[Pixman] [PATCH 1/2] sse2: if width == stride, convert to linear image

[Chris Wilson]

---
 pixman/pixman-sse2.c |   27 ++++++++++++++++-----------
 1 files changed, 16 insertions(+), 11 deletions(-)

diff --git a/pixman/pixman-sse2.c b/pixman/pixman-sse2.c
index 533b858..c39930e 100644
--- a/pixman/pixman-sse2.c
+++ b/pixman/pixman-sse2.c
@@ -3345,7 +3345,6 @@ pixman_fill_sse2 (uint32_t *bits,
                   int       height,
                   uint32_t  data)
 {
-    uint32_t byte_width;
     uint8_t         *byte_line;
 
     __m128i xmm_def;
@@ -3357,8 +3356,6 @@ pixman_fill_sse2 (uint32_t *bits,
 
 	stride = stride * (int) sizeof (uint32_t) / 1;
 	byte_line = (uint8_t *)(((uint8_t *)bits) + stride * y + x);
-	byte_width = width;
-	stride *= 1;
 
 	b = data & 0xff;
 	w = (b << 8) | b;
@@ -3368,7 +3365,7 @@ pixman_fill_sse2 (uint32_t *bits,
     {
 	stride = stride * (int) sizeof (uint32_t) / 2;
 	byte_line = (uint8_t *)(((uint16_t *)bits) + stride * y + x);
-	byte_width = 2 * width;
+	width *= 2;
 	stride *= 2;
 
         data = (data & 0xffff) * 0x00010001;
@@ -3377,7 +3374,7 @@ pixman_fill_sse2 (uint32_t *bits,
     {
 	stride = stride * (int) sizeof (uint32_t) / 4;
 	byte_line = (uint8_t *)(((uint32_t *)bits) + stride * y + x);
-	byte_width = 4 * width;
+	width *= 4;
 	stride *= 4;
     }
     else
@@ -3385,14 +3382,18 @@ pixman_fill_sse2 (uint32_t *bits,
 	return FALSE;
     }
 
+    if (width == stride) {
+	width *= height;
+	height = 1;
+    }
+
     xmm_def = create_mask_2x32_128 (data, data);
 
     while (height--)
     {
-	int w;
+	int w = width;
 	uint8_t *d = byte_line;
 	byte_line += stride;
-	w = byte_width;
 
 	while (w >= 1 && ((unsigned long)d & 1))
 	{
@@ -4691,7 +4692,6 @@ pixman_blt_sse2 (uint32_t *src_bits,
 {
     uint8_t *   src_bytes;
     uint8_t *   dst_bytes;
-    int byte_width;
 
     if (src_bpp != dst_bpp)
 	return FALSE;
@@ -4702,7 +4702,7 @@ pixman_blt_sse2 (uint32_t *src_bits,
 	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;
+	width *= 2;
 	src_stride *= 2;
 	dst_stride *= 2;
     }
@@ -4712,7 +4712,7 @@ pixman_blt_sse2 (uint32_t *src_bits,
 	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;
+	width *= 4;
 	src_stride *= 4;
 	dst_stride *= 4;
     }
@@ -4721,6 +4721,11 @@ pixman_blt_sse2 (uint32_t *src_bits,
 	return FALSE;
     }
 
+    if (width == src_stride && width == dst_stride) {
+	    width *= height;
+	    height = 1;
+    }
+
     while (height--)
     {
 	int w;
@@ -4728,7 +4733,7 @@ pixman_blt_sse2 (uint32_t *src_bits,
 	uint8_t *d = dst_bytes;
 	src_bytes += src_stride;
 	dst_bytes += dst_stride;
-	w = byte_width;
+	w = width;
 
 	while (w >= 2 && ((unsigned long)d & 3))
 	{
-- 
1.7.4.4