Files
third_party_mesa3d/src/mesa/pipe/softpipe/sp_surface.c
Brian 59e7bfa1de Add support for more surface types in sp_surface.c
replace PIPE_FORMAT_U_L8_A8 with PIPE_FORMAT_U_A8_L8
2007-08-24 18:08:23 -06:00

821 lines
22 KiB
C
Executable File

/**************************************************************************
*
* Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
* 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"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sub license, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice (including the
* next paragraph) shall be included in all copies or substantial portions
* of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
* IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
**************************************************************************/
#include "sp_context.h"
#include "sp_state.h"
#include "sp_surface.h"
#include "pipe/p_defines.h"
#include "pipe/p_util.h"
/**
* Softpipe surface functions.
* Basically, create surface of a particular type, then plug in default
* read/write_quad and get/put_tile() functions.
* Note that these quad funcs assume the buffer/region is in a linear
* layout with Y=0=top.
* If we had swizzled/AOS buffers the read/write quad functions could be
* simplified a lot....
*/
/*** PIPE_FORMAT_U_A8_R8_G8_B8 ***/
static void
a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
const unsigned *src
= ((const unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8);
assert(x < (int) sps->surface.width - 1);
assert(y < (int) sps->surface.height - 1);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
const unsigned p = src[j];
rrrr[0][i * 2 + j] = UBYTE_TO_FLOAT((p >> 16) & 0xff); /*R*/
rrrr[1][i * 2 + j] = UBYTE_TO_FLOAT((p >> 8) & 0xff); /*G*/
rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT((p ) & 0xff); /*B*/
rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT((p >> 24) & 0xff); /*A*/
}
src += sps->surface.region->pitch;
}
}
static void
a8r8g8b8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
unsigned *dst
= ((unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
ubyte r, g, b, a;
UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/
UNCLAMPED_FLOAT_TO_UBYTE(g, rrrr[1][i * 2 + j]); /*G*/
UNCLAMPED_FLOAT_TO_UBYTE(b, rrrr[2][i * 2 + j]); /*B*/
UNCLAMPED_FLOAT_TO_UBYTE(a, rrrr[3][i * 2 + j]); /*A*/
dst[j] = (a << 24) | (r << 16) | (g << 8) | b;
}
dst += sps->surface.region->pitch;
}
}
static void
a8r8g8b8_get_tile(struct pipe_surface *ps,
unsigned x, unsigned y, unsigned w, unsigned h, float *p)
{
const unsigned *src
= ((const unsigned *) (ps->region->map + ps->offset))
+ y * ps->region->pitch + x;
unsigned i, j;
unsigned w0 = w;
assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8);
#if 0
assert(x + w <= ps->width);
assert(y + h <= ps->height);
#else
/* temp clipping hack */
if (x + w > ps->width)
w = ps->width - x;
if (y + h > ps->height)
h = ps->height -y;
#endif
for (i = 0; i < h; i++) {
float *pRow = p;
for (j = 0; j < w; j++) {
const unsigned pixel = src[j];
pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff);
pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff);
pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff);
pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff);
pRow += 4;
}
src += ps->region->pitch;
p += w0 * 4;
}
}
/*** PIPE_FORMAT_U_A1_R5_G5_B5 ***/
static void
a1r5g5b5_get_tile(struct pipe_surface *ps,
unsigned x, unsigned y, unsigned w, unsigned h, float *p)
{
const ushort *src
= ((const ushort *) (ps->region->map + ps->offset))
+ y * ps->region->pitch + x;
unsigned i, j;
assert(ps->format == PIPE_FORMAT_U_A1_R5_G5_B5);
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
const ushort pixel = src[j];
p[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f);
p[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f);
p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f);
p[3] = ((pixel >> 15) ) * 1.0f;
p += 4;
}
src += ps->region->pitch;
}
}
/*** PIPE_FORMAT_U_Z16 ***/
static void
z16_read_quad_z(struct softpipe_surface *sps,
int x, int y, unsigned zzzz[QUAD_SIZE])
{
const ushort *src
= ((const ushort *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_U_Z16);
/* converting ushort to unsigned: */
zzzz[0] = src[0];
zzzz[1] = src[1];
src += sps->surface.region->pitch;
zzzz[2] = src[0];
zzzz[3] = src[1];
}
static void
z16_write_quad_z(struct softpipe_surface *sps,
int x, int y, const unsigned zzzz[QUAD_SIZE])
{
ushort *dst
= ((ushort *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_U_Z16);
/* converting unsigned to ushort: */
