pixman: Branch 'master' - 2 commits
Søren Sandmann Pedersen
sandmann at kemper.freedesktop.org
Tue Sep 10 06:36:57 PDT 2013
pixman/pixman-access.c | 50 ++++++++++++++++++++++-----------------------
pixman/pixman-bits-image.c | 18 +++++++---------
pixman/pixman-mmx.c | 35 ++++++-------------------------
pixman/pixman-private.h | 2 -
pixman/pixman-sse2.c | 43 +++++++-------------------------------
test/affine-test.c | 4 ---
test/scaling-test.c | 4 ---
7 files changed, 51 insertions(+), 105 deletions(-)
New commits:
commit a4c79d695d52c94647b1aff78548e5892d616b70
Author: Søren Sandmann Pedersen <ssp at redhat.com>
Date: Wed Sep 4 22:32:15 2013 -0400
Drop support for 8-bit precision in bilinear filtering
The default has been 7-bit for a while now, and the quality
improvement with 8-bit precision is not enough to justify keeping the
code around as a compile-time option.
diff --git a/pixman/pixman-mmx.c b/pixman/pixman-mmx.c
index c94d282..a0f59ef 100644
--- a/pixman/pixman-mmx.c
+++ b/pixman/pixman-mmx.c
@@ -3561,7 +3561,6 @@ mmx_composite_over_reverse_n_8888 (pixman_implementation_t *imp,
#define BILINEAR_DECLARE_VARIABLES \
const __m64 mm_wt = _mm_set_pi16 (wt, wt, wt, wt); \
const __m64 mm_wb = _mm_set_pi16 (wb, wb, wb, wb); \
- const __m64 mm_BSHIFT = _mm_set_pi16 (BSHIFT, BSHIFT, BSHIFT, BSHIFT); \
const __m64 mm_addc7 = _mm_set_pi16 (0, 1, 0, 1); \
const __m64 mm_xorc7 = _mm_set_pi16 (0, BMSK, 0, BMSK); \
const __m64 mm_ux = _mm_set_pi16 (unit_x, unit_x, unit_x, unit_x); \
@@ -3581,35 +3580,15 @@ do { \
__m64 hi = _mm_add_pi16 (t_hi, b_hi); \
__m64 lo = _mm_add_pi16 (t_lo, b_lo); \
vx += unit_x; \
- if (BILINEAR_INTERPOLATION_BITS < 8) \
- { \
- /* calculate horizontal weights */ \
- __m64 mm_wh = _mm_add_pi16 (mm_addc7, _mm_xor_si64 (mm_xorc7, \
+ /* calculate horizontal weights */ \
+ __m64 mm_wh = _mm_add_pi16 (mm_addc7, _mm_xor_si64 (mm_xorc7, \
_mm_srli_pi16 (mm_x, \
16 - BILINEAR_INTERPOLATION_BITS))); \
- /* horizontal interpolation */ \
- __m64 p = _mm_unpacklo_pi16 (lo, hi); \
- __m64 q = _mm_unpackhi_pi16 (lo, hi); \
- lo = _mm_madd_pi16 (p, mm_wh); \
- hi = _mm_madd_pi16 (q, mm_wh); \
- } \
- else \
- { \
- /* calculate horizontal weights */ \
- __m64 mm_wh_lo = _mm_sub_pi16 (mm_BSHIFT, _mm_srli_pi16 (mm_x, \
- 16 - BILINEAR_INTERPOLATION_BITS)); \
- __m64 mm_wh_hi = _mm_srli_pi16 (mm_x, \
- 16 - BILINEAR_INTERPOLATION_BITS); \
- /* horizontal interpolation */ \
- __m64 mm_lo_lo = _mm_mullo_pi16 (lo, mm_wh_lo); \
- __m64 mm_lo_hi = _mm_mullo_pi16 (hi, mm_wh_hi); \
- __m64 mm_hi_lo = _mm_mulhi_pu16 (lo, mm_wh_lo); \
- __m64 mm_hi_hi = _mm_mulhi_pu16 (hi, mm_wh_hi); \
- lo = _mm_add_pi32 (_mm_unpacklo_pi16 (mm_lo_lo, mm_hi_lo), \
- _mm_unpacklo_pi16 (mm_lo_hi, mm_hi_hi)); \
- hi = _mm_add_pi32 (_mm_unpackhi_pi16 (mm_lo_lo, mm_hi_lo), \
- _mm_unpackhi_pi16 (mm_lo_hi, mm_hi_hi)); \
- } \
+ /* horizontal interpolation */ \
+ __m64 p = _mm_unpacklo_pi16 (lo, hi); \
+ __m64 q = _mm_unpackhi_pi16 (lo, hi); \
+ lo = _mm_madd_pi16 (p, mm_wh); \
+ hi = _mm_madd_pi16 (q, mm_wh); \
mm_x = _mm_add_pi16 (mm_x, mm_ux); \
/* shift and pack the result */ \
hi = _mm_srli_pi32 (hi, BILINEAR_INTERPOLATION_BITS * 2); \
diff --git a/pixman/pixman-sse2.c b/pixman/pixman-sse2.c
index dde9235..a629565 100644
