summaryrefslogtreecommitdiff
path: root/src/gallium
diff options
context:
space:
mode:
Diffstat (limited to 'src/gallium')
-rw-r--r--src/gallium/drivers/llvmpipe/lp_quad_blend.c5
-rw-r--r--src/gallium/drivers/llvmpipe/lp_tile_cache.c11
-rw-r--r--src/gallium/drivers/llvmpipe/lp_tile_cache.h11
-rw-r--r--src/gallium/drivers/llvmpipe/lp_tile_soa.c238
-rw-r--r--src/gallium/drivers/llvmpipe/lp_tile_soa.h30
5 files changed, 161 insertions, 134 deletions
diff --git a/src/gallium/drivers/llvmpipe/lp_quad_blend.c b/src/gallium/drivers/llvmpipe/lp_quad_blend.c
index 2b060574ae..4312753eef 100644
--- a/src/gallium/drivers/llvmpipe/lp_quad_blend.c
+++ b/src/gallium/drivers/llvmpipe/lp_quad_blend.c
@@ -41,6 +41,7 @@
#include "lp_quad.h"
#include "lp_surface.h"
#include "lp_tile_cache.h"
+#include "lp_tile_soa.h"
#include "lp_quad_pipe.h"
@@ -163,7 +164,7 @@ blend_run(struct quad_stage *qs,
int y = ity + (j >> 1);
for (i = 0; i < 4; i++) {
src[i][4*k + j] = float_to_ubyte(quad->output.color[cbuf][i][j]);
- dst[i][4*k + j] = tile->data.color[i][y][x];
+ dst[i][4*k + j] = TILE_PIXEL(tile->data.color, x, y, i);
}
}
}
@@ -193,7 +194,7 @@ blend_run(struct quad_stage *qs,
int x = itx + (j & 1);
int y = ity + (j >> 1);
for (i = 0; i < 4; i++) { /* loop over color chans */
- tile->data.color[i][y][x] = src[i][4*k + j];
+ TILE_PIXEL(tile->data.color, x, y, i) = src[i][4*k + j];
}
}
}
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.c b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
index 213896ec10..94908f601c 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_cache.c
+++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
@@ -218,11 +218,14 @@ clear_tile_rgba(struct llvmpipe_cached_tile *tile,
memset(tile->data.color, 0, sizeof(tile->data.color));
}
else {
- uint i, x, y;
+ uint8_t c[4];
+ uint x, y, i;
for (i = 0; i < 4; ++i)
- for (y = 0; y < TILE_SIZE; y++)
- for (x = 0; x < TILE_SIZE; x++)
- tile->data.color[i][y][x] = float_to_ubyte(clear_value[i]);
+ c[i] = float_to_ubyte(clear_value[i]);
+ for (y = 0; y < TILE_SIZE; y++)
+ for (x = 0; x < TILE_SIZE; x++)
+ for (i = 0; i < 4; ++i)
+ TILE_PIXEL(tile->data.color, x, y, i) = c[i];
}
}
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.h b/src/gallium/drivers/llvmpipe/lp_tile_cache.h
index df63d47e1b..20d4137696 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_cache.h
+++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.h
@@ -32,18 +32,13 @@
#include "pipe/p_compiler.h"
+#include "lp_tile_soa.h"
struct llvmpipe_context;
struct llvmpipe_tile_cache;
-/**
- * Cache tile size (width and height). This needs to be a power of two.
- */
-#define TILE_SIZE 64
-
-
/* If we need to support > 4096, just expand this to be a 64 bit
* union, or consider tiling in Z as well.
