diff options
author | Koen Kooi <koen@dominion.thruhere.net> | 2010-11-01 14:39:20 +0100 |
---|---|---|
committer | Koen Kooi <koen@dominion.thruhere.net> | 2010-11-01 20:07:13 +0100 |
commit | 2b41bf1cc71b52e09573b130a568100e18bef12b (patch) | |
tree | 20ed56e64ecf1000f65fdae1e5ab54118ccf290d /recipes-graphics | |
download | meta-openembedded-2b41bf1cc71b52e09573b130a568100e18bef12b.tar.gz |
angstrom-layers: add 'meta-openembedded' layer to merge in cleaned up OE recipes
Signed-off-by: Koen Kooi <k-kooi@ti.com>
Diffstat (limited to 'recipes-graphics')
11 files changed, 1010 insertions, 0 deletions
diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0000-Add-pixman_bits_override_accessors.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0000-Add-pixman_bits_override_accessors.patch new file mode 100644 index 0000000000..43cd44268e --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0000-Add-pixman_bits_override_accessors.patch @@ -0,0 +1,77 @@ +From 38aabb3be87ea68e37f34256c778d07f62680ec6 Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Thu, 10 Dec 2009 00:51:50 +0200 +Subject: [PATCH 1/9] add _pixman_bits_override_accessors + +* from patch ARM: HACK: added NEON optimizations for fetch/store r5g6b5 scanline +* used in + 0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch + 0006-ARM-added-NEON-optimizations-for-fetch-store-a8-scan.patch + 0007-ARM-added-NEON-optimizations-for-fetching-x8r8g8b8-s.patch + + +--- + pixman/pixman-access.c | 23 ++++++++++++++++++++++- + pixman/pixman-private.h | 5 +++++ + 4 files changed, 87 insertions(+), 1 deletions(-) + +diff --git a/pixman/pixman-access.c b/pixman/pixman-access.c +index f1ce0ba..b33da29 100644 +--- a/pixman/pixman-access.c ++++ b/pixman/pixman-access.c +@@ -2836,7 +2836,7 @@ 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), +@@ -2978,6 +2978,27 @@ _pixman_bits_image_setup_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_32 = fetch_func; ++ if (store_func) ++ info->store_scanline_32 = store_func; ++ return; ++ } ++ info++; ++ } ++} ++ + #else + + void +diff --git a/pixman/pixman-private.h b/pixman/pixman-private.h +index d85868f..564f8f0 100644 +--- a/pixman/pixman-private.h ++++ b/pixman/pixman-private.h +@@ -206,6 +206,11 @@ void + _pixman_bits_image_setup_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 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch new file mode 100644 index 0000000000..795cce5bad --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch @@ -0,0 +1,114 @@ +From 97b2bb933455f222b392b5c60a8bde82d7d6329f Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Tue, 16 Mar 2010 16:55:28 +0100 +Subject: [PATCH 1/8] Generic C implementation of pixman_blt with overlapping support + +Uses memcpy/memmove functions to copy pixels, can handle the +case when both source and destination areas are in the same +image (this is useful for scrolling). + +It is assumed that copying direction is only important when +using the same image for both source and destination (and +src_stride == dst_stride). Copying direction is undefined +for the images with different source and destination stride +which happen to be in the overlapped areas (but this is an +unrealistic case anyway). +--- + pixman/pixman-general.c | 21 ++++++++++++++++++--- + pixman/pixman-private.h | 43 +++++++++++++++++++++++++++++++++++++++++++ + 2 files changed, 61 insertions(+), 3 deletions(-) + +diff --git a/pixman/pixman-general.c b/pixman/pixman-general.c +index 4d234a0..c4d2c14 100644 +--- a/pixman/pixman-general.c ++++ b/pixman/pixman-general.c +@@ -280,9 +280,24 @@ general_blt (pixman_implementation_t *imp, + int width, + int height) + { +- /* We can't blit unless we have sse2 or mmx */ +- +- return FALSE; ++ uint8_t *dst_bytes = (uint8_t *)dst_bits; ++ uint8_t *src_bytes = (uint8_t *)src_bits; ++ int bpp; ++ ++ if (src_bpp != dst_bpp || src_bpp & 7) ++ return FALSE; ++ ++ bpp = src_bpp >> 3; ++ width *= bpp; ++ src_stride *= 4; ++ dst_stride *= 4; ++ pixman_blt_helper (src_bytes + src_y * src_stride + src_x * bpp, ++ dst_bytes + dst_y * dst_stride + dst_x * bpp, ++ src_stride, ++ dst_stride, ++ width, ++ height); ++ return TRUE; + } + + static pixman_bool_t +diff --git a/pixman/pixman-private.h b/pixman/pixman-private.h +index c43172b..f980454 100644 +--- a/pixman/pixman-private.h ++++ b/pixman/pixman-private.h +@@ -10,6 +10,7 @@ + + #include "pixman.h" + #include <time.h> ++#include <string.h> + #include <assert.h> + #include <stdio.h> + #include <string.h> +@@ -873,4 +874,46 @@ void pixman_timer_register (pixman_timer_t *timer); + + #endif /* PIXMAN_TIMERS */ + ++/* a helper function, can blit 8-bit images with src/dst overlapping support */ ++static inline void ++pixman_blt_helper (uint8_t *src_bytes, ++ uint8_t *dst_bytes, ++ int src_stride, ++ int dst_stride, ++ int width, ++ int height) ++{ ++ /* ++ * The second part of this check is not strictly needed, but it prevents ++ * unnecessary upside-down processing of areas which belong to different ++ * images. Upside-down processing can be slower with fixed-distance-ahead ++ * prefetch and perceived as having more tearing. ++ */ ++ if (src_bytes < dst_bytes + width && ++ src_bytes + src_stride * height > dst_bytes) ++ { ++ src_bytes += src_stride * height - src_stride; ++ dst_bytes += dst_stride * height - dst_stride; ++ dst_stride = -dst_stride; ++ src_stride = -src_stride; ++ /* Horizontal scrolling to the left needs memmove */ ++ if (src_bytes + width > dst_bytes) ++ { ++ while (--height >= 0) ++ { ++ memmove (dst_bytes, src_bytes, width); ++ dst_bytes += dst_stride; ++ src_bytes += src_stride; ++ } ++ return; ++ } ++ } ++ while (--height >= 0) ++ { ++ memcpy (dst_bytes, src_bytes, width); ++ dst_bytes += dst_stride; ++ src_bytes += src_stride; ++ } ++} ++ + #endif /* PIXMAN_PRIVATE_H */ +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch new file mode 100644 index 0000000000..6e2d492aaf --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch @@ -0,0 +1,91 @@ +From 47b31f936641da07431093ede340465625bfcb3d Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Thu, 22 Oct 2009 05:45:47 +0300 +Subject: [PATCH 2/8] Support of overlapping src/dst for pixman_blt_mmx + +--- + pixman/pixman-mmx.c | 55 +++++++++++++++++++++++++++++--------------------- + 1 files changed, 32 insertions(+), 23 deletions(-) + +diff --git a/pixman/pixman-mmx.c b/pixman/pixman-mmx.c +index e936c4c..2413197 100644 +--- a/pixman/pixman-mmx.c ++++ b/pixman/pixman-mmx.c +@@ -2996,34 +2996,43 @@ pixman_blt_mmx (uint32_t *src_bits, + { + uint8_t * src_bytes; + uint8_t * dst_bytes; +- int byte_width; ++ int bpp; + +- if (src_bpp != dst_bpp) ++ if (src_bpp != dst_bpp || src_bpp & 7) + return FALSE; + +- if (src_bpp == 16) +- { +- src_stride = src_stride * (int) sizeof (uint32_t) / 2; +- dst_stride = dst_stride * (int) sizeof (uint32_t) / 2; +- src_bytes = (uint8_t *)(((uint16_t *)src_bits) + src_stride * (src_y) + (src_x)); +- dst_bytes = (uint8_t *)(((uint16_t *)dst_bits) + dst_stride * (dst_y) + (dst_x)); +- byte_width = 2 * width; +- src_stride *= 2; +- dst_stride *= 2; +- } +- else if (src_bpp == 32) ++ bpp = src_bpp >> 3; ++ width *= bpp; ++ src_stride *= 4; ++ dst_stride *= 4; ++ src_bytes = (uint8_t *)src_bits + src_y * src_stride + src_x * bpp; ++ dst_bytes = (uint8_t *)dst_bits + dst_y * dst_stride + dst_x * bpp; ++ ++ if (src_bpp != 16 && src_bpp != 32) + { +- src_stride = src_stride * (int) sizeof (uint32_t) / 4; +- dst_stride = dst_stride * (int) sizeof (uint32_t) / 4; +- src_bytes = (uint8_t *)(((uint32_t *)src_bits) + src_stride * (src_y) + (src_x)); +- dst_bytes = (uint8_t *)(((uint32_t *)dst_bits) + dst_stride * (dst_y) + (dst_x)); +- byte_width = 