--- a/pixman/pixman-sse2.c
+++ b/pixman/pixman-sse2.c
@@ -5554,7 +5554,6 @@ FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_normal_OVER,
scaled_nearest_scanline_sse2_8888_n_8888_OVER,
uint32_t, uint32_t, uint32_t, NORMAL, TRUE, TRUE)
-#if BILINEAR_INTERPOLATION_BITS < 8
# define BILINEAR_DECLARE_VARIABLES \
const __m128i xmm_wt = _mm_set_epi16 (wt, wt, wt, wt, wt, wt, wt, wt); \
const __m128i xmm_wb = _mm_set_epi16 (wb, wb, wb, wb, wb, wb, wb, wb); \
@@ -5564,21 +5563,10 @@ FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_normal_OVER,
const __m128i xmm_zero = _mm_setzero_si128 (); \
__m128i xmm_x = _mm_set_epi16 (vx, -(vx + 1), vx, -(vx + 1), \
vx, -(vx + 1), vx, -(vx + 1))
-#else
-# define BILINEAR_DECLARE_VARIABLES \
- const __m128i xmm_wt = _mm_set_epi16 (wt, wt, wt, wt, wt, wt, wt, wt); \
- const __m128i xmm_wb = _mm_set_epi16 (wb, wb, wb, wb, wb, wb, wb, wb); \
- const __m128i xmm_addc = _mm_set_epi16 (0, 0, 0, 0, 1, 1, 1, 1); \
- const __m128i xmm_ux = _mm_set_epi16 (unit_x, unit_x, unit_x, unit_x, \
- -unit_x, -unit_x, -unit_x, -unit_x); \
- const __m128i xmm_zero = _mm_setzero_si128 (); \
- __m128i xmm_x = _mm_set_epi16 (vx, vx, vx, vx, \
- -(vx + 1), -(vx + 1), -(vx + 1), -(vx + 1))
-#endif
#define BILINEAR_INTERPOLATE_ONE_PIXEL(pix) \
do { \
- __m128i xmm_wh, xmm_lo, xmm_hi, a; \
+ __m128i xmm_wh, a; \
/* fetch 2x2 pixel block into sse2 registers */ \
__m128i tltr = _mm_loadl_epi64 ( \
(__m128i *)&src_top[pixman_fixed_to_int (vx)]); \
@@ -5590,28 +5578,13 @@ do { \
xmm_wt), \
_mm_mullo_epi16 (_mm_unpacklo_epi8 (blbr, xmm_zero), \
xmm_wb)); \
- if (BILINEAR_INTERPOLATION_BITS < 8) \
- { \
- /* calculate horizontal weights */ \
- xmm_wh = _mm_add_epi16 (xmm_addc, _mm_srli_epi16 (xmm_x, \
- 16 - BILINEAR_INTERPOLATION_BITS)); \
- xmm_x = _mm_add_epi16 (xmm_x, xmm_ux); \
- /* horizontal interpolation */ \
- a = _mm_madd_epi16 (_mm_unpackhi_epi16 (_mm_shuffle_epi32 ( \
- a, _MM_SHUFFLE (1, 0, 3, 2)), a), xmm_wh); \
- } \
- else \
- { \
- /* calculate horizontal weights */ \
- xmm_wh = _mm_add_epi16 (xmm_addc, _mm_srli_epi16 (xmm_x, \
- 16 - BILINEAR_INTERPOLATION_BITS)); \
- xmm_x = _mm_add_epi16 (xmm_x, xmm_ux); \
- /* horizontal interpolation */ \
- xmm_lo = _mm_mullo_epi16 (a, xmm_wh); \
- xmm_hi = _mm_mulhi_epu16 (a, xmm_wh); \
- a = _mm_add_epi32 (_mm_unpacklo_epi16 (xmm_lo, xmm_hi), \
- _mm_unpackhi_epi16 (xmm_lo, xmm_hi)); \
- } \
+ /* calculate horizontal weights */ \
+ xmm_wh = _mm_add_epi16 (xmm_addc, _mm_srli_epi16 (xmm_x, \
+ 16 - BILINEAR_INTERPOLATION_BITS)); \
+ xmm_x = _mm_add_epi16 (xmm_x, xmm_ux); \
+ /* horizontal interpolation */ \
+ a = _mm_madd_epi16 (_mm_unpackhi_epi16 (_mm_shuffle_epi32 ( \
+ a, _MM_SHUFFLE (1, 0, 3, 2)), a), xmm_wh); \
/* shift and pack the result */ \
a = _mm_srli_epi32 (a, BILINEAR_INTERPOLATION_BITS * 2); \
a = _mm_packs_epi32 (a, a); \
diff --git a/test/affine-test.c b/test/affine-test.c
index 2506250..c1649ed 100644
--- a/test/affine-test.c
+++ b/test/affine-test.c
@@ -306,9 +306,7 @@ test_composite (int testnum,
return crc32;
}
-#if BILINEAR_INTERPOLATION_BITS == 8
-#define CHECKSUM 0x2CDF1F07
-#elif BILINEAR_INTERPOLATION_BITS == 7
+#if BILINEAR_INTERPOLATION_BITS == 7
#define CHECKSUM 0xBC00B1DF
#elif BILINEAR_INTERPOLATION_BITS == 4
