--- /dev/null
+From e7ee43c39d2370716a4d011afa8f5067eced9899 Mon Sep 17 00:00:00 2001
+From: Cyril Brulebois <kibi@debian.org>
+Date: Wed, 17 Nov 2010 16:16:56 +0100
+Subject: [PATCH 02/24] Fix argument quoting for AC_INIT.
+
+One gets rid of this accordingly:
+| autoreconf -vfi
+| autoreconf: Entering directory `.'
+| autoreconf: configure.ac: not using Gettext
+| autoreconf: running: aclocal --force
+| configure.ac:61: warning: AC_INIT: not a literal: "pixman@lists.freedesktop.org"
+| autoreconf: configure.ac: tracing
+| configure.ac:61: warning: AC_INIT: not a literal: "pixman@lists.freedesktop.org"
+
+Signed-off-by: Cyril Brulebois <kibi@debian.org>
+---
+ configure.ac | 2 +-
+ 1 files changed, 1 insertions(+), 1 deletions(-)
+
+diff --git a/configure.ac b/configure.ac
+index db1da21..147e1bf 100644
+--- a/configure.ac
++++ b/configure.ac
+@@ -58,7 +58,7 @@ m4_define([pixman_micro], 3)
+
+ m4_define([pixman_version],[pixman_major.pixman_minor.pixman_micro])
+
+-AC_INIT(pixman, pixman_version, "pixman@lists.freedesktop.org", pixman)
++AC_INIT(pixman, pixman_version, [pixman@lists.freedesktop.org], pixman)
+ AM_INIT_AUTOMAKE([foreign dist-bzip2])
+
+ # Suppress verbose compile lines
+--
+1.6.6.1
+
--- /dev/null
+From 654961efe405ad1a7e54a77548ca8af322ecc1f8 Mon Sep 17 00:00:00 2001
+From: Alan Coopersmith <alan.coopersmith@oracle.com>
+Date: Sun, 21 Nov 2010 11:42:22 -0800
+Subject: [PATCH 03/24] Sun's copyrights belong to Oracle now
+
+Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
+---
+ COPYING | 2 +-
+ pixman/solaris-hwcap.mapfile | 2 +-
+ 2 files changed, 2 insertions(+), 2 deletions(-)
+
+diff --git a/COPYING b/COPYING
+index 3092a34..15f9517 100644
+--- a/COPYING
++++ b/COPYING
+@@ -18,7 +18,7 @@ possible. They may also add themselves to the list below.
+ * Copyright 2008 André Tupinambá
+ * Copyright 2008 Mozilla Corporation
+ * Copyright 2008 Frederic Plourde
+- * Copyright 2009 Sun Microsystems, Inc.
++ * Copyright 2009, Oracle and/or its affiliates. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+diff --git a/pixman/solaris-hwcap.mapfile b/pixman/solaris-hwcap.mapfile
+index 3605ca7..87efce1 100644
+--- a/pixman/solaris-hwcap.mapfile
++++ b/pixman/solaris-hwcap.mapfile
+@@ -1,6 +1,6 @@
+ ###############################################################################
+ #
+-# Copyright 2009 Sun Microsystems, Inc. All rights reserved.
++# Copyright 2009, Oracle and/or its affiliates. All rights reserved.
+ #
+ # Permission is hereby granted, free of charge, to any person obtaining a
+ # copy of this software and associated documentation files (the "Software"),
+--
+1.6.6.1
+
--- /dev/null
+From 4b5b5a2a832cd67f2a0ec231f75a2825b45571fa Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Mon, 15 Nov 2010 18:26:43 +0200
+Subject: [PATCH 04/24] C fast path for a1 fill operation
+
+Can be used as one of the solutions to fix bug
+https://bugs.freedesktop.org/show_bug.cgi?id=31604
+---
+ pixman/pixman-fast-path.c | 87 ++++++++++++++++++++++++++++++++++++++++++++-
+ pixman/pixman.c | 7 +++-
+ 2 files changed, 91 insertions(+), 3 deletions(-)
+
+diff --git a/pixman/pixman-fast-path.c b/pixman/pixman-fast-path.c
+index 5d5fa95..37dfbae 100644
+--- a/pixman/pixman-fast-path.c
++++ b/pixman/pixman-fast-path.c
+@@ -1334,7 +1334,11 @@ fast_composite_solid_fill (pixman_implementation_t *imp,
+
+ src = _pixman_image_get_solid (src_image, dst_image->bits.format);
+
+- if (dst_image->bits.format == PIXMAN_a8)
++ if (dst_image->bits.format == PIXMAN_a1)
++ {
++ src = src >> 31;
++ }
++ else if (dst_image->bits.format == PIXMAN_a8)
+ {
+ src = src >> 24;
+ }
+@@ -1655,6 +1659,7 @@ static const pixman_fast_path_t c_fast_paths[] =
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, x8r8g8b8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, a8b8g8r8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, x8b8g8r8, fast_composite_solid_fill),
++ PIXMAN_STD_FAST_PATH (SRC, solid, null, a1, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, a8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, r5g6b5, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, a8r8g8b8, fast_composite_src_x888_8888),
+@@ -1733,6 +1738,82 @@ static const pixman_fast_path_t c_fast_paths[] =
+ { PIXMAN_OP_NONE },
+ };
+
++#ifdef WORDS_BIGENDIAN
++#define A1_FILL_MASK(n, offs) (((1 << (n)) - 1) << (32 - (offs) - (n)))
++#else
++#define A1_FILL_MASK(n, offs) (((1 << (n)) - 1) << (offs))
++#endif
++
++static force_inline void
++pixman_fill1_line (uint32_t *dst, int offs, int width, int v)
++{
++ if (offs)
++ {
++ int leading_pixels = 32 - offs;
++ if (leading_pixels >= width)
++ {
++ if (v)
++ *dst |= A1_FILL_MASK (width, offs);
++ else
++ *dst &= ~A1_FILL_MASK (width, offs);
++ return;
++ }
++ else
++ {
++ if (v)
++ *dst++ |= A1_FILL_MASK (leading_pixels, offs);
++ else
++ *dst++ &= ~A1_FILL_MASK (leading_pixels, offs);
++ width -= leading_pixels;
++ }
++ }
++ while (width >= 32)
++ {
++ if (v)
++ *dst++ = 0xFFFFFFFF;
++ else
++ *dst++ = 0;
++ width -= 32;
++ }
++ if (width > 0)
++ {
++ if (v)
++ *dst |= A1_FILL_MASK (width, 0);
++ else
++ *dst &= ~A1_FILL_MASK (width, 0);
++ }
++}
++
++static void
++pixman_fill1 (uint32_t *bits,
++ int stride,
++ int x,
++ int y,
++ int width,
++ int height,
++ uint32_t xor)
++{
++ uint32_t *dst = bits + y * stride + (x >> 5);
++ int offs = x & 31;
++
++ if (xor & 1)
++ {
++ while (height--)
++ {
++ pixman_fill1_line (dst, offs, width, 1);
++ dst += stride;
++ }
++ }
++ else
++ {
++ while (height--)
++ {
++ pixman_fill1_line (dst, offs, width, 0);
++ dst += stride;
++ }
++ }
++}
++
+ static void
+ pixman_fill8 (uint32_t *bits,
+ int stride,
+@@ -1819,6 +1900,10 @@ fast_path_fill (pixman_implementation_t *imp,
+ {
+ switch (bpp)
+ {
++ case 1:
++ pixman_fill1 (bits, stride, x, y, width, height, xor);
++ break;
++
+ case 8:
+ pixman_fill8 (bits, stride, x, y, width, height, xor);
+ break;
+diff --git a/pixman/pixman.c b/pixman/pixman.c
+index 045c556..ec565f9 100644
+--- a/pixman/pixman.c
++++ b/pixman/pixman.c
+@@ -875,7 +875,8 @@ color_to_pixel (pixman_color_t * color,
+ format == PIXMAN_b8g8r8x8 ||
+ format == PIXMAN_r5g6b5 ||
+ format == PIXMAN_b5g6r5 ||
+- format == PIXMAN_a8))
++ format == PIXMAN_a8 ||
++ format == PIXMAN_a1))
+ {
+ return FALSE;
+ }
+@@ -895,7 +896,9 @@ color_to_pixel (pixman_color_t * color,
+ ((c & 0x000000ff) << 24);
+ }
+
+- if (format == PIXMAN_a8)
++ if (format == PIXMAN_a1)
++ c = c >> 31;
++ else if (format == PIXMAN_a8)
+ c = c >> 24;
+ else if (format == PIXMAN_r5g6b5 ||
+ format == PIXMAN_b5g6r5)
+--
+1.6.6.1
+
--- /dev/null
+From 98d08b37f17a3379d0ceff8bb7de8f943873fbd8 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Fri, 26 Nov 2010 08:55:49 +0200
+Subject: [PATCH 05/24] ARM: added 'neon_composite_over_n_8_8' fast path
+
+---
+ pixman/pixman-arm-neon-asm.S | 68 ++++++++++++++++++++++++++++++++++++++++++
+ pixman/pixman-arm-neon.c | 3 ++
+ 2 files changed, 71 insertions(+), 0 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 91ec27d..a3875ee 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -1203,6 +1203,74 @@ generate_composite_function \
+
+ /******************************************************************************/
+
++.macro pixman_composite_over_n_8_8_process_pixblock_head
++ vmull.u8 q0, d24, d8
++ vmull.u8 q1, d25, d8
++ vmull.u8 q6, d26, d8
++ vmull.u8 q7, d27, d8
++ vrshr.u16 q10, q0, #8
++ vrshr.u16 q11, q1, #8
++ vrshr.u16 q12, q6, #8
++ vrshr.u16 q13, q7, #8
++ vraddhn.u16 d0, q0, q10
++ vraddhn.u16 d1, q1, q11
++ vraddhn.u16 d2, q6, q12
++ vraddhn.u16 d3, q7, q13
++ vmvn.8 q12, q0
++ vmvn.8 q13, q1
++ vmull.u8 q8, d24, d4
++ vmull.u8 q9, d25, d5
++ vmull.u8 q10, d26, d6
++ vmull.u8 q11, d27, d7
++.endm
++
++.macro pixman_composite_over_n_8_8_process_pixblock_tail
++ vrshr.u16 q14, q8, #8
++ vrshr.u16 q15, q9, #8
++ vrshr.u16 q12, q10, #8
++ vrshr.u16 q13, q11, #8
++ vraddhn.u16 d28, q14, q8
++ vraddhn.u16 d29, q15, q9
++ vraddhn.u16 d30, q12, q10
++ vraddhn.u16 d31, q13, q11
++ vqadd.u8 q14, q0, q14
++ vqadd.u8 q15, q1, q15
++.endm
++
++/* TODO: expand macros and do better instructions scheduling */
++.macro pixman_composite_over_n_8_8_process_pixblock_tail_head
++ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
++ pixman_composite_over_n_8_8_process_pixblock_tail
++ vld1.8 {d24, d25, d26, d27}, [MASK]!
