++++ 4 files changed, 88 insertions(+), 1 deletions(-) diff --git a/pixman/pixman-access.c b/pixman/pixman-access.c index fa0a267..5bb3e09 100644 --- a/pixman/pixman-access.c +++ b/pixman/pixman-access.c @@ -2748,7 +2748,7 @@ typedef struct store_scanline_ ## format, store_scanline_generic_64 \ } -static const format_info_t accessors[] = +static format_info_t accessors[] = { /* 32 bpp formats */ FORMAT_INFO (a8r8g8b8), @@ -2891,6 +2891,27 @@ _pixman_bits_image_setup_raw_accessors (bits_image_t *image) setup_accessors (image); } +void +_pixman_bits_override_accessors (pixman_format_code_t format, + fetch_scanline_t fetch_func, + store_scanline_t store_func) +{ + format_info_t *info = accessors; + + while (info->format != PIXMAN_null) + { + if (info->format == format) + { + if (fetch_func) + info->fetch_scanline_raw_32 = fetch_func; + if (store_func) + info->store_scanline_raw_32 = store_func; + return; + } + info++; + } +} + #else void diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S index 51bc347..f30869e 100644 --- a/pixman/pixman-arm-neon-asm.S +++ b/pixman/pixman-arm-neon-asm.S @@ -458,6 +458,16 @@ generate_composite_function \ pixman_composite_src_8888_0565_process_pixblock_tail, \ pixman_composite_src_8888_0565_process_pixblock_tail_head +generate_composite_function_single_scanline \ + pixman_store_scanline_r5g6b5_asm_neon, 32, 0, 16, \ + FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ + 8, /* number of pixels, processed in a single block */ \ + default_init, \ + default_cleanup, \ + pixman_composite_src_8888_0565_process_pixblock_head, \ + pixman_composite_src_8888_0565_process_pixblock_tail, \ + pixman_composite_src_8888_0565_process_pixblock_tail_head + /******************************************************************************/ .macro pixman_composite_src_0565_8888_process_pixblock_head @@ -493,6 +503,16 @@ generate_composite_function \ pixman_composite_src_0565_8888_process_pixblock_tail, \ pixman_composite_src_0565_8888_process_pixblock_tail_head +generate_composite_function_single_scanline \ + pixman_fetch_scanline_r5g6b5_asm_neon, 16, 0, 32, \ + FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ + 8, /* number of pixels, processed in a single block */ \ + 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 diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c index 7feee1d..fda7a09 100644 --- a/pixman/pixman-arm-neon.c +++ b/pixman/pixman-arm-neon.c @@ -375,6 +375,43 @@ neon_combine_##name##_u (pixman_implementation_t *imp, \ BIND_COMBINE_U (over) BIND_COMBINE_U (add) +void +pixman_fetch_scanline_r5g6b5_asm_neon (int width, + uint32_t *buffer, + const uint16_t *pixel); +void +pixman_store_scanline_r5g6b5_asm_neon (int width, + uint16_t *pixel, + const uint32_t *values); + +static void +neon_fetch_scanline_r5g6b5 (pixman_image_t *image, + int x, + int y, + int width, + uint32_t * buffer, + const uint32_t *mask, + uint32_t mask_bits) +{ + const uint32_t *bits = image->bits.bits + y * image->bits.rowstride; + const uint16_t *pixel = (const uint16_t *)bits + x; + + pixman_fetch_scanline_r5g6b5_asm_neon (width, buffer, pixel); +} + +static void +neon_store_scanline_r5g6b5 (bits_image_t * image, + int x, + int y, + int width, + const uint32_t *values) +{ + uint32_t *bits = image->bits + image->rowstride * y; + uint16_t *pixel = ((uint16_t *) bits) + x; + + pixman_store_scanline_r5g6b5_asm_neon (width, pixel, values); +} + pixman_implementation_t * _pixman_implementation_create_arm_neon (void) { @@ -385,6 +422,10 @@ _pixman_implementation_create_arm_neon (void) imp->combine_32[PIXMAN_OP_OVER] = neon_combine_over_u; imp->combine_32[PIXMAN_OP_ADD] = neon_combine_add_u; + _pixman_bits_override_accessors (PIXMAN_r5g6b5, + neon_fetch_scanline_r5g6b5, + neon_store_scanline_r5g6b5); + imp->blt = arm_neon_blt; imp->fill = arm_neon_fill; diff --git a/pixman/pixman-private.h b/pixman/pixman-private.h index eeb677d..ba2d401 100644 --- a/pixman/pixman-private.h +++ b/pixman/pixman-private.h @@ -220,6 +220,11 @@ void _pixman_bits_image_setup_raw_accessors (bits_image_t *image); void +_pixman_bits_override_accessors (pixman_format_code_t format, + fetch_scanline_t fetch_func, + store_scanline_t store_func); + +void _pixman_image_get_scanline_generic_64 (pixman_image_t *image, int x, int y, -- 1.6.6.1