#define CHECKSUM 0xA227306B
diff --git a/test/scaling-test.c b/test/scaling-test.c
index a8cb4c4..b4142a7 100644
--- a/test/scaling-test.c
+++ b/test/scaling-test.c
@@ -375,9 +375,7 @@ test_composite (int testnum,
return crc32;
}
-#if BILINEAR_INTERPOLATION_BITS == 8
-#define CHECKSUM 0x9096E6B6
-#elif BILINEAR_INTERPOLATION_BITS == 7
+#if BILINEAR_INTERPOLATION_BITS == 7
#define CHECKSUM 0xCE8EC6BA
#elif BILINEAR_INTERPOLATION_BITS == 4
#define CHECKSUM 0xAB1D39BE
commit 80a232db68fcff6460e8704da6a97ee2e6971e99
Author: Søren Sandmann Pedersen <ssp at redhat.com>
Date: Sat Aug 31 22:59:53 2013 -0400
Make the first argument to scanline fetchers have type bits_image_t *
Scanline fetchers haven't been used for images other than bits for a
long time, so by making the type reflect this fact, a bit of casting
can be saved in various places.
diff --git a/pixman/pixman-access.c b/pixman/pixman-access.c
index b5c8e40..4f0642d 100644
--- a/pixman/pixman-access.c
+++ b/pixman/pixman-access.c
@@ -294,14 +294,14 @@ convert_pixel (pixman_format_code_t from, pixman_format_code_t to, uint32_t pixe
}
static force_inline uint32_t
-convert_pixel_to_a8r8g8b8 (pixman_image_t *image,
+convert_pixel_to_a8r8g8b8 (bits_image_t *image,
pixman_format_code_t format,
uint32_t pixel)
{
if (PIXMAN_FORMAT_TYPE (format) == PIXMAN_TYPE_GRAY ||
PIXMAN_FORMAT_TYPE (format) == PIXMAN_TYPE_COLOR)
{
- return image->bits.indexed->rgba[pixel];
+ return image->indexed->rgba[pixel];
}
else
{
@@ -332,7 +332,7 @@ convert_pixel_from_a8r8g8b8 (pixman_image_t *image,
}
static force_inline uint32_t
-fetch_and_convert_pixel (pixman_image_t * image,
+fetch_and_convert_pixel (bits_image_t * image,
const uint8_t * bits,
int offset,
pixman_format_code_t format)
@@ -417,7 +417,7 @@ convert_and_store_pixel (bits_image_t * image,
#define MAKE_ACCESSORS(format) \
static void \
- fetch_scanline_ ## format (pixman_image_t *image, \
+ fetch_scanline_ ## format (bits_image_t *image, \
int x, \
int y, \
int width, \
@@ -425,7 +425,7 @@ convert_and_store_pixel (bits_image_t * image,
const uint32_t *mask) \
{ \
uint8_t *bits = \
- (uint8_t *)(image->bits.bits + y * image->bits.rowstride); \
+ (uint8_t *)(image->bits + y * image->rowstride); \
int i; \
\
for (i = 0; i < width; ++i) \
@@ -461,8 +461,8 @@ convert_and_store_pixel (bits_image_t * image,
uint8_t *bits = \
(uint8_t *)(image->bits + line * image->rowstride); \
\
- return fetch_and_convert_pixel ((pixman_image_t *)image, \
- bits, offset, PIXMAN_ ## format); \
+ return fetch_and_convert_pixel ( \
+ image, bits, offset, PIXMAN_ ## format); \
} \
\
static const void *const __dummy__ ## format
@@ -583,14 +583,14 @@ to_srgb (float f)
}
static void
-fetch_scanline_a8r8g8b8_sRGB_float (pixman_image_t *image,
+fetch_scanline_a8r8g8b8_sRGB_float (bits_image_t * image,
int x,
int y,
int width,
uint32_t * b,
const uint32_t *mask)
{
- const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
+ const uint32_t *bits = image->bits + y * image->rowstride;
const uint32_t *pixel = bits + x;
const uint32_t *end = pixel + width;
argb_t *buffer = (argb_t *)b;
@@ -612,14 +612,14 @@ fetch_scanline_a8r8g8b8_sRGB_float (pixman_image_t *image,
/* Expects a float buffer */
static void