++ cache_preload 32, 32
++ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
++ pixman_composite_over_n_8_8_process_pixblock_head
++.endm
++
++.macro pixman_composite_over_n_8_8_init
++ add DUMMY, sp, #ARGS_STACK_OFFSET
++ vpush {d8-d15}
++ vld1.32 {d8[0]}, [DUMMY]
++ vdup.8 d8, d8[3]
++.endm
++
++.macro pixman_composite_over_n_8_8_cleanup
++ vpop {d8-d15}
++.endm
++
++generate_composite_function \
++ pixman_composite_over_n_8_8_asm_neon, 0, 8, 8, \
++ FLAG_DST_READWRITE, \
++ 32, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ pixman_composite_over_n_8_8_init, \
++ pixman_composite_over_n_8_8_cleanup, \
++ pixman_composite_over_n_8_8_process_pixblock_head, \
++ pixman_composite_over_n_8_8_process_pixblock_tail, \
++ pixman_composite_over_n_8_8_process_pixblock_tail_head
++
++/******************************************************************************/
++
+ .macro pixman_composite_over_n_8888_8888_ca_process_pixblock_head
+ /*
+ * 'combine_mask_ca' replacement
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index 2f82069..72ef75e 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -76,6 +76,8 @@ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_8888,
+ uint8_t, 1, uint32_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8888_8888_ca,
+ uint32_t, 1, uint32_t, 1)
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_8,
++ uint8_t, 1, uint8_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8,
+ uint8_t, 1, uint8_t, 1)
+
+@@ -235,6 +237,7 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, x8r8g8b8, neon_composite_src_0888_8888_rev),
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, r5g6b5, neon_composite_src_0888_0565_rev),
+ PIXMAN_STD_FAST_PATH (SRC, pixbuf, pixbuf, a8r8g8b8, neon_composite_src_pixbuf_8888),
++ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8, neon_composite_over_n_8_8),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, r5g6b5, neon_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, b5g6r5, neon_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8r8g8b8, neon_composite_over_n_8_8888),
+--
+1.6.6.1
+
--- /dev/null
+From 3be86a92ccab240859062a541cdb871d81c9501a Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Sun, 28 Nov 2010 21:45:06 +0200
+Subject: [PATCH 06/24] ARM: introduced 'fetch_mask_pixblock' macro to simplify code
+
+This macro hides the implementation details of pixels fetching
+for the mask image just like 'fetch_src_pixblock' does for the
+source image. This provides more possibilities for reusing the
+same code blocks in different compositing functions.
+
+This patch does not introduce any functional changes and the
+resulting code in the compiled object file is exactly the same.
+---
+ pixman/pixman-arm-neon-asm.S | 26 +++++++++++++-------------
+ pixman/pixman-arm-neon-asm.h | 5 +++++
+ 2 files changed, 18 insertions(+), 13 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index a3875ee..155a236 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -841,7 +841,7 @@ generate_composite_function \
+ pixman_composite_over_n_8_0565_process_pixblock_tail
+ vst1.16 {d28, d29}, [DST_W, :128]!
+ vld1.16 {d4, d5}, [DST_R, :128]!
+- vld1.8 {d24}, [MASK]!
++ fetch_mask_pixblock
+ cache_preload 8, 8
+ pixman_composite_over_n_8_0565_process_pixblock_head
+ .endm
+@@ -889,7 +889,7 @@ generate_composite_function \
+ pixman_composite_over_n_8_0565_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+- vld1.8 {d24}, [MASK]!
++ fetch_mask_pixblock
+ pixman_composite_over_n_8_0565_process_pixblock_head
+ vst1.16 {d28, d29}, [DST_W, :128]!
+ .endm
+@@ -1171,7 +1171,7 @@ generate_composite_function \
+ pixman_composite_over_n_8_8888_process_pixblock_tail
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+- vld1.8 {d24}, [MASK]!
++ fetch_mask_pixblock
+ cache_preload 8, 8
+ pixman_composite_over_n_8_8888_process_pixblock_head
+ .endm
+@@ -1241,7 +1241,7 @@ generate_composite_function \
+ .macro pixman_composite_over_n_8_8_process_pixblock_tail_head
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ pixman_composite_over_n_8_8_process_pixblock_tail
+- vld1.8 {d24, d25, d26, d27}, [MASK]!
++ fetch_mask_pixblock
+ cache_preload 32, 32
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ pixman_composite_over_n_8_8_process_pixblock_head
+@@ -1341,7 +1341,7 @@ generate_composite_function \
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q6, q10
+ vraddhn.u16 d31, q7, q11
+- vld4.8 {d24, d25, d26, d27}, [MASK]!
++ fetch_mask_pixblock
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+ cache_preload 8, 8
+@@ -1405,7 +1405,7 @@ generate_composite_function \
+ pixman_composite_add_n_8_8_process_pixblock_tail
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+- vld1.8 {d24, d25, d26, d27}, [MASK]!
++ fetch_mask_pixblock
+ cache_preload 32, 32
+ pixman_composite_add_n_8_8_process_pixblock_head
+ .endm
+@@ -1462,7 +1462,7 @@ generate_composite_function \
+ pixman_composite_add_8_8_8_process_pixblock_tail
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+- vld1.8 {d24, d25, d26, d27}, [MASK]!
++ fetch_mask_pixblock
+ fetch_src_pixblock
+ cache_preload 32, 32
+ pixman_composite_add_8_8_8_process_pixblock_head
+@@ -1515,7 +1515,7 @@ generate_composite_function \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+- vld4.8 {d24, d25, d26, d27}, [MASK]!
++ fetch_mask_pixblock
+ fetch_src_pixblock
+ cache_preload 8, 8
+ pixman_composite_add_8888_8888_8888_process_pixblock_head
+@@ -1587,7 +1587,7 @@ generate_composite_function_single_scanline \
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+- vld4.8 {d12, d13, d14, d15}, [MASK]!
++ fetch_mask_pixblock
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ .endm
+@@ -1658,7 +1658,7 @@ generate_composite_function \
+ pixman_composite_over_8888_n_8888_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+- vld4.8 {d12, d13, d14, d15}, [MASK]!
