diff options
10 files changed, 905 insertions, 2 deletions
diff --git a/recipes/xorg-lib/pixman-0.18.0/missing-cache-preload.diff b/recipes/xorg-lib/pixman-0.18.0/missing-cache-preload.diff new file mode 100644 index 0000000000..475bb35495 --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.0/missing-cache-preload.diff @@ -0,0 +1,28 @@ +From 4f45c7688e6b767a5f7b7f5cced51b4797a3c288 Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Fri, 21 May 2010 13:31:03 +0000 +Subject: ARM: added missing cache preload + +--- +diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S +index 51bc347..a99555a 100644 +--- a/pixman/pixman-arm-neon-asm.S ++++ b/pixman/pixman-arm-neon-asm.S +@@ -388,6 +388,7 @@ generate_composite_function \ + vld1.16 {d4, d5}, [DST_R, :128]! + vst1.16 {d28, d29}, [DST_W, :128]! + pixman_composite_over_n_0565_process_pixblock_head ++ cache_preload 8, 8 + .endm + + .macro pixman_composite_over_n_0565_init +@@ -660,6 +661,7 @@ generate_composite_function_single_scanline \ + vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! + vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! + pixman_composite_over_8888_8888_process_pixblock_head ++ cache_preload 8, 8 + .endm + + .macro pixman_composite_over_n_8888_init +-- +cgit v0.8.3-6-g21f6 diff --git a/recipes/xorg-lib/pixman-0.18.2/0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch b/recipes/xorg-lib/pixman-0.18.2/0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch new file mode 100644 index 0000000000..b2488de4dd --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.2/0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch @@ -0,0 +1,114 @@ +From 05875eca09ee23ea04ccb32f87c7c355fd1b88f3 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 1/7] 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 bddf79a..f525744 100644 +--- a/pixman/pixman-general.c ++++ b/pixman/pixman-general.c +@@ -285,9 +285,24 @@ 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 d5767af..eeb677d 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> +@@ -867,4 +868,46 @@ 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/xorg-lib/pixman-0.18.2/0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch b/recipes/xorg-lib/pixman-0.18.2/0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch new file mode 100644 index 0000000000..ca6a3c5220 --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.2/0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch @@ -0,0 +1,91 @@ +From 8aca14948a43f5cfd478e22cef745e55960d68cb 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 2/7] 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 d51b40c..6e0296b 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/xorg-lib/pixman-0.18.2/0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch b/recipes/xorg-lib/pixman-0.18.2/0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch new file mode 100644 index 0000000000..faeb6e67ab --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.2/0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch @@ -0,0 +1,91 @@ +From 56c446d6912936eabcc6bdd583f37ee1209eff8c 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 3/7] 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 d5349d7..aa34012 100644 +--- a/pixman/pixman-sse2.c ++++ b/pixman/pixman-sse2.c +@@ -5328,34 +5328,43 @@ 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; ++ } + } + + cache_prefetch ((__m128i*)src_bytes); +@@ -5368,7 +5377,7 @@ 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; + + cache_prefetch_next ((__m128i*)s); + cache_prefetch_next ((__m128i*)d); +-- +1.6.6.1 + diff --git a/recipes/xorg-lib/pixman-0.18.2/0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch b/recipes/xorg-lib/pixman-0.18.2/0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch new file mode 100644 index 0000000000..d56fd99c6f --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.2/0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch @@ -0,0 +1,94 @@ +From 80c1eae677a6ea10d67bb7328230626cd8fce08e 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 4/7] 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 6808b36..7feee1d 100644 +--- a/pixman/pixman-arm-neon.c ++++ b/pixman/pixman-arm-neon.c +@@ -168,26 +168,66 @@ 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/xorg-lib/pixman-0.18.2/0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch b/recipes/xorg-lib/pixman-0.18.2/0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch new file mode 100644 index 0000000000..39eda8b546 --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.2/0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch @@ -0,0 +1,169 @@ +From 2f5774db49521e990c5e9f7cac684a06f4e67a43 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 5/7] 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 + diff --git a/recipes/xorg-lib/pixman-0.18.2/0006-A-copy-paste-version-of-16bpp-bilinear-scanline-fetc.patch b/recipes/xorg-lib/pixman-0.18.2/0006-A-copy-paste-version-of-16bpp-bilinear-scanline-fetc.patch new file mode 100644 index 0000000000..dd0e0f6d4e --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.2/0006-A-copy-paste-version-of-16bpp-bilinear-scanline-fetc.patch @@ -0,0 +1,261 @@ +From fc24a2d8d0039cd10de4175da40bb784e2c3bf49 Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Fri, 4 Dec 2009 16:49:19 +0000 +Subject: [PATCH 6/7] A copy-paste version of 16bpp bilinear scanline fetcher + +--- + pixman/pixman-bits-image.c | 228 +++++++++++++++++++++++++++++++++++++++++++- + 1 files changed, 223 insertions(+), 5 deletions(-) + +diff --git a/pixman/pixman-bits-image.c b/pixman/pixman-bits-image.c +index 3d78ff0..1656975 100644 +--- a/pixman/pixman-bits-image.c ++++ b/pixman/pixman-bits-image.c +@@ -535,6 +535,212 @@ bits_image_fetch_bilinear_no_repeat_8888 (pixman_image_t * ima, + *buffer++ = 0; + } + ++static void ++bits_image_fetch_bilinear_no_repeat_0565 (pixman_image_t * ima, ++ int offset, ++ int line, ++ int width, ++ uint32_t * buffer, ++ const uint32_t * mask, ++ uint32_t mask_bits) ++{ ++ bits_image_t *bits = &ima->bits; ++ pixman_fixed_t x_top, x_bottom, x; ++ pixman_fixed_t ux_top, ux_bottom, ux; ++ pixman_vector_t v; ++ uint32_t top_mask, bottom_mask; ++ uint16_t *top_row; ++ uint16_t *bottom_row; ++ uint32_t *end; ++ uint16_t zero[2] = { 0, 0 }; ++ int y, y1, y2; ++ int disty; ++ int mask_inc; ++ int w; ++ ++ /* reference point is the center of the pixel */ ++ v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2; ++ v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2; ++ v.vector[2] = pixman_fixed_1; ++ ++ if (!pixman_transform_point_3d (bits->common.transform, &v)) ++ return; ++ ++ ux = ux_top = ux_bottom = bits->common.transform->matrix[0][0]; ++ x = x_top = x_bottom = v.vector[0] - pixman_fixed_1/2; ++ ++ y = v.vector[1] - pixman_fixed_1/2; ++ disty = (y >> 8) & 0xff; ++ ++ /* Load the pointers to the first and second lines from the source ++ * image that bilinear code must read. ++ * ++ * The main trick in this code is about the check if any line are ++ * outside of the image; ++ * ++ * When I realize that a line (any one) is outside, I change ++ * the pointer to a dummy area with zeros. Once I change this, I ++ * must be sure the pointer will not change, so I set the ++ * variables to each pointer increments inside the loop. ++ */ ++ y1 = pixman_fixed_to_int (y); ++ y2 = y1 + 1; ++ ++ if (y1 < 0 || y1 >= bits->height) ++ { ++ top_row = zero; ++ x_top = 0; ++ ux_top = 0; ++ } ++ else ++ { ++ top_row = bits->bits + y1 * bits->rowstride; ++ x_top = x; ++ ux_top = ux; ++ } ++ ++ if (y2 < 0 || y2 >= bits->height) ++ { ++ bottom_row = zero; ++ x_bottom = 0; ++ ux_bottom = 0; ++ } ++ else ++ { ++ bottom_row = bits->bits + y2 * bits->rowstride; ++ x_bottom = x; ++ ux_bottom = ux; ++ } ++ ++ /* Instead of checking whether the operation uses the mast in ++ * each loop iteration, verify this only once and prepare the ++ * variables to make the code smaller inside the loop. ++ */ ++ if (!mask) ++ { ++ mask_inc = 0; ++ mask_bits = 1; ++ mask = &mask_bits; ++ } ++ else ++ { ++ /* If have a mask, prepare the variables to check it */ ++ mask_inc = 1; ++ } ++ ++ /* If both are zero, then the whole thing is zero */ ++ if (top_row == zero && bottom_row == zero) ++ { ++ memset (buffer, 0, width * sizeof (uint32_t)); ++ return; ++ } ++ else ++ { ++ if (top_row == zero) ++ { ++ top_mask = 0; ++ bottom_mask = 0xff000000; ++ } ++ else if (bottom_row == zero) ++ { ++ top_mask = 0xff000000; ++ bottom_mask = 0; ++ } ++ else ++ { ++ top_mask = 0xff000000; ++ bottom_mask = 0xff000000; ++ } ++ } ++ ++ end = buffer + width; ++ ++ /* Zero fill to the left of the image */ ++ while (buffer < end && x < pixman_fixed_minus_1) ++ { ++ *buffer++ = 0; ++ x += ux; ++ x_top += ux_top; ++ x_bottom += ux_bottom; ++ mask += mask_inc; ++ } ++ ++ /* Left edge ++ */ ++ while (buffer < end && x < 0) ++ { ++ uint32_t tr, br; ++ int32_t distx; ++ ++ tr = CONVERT_0565_TO_0888 (top_row[pixman_fixed_to_int (x_top) + 1]) | top_mask; ++ br = CONVERT_0565_TO_0888 (bottom_row[pixman_fixed_to_int (x_bottom) + 1]) | bottom_mask; ++ ++ distx = (x >> 8) & 0xff; ++ ++ *buffer++ = bilinear_interpolation (0, tr, 0, br, distx, disty); ++ ++ x += ux; ++ x_top += ux_top; ++ x_bottom += ux_bottom; ++ mask += mask_inc; ++ } ++ ++ /* Main part */ ++ w = pixman_int_to_fixed (bits->width - 1); ++ ++ while (buffer < end && x < w) ++ { ++ if (*mask) ++ { ++ uint32_t tl, tr, bl, br; ++ int32_t distx; ++ ++ tl = CONVERT_0565_TO_0888 (top_row [pixman_fixed_to_int (x_top)]) | top_mask; ++ tr = CONVERT_0565_TO_0888 (top_row [pixman_fixed_to_int (x_top) + 1]) | top_mask; ++ bl = CONVERT_0565_TO_0888 (bottom_row [pixman_fixed_to_int (x_bottom)]) | bottom_mask; ++ br = CONVERT_0565_TO_0888 (bottom_row [pixman_fixed_to_int (x_bottom) + 1]) | bottom_mask; ++ ++ distx = (x >> 8) & 0xff; ++ ++ *buffer = bilinear_interpolation (tl, tr, bl, br, distx, disty); ++ } ++ ++ buffer++; ++ x += ux; ++ x_top += ux_top; ++ x_bottom += ux_bottom; ++ mask += mask_inc; ++ } ++ ++ /* Right Edge */ ++ w = pixman_int_to_fixed (bits->width); ++ while (buffer < end && x < w) ++ { ++ if (*mask) ++ { ++ uint32_t tl, bl; ++ int32_t distx; ++ ++ tl = CONVERT_0565_TO_0888 (top_row [pixman_fixed_to_int (x_top)]) | top_mask; ++ bl = CONVERT_0565_TO_0888 (bottom_row [pixman_fixed_to_int (x_bottom)]) | bottom_mask; ++ ++ distx = (x >> 8) & 0xff; ++ ++ *buffer = bilinear_interpolation (tl, 0, bl, 0, distx, disty); ++ } ++ ++ buffer++; ++ x += ux; ++ x_top += ux_top; ++ x_bottom += ux_bottom; ++ mask += mask_inc; ++ } ++ ++ /* Zero fill to the left of the image */ ++ while (buffer < end) ++ *buffer++ = 0; ++} ++ + static force_inline uint32_t + bits_image_fetch_pixel_convolution (bits_image_t *image, + pixman_fixed_t x, +@@ -917,14 +1123,26 @@ bits_image_property_changed (pixman_image_t *image) + (bits->common.filter == PIXMAN_FILTER_BILINEAR || + bits->common.filter == PIXMAN_FILTER_GOOD || + bits->common.filter == PIXMAN_FILTER_BEST) && +- bits->common.repeat == PIXMAN_REPEAT_NONE && +- (bits->format == PIXMAN_a8r8g8b8 || +- bits->format == PIXMAN_x8r8g8b8)) ++ bits->common.repeat == PIXMAN_REPEAT_NONE) + { + image->common.get_scanline_64 = + _pixman_image_get_scanline_generic_64; +- image->common.get_scanline_32 = +- bits_image_fetch_bilinear_no_repeat_8888; ++ ++ if (bits->format == PIXMAN_a8r8g8b8 || bits->format == PIXMAN_x8r8g8b8) ++ { ++ image->common.get_scanline_32 = ++ bits_image_fetch_bilinear_no_repeat_8888; ++ } ++ else if (bits->format == PIXMAN_r5g6b5) ++ { ++ image->common.get_scanline_32 = ++ bits_image_fetch_bilinear_no_repeat_0565; ++ } ++ else ++ { ++ image->common.get_scanline_32 = ++ bits_image_fetch_transformed; ++ } + } + else + { +-- +1.6.6.1 + diff --git a/recipes/xorg-lib/pixman-0.18.2/0007-ARM-added-missing-cache-preload.patch b/recipes/xorg-lib/pixman-0.18.2/0007-ARM-added-missing-cache-preload.patch new file mode 100644 index 0000000000..f601d2ce26 --- /dev/null +++ b/recipes/xorg-lib/pixman-0.18.2/0007-ARM-added-missing-cache-preload.patch @@ -0,0 +1,32 @@ +From af8a33d3b132c0192d59c47426bb3a201cb64b32 Mon Sep 17 00:00:00 2001 +From: Siarhei Siamashka <siarhei.siamashka@nokia.com> +Date: Fri, 21 May 2010 13:31:03 +0000 +Subject: [PATCH 7/7] ARM: added missing cache preload + +--- + pixman/pixman-arm-neon-asm.S | 2 ++ + 1 files changed, 2 insertions(+), 0 deletions(-) + +diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S +index f30869e..dc7fd69 100644 +--- a/pixman/pixman-arm-neon-asm.S ++++ b/pixman/pixman-arm-neon-asm.S +@@ -388,6 +388,7 @@ generate_composite_function \ + vld1.16 {d4, d5}, [DST_R, :128]! + vst1.16 {d28, d29}, [DST_W, :128]! + pixman_composite_over_n_0565_process_pixblock_head ++ cache_preload 8, 8 + .endm + + .macro pixman_composite_over_n_0565_init +@@ -680,6 +681,7 @@ generate_composite_function_single_scanline \ + vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! + vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! + pixman_composite_over_8888_8888_process_pixblock_head ++ cache_preload 8, 8 + .endm + + .macro pixman_composite_over_n_8888_init +-- +1.6.6.1 + diff --git a/recipes/xorg-lib/pixman_0.18.0.bb b/recipes/xorg-lib/pixman_0.18.0.bb index b4fba422ef..783625bfa8 100644 --- a/recipes/xorg-lib/pixman_0.18.0.bb +++ b/recipes/xorg-lib/pixman_0.18.0.bb @@ -1,9 +1,8 @@ require pixman.inc -PR = "${INC_PR}.2" +PR = "${INC_PR}.3" DEFAULT_PREFERENCE = "-1" DEFAULT_PREFERENCE_shr = "2" -DEFAULT_PREFERENCE_angstrom = "2" SRC_URI += "\ file://0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch;patch=1 \ @@ -14,6 +13,7 @@ SRC_URI += "\ file://calloc.patch;patch=1 \ file://tls.patch;patch=1 \ file://565-scanline.patch;patch=1 \ + file://missing-cache-preload.diff;patch=1 \ " SRC_URI[archive.md5sum] = "a4fb870fc325be258089f1683642e976" diff --git a/recipes/xorg-lib/pixman_0.18.2.bb b/recipes/xorg-lib/pixman_0.18.2.bb new file mode 100644 index 0000000000..caac38b0af --- /dev/null +++ b/recipes/xorg-lib/pixman_0.18.2.bb @@ -0,0 +1,23 @@ +require pixman.inc +PR = "${INC_PR}.0" + +DEFAULT_PREFERENCE = "-1" +DEFAULT_PREFERENCE_angstrom = "2" + +SRC_URI[archive.md5sum] = "5d1378fa61610dd5d3c7e0111b2c5253" +SRC_URI[archive.sha256sum] = "80aee833b429d105d2c7593ef96993da04441b3b747084f1f3bfd7be594e1c45" + +SRC_URI += "\ + file://0001-Generic-C-implementation-of-pixman_blt-with-overlapp.patch;patch=1\ + file://0002-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch;patch=1\ + file://0003-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch;patch=1\ + file://0004-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch;patch=1\ + file://0005-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch;patch=1\ + file://0006-A-copy-paste-version-of-16bpp-bilinear-scanline-fetc.patch;patch=1\ + file://0007-ARM-added-missing-cache-preload.patch;patch=1\ +" + +NEON = " --disable-arm-neon " +NEON_armv7a = " " + +EXTRA_OECONF = "${NEON} --disable-gtk" |