-fetch_scanline_a2r10g10b10_float (pixman_image_t *image,
+fetch_scanline_a2r10g10b10_float (bits_image_t * image,
int x,
int y,
int width,
uint32_t * b,
const uint32_t *mask)
{
- const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
+ const uint32_t *bits = image->bits + y * image->rowstride;
const uint32_t *pixel = bits + x;
const uint32_t *end = pixel + width;
argb_t *buffer = (argb_t *)b;
@@ -643,14 +643,14 @@ fetch_scanline_a2r10g10b10_float (pixman_image_t *image,
/* Expects a float buffer */
static void
-fetch_scanline_x2r10g10b10_float (pixman_image_t *image,
+fetch_scanline_x2r10g10b10_float (bits_image_t *image,
int x,
int y,
int width,
uint32_t * b,
const uint32_t *mask)
{
- const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
+ const uint32_t *bits = image->bits + y * image->rowstride;
const uint32_t *pixel = (uint32_t *)bits + x;
const uint32_t *end = pixel + width;
argb_t *buffer = (argb_t *)b;
@@ -673,14 +673,14 @@ fetch_scanline_x2r10g10b10_float (pixman_image_t *image,
/* Expects a float buffer */
static void
-fetch_scanline_a2b10g10r10_float (pixman_image_t *image,
+fetch_scanline_a2b10g10r10_float (bits_image_t *image,
int x,
int y,
int width,
uint32_t * b,
const uint32_t *mask)
{
- const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
+ const uint32_t *bits = image->bits + y * image->rowstride;
const uint32_t *pixel = bits + x;
const uint32_t *end = pixel + width;
argb_t *buffer = (argb_t *)b;
@@ -704,14 +704,14 @@ fetch_scanline_a2b10g10r10_float (pixman_image_t *image,
/* Expects a float buffer */
static void
-fetch_scanline_x2b10g10r10_float (pixman_image_t *image,
+fetch_scanline_x2b10g10r10_float (bits_image_t *image,
int x,
int y,
int width,
uint32_t * b,
const uint32_t *mask)
{
- const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
+ const uint32_t *bits = image->bits + y * image->rowstride;
const uint32_t *pixel = (uint32_t *)bits + x;
const uint32_t *end = pixel + width;
argb_t *buffer = (argb_t *)b;
@@ -733,14 +733,14 @@ fetch_scanline_x2b10g10r10_float (pixman_image_t *image,
}
static void
-fetch_scanline_yuy2 (pixman_image_t *image,
+fetch_scanline_yuy2 (bits_image_t *image,
int x,
int line,
int width,
uint32_t * buffer,
const uint32_t *mask)
{
- const uint32_t *bits = image->bits.bits + image->bits.rowstride * line;
+ const uint32_t *bits = image->bits + image->rowstride * line;
int i;
for (i = 0; i < width; i++)
@@ -767,7 +767,7 @@ fetch_scanline_yuy2 (pixman_image_t *image,
}
static void
-fetch_scanline_yv12 (pixman_image_t *image,
+fetch_scanline_yv12 (bits_image_t *image,
int x,
int line,
int width,
@@ -1121,30 +1121,30 @@ store_scanline_generic_float (bits_image_t * image,
}
static void
-fetch_scanline_generic_float (pixman_image_t *image,
+fetch_scanline_generic_float (bits_image_t * image,
int x,
int y,
int width,
uint32_t * buffer,
const uint32_t *mask)
{
- image->bits.fetch_scanline_32 (image, x, y, width, buffer, NULL);
+ image->fetch_scanline_32 (image, x, y, width, buffer, NULL);
- pixman_expand_to_float ((argb_t *)buffer, buffer, image->bits.format, width);
+ pixman_expand_to_float ((argb_t *)buffer, buffer, image->format, width);
}
/* The 32_sRGB paths should be deleted after narrow processing
* is no longer invoked for formats that are considered wide.