++ fetch_mask_pixblock
+ pixman_composite_over_8888_n_8888_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ .endm
+@@ -1700,7 +1700,7 @@ generate_composite_function_single_scanline \
+ pixman_composite_over_8888_n_8888_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+- vld1.8 {d15}, [MASK]!
++ fetch_mask_pixblock
+ pixman_composite_over_8888_n_8888_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ .endm
+@@ -1917,7 +1917,7 @@ generate_composite_function \
+
+ /* TODO: expand macros and do better instructions scheduling */
+ .macro pixman_composite_over_0565_8_0565_process_pixblock_tail_head
+- vld1.8 {d15}, [MASK]!
++ fetch_mask_pixblock
+ pixman_composite_over_0565_8_0565_process_pixblock_tail
+ fetch_src_pixblock
+ vld1.16 {d10, d11}, [DST_R, :128]!
+@@ -1969,7 +1969,7 @@ generate_composite_function \
+
+ /* TODO: expand macros and do better instructions scheduling */
+ .macro pixman_composite_add_0565_8_0565_process_pixblock_tail_head
+- vld1.8 {d15}, [MASK]!
++ fetch_mask_pixblock
+ pixman_composite_add_0565_8_0565_process_pixblock_tail
+ fetch_src_pixblock
+ vld1.16 {d10, d11}, [DST_R, :128]!
+diff --git a/pixman/pixman-arm-neon-asm.h b/pixman/pixman-arm-neon-asm.h
+index c75bdc3..24fa361 100644
+--- a/pixman/pixman-arm-neon-asm.h
++++ b/pixman/pixman-arm-neon-asm.h
+@@ -431,6 +431,11 @@
+ .endif
+ .endm
+
++.macro fetch_mask_pixblock
++ pixld pixblock_size, mask_bpp, \
++ (mask_basereg - pixblock_size * mask_bpp / 64), MASK
++.endm
++
+ /*
+ * Macro which is used to process leading pixels until destination
+ * pointer is properly aligned (at 16 bytes boundary). When destination
+--
+1.6.6.1
+
--- /dev/null
+From e6814837a6ccd3e4db329e0131eaf2055d2c864b Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Fri, 26 Nov 2010 17:06:58 +0200
+Subject: [PATCH 07/24] ARM: better NEON instructions scheduling for over_n_8_0565
+
+Code rearranged to get better instructions scheduling for ARM Cortex-A8/A9.
+Now it is ~30% faster for the pixel data in L1 cache and makes better use
+of memory bandwidth when running at lower clock frequencies (ex. 500MHz).
+Also register d24 (pixels from the mask image) is now not clobbered by
+supplementary macros, which allows to reuse them for the other variants
+of compositing operations later.
+
+Benchmark from ARM Cortex-A8 @500MHz:
+
+== before ==
+
+ over_n_8_0565 = L1: 63.90 L2: 63.15 M: 60.97 ( 73.53%)
+ HT: 28.89 VT: 24.14 R: 21.33 RT: 6.78 ( 67Kops/s)
+
+== after ==
+
+ over_n_8_0565 = L1: 82.64 L2: 75.19 M: 71.52 ( 84.14%)
+ HT: 30.49 VT: 25.56 R: 22.36 RT: 6.89 ( 68Kops/s)
+---
+ pixman/pixman-arm-neon-asm.S | 120 +++++++++++++++++++++++++++---------------
+ 1 files changed, 77 insertions(+), 43 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 155a236..ffffc1c 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -792,58 +792,92 @@ generate_composite_function \
+ /******************************************************************************/
+
+ .macro pixman_composite_over_n_8_0565_process_pixblock_head
+- /* in */
+- vmull.u8 q0, d24, d8
+- vmull.u8 q1, d24, d9
+- vmull.u8 q6, d24, d10
+- vmull.u8 q7, d24, d11
+- vrshr.u16 q10, q0, #8
+- vrshr.u16 q11, q1, #8
+- vrshr.u16 q12, q6, #8
+- vrshr.u16 q13, q7, #8
+- vraddhn.u16 d0, q0, q10
+- vraddhn.u16 d1, q1, q11
+- vraddhn.u16 d2, q6, q12
+- vraddhn.u16 d3, q7, q13
+-
+- vshrn.u16 d6, q2, #8
+- vshrn.u16 d7, q2, #3
+- vsli.u16 q2, q2, #5
+- vsri.u8 d6, d6, #5
+- vmvn.8 d3, d3
+- vsri.u8 d7, d7, #6
+- vshrn.u16 d30, q2, #2
+- /* now do alpha blending */
+- vmull.u8 q10, d3, d6
+- vmull.u8 q11, d3, d7
+- vmull.u8 q12, d3, d30
+- vrshr.u16 q13, q10, #8
+- vrshr.u16 q3, q11, #8
+- vrshr.u16 q15, q12, #8
+- vraddhn.u16 d20, q10, q13
+- vraddhn.u16 d23, q11, q3
+- vraddhn.u16 d22, q12, q15
++ vmull.u8 q0, d24, d8 /* IN for SRC pixels (part1) */
++ vmull.u8 q1, d24, d9
++ vmull.u8 q6, d24, d10
++ vmull.u8 q7, d24, d11
++ vshrn.u16 d6, q2, #8 /* convert DST_R data to 32-bpp (part1) */
++ vshrn.u16 d7, q2, #3
++ vsli.u16 q2, q2, #5
++ vrshr.u16 q8, q0, #8 /* IN for SRC pixels (part2) */
++ vrshr.u16 q9, q1, #8
++ vrshr.u16 q10, q6, #8
++ vrshr.u16 q11, q7, #8
++ vraddhn.u16 d0, q0, q8
++ vraddhn.u16 d1, q1, q9
++ vraddhn.u16 d2, q6, q10
++ vraddhn.u16 d3, q7, q11
++ vsri.u8 d6, d6, #5 /* convert DST_R data to 32-bpp (part2) */
++ vsri.u8 d7, d7, #6
++ vmvn.8 d3, d3
++ vshrn.u16 d30, q2, #2
++ vmull.u8 q8, d3, d6 /* now do alpha blending */
++ vmull.u8 q9, d3, d7
++ vmull.u8 q10, d3, d30
+ .endm
+
+ .macro pixman_composite_over_n_8_0565_process_pixblock_tail
+- vqadd.u8 d16, d2, d20
+- vqadd.u8 q9, q0, q11
+- /* convert to r5g6b5 */
+- vshll.u8 q14, d16, #8
+- vshll.u8 q8, d19, #8
+- vshll.u8 q9, d18, #8
+- vsri.u16 q14, q8, #5
+- vsri.u16 q14, q9, #11
++ /* 3 cycle bubble (after vmull.u8) */
++ vrshr.u16 q13, q8, #8
++ vrshr.u16 q11, q9, #8
++ vrshr.u16 q15, q10, #8
++ vraddhn.u16 d16, q8, q13
++ vraddhn.u16 d27, q9, q11
++ vraddhn.u16 d26, q10, q15
++ vqadd.u8 d16, d2, d16
++ /* 1 cycle bubble */
++ vqadd.u8 q9, q0, q13
++ vshll.u8 q14, d16, #8 /* convert to 16bpp */
++ vshll.u8 q8, d19, #8
++ vshll.u8 q9, d18, #8
++ vsri.u16 q14, q8, #5
++ /* 1 cycle bubble */
++ vsri.u16 q14, q9, #11
+ .endm
+
+-/* TODO: expand macros and do better instructions scheduling */
+ .macro pixman_composite_over_n_8_0565_process_pixblock_tail_head
+- pixman_composite_over_n_8_0565_process_pixblock_tail
+- vst1.16 {d28, d29}, [DST_W, :128]!
+ vld1.16 {d4, d5}, [DST_R, :128]!
++ vshrn.u16 d6, q2, #8
+ fetch_mask_pixblock
++ vshrn.u16 d7, q2, #3
++ fetch_src_pixblock
++ vmull.u8 q6, d24, d10
++ vrshr.u16 q13, q8, #8
++ vrshr.u16 q11, q9, #8
++ vrshr.u16 q15, q10, #8
++ vraddhn.u16 d16, q8, q13
++ vraddhn.u16 d27, q9, q11
++ vraddhn.u16 d26, q10, q15
++ vqadd.u8 d16, d2, d16
++ vmull.u8 q1, d24, d9
++ vqadd.u8 q9, q0, q13
++ vshll.u8 q14, d16, #8
++ vmull.u8 q0, d24, d8
++ vshll.u8 q8, d19, #8
++ vshll.u8 q9, d18, #8
++ vsri.u16 q14, q8, #5
++ vmull.u8 q7, d24, d11
++ vsri.u16 q14, q9, #11
++
+ cache_preload 8, 8
+- pixman_composite_over_n_8_0565_process_pixblock_head
++
++ vsli.u16 q2, q2, #5
++ vrshr.u16 q8, q0, #8
++ vrshr.u16 q9, q1, #8
++ vrshr.u16 q10, q6, #8
++ vrshr.u16 q11, q7, #8
++ vraddhn.u16 d0, q0, q8
++ vraddhn.u16 d1, q1, q9
++ vraddhn.u16 d2, q6, q10
++ vraddhn.u16 d3, q7, q11
++ vsri.u8 d6, d6, #5
++ vsri.u8 d7, d7, #6
++ vmvn.8 d3, d3
++ vshrn.u16 d30, q2, #2
++ vst1.16 {d28, d29}, [DST_W, :128]!