4 * width; +- src_stride *= 4; +- dst_stride *= 4; ++ pixman_blt_helper (src_bytes, dst_bytes, src_stride, dst_stride, ++ width, height); ++ return TRUE; + } +- else ++ ++ if (src_bytes < dst_bytes && src_bytes + src_stride * height > dst_bytes) + { +- return FALSE; ++ src_bytes += src_stride * height - src_stride; ++ dst_bytes += dst_stride * height - dst_stride; ++ dst_stride = -dst_stride; ++ src_stride = -src_stride; ++ ++ if (src_bytes + width > dst_bytes) ++ { ++ /* TODO: reverse scanline copy using MMX */ ++ while (--height >= 0) ++ { ++ memmove (dst_bytes, src_bytes, width); ++ dst_bytes += dst_stride; ++ src_bytes += src_stride; ++ } ++ return TRUE; ++ } + } + + while (height--) +@@ -3033,7 +3042,7 @@ pixman_blt_mmx (uint32_t *src_bits, + uint8_t *d = dst_bytes; + src_bytes += src_stride; + dst_bytes += dst_stride; +- w = byte_width; ++ w = width; + + while (w >= 2 && ((unsigned long)d & 3)) + { +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch new file mode 100644 index 0000000000..910f62e6dd --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch @@ -0,0 +1,91 @@ +From 13be027637602fffda3b3cb6e171d8d6a67b3b4b Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Thu, 22 Oct 2009 05:45:54 +0300 +Subject: [PATCH 3/8] Support of overlapping src/dst for pixman_blt_sse2 + +--- + pixman/pixman-sse2.c | 55 +++++++++++++++++++++++++++++-------------------- + 1 files changed, 32 insertions(+), 23 deletions(-) + +diff --git a/pixman/pixman-sse2.c b/pixman/pixman-sse2.c +index 5907de0..25015ae 100644 +--- a/pixman/pixman-sse2.c ++++ b/pixman/pixman-sse2.c +@@ -5027,34 +5027,43 @@ pixman_blt_sse2 (uint32_t *src_bits, + { + uint8_t * src_bytes; + uint8_t * dst_bytes; +- int byte_width; ++ int bpp; + +- if (src_bpp != dst_bpp) ++ if (src_bpp != dst_bpp || src_bpp & 7) + return FALSE; + +- if (src_bpp == 16) +- { +- src_stride = src_stride * (int) sizeof (uint32_t) / 2; +- dst_stride = dst_stride * (int) sizeof (uint32_t) / 2; +- src_bytes =(uint8_t *)(((uint16_t *)src_bits) + src_stride * (src_y) + (src_x)); +- dst_bytes = (uint8_t *)(((uint16_t *)dst_bits) + dst_stride * (dst_y) + (dst_x)); +- byte_width = 2 * width; +- src_stride *= 2; +- dst_stride *= 2; +- } +- else if (src_bpp == 32) ++ bpp = src_bpp >> 3; ++ width *= bpp; ++ src_stride *= 4; ++ dst_stride *= 4; ++ src_bytes = (uint8_t *)src_bits + src_y * src_stride + src_x * bpp; ++ dst_bytes = (uint8_t *)dst_bits + dst_y * dst_stride + dst_x * bpp; ++ ++ if (src_bpp != 16 && src_bpp != 32) + { +- src_stride = src_stride * (int) sizeof (uint32_t) / 4; +- dst_stride = dst_stride * (int) sizeof (uint32_t) / 4; +- src_bytes = (uint8_t *)(((uint32_t *)src_bits) + src_stride * (src_y) + (src_x)); +- dst_bytes = (uint8_t *)(((uint32_t *)dst_bits) + dst_stride * (dst_y) + (dst_x)); +- byte_width = 4 * width; +- src_stride *= 4; +- dst_stride *= 4; ++ pixman_blt_helper (src_bytes, dst_bytes, src_stride, dst_stride, ++ width, height); ++ return TRUE; + } +- else ++ ++ if (src_bytes < dst_bytes && src_bytes + src_stride * height > dst_bytes) + { +- return FALSE; ++ src_bytes += src_stride * height - src_stride; ++ dst_bytes += dst_stride * height - dst_stride; ++ dst_stride = -dst_stride; ++ src_stride = -src_stride; ++ ++ if (src_bytes + width > dst_bytes) ++ { ++ /* TODO: reverse scanline copy using SSE2 */ ++ while (--height >= 0) ++ { ++ memmove (dst_bytes, src_bytes, width); ++ dst_bytes += dst_stride; ++ src_bytes += src_stride; ++ } ++ return TRUE; ++ } + } + + while (height--) +@@ -5064,7 +5073,7 @@ pixman_blt_sse2 (uint32_t *src_bits, + uint8_t *d = dst_bytes; + src_bytes += src_stride; + dst_bytes += dst_stride; +- w = byte_width; ++ w = width; + + while (w >= 2 && ((unsigned long)d & 3)) + { +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch new file mode 100644 index 0000000000..f6aa9792b8 --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch @@ -0,0 +1,94 @@ +From a913cc05a1a1c5a813cf06d248334edede9caab7 Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Wed, 18 Nov 2009 06:08:48 +0200 +Subject: [PATCH 4/8] Support of overlapping src/dst for pixman_blt_neon + +--- + pixman/pixman-arm-neon.c | 62 +++++++++++++++++++++++++++++++++++++-------- + 1 files changed, 51 insertions(+), 11 deletions(-) + +diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c +index be5d403..cbfd7cf 100644 +--- a/pixman/pixman-arm-neon.c ++++ b/pixman/pixman-arm-neon.c +@@ -176,26 +176,66 @@ pixman_blt_neon (uint32_t *src_bits, + int width, + int height) + { +- if (src_bpp != dst_bpp) ++ uint8_t * src_bytes; ++ uint8_t * dst_bytes; ++ int bpp; ++ ++ if (src_bpp != dst_bpp || src_bpp & 7) + return FALSE; + ++ bpp = src_bpp >> 3; ++ width *= bpp; ++ src_stride *= 4; ++ dst_stride *= 4; ++ src_bytes = (uint8_t *)src_bits + src_y * src_stride + src_x * bpp; ++ dst_bytes = (uint8_t *)dst_bits + dst_y * dst_stride + dst_x * bpp; ++ ++ if (src_bpp != 16 && src_bpp != 32) ++ { ++ pixman_blt_helper (src_bytes, dst_bytes, src_stride, dst_stride, ++ width, height); ++ return TRUE; ++ } ++ ++ if (src_bytes < dst_bytes && src_bytes + src_stride * height > dst_bytes) ++ { ++ src_bytes += src_stride * height - src_stride; ++ dst_bytes += dst_stride * height - dst_stride; ++ dst_stride = -dst_stride; ++ src_stride = -src_stride; ++ ++ if (src_bytes + width > dst_bytes) ++ { ++ /* TODO: reverse scanline copy using NEON */ ++ while (--height >= 0) ++ { ++ memmove (dst_bytes, src_bytes, width); ++ dst_bytes += dst_stride; ++ src_bytes += src_stride; ++ } ++ return TRUE; ++ } ++ } ++ + switch (src_bpp) + { + case 16: + pixman_composite_src_0565_0565_asm_neon ( +- width, height, +- (uint16_t *)(((char *) dst_bits) + +- dst_y * dst_stride * 4 + dst_x * 2), dst_stride * 2, +- (uint16_t *)(((char *) src_bits) + +- src_y * src_stride * 4 + src_x * 2), src_stride * 2); ++ width >> 1, ++ height, ++ (uint16_t *) dst_bytes, ++ dst_stride >> 1, ++ (uint16_t *) src_bytes, ++ src_stride >> 1); + return TRUE; + case 32: + pixman_composite_src_8888_8888_asm_neon ( +- width, height, +- (uint32_t *)(((char *) dst_bits) + +- dst_y * dst_stride * 4 + dst_x * 4), dst_stride, +- (uint32_t *)(((char *) src_bits) + +- src_y * src_stride * 4 + src_x * 4), src_stride); ++ width >> 2, ++ height, ++ (uint32_t *) dst_bytes, ++ dst_stride >> 2, ++ (uint32_t *) src_bytes, ++ src_stride >> 2); + return TRUE; + default: + return FALSE; +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch new file mode 100644 index 0000000000..dbe98b38ce --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch @@ -0,0 +1,109 @@ +From f75e9d1868e21dd75ff3a2ca3561546d23877ddb Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Thu, 10 Dec 2009 00:51:50 +0200 +Subject: [PATCH 5/8] ARM: added NEON optimizations for fetch/store r5g6b5 scanline + +--- + pixman/pixman-arm-neon-asm.S | 20 ++++++++++++++++++++ + pixman/pixman-arm-neon.c | 40 ++++++++++++++++++++++++++++++++++++++++ + 2 files changed, 60 insertions(+), 0 deletions(-) + +diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S +index e4db5cd..c79ba81 100644 +--- a/pixman/pixman-arm-neon-asm.S ++++ b/pixman/pixman-arm-neon-asm.S +@@ -459,6 +459,16 @@ 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 +@@ -494,6 +504,16 @@ 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_8_8_process_pixblock_head +diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c +index cbfd7cf..f88c8f8 100644 +--- a/pixman/pixman-arm-neon.c ++++ b/pixman/pixman-arm-neon.c +@@ -392,6 +392,42 @@ BIND_COMBINE_U (over) + BIND_COMBINE_U (add) + BIND_COMBINE_U (out_reverse) + ++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) ++{ ++ 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) + { +@@ -407,6 +443,10 @@ _pixman_implementation_create_arm_neon (void) + imp->combine_32[PIXMAN_OP_ADD] = neon_combine_add_u; + imp->combine_32[PIXMAN_OP_OUT_REVERSE] = neon_combine_out_reverse_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; + +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0006-ARM-added-NEON-optimizations-for-fetch-store-a8-scan.