* (Also see fetch_pixel_generic_lossy_32) */
static void
-fetch_scanline_a8r8g8b8_32_sRGB (pixman_image_t *image,
+fetch_scanline_a8r8g8b8_32_sRGB (bits_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 *bits = image->bits + y * image->rowstride;
const uint32_t *pixel = (uint32_t *)bits + x;
const uint32_t *end = pixel + width;
uint32_t tmp;
diff --git a/pixman/pixman-bits-image.c b/pixman/pixman-bits-image.c
index 75a39a1..7b9a973 100644
--- a/pixman/pixman-bits-image.c
+++ b/pixman/pixman-bits-image.c
@@ -1253,9 +1253,9 @@ bits_image_fetch_untransformed_repeat_none (bits_image_t *image,
w = MIN (width, image->width - x);
if (wide)
- image->fetch_scanline_float ((pixman_image_t *)image, x, y, w, buffer, NULL);
+ image->fetch_scanline_float (image, x, y, w, buffer, NULL);
else
- image->fetch_scanline_32 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+ image->fetch_scanline_32 (image, x, y, w, buffer, NULL);
width -= w;
buffer += w * (wide? 4 : 1);
@@ -1301,9 +1301,9 @@ bits_image_fetch_untransformed_repeat_normal (bits_image_t *image,
w = MIN (width, image->width - x);
if (wide)
- image->fetch_scanline_float ((pixman_image_t *)image, x, y, w, buffer, NULL);
+ image->fetch_scanline_float (image, x, y, w, buffer, NULL);
else
- image->fetch_scanline_32 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+ image->fetch_scanline_32 (image, x, y, w, buffer, NULL);
buffer += w * (wide? 4 : 1);
x += w;
@@ -1528,7 +1528,7 @@ dest_get_scanline_narrow (pixman_iter_t *iter, const uint32_t *mask)
int width = iter->width;
uint32_t * buffer = iter->buffer;
- image->bits.fetch_scanline_32 (image, x, y, width, buffer, mask);
+ image->bits.fetch_scanline_32 (&image->bits, x, y, width, buffer, mask);
if (image->common.alpha_map)
{
uint32_t *alpha;
@@ -1541,8 +1541,7 @@ dest_get_scanline_narrow (pixman_iter_t *iter, const uint32_t *mask)
y -= image->common.alpha_origin_y;
image->common.alpha_map->fetch_scanline_32 (
- (pixman_image_t *)image->common.alpha_map,
- x, y, width, alpha, mask);
+ image->common.alpha_map, x, y, width, alpha, mask);
for (i = 0; i < width; ++i)
{
@@ -1567,7 +1566,7 @@ dest_get_scanline_wide (pixman_iter_t *iter, const uint32_t *mask)
argb_t * buffer = (argb_t *)iter->buffer;
image->fetch_scanline_float (
- (pixman_image_t *)image, x, y, width, (uint32_t *)buffer, mask);
+ image, x, y, width, (uint32_t *)buffer, mask);
if (image->common.alpha_map)
{
argb_t *alpha;
@@ -1580,8 +1579,7 @@ dest_get_scanline_wide (pixman_iter_t *iter, const uint32_t *mask)
y -= image->common.alpha_origin_y;
image->common.alpha_map->fetch_scanline_float (
- (pixman_image_t *)image->common.alpha_map,
- x, y, width, (uint32_t *)alpha, mask);
+ image->common.alpha_map, x, y, width, (uint32_t *)alpha, mask);
for (i = 0; i < width; ++i)
buffer[i].a = alpha[i].a;
diff --git a/pixman/pixman-private.h b/pixman/pixman-private.h
index 9646605..120196d 100644
--- a/pixman/pixman-private.h
+++ b/pixman/pixman-private.h
@@ -57,7 +57,7 @@ struct argb_t
float b;
};
-typedef void (*fetch_scanline_t) (pixman_image_t *image,
+typedef void (*fetch_scanline_t) (bits_image_t *image,
int x,
int y,
int width,
More information about the xorg-commit
mailing list