++ vmull.u8 q8, d3, d6
++ vmull.u8 q9, d3, d7
++ vmull.u8 q10, d3, d30
+ .endm
+
+ /*
+--
+1.6.6.1
+
--- /dev/null
+From a7c36681c0c1955ff9110b81f1789e56abb10a95 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Sat, 27 Nov 2010 03:53:12 +0200
+Subject: [PATCH 08/24] ARM: added 'neon_composite_over_8888_n_0565' fast path
+
+---
+ pixman/pixman-arm-neon-asm.S | 28 ++++++++++++++++++++++++++++
+ pixman/pixman-arm-neon.c | 4 ++++
+ 2 files changed, 32 insertions(+), 0 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index ffffc1c..3e52a49 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -917,6 +917,34 @@ generate_composite_function \
+
+ /******************************************************************************/
+
++.macro pixman_composite_over_8888_n_0565_init
++ add DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
++ vpush {d8-d15}
++ vld1.32 {d24[0]}, [DUMMY]
++ vdup.8 d24, d24[3]
++.endm
++
++.macro pixman_composite_over_8888_n_0565_cleanup
++ vpop {d8-d15}
++.endm
++
++generate_composite_function \
++ pixman_composite_over_8888_n_0565_asm_neon, 32, 0, 16, \
++ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
++ 8, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ pixman_composite_over_8888_n_0565_init, \
++ pixman_composite_over_8888_n_0565_cleanup, \
++ pixman_composite_over_n_8_0565_process_pixblock_head, \
++ pixman_composite_over_n_8_0565_process_pixblock_tail, \
++ pixman_composite_over_n_8_0565_process_pixblock_tail_head, \
++ 28, /* dst_w_basereg */ \
++ 4, /* dst_r_basereg */ \
++ 8, /* src_basereg */ \
++ 24 /* mask_basereg */
++
++/******************************************************************************/
++
+ /* TODO: expand macros and do better instructions scheduling */
+ .macro pixman_composite_over_8888_8_0565_process_pixblock_tail_head
+ vld1.16 {d4, d5}, [DST_R, :128]!
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index 72ef75e..8156bbb 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -83,6 +83,8 @@ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8,
+
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_0565,
++ uint32_t, 1, uint16_t, 1)
+
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8_8_8,
+ uint8_t, 1, uint8_t, 1, uint8_t, 1)
+@@ -253,6 +255,8 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, x8b8g8r8, neon_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, a8r8g8b8, neon_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, x8r8g8b8, neon_composite_over_8888_n_8888),
++ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, r5g6b5, neon_composite_over_8888_n_0565),
++ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, b5g6r5, neon_composite_over_8888_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, a8r8g8b8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, x8r8g8b8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, a8b8g8r8, neon_composite_over_8888_8_8888),
+--
+1.6.6.1
+
--- /dev/null
+From 3990931bf6197eff1cec06cf24bce53ddf9a539a Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Sat, 27 Nov 2010 04:47:39 +0200
+Subject: [PATCH 09/24] ARM: reuse common NEON code for over_{n_8|8888_n|8888_8}_0565
+
+Renamed suppementary macros from 'over_n_8_0565' to 'over_8888_8_0565',
+because they can actually support all variants of this operation:
+over_8888_8_0565/over_n_8_0565/over_8888_n_0565.
+
+Also 'over_8888_8_0565' now uses more optimized common code instead of its
+own variant, improving performance a bit. Even though this operation is
+still memory bandwidth limited, scaled variants of these fast paths may
+put more stress on CPU later.
+
+Benchmarked on ARM Cortex-A8 @500MHz:
+
+== before ==
+
+ over_8888_8_0565 = L1: 67.10 L2: 53.82 M: 44.70 (105.17%)
+ HT: 18.73 VT: 16.91 R: 14.25 RT: 4.80 (52Kops/s)
+
+== after ==
+
+ over_8888_8_0565 = L1: 77.83 L2: 58.14 M: 44.82 (105.52%)
+ HT: 20.58 VT: 17.44 R: 15.05 RT: 4.88 (52Kops/s)
+---
+ pixman/pixman-arm-neon-asm.S | 61 +++++++++++++++++------------------------
+ 1 files changed, 25 insertions(+), 36 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 3e52a49..4175144 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -791,7 +791,7 @@ generate_composite_function \
+
+ /******************************************************************************/
+
+-.macro pixman_composite_over_n_8_0565_process_pixblock_head
++.macro pixman_composite_over_8888_8_0565_process_pixblock_head
+ vmull.u8 q0, d24, d8 /* IN for SRC pixels (part1) */
+ vmull.u8 q1, d24, d9
+ vmull.u8 q6, d24, d10
+@@ -816,7 +816,7 @@ generate_composite_function \
+ vmull.u8 q10, d3, d30
+ .endm
+
+-.macro pixman_composite_over_n_8_0565_process_pixblock_tail
++.macro pixman_composite_over_8888_8_0565_process_pixblock_tail
+ /* 3 cycle bubble (after vmull.u8) */
+ vrshr.u16 q13, q8, #8
+ vrshr.u16 q11, q9, #8
+@@ -835,7 +835,7 @@ generate_composite_function \
+ vsri.u16 q14, q9, #11
+ .endm
+
+-.macro pixman_composite_over_n_8_0565_process_pixblock_tail_head
++.macro pixman_composite_over_8888_8_0565_process_pixblock_tail_head
+ vld1.16 {d4, d5}, [DST_R, :128]!
+ vshrn.u16 d6, q2, #8
+ fetch_mask_pixblock
+@@ -880,6 +880,23 @@ generate_composite_function \
+ vmull.u8 q10, d3, d30
+ .endm
+
++generate_composite_function \
++ pixman_composite_over_8888_8_0565_asm_neon, 32, 8, 16, \
++ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
++ 8, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ default_init_need_all_regs, \
++ default_cleanup_need_all_regs, \
++ pixman_composite_over_8888_8_0565_process_pixblock_head, \
++ pixman_composite_over_8888_8_0565_process_pixblock_tail, \
++ pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \
++ 28, /* dst_w_basereg */ \
++ 4, /* dst_r_basereg */ \
++ 8, /* src_basereg */ \
++ 24 /* mask_basereg */
++
++/******************************************************************************/
++
+ /*
+ * This function needs a special initialization of solid mask.
+ * Solid source pixel data is fetched from stack at ARGS_STACK_OFFSET
+@@ -911,9 +928,9 @@ generate_composite_function \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_n_8_0565_init, \
+ pixman_composite_over_n_8_0565_cleanup, \
+- pixman_composite_over_n_8_0565_process_pixblock_head, \
+- pixman_composite_over_n_8_0565_process_pixblock_tail, \
+- pixman_composite_over_n_8_0565_process_pixblock_tail_head
++ pixman_composite_over_8888_8_0565_process_pixblock_head, \
++ pixman_composite_over_8888_8_0565_process_pixblock_tail, \
++ pixman_composite_over_8888_8_0565_process_pixblock_tail_head
+
+ /******************************************************************************/
+
+@@ -935,36 +952,8 @@ generate_composite_function \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_8888_n_0565_init, \
+ pixman_composite_over_8888_n_0565_cleanup, \
+- pixman_composite_over_n_8_0565_process_pixblock_head, \
+- pixman_composite_over_n_8_0565_process_pixblock_tail, \
+- pixman_composite_over_n_8_0565_process_pixblock_tail_head, \
+- 28, /* dst_w_basereg */ \
+- 4, /* dst_r_basereg */ \
+- 8, /* src_basereg */ \
+- 24 /* mask_basereg */
+-
+-/******************************************************************************/
+-
+-/* TODO: expand macros and do better instructions scheduling */
+-.macro pixman_composite_over_8888_8_0565_process_pixblock_tail_head
+- vld1.16 {d4, d5}, [DST_R, :128]!