*/
@@ -65,8 +60,8 @@ struct llvmpipe_cached_tile
union tile_address addr;
union {
- /** color in SOA format (rrrr, gggg, bbbb, aaaa) */
- uint8_t color[4][TILE_SIZE][TILE_SIZE];
+ /** color in SOA format */
+ uint8_t color[TILE_SIZE*TILE_SIZE*NUM_CHANNELS];
uint color32[TILE_SIZE][TILE_SIZE];
uint depth32[TILE_SIZE][TILE_SIZE];
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_soa.c b/src/gallium/drivers/llvmpipe/lp_tile_soa.c
index 05b95e10bd..4e4ccb31cc 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_soa.c
+++ b/src/gallium/drivers/llvmpipe/lp_tile_soa.c
@@ -42,7 +42,11 @@
#include "lp_tile_soa.h"
-#define PIXEL(_p, _x, _y, _c) ((_p)[(_c)*TILE_SIZE*TILE_SIZE + (_y)*TILE_SIZE + (_x)])
+const unsigned char
+tile_offset[TILE_VECTOR_HEIGHT][TILE_VECTOR_WIDTH] = {
+ { 0, 1, 4, 5, 8, 9, 12, 13},
+ { 2, 3, 6, 7, 10, 11, 14, 15}
+};
@@ -58,10 +62,10 @@ a8r8g8b8_get_tile_rgba(const unsigned *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
const unsigned pixel = *src++;
- PIXEL(p, j, i, 0) = (pixel >> 16) & 0xff;
- PIXEL(p, j, i, 1) = (pixel >> 8) & 0xff;
- PIXEL(p, j, i, 2) = (pixel >> 0) & 0xff;
- PIXEL(p, j, i, 3) = (pixel >> 24) & 0xff;
+ TILE_PIXEL(p, j, i, 0) = (pixel >> 16) & 0xff;
+ TILE_PIXEL(p, j, i, 1) = (pixel >> 8) & 0xff;
+ TILE_PIXEL(p, j, i, 2) = (pixel >> 0) & 0xff;
+ TILE_PIXEL(p, j, i, 3) = (pixel >> 24) & 0xff;
}
}
}
@@ -77,10 +81,10 @@ a8r8g8b8_put_tile_rgba(unsigned *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r, g, b, a;
- r = PIXEL(p, j, i, 0);
- g = PIXEL(p, j, i, 1);
- b = PIXEL(p, j, i, 2);
- a = PIXEL(p, j, i, 3);
+ r = TILE_PIXEL(p, j, i, 0);
+ g = TILE_PIXEL(p, j, i, 1);
+ b = TILE_PIXEL(p, j, i, 2);
+ a = TILE_PIXEL(p, j, i, 3);
*dst++ = (a << 24) | (r << 16) | (g << 8) | b;
}
}
@@ -99,10 +103,10 @@ x8r8g8b8_get_tile_rgba(const unsigned *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
const unsigned pixel = *src++;
- PIXEL(p, j, i, 0) = (pixel >> 16) & 0xff;
- PIXEL(p, j, i, 1) = (pixel >> 8) & 0xff;
- PIXEL(p, j, i, 2) = (pixel >> 0) & 0xff;
- PIXEL(p, j, i, 3) = 0xff;
+ TILE_PIXEL(p, j, i, 0) = (pixel >> 16) & 0xff;
+ TILE_PIXEL(p, j, i, 1) = (pixel >> 8) & 0xff;
+ TILE_PIXEL(p, j, i, 2) = (pixel >> 0) & 0xff;
+ TILE_PIXEL(p, j, i, 3) = 0xff;
}
}
}
@@ -118,9 +122,9 @@ x8r8g8b8_put_tile_rgba(unsigned *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r, g, b;
- r = PIXEL(p, j, i, 0);
- g = PIXEL(p, j, i, 1);
- b = PIXEL(p, j, i, 2);
+ r = TILE_PIXEL(p, j, i, 0);
+ g = TILE_PIXEL(p, j, i, 1);
+ b = TILE_PIXEL(p, j, i, 2);
*dst++ = (0xff << 24) | (r << 16) | (g << 8) | b;
}
}
@@ -139,10 +143,10 @@ b8g8r8a8_get_tile_rgba(const unsigned *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
const unsigned pixel = *src++;
- PIXEL(p, j, i, 0) = (pixel >> 8) & 0xff;
- PIXEL(p, j, i, 1) = (pixel >> 16) & 0xff;
- PIXEL(p, j, i, 2) = (pixel >> 24) & 0xff;