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0006-ARM-added-NEON-optimizations-for-fetch-store-a8-scan.patch new file mode 100644 index 0000000000..d050646fa7 --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0006-ARM-added-NEON-optimizations-for-fetch-store-a8-scan.patch @@ -0,0 +1,148 @@ +From a1cd695c5e22f0f4a2b7272fab675a3cc510bacb Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Thu, 23 Sep 2010 21:10:56 +0300 +Subject: [PATCH 6/8] ARM: added NEON optimizations for fetch/store a8 scanline + +--- + pixman/pixman-arm-neon-asm.S | 64 ++++++++++++++++++++++++++++++++++++++++++ + pixman/pixman-arm-neon.c | 42 +++++++++++++++++++++++++++ + 2 files changed, 106 insertions(+), 0 deletions(-) + +diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S +index c79ba81..ca0825c 100644 +--- a/pixman/pixman-arm-neon-asm.S ++++ b/pixman/pixman-arm-neon-asm.S +@@ -418,6 +418,70 @@ generate_composite_function \ + + /******************************************************************************/ + ++.macro pixman_composite_src_8_8888_process_pixblock_head ++ /* This is tricky part: we can't set these values just once in 'init' macro ++ * because leading/trailing pixels handling part uses VZIP.8 instructions, ++ * and they operate on values in-place and destroy original registers ++ * content. Think about it like VST4.8 instruction corrupting NEON ++ * registers after write in 'tail_head' macro. Except that 'tail_head' ++ * macro itself actually does not need these extra VMOVs because it uses ++ * real VST4.8 instruction. ++ */ ++ vmov.u8 q0, #0 ++ vmov.u8 d2, #0 ++.endm ++ ++.macro pixman_composite_src_8_8888_process_pixblock_tail ++.endm ++ ++.macro pixman_composite_src_8_8888_process_pixblock_tail_head ++ vst4.8 {d0, d1, d2, d3}, [DST_W, :128]! ++ vld1.8 {d3}, [SRC]! ++.endm ++ ++generate_composite_function_single_scanline \ ++ pixman_fetch_scanline_a8_asm_neon, 8, 0, 32, \ ++ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ ++ 8, /* number of pixels, processed in a single block */ \ ++ default_init, \ ++ default_cleanup, \ ++ pixman_composite_src_8_8888_process_pixblock_head, \ ++ pixman_composite_src_8_8888_process_pixblock_tail, \ ++ pixman_composite_src_8_8888_process_pixblock_tail_head, \ ++ 0, /* dst_w_basereg */ \ ++ 0, /* dst_r_basereg */ \ ++ 3, /* src_basereg */ \ ++ 0 /* mask_basereg */ ++ ++/******************************************************************************/ ++ ++.macro pixman_composite_src_8888_8_process_pixblock_head ++.endm ++ ++.macro pixman_composite_src_8888_8_process_pixblock_tail ++.endm ++ ++.macro pixman_composite_src_8888_8_process_pixblock_tail_head ++ vst1.8 {d3}, [DST_W, :64]! ++ vld4.8 {d0, d1, d2, d3}, [SRC]! ++.endm ++ ++generate_composite_function_single_scanline \ ++ pixman_store_scanline_a8_asm_neon, 32, 0, 8, \ ++ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ ++ 8, /* number of pixels, processed in a single block */ \ ++ default_init, \ ++ default_cleanup, \ ++ pixman_composite_src_8888_8_process_pixblock_head, \ ++ pixman_composite_src_8888_8_process_pixblock_tail, \ ++ pixman_composite_src_8888_8_process_pixblock_tail_head, \ ++ 3, /* dst_w_basereg */ \ ++ 0, /* dst_r_basereg */ \ ++ 0, /* src_basereg */ \ ++ 0 /* mask_basereg */ ++ ++/******************************************************************************/ ++ + .macro pixman_composite_src_8888_0565_process_pixblock_head + vshll.u8 q8, d1, #8 + vshll.u8 q14, d2, #8 +diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c +index