+- pixman_composite_over_n_8_0565_process_pixblock_tail
+- fetch_src_pixblock
+- cache_preload 8, 8
+- fetch_mask_pixblock
+- pixman_composite_over_n_8_0565_process_pixblock_head
+- vst1.16 {d28, d29}, [DST_W, :128]!
+-.endm
+-
+-generate_composite_function \
+- pixman_composite_over_8888_8_0565_asm_neon, 32, 8, 16, \
+- FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+- 8, /* number of pixels, processed in a single block */ \
+- 5, /* prefetch distance */ \
+- default_init_need_all_regs, \
+- default_cleanup_need_all_regs, \
+- pixman_composite_over_n_8_0565_process_pixblock_head, \
+- pixman_composite_over_n_8_0565_process_pixblock_tail, \
++ pixman_composite_over_8888_8_0565_process_pixblock_head, \
++ pixman_composite_over_8888_8_0565_process_pixblock_tail, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+--
+1.6.6.1
+
--- /dev/null
+From 6d2f7f981b52b41f4321071c325babcf792bd666 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Sat, 27 Nov 2010 15:53:54 +0200
+Subject: [PATCH 10/24] ARM: added 'neon_composite_over_0565_n_0565' fast path
+
+---
+ pixman/pixman-arm-neon-asm.S | 28 ++++++++++++++++++++++++++++
+ pixman/pixman-arm-neon.c | 4 ++++
+ 2 files changed, 32 insertions(+), 0 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 4175144..81c0a34 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -1994,6 +1994,34 @@ generate_composite_function \
+
+ /******************************************************************************/
+
++.macro pixman_composite_over_0565_n_0565_init
++ add DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
++ vpush {d8-d15}
++ vld1.32 {d15[0]}, [DUMMY]
++ vdup.8 d15, d15[3]
++.endm
++
++.macro pixman_composite_over_0565_n_0565_cleanup
++ vpop {d8-d15}
++.endm
++
++generate_composite_function \
++ pixman_composite_over_0565_n_0565_asm_neon, 16, 0, 16, \
++ FLAG_DST_READWRITE, \
++ 8, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ pixman_composite_over_0565_n_0565_init, \
++ pixman_composite_over_0565_n_0565_cleanup, \
++ pixman_composite_over_0565_8_0565_process_pixblock_head, \
++ pixman_composite_over_0565_8_0565_process_pixblock_tail, \
++ pixman_composite_over_0565_8_0565_process_pixblock_tail_head, \
++ 28, /* dst_w_basereg */ \
++ 10, /* dst_r_basereg */ \
++ 8, /* src_basereg */ \
++ 15 /* mask_basereg */
++
++/******************************************************************************/
++
+ .macro pixman_composite_add_0565_8_0565_process_pixblock_head
+ /* mask is in d15 */
+ convert_0565_to_x888 q4, d2, d1, d0
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index 8156bbb..b01c3e0 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -85,6 +85,8 @@ PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_0565,
+ uint32_t, 1, uint16_t, 1)
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_0565_n_0565,
++ uint16_t, 1, uint16_t, 1)
+
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8_8_8,
+ uint8_t, 1, uint8_t, 1, uint8_t, 1)
+@@ -257,6 +259,8 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, x8r8g8b8, neon_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, r5g6b5, neon_composite_over_8888_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, b5g6r5, neon_composite_over_8888_n_0565),
++ PIXMAN_STD_FAST_PATH (OVER, r5g6b5, solid, r5g6b5, neon_composite_over_0565_n_0565),
++ PIXMAN_STD_FAST_PATH (OVER, b5g6r5, solid, b5g6r5, neon_composite_over_0565_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, a8r8g8b8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, x8r8g8b8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, a8b8g8r8, neon_composite_over_8888_8_8888),
+--
+1.6.6.1
+
--- /dev/null
+From c3f48b6aa2f9354af02ffc8c938ec6753fdcbde3 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Sun, 28 Nov 2010 22:05:53 +0200
+Subject: [PATCH 11/24] ARM: added 'neon_composite_add_8888_8_8888' fast path
+
+---
+ pixman/pixman-arm-neon-asm.S | 17 +++++++++++++++++
+ pixman/pixman-arm-neon.c | 4 ++++
+ 2 files changed, 21 insertions(+), 0 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 81c0a34..11ef166 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -1595,6 +1595,23 @@ generate_composite_function_single_scanline \
+
+ /******************************************************************************/
+
++generate_composite_function \
++ pixman_composite_add_8888_8_8888_asm_neon, 32, 8, 32, \
++ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
++ 8, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ default_init, \
++ default_cleanup, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \
++ 28, /* dst_w_basereg */ \
++ 4, /* dst_r_basereg */ \
++ 0, /* src_basereg */ \
++ 27 /* mask_basereg */
++
++/******************************************************************************/
++
+ .macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* destination data in {d4, d5, d6, d7} */
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index b01c3e0..eaf9787 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -92,6 +92,8 @@ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8_8_8,
+ uint8_t, 1, uint8_t, 1, uint8_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_0565_8_0565,
+ uint16_t, 1, uint8_t, 1, uint16_t, 1)
++PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8888_8_8888,
++ uint32_t, 1, uint8_t, 1, uint32_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8888_8888_8888,
+ uint32_t, 1, uint32_t, 1, uint32_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_8888_8_8888,
+@@ -282,6 +284,8 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
+ PIXMAN_STD_FAST_PATH (ADD, a8, a8, a8, neon_composite_add_8_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, r5g6b5, a8, r5g6b5, neon_composite_add_0565_8_0565),
+ PIXMAN_STD_FAST_PATH (ADD, b5g6r5, a8, b5g6r5, neon_composite_add_0565_8_0565),
++ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8, a8r8g8b8, neon_composite_add_8888_8_8888),
++ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, a8, a8b8g8r8, neon_composite_add_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8r8g8b8, a8r8g8b8, neon_composite_add_8888_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, neon_composite_add_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, neon_composite_add_8888_8888),
+--
+1.6.6.1
+
--- /dev/null
+From 1fba7790367d7b726d05a33bbbcebe10b9280a31 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Mon, 29 Nov 2010 02:10:22 +0200
+Subject: [PATCH 12/24] ARM: better NEON instructions scheduling for add_8888_8888_8888
+
+Provides a minor performance improvement by using pipelining and hiding
+instructions latencies. Also do not clobber d0-d3 registers (source
+image pixels) while doing calculations in order to allow the use of
+the same macro for add_n_8_8888 fast path later.