- PIXEL(p, j, i, 3) = (pixel >> 0) & 0xff;
+ TILE_PIXEL(p, j, i, 0) = (pixel >> 8) & 0xff;
+ TILE_PIXEL(p, j, i, 1) = (pixel >> 16) & 0xff;
+ TILE_PIXEL(p, j, i, 2) = (pixel >> 24) & 0xff;
+ TILE_PIXEL(p, j, i, 3) = (pixel >> 0) & 0xff;
}
}
}
@@ -158,10 +162,10 @@ b8g8r8a8_put_tile_rgba(unsigned *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r, g, b, a;
- r = PIXEL(p, j, i, 0);
- g = PIXEL(p, j, i, 1);
- b = PIXEL(p, j, i, 2);
- a = PIXEL(p, j, i, 3);
+ r = TILE_PIXEL(p, j, i, 0);
+ g = TILE_PIXEL(p, j, i, 1);
+ b = TILE_PIXEL(p, j, i, 2);
+ a = TILE_PIXEL(p, j, i, 3);
*dst++ = (b << 24) | (g << 16) | (r << 8) | a;
}
}
@@ -180,10 +184,10 @@ a1r5g5b5_get_tile_rgba(const ushort *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
const ushort pixel = *src++;
- PIXEL(p, j, i, 0) = ((pixel >> 10) & 0x1f) * 255 / 31;
- PIXEL(p, j, i, 1) = ((pixel >> 5) & 0x1f) * 255 / 31;
- PIXEL(p, j, i, 2) = ((pixel ) & 0x1f) * 255 / 31;
- PIXEL(p, j, i, 3) = ((pixel >> 15) ) * 255;
+ TILE_PIXEL(p, j, i, 0) = ((pixel >> 10) & 0x1f) * 255 / 31;
+ TILE_PIXEL(p, j, i, 1) = ((pixel >> 5) & 0x1f) * 255 / 31;
+ TILE_PIXEL(p, j, i, 2) = ((pixel ) & 0x1f) * 255 / 31;
+ TILE_PIXEL(p, j, i, 3) = ((pixel >> 15) ) * 255;
}
}
}
@@ -199,10 +203,10 @@ a1r5g5b5_put_tile_rgba(ushort *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r, g, b, a;
- r = PIXEL(p, j, i, 0);
- g = PIXEL(p, j, i, 1);
- b = PIXEL(p, j, i, 2);
- a = PIXEL(p, j, i, 3);
+ r = TILE_PIXEL(p, j, i, 0);
+ g = TILE_PIXEL(p, j, i, 1);
+ b = TILE_PIXEL(p, j, i, 2);
+ a = TILE_PIXEL(p, j, i, 3);
r = r >> 3; /* 5 bits */
g = g >> 3; /* 5 bits */
b = b >> 3; /* 5 bits */
@@ -225,10 +229,10 @@ a4r4g4b4_get_tile_rgba(const ushort *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
const ushort pixel = *src++;
- PIXEL(p, j, i, 0) = ((pixel >> 8) & 0xf) * 255 / 15;
- PIXEL(p, j, i, 1) = ((pixel >> 4) & 0xf) * 255 / 15;
- PIXEL(p, j, i, 2) = ((pixel ) & 0xf) * 255 / 15;
- PIXEL(p, j, i, 3) = ((pixel >> 12) ) * 255 / 15;
+ TILE_PIXEL(p, j, i, 0) = ((pixel >> 8) & 0xf) * 255 / 15;
+ TILE_PIXEL(p, j, i, 1) = ((pixel >> 4) & 0xf) * 255 / 15;
+ TILE_PIXEL(p, j, i, 2) = ((pixel ) & 0xf) * 255 / 15;
+ TILE_PIXEL(p, j, i, 3) = ((pixel >> 12) ) * 255 / 15;
}
}
}
@@ -244,10 +248,10 @@ a4r4g4b4_put_tile_rgba(ushort *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r, g, b, a;
- r = PIXEL(p, j, i, 0);
- g = PIXEL(p, j, i, 1);
- b = PIXEL(p, j, i, 2);
- a = PIXEL(p, j, i, 3);
+ r = TILE_PIXEL(p, j, i, 0);
+ g = TILE_PIXEL(p, j, i, 1);
+ b = TILE_PIXEL(p, j, i, 2);
+ a = TILE_PIXEL(p, j, i, 3);
r >>= 4;
g >>= 4;
b >>= 4;
@@ -270,10 +274,10 @@ r5g6b5_get_tile_rgba(const ushort *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
const ushort pixel = *src++;
- PIXEL(p, j, i, 0) = ((pixel >> 11) & 0x1f) * 255 / 31;
- PIXEL(p, j, i, 1) = ((pixel >> 5) & 0x3f) * 255 / 63;