f88c8f8..43091d2 100644 +--- a/pixman/pixman-arm-neon.c ++++ b/pixman/pixman-arm-neon.c +@@ -428,6 +428,45 @@ neon_store_scanline_r5g6b5 (bits_image_t * image, + pixman_store_scanline_r5g6b5_asm_neon (width, pixel, values); + } + ++void ++pixman_fetch_scanline_a8_asm_neon (int width, ++ uint32_t *buffer, ++ const uint8_t *pixel); ++ ++ ++void ++pixman_store_scanline_a8_asm_neon (int width, ++ uint8_t *pixel, ++ const uint32_t *values); ++ ++static void ++neon_fetch_scanline_a8 (pixman_image_t *image, ++ int x, ++ int y, ++ int width, ++ uint32_t * buffer, ++ const uint32_t *mask) ++{ ++ const uint32_t *bits = image->bits.bits + y * image->bits.rowstride; ++ const uint8_t *pixel = (const uint8_t *) bits + x; ++ ++ pixman_fetch_scanline_a8_asm_neon (width, buffer, pixel); ++} ++ ++static void ++neon_store_scanline_a8 (bits_image_t * image, ++ int x, ++ int y, ++ int width, ++ const uint32_t *values) ++{ ++ uint32_t *bits = image->bits + image->rowstride * y; ++ uint8_t *pixel = (uint8_t *) bits + x; ++ ++ pixman_store_scanline_a8_asm_neon (width, pixel, values); ++} ++ ++ + pixman_implementation_t * + _pixman_implementation_create_arm_neon (void) + { +@@ -446,6 +485,9 @@ _pixman_implementation_create_arm_neon (void) + _pixman_bits_override_accessors (PIXMAN_r5g6b5, + neon_fetch_scanline_r5g6b5, + neon_store_scanline_r5g6b5); ++ _pixman_bits_override_accessors (PIXMAN_a8, ++ neon_fetch_scanline_a8, ++ neon_store_scanline_a8); + + imp->blt = arm_neon_blt; + imp->fill = arm_neon_fill; +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0007-ARM-added-NEON-optimizations-for-fetching-x8r8g8b8-s.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0007-ARM-added-NEON-optimizations-for-fetching-x8r8g8b8-s.patch new file mode 100644 index 0000000000..7f28f47cff --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0007-ARM-added-NEON-optimizations-for-fetching-x8r8g8b8-s.patch @@ -0,0 +1,77 @@ +From d6ae7da60cc797900b5eff0786536c4a11ab0f50 Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Fri, 24 Sep 2010 18:22:44 +0300 +Subject: [PATCH 7/8] ARM: added NEON optimizations for fetching x8r8g8b8 scanline + +--- + pixman/pixman-arm-neon-asm.S | 14 ++++++++++++++ + pixman/pixman-arm-neon.c | 21 +++++++++++++++++++++ + 2 files changed, 35 insertions(+), 0 deletions(-) + +diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S +index ca0825c..ffd0b83 100644 +--- a/pixman/pixman-arm-neon-asm.S ++++ b/pixman/pixman-arm-neon-asm.S +@@ -1206,6 +1206,20 @@ generate_composite_function \ + 0, /* src_basereg */ \ + 0 /* mask_basereg */ + ++generate_composite_function_single_scanline \ ++ pixman_fetch_scanline_x888_asm_neon, 32, 0, 32, \ ++ FLAG_DST_WRITEONLY, \ ++ 8, /* number of pixels, processed in a single block */ \ ++ pixman_composite_src_x888_8888_init, \ ++ default_cleanup, \ ++ pixman_composite_src_x888_8888_process_pixblock_head, \ ++ pixman_composite_src_x888_8888_process_pixblock_tail, \ ++ pixman_composite_src_x888_8888_process_pixblock_tail_head, \ ++ 0, /* dst_w_basereg */ \ ++ 0, /* dst_r_basereg */ \ ++ 0, /* src_basereg */ \ ++ 0 /* mask_basereg */ ++ + /******************************************************************************/ + + .macro pixman_composite_over_n_8_8888_process_pixblock_head +diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c +index 43091d2..f84b5e6 100644 +--- a/pixman/pixman-arm-neon.c ++++ b/pixman/pixman-arm-neon.c +@@ -466,6 +466,24 @@ neon_store_scanline_a8 (bits_image_t * image, + pixman_store_scanline_a8_asm_neon (width, pixel, values); + } + ++void ++pixman_fetch_scanline_x888_asm_neon (int width, ++ uint32_t *buffer, ++ const uint32_t *pixel); ++ ++static void ++neon_fetch_scanline_x888 (pixman_image_t *image, ++ int x, ++ int y, ++ int width, ++ uint32_t * buffer, ++ const uint32_t *mask) ++{ ++ const uint32_t *bits = image->bits.bits + y * image->bits.rowstride; ++ const uint32_t *pixel = (const uint32_t *) bits + x; ++ ++ pixman_fetch_scanline_x888_asm_neon (width, buffer, pixel); ++} + + pixman_implementation_t * + _pixman_implementation_create_arm_neon (void) +@@ -488,6 +506,9 @@ _pixman_implementation_create_arm_neon (void) + _pixman_bits_override_accessors (PIXMAN_a8, + neon_fetch_scanline_a8, + neon_store_scanline_a8); ++ _pixman_bits_override_accessors (PIXMAN_x8r8g8b8, ++ neon_fetch_scanline_x888, ++ NULL); + + imp->blt = arm_neon_blt; + imp->fill = arm_neon_fill; +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman-0.20.0/0008-ARM-optimization-for-scaled-src_0565_0565-operation-.patch b/recipes-graphics/xorg-lib/pixman-0.20.0/0008-ARM-optimization-for-scaled-src_0565_0565-operation-.patch new file mode 100644 index 0000000000..6efdb621ad --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman-0.20.0/0008-ARM-optimization-for-scaled-src_0565_0565-operation-.patch @@ -0,0 +1,172 @@ +From e1191ad6563a1fb02a45982b1c4d7fed3c655e97 Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Mon, 4 Oct 2010 01:56:59 +0300 +Subject: [PATCH 8/8] ARM optimization for scaled src_0565_0565 operation with nearest filter + +The code actually uses only armv4t instructions. + +Benchmark from ARM11: + + == before == + op=1, src_fmt=10020565, dst_fmt=10020565, speed=34.86 MPix/s + + == after == + op=1, src_fmt=10020565, dst_fmt=10020565, speed=36.62 MPix/s + +Benchmark from ARM Cortex-A8: + + == before == + op=1, src_fmt=10020565, dst_fmt=10020565, speed=89.55 MPix/s + + == after == + op=1, src_fmt=10020565, dst_fmt=10020565, speed=94.91 MPix/s +--- + pixman/pixman-arm-simd-asm.S | 66 ++++++++++++++++++++++++++++++++++++++++++ + pixman/pixman-arm-simd.c | 37 +++++++++++++++++++++++ + 2 files changed, 103 insertions(+), 0 deletions(-) + +diff --git a/pixman/pixman-arm-simd-asm.S b/pixman/pixman-arm-simd-asm.S +index a3d2d40..b6f69db 100644 +--- a/pixman/pixman-arm-simd-asm.S ++++ b/pixman/pixman-arm-simd-asm.S +@@ -1,5 +1,6 @@ + /* + * Copyright © 2008 Mozilla Corporation ++ * Copyright © 2010 Nokia Corporation + * + * Permission to use, copy, modify, distribute, and sell this software and its + * documentation for any purpose is hereby granted without fee, provided that +@@ -328,3 +329,68 @@ pixman_asm_function pixman_composite_over_n_8_8888_asm_armv6 + pop {r4, r5, r6, r7, r8, r9, r10, r11} + bx lr + .endfunc ++ ++/* ++ * Note: This function is actually primarily optimized for ARM Cortex-A8 ++ * pipeline. In order to get good performance on ARM9/ARM11 cores (which ++ * don't have efficient write combining), it needs to be changed to use ++ * 16-byte aligned writes using STM instruction. ++ */ ++pixman_asm_function pixman_scaled_nearest_scanline_565_565_SRC_asm_armv6 ++ DST .req r0 ++ SRC .req r1 ++ W .req r2 ++ VX .req r3 ++ UNIT_X .req r12 ++ TMP1 .req r4 ++ TMP2 .req r5 ++ MASK .req r6 ++ ldr UNIT_X, [sp] ++ push {r4, r5, r6, r7} ++ mvn MASK, #1 ++ ++ /* define helper macro */ ++ .macro scale_2_pixels ++ ldrh TMP1, [SRC, TMP1] ++ and TMP2, MASK, VX, lsr #15 ++ add VX, VX, UNIT_X ++ strh TMP1, [DST], #2 ++ ++ ldrh TMP2, [SRC, TMP2] ++ and TMP1, MASK, VX, lsr #15 ++ add VX, VX, UNIT_X ++ strh TMP2, [DST], #2 ++ .endm ++ ++ /* now do the scaling */ ++ and TMP1, MASK, VX, lsr #15 ++ add VX, VX, UNIT_X ++ subs W, #4 ++ blt 2f ++1: /* main loop, process 4 pixels per iteration */ ++ scale_2_pixels ++ scale_2_pixels ++ subs W, W, #4 ++ bge 1b ++2: ++ tst W, #2 ++ beq 2f ++ scale_2_pixels ++2: ++ tst W, #1 ++ ldrneh TMP1, [SRC, TMP1] ++ strneh TMP1, [DST], #2 ++ /* cleanup helper macro */ ++ .purgem scale_2_pixels ++ .unreq DST ++ .unreq SRC ++ .unreq W ++ .unreq VX ++ .unreq UNIT_X ++ .unreq TMP1 ++ .unreq TMP2 ++ .unreq MASK ++ /* return */ ++ pop {r4, r5, r6, r7} ++ bx lr ++.endfunc +diff --git a/pixman/pixman-arm-simd.c