+
+Benchmark from ARM Cortex-A8 @500MHz:
+
+== before ==
+
+ add_8888_8888_8888 = L1: 95.94 L2: 42.27 M: 25.60 (121.09%)
+ HT: 14.54 VT: 13.13 R: 12.77 RT: 4.49 (48Kops/s)
+ add_8888_8_8888 = L1: 104.51 L2: 57.81 M: 36.06 (106.62%)
+ HT: 19.24 VT: 16.45 R: 14.71 RT: 4.80 (51Kops/s)
+
+== after ==
+
+ add_8888_8888_8888 = L1: 106.66 L2: 47.82 M: 27.32 (129.30%)
+ HT: 15.44 VT: 13.96 R: 12.86 RT: 4.48 (48Kops/s)
+ add_8888_8_8888 = L1: 107.72 L2: 61.02 M: 38.26 (113.16%)
+ HT: 19.48 VT: 16.72 R: 14.82 RT: 4.80 (51Kops/s)
+---
+ pixman/pixman-arm-neon-asm.S | 52 +++++++++++++++++++++++++++--------------
+ 1 files changed, 34 insertions(+), 18 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 11ef166..829ef84 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -1542,34 +1542,50 @@ generate_composite_function \
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* destination data in {d4, d5, d6, d7} */
+ /* mask in {d24, d25, d26, d27} */
+- vmull.u8 q8, d27, d0
+- vmull.u8 q9, d27, d1
++ vmull.u8 q8, d27, d0
++ vmull.u8 q9, d27, d1
+ vmull.u8 q10, d27, d2
+ vmull.u8 q11, d27, d3
+- vrshr.u16 q0, q8, #8
+- vrshr.u16 q1, q9, #8
+- vrshr.u16 q12, q10, #8
+- vrshr.u16 q13, q11, #8
+- vraddhn.u16 d0, q0, q8
+- vraddhn.u16 d1, q1, q9
+- vraddhn.u16 d2, q12, q10
+- vraddhn.u16 d3, q13, q11
+- vqadd.u8 q14, q0, q2
+- vqadd.u8 q15, q1, q3
++ /* 1 cycle bubble */
++ vrsra.u16 q8, q8, #8
++ vrsra.u16 q9, q9, #8
++ vrsra.u16 q10, q10, #8
++ vrsra.u16 q11, q11, #8
+ .endm
+
+ .macro pixman_composite_add_8888_8888_8888_process_pixblock_tail
++ /* 2 cycle bubble */
++ vrshrn.u16 d28, q8, #8
++ vrshrn.u16 d29, q9, #8
++ vrshrn.u16 d30, q10, #8
++ vrshrn.u16 d31, q11, #8
++ vqadd.u8 q14, q2, q14
++ /* 1 cycle bubble */
++ vqadd.u8 q15, q3, q15
+ .endm
+
+-/* TODO: expand macros and do better instructions scheduling */
+ .macro pixman_composite_add_8888_8888_8888_process_pixblock_tail_head
+- pixman_composite_add_8888_8888_8888_process_pixblock_tail
+- vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+- vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+- fetch_mask_pixblock
+ fetch_src_pixblock
++ vrshrn.u16 d28, q8, #8
++ fetch_mask_pixblock
++ vrshrn.u16 d29, q9, #8
++ vmull.u8 q8, d27, d0
++ vrshrn.u16 d30, q10, #8
++ vmull.u8 q9, d27, d1
++ vrshrn.u16 d31, q11, #8
++ vmull.u8 q10, d27, d2
++ vqadd.u8 q14, q2, q14
++ vmull.u8 q11, d27, d3
++ vqadd.u8 q15, q3, q15
++ vrsra.u16 q8, q8, #8
++ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
++ vrsra.u16 q9, q9, #8
++ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
++ vrsra.u16 q10, q10, #8
++
+ cache_preload 8, 8
+- pixman_composite_add_8888_8888_8888_process_pixblock_head
++
++ vrsra.u16 q11, q11, #8
+ .endm
+
+ generate_composite_function \
+--
+1.6.6.1
+
--- /dev/null
+From b066b520dfaf0a9f4d1bc9a73c789091e9ce7cc8 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Mon, 29 Nov 2010 02:38:52 +0200
+Subject: [PATCH 13/24] ARM: added 'neon_composite_add_n_8_8888' fast path
+
+---
+ pixman/pixman-arm-neon-asm.S | 29 +++++++++++++++++++++++++++++
+ pixman/pixman-arm-neon.c | 4 ++++
+ 2 files changed, 33 insertions(+), 0 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 829ef84..dd6f2c5 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -1628,6 +1628,35 @@ generate_composite_function \
+
+ /******************************************************************************/
+
++.macro pixman_composite_add_n_8_8888_init
++ add DUMMY, sp, #ARGS_STACK_OFFSET
++ vld1.32 {d3[0]}, [DUMMY]
++ vdup.8 d0, d3[0]
++ vdup.8 d1, d3[1]
++ vdup.8 d2, d3[2]
++ vdup.8 d3, d3[3]
++.endm
++
++.macro pixman_composite_add_n_8_8888_cleanup
++.endm
++
++generate_composite_function \
++ pixman_composite_add_n_8_8888_asm_neon, 0, 8, 32, \
++ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
++ 8, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ pixman_composite_add_n_8_8888_init, \
++ pixman_composite_add_n_8_8888_cleanup, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \
++ 28, /* dst_w_basereg */ \
++ 4, /* dst_r_basereg */ \
++ 0, /* src_basereg */ \
++ 27 /* mask_basereg */
++
++/******************************************************************************/
++
+ .macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* destination data in {d4, d5, d6, d7} */
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index eaf9787..5ad58bd 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -80,6 +80,8 @@ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_8,
+ uint8_t, 1, uint8_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8,
+ uint8_t, 1, uint8_t, 1)
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8888,
++ uint8_t, 1, uint32_t, 1)
+
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
+@@ -281,6 +283,8 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, null, a8r8g8b8, neon_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, null, a8b8g8r8, neon_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8, neon_composite_add_n_8_8),
++ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8r8g8b8, neon_composite_add_n_8_8888),
++ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8b8g8r8, neon_composite_add_n_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8, a8, a8, neon_composite_add_8_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, r5g6b5, a8, r5g6b5, neon_composite_add_0565_8_0565),
+ PIXMAN_STD_FAST_PATH (ADD, b5g6r5, a8, b5g6r5, neon_composite_add_0565_8_0565),
+--
+1.6.6.1
+
--- /dev/null
+From f6843e3797eea7e4aed7614b1086f5cefc06c0f9 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Mon, 29 Nov 2010 03:31:32 +0200
+Subject: [PATCH 14/24] ARM: added 'neon_composite_add_8888_n_8888' fast path
+
+---
+ pixman/pixman-arm-neon-asm.S | 26 ++++++++++++++++++++++++++
+ pixman/pixman-arm-neon.c | 4 ++++
+ 2 files changed, 30 insertions(+), 0 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index dd6f2c5..2c0fd37 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -1657,6 +1657,32 @@ generate_composite_function \
+
+ /******************************************************************************/
+
++.macro pixman_composite_add_8888_n_8888_init
++ add DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
++ vld1.32 {d27[0]}, [DUMMY]
++ vdup.8 d27, d27[3]
++.endm
++
++.macro pixman_composite_add_8888_n_8888_cleanup
++.endm
++
++generate_composite_function \
++ pixman_composite_add_8888_n_8888_asm_neon, 32, 0, 32, \
++ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
++ 8, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ pixman_composite_add_8888_n_8888_init, \
++ pixman_composite_add_8888_n_8888_cleanup, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
++ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \
++ 28, /* dst_w_basereg */ \
++ 4, /* dst_r_basereg */ \
++ 0, /* src_basereg */ \
++ 27 /* mask_basereg */
++
++/******************************************************************************/
++
+ .macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* destination data in {d4, d5, d6, d7} */
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index 5ad58bd..f0dc111 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -89,6 +89,8 @@ PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_0565,
+ uint32_t, 1, uint16_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_0565_n_0565,
+ uint16_t, 1, uint16_t, 1)
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, add_8888_n_8888,
++ uint32_t, 1, uint32_t, 1)
+
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8_8_8,
+ uint8_t, 1, uint8_t, 1, uint8_t, 1)
+@@ -291,6 +293,8 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8, a8r8g8b8, neon_composite_add_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, a8, a8b8g8r8, neon_composite_add_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8r8g8b8, a8r8g8b8, neon_composite_add_8888_8888_8888),
++ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, solid, a8r8g8b8, neon_composite_add_8888_n_8888),
++ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, solid, a8b8g8r8, neon_composite_add_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, neon_composite_add_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, neon_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, neon_composite_add_8888_8888),
+--
+1.6.6.1
+
--- /dev/null
+From af7a69d90ea2b43a4e850870727723d719f09a1c Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Mon, 29 Nov 2010 09:00:46 +0200
+Subject: [PATCH 15/24] ARM: added flags parameter to some asm fast path wrapper macros
+
+Not all types of operations can be skipped when having transparent
+solid source or transparent solid mask. Add an extra flags parameter
+for providing this information to the wrappers.
+---
+ pixman/pixman-arm-common.h | 15 +++++++++------
+ pixman/pixman-arm-neon.c | 26 +++++++++++++-------------
+ pixman/pixman-arm-simd.c | 4 ++--
+ 3 files changed, 24 insertions(+), 21 deletions(-)
+
+diff --git a/pixman/pixman-arm-common.h b/pixman/pixman-arm-common.h
+index 2cff6c8..66f448d 100644
+--- a/pixman/pixman-arm-common.h
++++ b/pixman/pixman-arm-common.h
+@@ -47,6 +47,9 @@
+ * or mask), the corresponding stride argument is unused.