- PIXEL(p, j, i, 2) = ((pixel ) & 0x1f) * 255 / 31;
- PIXEL(p, j, i, 3) = 255;
+ TILE_PIXEL(p, j, i, 0) = ((pixel >> 11) & 0x1f) * 255 / 31;
+ TILE_PIXEL(p, j, i, 1) = ((pixel >> 5) & 0x3f) * 255 / 63;
+ TILE_PIXEL(p, j, i, 2) = ((pixel ) & 0x1f) * 255 / 31;
+ TILE_PIXEL(p, j, i, 3) = 255;
}
}
}
@@ -288,9 +292,9 @@ r5g6b5_put_tile_rgba(ushort *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
- uint r = (uint) PIXEL(p, j, i, 0) * 31 / 255;
- uint g = (uint) PIXEL(p, j, i, 1) * 63 / 255;
- uint b = (uint) PIXEL(p, j, i, 2) * 31 / 255;
+ uint r = (uint) TILE_PIXEL(p, j, i, 0) * 31 / 255;
+ uint g = (uint) TILE_PIXEL(p, j, i, 1) * 63 / 255;
+ uint b = (uint) TILE_PIXEL(p, j, i, 2) * 31 / 255;
*dst++ = (r << 11) | (g << 5) | (b);
}
}
@@ -313,10 +317,10 @@ z16_get_tile_rgba(const ushort *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) =
- PIXEL(p, j, i, 3) = *src++ * scale;
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) =
+ TILE_PIXEL(p, j, i, 3) = *src++ * scale;
}
}
}
@@ -335,10 +339,10 @@ l8_get_tile_rgba(const ubyte *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++, src++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) = *src;
- PIXEL(p, j, i, 3) = 255;
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) = *src;
+ TILE_PIXEL(p, j, i, 3) = 255;
}
}
}
@@ -354,7 +358,7 @@ l8_put_tile_rgba(ubyte *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r;
- r = PIXEL(p, j, i, 0);
+ r = TILE_PIXEL(p, j, i, 0);
*dst++ = (ubyte) r;
}
}
@@ -373,10 +377,10 @@ a8_get_tile_rgba(const ubyte *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++, src++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) = 0;
- PIXEL(p, j, i, 3) = *src;
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) = 0;
+ TILE_PIXEL(p, j, i, 3) = *src;
}
}
}
@@ -392,7 +396,7 @@ a8_put_tile_rgba(ubyte *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned a;
- a = PIXEL(p, j, i, 3);
+ a = TILE_PIXEL(p, j, i, 3);
*dst++ = (ubyte) a;
}
}
@@ -411,10 +415,10 @@ r16_get_tile_rgba(const short *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++, src++) {
- PIXEL(p, j, i, 0) = MAX2(src[0] >> 7, 0);
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) = 0;
- PIXEL(p, j, i, 3) = 255;
+ TILE_PIXEL(p, j, i, 0) = MAX2(src[0] >> 7, 0);
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) = 0;
+ TILE_PIXEL(p, j, i, 3) = 255;
}
}
}
@@ -429,7 +433,7 @@ r16_put_tile_rgba(short *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++, dst++) {
- dst[0] = PIXEL(p, j, i, 0) << 7;
+ dst[0] = TILE_PIXEL(p, j, i, 0) << 7;
}
}
}
@@ -446,10 +450,10 @@ r16g16b16a16_get_tile_rgba(const short *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++, src += 4) {
- PIXEL(p, j, i, 0) = src[0] >> 8;
- PIXEL(p, j, i, 1) = src[1] >> 8;
- PIXEL(p, j, i, 2) = src[2] >> 8;
- PIXEL(p, j, i, 3) = src[3] >> 8;
+ TILE_PIXEL(p, j, i, 0) = src[0] >> 8;