b/pixman/pixman-arm-simd.c +index d466a31..f6f464c 100644 +--- a/pixman/pixman-arm-simd.c ++++ b/pixman/pixman-arm-simd.c +@@ -29,6 +29,7 @@ + + #include "pixman-private.h" + #include "pixman-arm-common.h" ++#include "pixman-fast-path.h" + + #if 0 /* This code was moved to 'pixman-arm-simd-asm.S' */ + +@@ -375,6 +376,35 @@ pixman_composite_over_n_8_8888_asm_armv6 (int32_t width, + + #endif + ++void ++pixman_scaled_nearest_scanline_565_565_SRC_asm_armv6 (uint16_t * dst, ++ uint16_t * src, ++ int32_t w, ++ pixman_fixed_t vx, ++ pixman_fixed_t unit_x); ++ ++static force_inline void ++scaled_nearest_scanline_armv6_565_565_SRC (uint16_t * dst, ++ uint16_t * src, ++ int32_t w, ++ pixman_fixed_t vx, ++ pixman_fixed_t unit_x, ++ pixman_fixed_t max_vx) ++{ ++ pixman_scaled_nearest_scanline_565_565_SRC_asm_armv6 (dst, src, w, ++ vx, unit_x); ++} ++ ++FAST_NEAREST_MAINLOOP (armv6_565_565_cover_SRC, ++ scaled_nearest_scanline_armv6_565_565_SRC, ++ uint16_t, uint16_t, COVER); ++FAST_NEAREST_MAINLOOP (armv6_565_565_none_SRC, ++ scaled_nearest_scanline_armv6_565_565_SRC, ++ uint16_t, uint16_t, NONE); ++FAST_NEAREST_MAINLOOP (armv6_565_565_pad_SRC, ++ scaled_nearest_scanline_armv6_565_565_SRC, ++ uint16_t, uint16_t, PAD); ++ + PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (armv6, add_8_8, + uint8_t, 1, uint8_t, 1) + PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (armv6, over_8888_8888, +@@ -404,6 +434,13 @@ static const pixman_fast_path_t arm_simd_fast_paths[] = + PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8b8g8r8, armv6_composite_over_n_8_8888), + PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8b8g8r8, armv6_composite_over_n_8_8888), + ++ SIMPLE_NEAREST_FAST_PATH_COVER (SRC, r5g6b5, r5g6b5, armv6_565_565), ++ SIMPLE_NEAREST_FAST_PATH_COVER (SRC, b5g6r5, b5g6r5, armv6_565_565), ++ SIMPLE_NEAREST_FAST_PATH_NONE (SRC, r5g6b5, r5g6b5, armv6_565_565), ++ SIMPLE_NEAREST_FAST_PATH_NONE (SRC, b5g6r5, b5g6r5, armv6_565_565), ++ SIMPLE_NEAREST_FAST_PATH_PAD (SRC, r5g6b5, r5g6b5, armv6_565_565), ++ SIMPLE_NEAREST_FAST_PATH_PAD (SRC, b5g6r5, b5g6r5, armv6_565_565), ++ + { PIXMAN_OP_NONE }, + }; + +-- +1.6.6.1 + diff --git a/recipes-graphics/xorg-lib/pixman.inc b/recipes-graphics/xorg-lib/pixman.inc new file mode 100644 index 0000000000..539a33a888 --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman.inc @@ -0,0 +1,14 @@ +SECTION = "libs" +PRIORITY = "optional" +DESCRIPTION = "Low-level pixel manipulation library." +LICENSE = "X11" + +DEPENDS = "perl-native" + +BBCLASSEXTEND = "native" + +SRC_URI = "http://cairographics.org/releases/${BPN}-${PV}.tar.gz;name=archive " + +INC_PR = "r5" + +inherit autotools pkgconfig diff --git a/recipes-graphics/xorg-lib/pixman_0.20.0.bb b/recipes-graphics/xorg-lib/pixman_0.20.0.bb new file mode 100644 index 0000000000..b3fbd4f155 --- /dev/null +++ b/recipes-graphics/xorg-lib/pixman_0.20.0.bb @@ -0,0 +1,23 @@ +require pixman.inc + +SRC_URI[archive.md5sum] = "c1a31d5cedfa97c5af7148a2d1fd4356" +SRC_URI[archive.sha256sum] = "9c02c22c6cc3f28f3633d02ef6f0cac130518f621edb011ebbbf08cd1a81251a" + +PR = "${INC_PR}.0" + +SRC_URI += "\ + file://0000-Add-pixman_bits_override_accessors.patch \ + file://0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch \ + file://0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch \ + file://0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch \ + file://0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch \ + file://0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch \ + file://0006-ARM-added-NEON-optimizations-for-fetch-store-a8-scan.patch \ + file://0007-ARM-added-NEON-optimizations-for-fetching-x8r8g8b8-s.patch \ + file://0008-ARM-optimization-for-scaled-src_0565_0565-operation-.patch \ +" + +NEON = " --disable-arm-neon " +NEON_armv7a = " " + +EXTRA_OECONF = "${NEON} --disable-gtk" |