+ */
+
++#define SKIP_ZERO_SRC 1
++#define SKIP_ZERO_MASK 2
++
+ #define PIXMAN_ARM_BIND_FAST_PATH_SRC_DST(cputype, name, \
+ src_type, src_cnt, \
+ dst_type, dst_cnt) \
+@@ -87,7 +90,7 @@ cputype##_composite_##name (pixman_implementation_t *imp, \
+ src_line, src_stride); \
+ }
+
+-#define PIXMAN_ARM_BIND_FAST_PATH_N_DST(cputype, name, \
++#define PIXMAN_ARM_BIND_FAST_PATH_N_DST(flags, cputype, name, \
+ dst_type, dst_cnt) \
+ void \
+ pixman_composite_##name##_asm_##cputype (int32_t w, \
+@@ -117,7 +120,7 @@ cputype##_composite_##name (pixman_implementation_t *imp, \
+ \
+ src = _pixman_image_get_solid (src_image, dst_image->bits.format); \
+ \
+- if (src == 0) \
++ if ((flags & SKIP_ZERO_SRC) && src == 0) \
+ return; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+@@ -128,7 +131,7 @@ cputype##_composite_##name (pixman_implementation_t *imp, \
+ src); \
+ }
+
+-#define PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST(cputype, name, \
++#define PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST(flags, cputype, name, \
+ mask_type, mask_cnt, \
+ dst_type, dst_cnt) \
+ void \
+@@ -163,7 +166,7 @@ cputype##_composite_##name (pixman_implementation_t *imp, \
+ \
+ src = _pixman_image_get_solid (src_image, dst_image->bits.format); \
+ \
+- if (src == 0) \
++ if ((flags & SKIP_ZERO_SRC) && src == 0) \
+ return; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+@@ -177,7 +180,7 @@ cputype##_composite_##name (pixman_implementation_t *imp, \
+ mask_line, mask_stride); \
+ }
+
+-#define PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST(cputype, name, \
++#define PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST(flags, cputype, name, \
+ src_type, src_cnt, \
+ dst_type, dst_cnt) \
+ void \
+@@ -211,7 +214,7 @@ cputype##_composite_##name (pixman_implementation_t *imp, \
+ \
+ mask = _pixman_image_get_solid (mask_image, dst_image->bits.format);\
+ \
+- if (mask == 0) \
++ if ((flags & SKIP_ZERO_MASK) && mask == 0) \
+ return; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index f0dc111..1a3741c 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -63,33 +63,33 @@ PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, over_8888_8888,
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, out_reverse_8_0565,
+ uint8_t, 1, uint16_t, 1)
+
+-PIXMAN_ARM_BIND_FAST_PATH_N_DST (neon, over_n_0565,
++PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_0565,
+ uint16_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_N_DST (neon, over_n_8888,
++PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_8888,
+ uint32_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_N_DST (neon, over_reverse_n_8888,
++PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_reverse_n_8888,
+ uint32_t, 1)
+
+-PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_0565,
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_0565,
+ uint8_t, 1, uint16_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_8888,
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_8888,
+ uint8_t, 1, uint32_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8888_8888_ca,
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8888_8888_ca,
+ uint32_t, 1, uint32_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, over_n_8_8,
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_8,
+ uint8_t, 1, uint8_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8,
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, add_n_8_8,
+ uint8_t, 1, uint8_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (neon, add_n_8_8888,
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, add_n_8_8888,
+ uint8_t, 1, uint32_t, 1)
+
+-PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_8888,
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_8888_n_0565,
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_8888_n_0565,
+ uint32_t, 1, uint16_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, over_0565_n_0565,
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_0565_n_0565,
+ uint16_t, 1, uint16_t, 1)
+-PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (neon, add_8888_n_8888,
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, add_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
+
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8_8_8,
+diff --git a/pixman/pixman-arm-simd.c b/pixman/pixman-arm-simd.c
+index 3b05007..dc2f471 100644
+--- a/pixman/pixman-arm-simd.c
++++ b/pixman/pixman-arm-simd.c
+@@ -381,10 +381,10 @@ PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (armv6, add_8_8,
+ PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (armv6, over_8888_8888,
+ uint32_t, 1, uint32_t, 1)
+
+-PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (armv6, over_8888_n_8888,
++PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, armv6, over_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
+
+-PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (armv6, over_n_8_8888,
++PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, armv6, over_n_8_8888,
+ uint8_t, 1, uint32_t, 1)
+
+ PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (armv6, 0565_0565, SRC,
+--
+1.6.6.1
+
--- /dev/null
+From 733f68912f4a44c24ad3973049a7e1d98f4c6ea8 Mon Sep 17 00:00:00 2001
+From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+Date: Mon, 29 Nov 2010 09:11:29 +0200
+Subject: [PATCH 16/24] ARM: added 'neon_composite_in_n_8' fast path
+
+---
+ pixman/pixman-arm-neon-asm.S | 52 ++++++++++++++++++++++++++++++++++++++++++
+ pixman/pixman-arm-neon.c | 3 ++
+ 2 files changed, 55 insertions(+), 0 deletions(-)
+
+diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
+index 2c0fd37..cf014fa 100644
+--- a/pixman/pixman-arm-neon-asm.S
++++ b/pixman/pixman-arm-neon-asm.S
+@@ -1427,6 +1427,58 @@ generate_composite_function \
+
+ /******************************************************************************/
+
++.macro pixman_composite_in_n_8_process_pixblock_head
++ /* expecting source data in {d0, d1, d2, d3} */
++ /* and destination data in {d4, d5, d6, d7} */
++ vmull.u8 q8, d4, d3
++ vmull.u8 q9, d5, d3
++ vmull.u8 q10, d6, d3
++ vmull.u8 q11, d7, d3
++.endm
++
++.macro pixman_composite_in_n_8_process_pixblock_tail
++ vrshr.u16 q14, q8, #8
++ vrshr.u16 q15, q9, #8
++ vrshr.u16 q12, q10, #8
++ vrshr.u16 q13, q11, #8
++ vraddhn.u16 d28, q8, q14
++ vraddhn.u16 d29, q9, q15
++ vraddhn.u16 d30, q10, q12
++ vraddhn.u16 d31, q11, q13
++.endm
++
++.macro pixman_composite_in_n_8_process_pixblock_tail_head
++ pixman_composite_in_n_8_process_pixblock_tail
++ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
++ cache_preload 32, 32
++ pixman_composite_in_n_8_process_pixblock_head
++ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
++.endm
++
++.macro pixman_composite_in_n_8_init
++ add DUMMY, sp, #ARGS_STACK_OFFSET
++ vld1.32 {d3[0]}, [DUMMY]
++ vdup.8 d3, d3[3]
++.endm
++
++.macro pixman_composite_in_n_8_cleanup
++.endm
++
++generate_composite_function \
++ pixman_composite_in_n_8_asm_neon, 0, 0, 8, \
++ FLAG_DST_READWRITE, \
++ 32, /* number of pixels, processed in a single block */ \
++ 5, /* prefetch distance */ \
++ pixman_composite_in_n_8_init, \
++ pixman_composite_in_n_8_cleanup, \
++ pixman_composite_in_n_8_process_pixblock_head, \
++ pixman_composite_in_n_8_process_pixblock_tail, \
++ pixman_composite_in_n_8_process_pixblock_tail_head, \
++ 28, /* dst_w_basereg */ \
++ 4, /* dst_r_basereg */ \
++ 0, /* src_basereg */ \
++ 24 /* mask_basereg */
++
+ .macro pixman_composite_add_n_8_8_process_pixblock_head
+ /* expecting source data in {d8, d9, d10, d11} */
+ /* d8 - blue, d9 - green, d10 - red, d11 - alpha */
+diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
+index 1a3741c..e3eca2b 100644
+--- a/pixman/pixman-arm-neon.c
++++ b/pixman/pixman-arm-neon.c
+@@ -69,6 +69,8 @@ PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_8888,
+ uint32_t, 1)
+ PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_reverse_n_8888,
+ uint32_t, 1)
++PIXMAN_ARM_BIND_FAST_PATH_N_DST (0, neon, in_n_8,
++ uint8_t, 1)
+
+ PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_0565,
+ uint8_t, 1, uint16_t, 1)
+@@ -298,6 +300,7 @@ static const pixman_fast_path_t arm_neon_fast_paths[] =
+ PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, neon_composite_add_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, neon_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, neon_composite_add_8888_8888),
