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 24/24] ARM: added NEON optimizations for fetching x8r8g8b8 scanline

---
 pixman/pixman-arm-neon-asm.S |   14 ++++++++++++++
 pixman/pixman-arm-neon.c     |   21 +++++++++++++++++++++
 2 files changed, 35 insertions(+), 0 deletions(-)

diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index 439b06b..3e0dcfe 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -1257,6 +1257,20 @@ generate_composite_function \
     0, /* src_basereg   */ \
     0  /* mask_basereg  */
 
+generate_composite_function_single_scanline \
+    pixman_fetch_scanline_x888_asm_neon, 32, 0, 32, \
+    FLAG_DST_WRITEONLY, \
+    8, /* number of pixels, processed in a single block */ \
+    pixman_composite_src_x888_8888_init, \
+    default_cleanup, \
+    pixman_composite_src_x888_8888_process_pixblock_head, \
+    pixman_composite_src_x888_8888_process_pixblock_tail, \
+    pixman_composite_src_x888_8888_process_pixblock_tail_head, \
+    0,  /* dst_w_basereg */ \
+    0,  /* dst_r_basereg */ \
+    0,  /* src_basereg   */ \
+    0   /* mask_basereg  */
+
 /******************************************************************************/
 
 .macro pixman_composite_over_n_8_8888_process_pixblock_head
diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index 55219b3..8cef414 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -522,6 +522,24 @@ neon_store_scanline_a8 (bits_image_t *  image,
     pixman_store_scanline_a8_asm_neon (width, pixel, values);
 }
 
+void
+pixman_fetch_scanline_x888_asm_neon (int             width,
+                                     uint32_t       *buffer,
+                                     const uint32_t *pixel);
+
+static void
+neon_fetch_scanline_x888 (pixman_image_t *image,
+                          int             x,
+                          int             y,
+                          int             width,
+                          uint32_t *      buffer,
+                          const uint32_t *mask)
+{
+    const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
+    const uint32_t *pixel = (const uint32_t *) bits + x;
+
+    pixman_fetch_scanline_x888_asm_neon (width, buffer, pixel);
+}
 
 pixman_implementation_t *
 _pixman_implementation_create_arm_neon (void)
@@ -544,6 +562,9 @@ _pixman_implementation_create_arm_neon (void)
     _pixman_bits_override_accessors (PIXMAN_a8,
                                      neon_fetch_scanline_a8,
                                      neon_store_scanline_a8);
+    _pixman_bits_override_accessors (PIXMAN_x8r8g8b8,
+                                     neon_fetch_scanline_x888,
+                                     NULL);
 
     imp->blt = arm_neon_blt;
     imp->fill = arm_neon_fill;
-- 
1.6.6.1