+ TILE_PIXEL(p, j, i, 1) = src[1] >> 8;
+ TILE_PIXEL(p, j, i, 2) = src[2] >> 8;
+ TILE_PIXEL(p, j, i, 3) = src[3] >> 8;
}
}
}
@@ -464,10 +468,10 @@ r16g16b16a16_put_tile_rgba(short *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++, dst += 4) {
- dst[0] = PIXEL(p, j, i, 0) << 8;
- dst[1] = PIXEL(p, j, i, 1) << 8;
- dst[2] = PIXEL(p, j, i, 2) << 8;
- dst[3] = PIXEL(p, j, i, 3) << 8;
+ dst[0] = TILE_PIXEL(p, j, i, 0) << 8;
+ dst[1] = TILE_PIXEL(p, j, i, 1) << 8;
+ dst[2] = TILE_PIXEL(p, j, i, 2) << 8;
+ dst[3] = TILE_PIXEL(p, j, i, 3) << 8;
}
}
}
@@ -485,10 +489,10 @@ i8_get_tile_rgba(const ubyte *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++, src++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) =
- PIXEL(p, j, i, 3) = *src;
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) =
+ TILE_PIXEL(p, j, i, 3) = *src;
}
}
}
@@ -504,7 +508,7 @@ i8_put_tile_rgba(ubyte *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r;
- r = PIXEL(p, j, i, 0);
+ r = TILE_PIXEL(p, j, i, 0);
*dst++ = (ubyte) r;
}
}
@@ -523,10 +527,10 @@ a8l8_get_tile_rgba(const ushort *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
ushort ra = *src++;
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) = ra & 0xff;
- PIXEL(p, j, i, 3) = ra >> 8;
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) = ra & 0xff;
+ TILE_PIXEL(p, j, i, 3) = ra >> 8;
}
}
}
@@ -542,8 +546,8 @@ a8l8_put_tile_rgba(ushort *dst,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
unsigned r, a;
- r = PIXEL(p, j, i, 0);
- a = PIXEL(p, j, i, 3);
+ r = TILE_PIXEL(p, j, i, 0);
+ a = TILE_PIXEL(p, j, i, 3);
*dst++ = (a << 8) | r;
}
}
@@ -567,10 +571,10 @@ z32_get_tile_rgba(const unsigned *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) =
- PIXEL(p, j, i, 3) = (float) (*src++ * scale);
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) =
+ TILE_PIXEL(p, j, i, 3) = (float) (*src++ * scale);
}
}
}
@@ -591,10 +595,10 @@ s8z24_get_tile_rgba(const unsigned *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) =
- PIXEL(p, j, i, 3) = (float) (scale * (*src++ & 0xffffff));
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) =
+ TILE_PIXEL(p, j, i, 3) = (float) (scale * (*src++ & 0xffffff));
}
}
}
@@ -615,10 +619,10 @@ z24s8_get_tile_rgba(const unsigned *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) =
- PIXEL(p, j, i, 3) = (float) (scale * (*src++ >> 8));
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) =
+ TILE_PIXEL(p, j, i, 3) = (float) (scale * (*src++ >> 8));
}
}
}
@@ -638,10 +642,10 @@ z32f_get_tile_rgba(const float *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) =
- PIXEL(p, j, i, 3) = *src++;
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) =
+ TILE_PIXEL(p, j, i, 3) = *src++;
}
}
}
@@ -683,19 +687,19 @@ ycbcr_get_tile_rgba(const ushort *src,
r = 1.164f * (y0-16) + 1.596f * (cr-128);