++ PIXMAN_STD_FAST_PATH (IN, solid, null, a8, neon_composite_in_n_8),
+ PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8r8g8b8, neon_composite_over_reverse_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8b8g8r8, neon_composite_over_reverse_n_8888),
+ PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8, null, r5g6b5, neon_composite_out_reverse_8_0565),
+--
+1.6.6.1
+
-From fff598814365037c8ffdd97afe10a14bb09558fc Mon Sep 17 00:00:00 2001
+From 6593d86679fde724e49efa96b16ca22d9521b288 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 1/8] add _pixman_bits_override_accessors
+Subject: [PATCH 17/24] add _pixman_bits_override_accessors
* from patch ARM: HACK: added NEON optimizations for fetch/store r5g6b5 scanline
* used in
int x,
int y,
--
-1.7.3.2
+1.6.6.1
-From 6b162fb9d4ede5faa25f24188964f31d7667e74e Mon Sep 17 00:00:00 2001
+From 8e8b2809b505486001dc213becab0d50bfd96c1b 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
+Subject: [PATCH 18/24] 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
+
#endif /* PIXMAN_PRIVATE_H */
--
-1.7.3.2
+1.6.6.1
-From 2d6cc769d233c0b1a391b501e84f5c3b0f1af4f8 Mon Sep 17 00:00:00 2001
+From f5a54f7d5eb1169bc79f0e445e2998e98080ef13 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
+Subject: [PATCH 19/24] Support of overlapping src/dst for pixman_blt_mmx
---
pixman/pixman-mmx.c | 55 +++++++++++++++++++++++++++++---------------------
while (w >= 2 && ((unsigned long)d & 3))
{
--
-1.7.3.2
+1.6.6.1
-From 532b8f45cee61ea2509a7f263dd30f40f3de29ba Mon Sep 17 00:00:00 2001
+From c8755294fa9ea396f7113370230b17c424a93be1 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
+Subject: [PATCH 20/24] Support of overlapping src/dst for pixman_blt_sse2
---
pixman/pixman-sse2.c | 55 +++++++++++++++++++++++++++++--------------------
while (w >= 2 && ((unsigned long)d & 3))
{
--
-1.7.3.2
+1.6.6.1
-From 4e101b976fa5fc72e44553a15516b804ffda0394 Mon Sep 17 00:00:00 2001
+From 86c8198598ef6d639e656c04644015795cc249aa 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
+Subject: [PATCH 21/24] 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 2f82069..6a6ed37 100644
+index e3eca2b..74316a8 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
-@@ -185,26 +185,66 @@ pixman_blt_neon (uint32_t *src_bits,
+@@ -199,26 +199,66 @@ pixman_blt_neon (uint32_t *src_bits,
int width,
int height)
{
default:
return FALSE;
--
-1.7.3.2
+1.6.6.1
-From 8d7a77b6780af1b96db32026fb8d79c5603f0fba Mon Sep 17 00:00:00 2001
+From 60d972afbae8613d700d3a6b3cb107429d7e11c6 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
+Subject: [PATCH 22/24] ARM: added NEON optimizations for fetch/store r5g6b5 scanline
---
pixman/pixman-arm-neon-asm.S | 20 ++++++++++++++++++++
2 files changed, 60 insertions(+), 0 deletions(-)
diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
-index 91ec27d..b838f92 100644
+index cf014fa..25f7bf0 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -459,6 +459,16 @@ generate_composite_function \
.macro pixman_composite_add_8_8_process_pixblock_head
diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
-index 6a6ed37..4b2bbea 100644
+index 74316a8..f773e92 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
-@@ -422,6 +422,42 @@ BIND_COMBINE_U (over)
+@@ -448,6 +448,42 @@ BIND_COMBINE_U (over)
BIND_COMBINE_U (add)
BIND_COMBINE_U (out_reverse)
pixman_implementation_t *
_pixman_implementation_create_arm_neon (void)
{
-@@ -437,6 +473,10 @@ _pixman_implementation_create_arm_neon (void)
+@@ -463,6 +499,10 @@ _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;
imp->fill = arm_neon_fill;
--
-1.7.3.2
+1.6.6.1
-From b689ddce66ce6391b6478d870f00fe21bbce944c Mon Sep 17 00:00:00 2001
+From cc99d8d6fcbabd7f9f3ed99e65c78a2fb71792fa 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 6/7] ARM: added NEON optimizations for fetch/store a8 scanline
+Subject: [PATCH 23/24] ARM: added NEON optimizations for fetch/store a8 scanline
---
pixman/pixman-arm-neon-asm.S | 64 ++++++++++++++++++++++++++++++++++++++++++
2 files changed, 106 insertions(+), 0 deletions(-)
diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
-index b838f92..8e43a3b 100644
+index 25f7bf0..439b06b 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -418,6 +418,70 @@ generate_composite_function \
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 4b2bbea..1c68d32 100644
+index f773e92..55219b3 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
-@@ -458,6 +458,45 @@ neon_store_scanline_r5g6b5 (bits_image_t * image,
+@@ -484,6 +484,45 @@ neon_store_scanline_r5g6b5 (bits_image_t * image,
pixman_store_scanline_r5g6b5_asm_neon (width, pixel, values);
}
pixman_implementation_t *
_pixman_implementation_create_arm_neon (void)
{
-@@ -476,6 +515,9 @@ _pixman_implementation_create_arm_neon (void)
+@@ -502,6 +541,9 @@ _pixman_implementation_create_arm_neon (void)
_pixman_bits_override_accessors (PIXMAN_r5g6b5,
neon_fetch_scanline_r5g6b5,
neon_store_scanline_r5g6b5);
imp->blt = arm_neon_blt;
imp->fill = arm_neon_fill;
--
-1.7.3.2
+1.6.6.1
-From 912d7b4f79cda5dd828f5db7608314057a39338e Mon Sep 17 00:00:00 2001
+From cf3b8fdc53144ff62c4054996559d3a1a4d62b75 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 7/7] ARM: added NEON optimizations for fetching x8r8g8b8 scanline
+Subject: [PATCH 24/24] ARM: added NEON optimizations for fetching x8r8g8b8 scanline
---
pixman/pixman-arm-neon-asm.S | 14 ++++++++++++++
2 files changed, 35 insertions(+), 0 deletions(-)
diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
-index 8e43a3b..5ebee5a 100644
+index 439b06b..3e0dcfe 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
-@@ -1206,6 +1206,20 @@ generate_composite_function \
+@@ -1257,6 +1257,20 @@ generate_composite_function \
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 1c68d32..0bcfc54 100644
+index 55219b3..8cef414 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
-@@ -496,6 +496,24 @@ neon_store_scanline_a8 (bits_image_t * image,
+@@ -522,6 +522,24 @@ neon_store_scanline_a8 (bits_image_t * image,
pixman_store_scanline_a8_asm_neon (width, pixel, values);
}
pixman_implementation_t *
_pixman_implementation_create_arm_neon (void)
-@@ -518,6 +536,9 @@ _pixman_implementation_create_arm_neon (void)
+@@ -544,6 +562,9 @@ _pixman_implementation_create_arm_neon (void)
_pixman_bits_override_accessors (PIXMAN_a8,
neon_fetch_scanline_a8,
neon_store_scanline_a8);
imp->blt = arm_neon_blt;
imp->fill = arm_neon_fill;
--
-1.7.3.2
+1.6.6.1
SRC_URI[archive.md5sum] = "9e09fd6e58cbf9717140891e0b7d4a7a"
SRC_URI[archive.sha256sum] = "295f51416caf307ff7caf1153ee9b1d86b9f7f02a7876d12db6538d80451c5de"
-PR = "${INC_PR}.0"
+PR = "${INC_PR}.1"
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://0002-Fix-argument-quoting-for-AC_INIT.patch \
+ file://0003-Sun-s-copyrights-belong-to-Oracle-now.patch \
+ file://0004-C-fast-path-for-a1-fill-operation.patch \
+ file://0005-ARM-added-neon_composite_over_n_8_8-fast-path.patch \
+ file://0006-ARM-introduced-fetch_mask_pixblock-macro-to-simplify.patch \
+ file://0007-ARM-better-NEON-instructions-scheduling-for-over_n_8.patch \
+ file://0008-ARM-added-neon_composite_over_8888_n_0565-fast-path.patch \
+ file://0009-ARM-reuse-common-NEON-code-for-over_-n_8-8888_n-8888.patch \
+ file://0010-ARM-added-neon_composite_over_0565_n_0565-fast-path.patch \
+ file://0011-ARM-added-neon_composite_add_8888_8_8888-fast-path.patch \
+ file://0012-ARM-better-NEON-instructions-scheduling-for-add_8888.patch \
+ file://0013-ARM-added-neon_composite_add_n_8_8888-fast-path.patch \
+ file://0014-ARM-added-neon_composite_add_8888_n_8888-fast-path.patch \
+ file://0015-ARM-added-flags-parameter-to-some-asm-fast-path-wrap.patch \
+ file://0016-ARM-added-neon_composite_in_n_8-fast-path.patch \
+ file://0017-add-_pixman_bits_override_accessors.patch \
+ file://0018-Generic-C-implementation-of-pixman_blt-with-overlapp.patch \
+ file://0019-Support-of-overlapping-src-dst-for-pixman_blt_mmx.patch \
+ file://0020-Support-of-overlapping-src-dst-for-pixman_blt_sse2.patch \
+ file://0021-Support-of-overlapping-src-dst-for-pixman_blt_neon.patch \
+ file://0022-ARM-added-NEON-optimizations-for-fetch-store-r5g6b5-.patch \
+ file://0023-ARM-added-NEON-optimizations-for-fetch-store-a8-scan.patch \
+ file://0024-ARM-added-NEON-optimizations-for-fetching-x8r8g8b8-s.patch \
"
NEON = " --disable-arm-neon "
git.openpandora.org / openembedded.git / commitdiff