g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
b = 1.164f * (y0-16) + 2.018f * (cb-128);
- PIXEL(p, j, i, 0) = r;
- PIXEL(p, j, i, 1) = g;
- PIXEL(p, j, i, 2) = b;
- PIXEL(p, j, i, 3) = 255;
+ TILE_PIXEL(p, j, i, 0) = r;
+ TILE_PIXEL(p, j, i, 1) = g;
+ TILE_PIXEL(p, j, i, 2) = b;
+ TILE_PIXEL(p, j, i, 3) = 255;
/* odd pixel: use y1,cr,cb */
r = 1.164f * (y1-16) + 1.596f * (cr-128);
g = 1.164f * (y1-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
b = 1.164f * (y1-16) + 2.018f * (cb-128);
- PIXEL(p, j + 1, i, 0) = r;
- PIXEL(p, j + 1, i, 1) = g;
- PIXEL(p, j + 1, i, 2) = b;
- PIXEL(p, j + 1, i, 3) = 255;
+ TILE_PIXEL(p, j + 1, i, 0) = r;
+ TILE_PIXEL(p, j + 1, i, 1) = g;
+ TILE_PIXEL(p, j + 1, i, 2) = b;
+ TILE_PIXEL(p, j + 1, i, 3) = 255;
}
/* do the last texel */
if (w & 1) {
@@ -718,10 +722,10 @@ ycbcr_get_tile_rgba(const ushort *src,
r = 1.164f * (y0-16) + 1.596f * (cr-128);
g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
b = 1.164f * (y0-16) + 2.018f * (cb-128);
- PIXEL(p, j, i, 0) = r;
- PIXEL(p, j, i, 1) = g;
- PIXEL(p, j, i, 2) = b;
- PIXEL(p, j, i, 3) = 255;
+ TILE_PIXEL(p, j, i, 0) = r;
+ TILE_PIXEL(p, j, i, 1) = g;
+ TILE_PIXEL(p, j, i, 2) = b;
+ TILE_PIXEL(p, j, i, 3) = 255;
}
}
}
@@ -736,10 +740,10 @@ fake_get_tile_rgba(const ushort *src,
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
- PIXEL(p, j, i, 0) =
- PIXEL(p, j, i, 1) =
- PIXEL(p, j, i, 2) =
- PIXEL(p, j, i, 3) = (i ^ j) & 1 ? 255 : 0;
+ TILE_PIXEL(p, j, i, 0) =
+ TILE_PIXEL(p, j, i, 1) =
+ TILE_PIXEL(p, j, i, 2) =
+ TILE_PIXEL(p, j, i, 3) = (i ^ j) & 1 ? 255 : 0;
}
}
}
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_soa.h b/src/gallium/drivers/llvmpipe/lp_tile_soa.h
index 96fe87245c..3d8c703b73 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_soa.h
+++ b/src/gallium/drivers/llvmpipe/lp_tile_soa.h
@@ -29,9 +29,7 @@
#define LP_TILE_SOA_H
#include "pipe/p_compiler.h"
-
-
-struct pipe_transfer;
+#include "tgsi/tgsi_exec.h" // for NUM_CHANNELS
#ifdef __cplusplus
@@ -39,6 +37,32 @@ extern "C" {
#endif
+struct pipe_transfer;
+
+
+/**
+ * Cache tile size (width and height). This needs to be a power of two.
+ */
+#define TILE_SIZE 64
+
+
+#define TILE_VECTOR_HEIGHT 2
+#define TILE_VECTOR_WIDTH 8
+
+extern const unsigned char
+tile_offset[TILE_VECTOR_HEIGHT][TILE_VECTOR_WIDTH];
+
+#define TILE_C_STRIDE (TILE_VECTOR_HEIGHT*TILE_VECTOR_WIDTH)
+#define TILE_X_STRIDE (NUM_CHANNELS*TILE_C_STRIDE)
+#define TILE_Y_STRIDE (TILE_VECTOR_HEIGHT*TILE_SIZE*NUM_CHANNELS)
+
+#define TILE_PIXEL(_p, _x, _y, _c) \
+ ((_p)[((_y)/TILE_VECTOR_HEIGHT)*TILE_Y_STRIDE + \
+ ((_x)/TILE_VECTOR_WIDTH)*TILE_X_STRIDE + \
+ (_c)*TILE_C_STRIDE + \
+ tile_offset[(_y) % TILE_VECTOR_HEIGHT][(_x) % TILE_VECTOR_WIDTH]])
+
+
void
lp_get_tile_rgba_soa(struct pipe_transfer *pt,
uint x, uint y,
generated by cgit v1.2.3 (git 2.39.1) at 2024-10-31 22:22:00 +0000