dst[0] = zzzz[0];
dst[1] = zzzz[1];
dst += sps->surface.region->pitch;
dst[0] = zzzz[2];
dst[1] = zzzz[3];
}
/*** PIPE_FORMAT_U_L8 ***/
static void
l8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
const ubyte *src
= ((const ubyte *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_L8);
assert(x < (int) sps->surface.width - 1);
assert(y < (int) sps->surface.height - 1);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
rrrr[0][i * 2 + j] =
rrrr[1][i * 2 + j] =
rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT(src[j]);
rrrr[3][i * 2 + j] = 1.0F;
}
src += sps->surface.region->pitch;
}
}
static void
l8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
ubyte *dst
= ((ubyte *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_L8);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
ubyte r;
UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/
dst[j] = r;
}
dst += sps->surface.region->pitch;
}
}
static void
l8_get_tile(struct pipe_surface *ps,
unsigned x, unsigned y, unsigned w, unsigned h, float *p)
{
const ubyte *src
= ((const ubyte *) (ps->region->map + ps->offset))
+ y * ps->region->pitch + x;
unsigned i, j;
unsigned w0 = w;
assert(ps->format == PIPE_FORMAT_U_L8);
#if 0
assert(x + w <= ps->width);
assert(y + h <= ps->height);
#else
/* temp clipping hack */
if (x + w > ps->width)
w = ps->width - x;
if (y + h > ps->height)
h = ps->height -y;
#endif
for (i = 0; i < h; i++) {
float *pRow = p;
for (j = 0; j < w; j++) {
pRow[0] =
pRow[1] =
pRow[2] = UBYTE_TO_FLOAT(src[j]);
pRow[3] = 1.0;
pRow += 4;
}
src += ps->region->pitch;
p += w0 * 4;
}
}
/*** PIPE_FORMAT_U_A8 ***/
static void
a8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
const ubyte *src
= ((const ubyte *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_A8);
assert(x < (int) sps->surface.width - 1);
assert(y < (int) sps->surface.height - 1);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
rrrr[0][i * 2 + j] =
rrrr[1][i * 2 + j] =
rrrr[2][i * 2 + j] = 0.0F;
rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(src[j]);
}
src += sps->surface.region->pitch;
}
}
static void
a8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
ubyte *dst
= ((ubyte *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_A8);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
ubyte r;
UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[3][i * 2 + j]); /*A*/
dst[j] = r;
}
dst += sps->surface.region->pitch;
}
}
static void
a8_get_tile(struct pipe_surface *ps,
unsigned x, unsigned y, unsigned w, unsigned h, float *p)
{
const ubyte *src
= ((const ubyte *) (ps->region->map + ps->offset))
+ y * ps->region->pitch + x;
unsigned i, j;
unsigned w0 = w;
assert(ps->format == PIPE_FORMAT_U_A8);
#if 0
assert(x + w <= ps->width);
assert(y + h <= ps->height);
#else
/* temp clipping hack */
if (x + w > ps->width)
w = ps->width - x;
if (y + h > ps->height)
h = ps->height -y;
#endif
for (i = 0; i < h; i++) {
float *pRow = p;
for (j = 0; j < w; j++) {
pRow[0] =
pRow[1] =
pRow[2] = 0.0;
pRow[3] = UBYTE_TO_FLOAT(src[j]);
pRow += 4;
}
src += ps->region->pitch;
p += w0 * 4;
}
}
/*** PIPE_FORMAT_U_I8 ***/
static void
i8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
const ubyte *src
= ((const ubyte *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_I8);
assert(x < (int) sps->surface.width - 1);
assert(y < (int) sps->surface.height - 1);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
rrrr[0][i * 2 + j] =
rrrr[1][i * 2 + j] =
rrrr[2][i * 2 + j] =
rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(src[j]);
}
src += sps->surface.region->pitch;
}
}
static void
i8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
ubyte *dst
= ((ubyte *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_I8);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
ubyte r;
UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/
dst[j] = r;
}
dst += sps->surface.region->pitch;
}
}
static void
i8_get_tile(struct pipe_surface *ps,
unsigned x, unsigned y, unsigned w, unsigned h, float *p)
{
const ubyte *src
= ((const ubyte *) (ps->region->map + ps->offset))
+ y * ps->region->pitch + x;
unsigned i, j;
unsigned w0 = w;
assert(ps->format == PIPE_FORMAT_U_I8);
#if 0
assert(x + w <= ps->width);
assert(y + h <= ps->height);
#else
/* temp clipping hack */
if (x + w > ps->width)
w = ps->width - x;
if (y + h > ps->height)
h = ps->height -y;
#endif
for (i = 0; i < h; i++) {
float *pRow = p;
for (j = 0; j < w; j++) {
pRow[0] =
pRow[1] =
pRow[2] =
pRow[3] = UBYTE_TO_FLOAT(src[j]);
pRow += 4;
}
src += ps->region->pitch;
p += w0 * 4;
}
}
/*** PIPE_FORMAT_U_A8_L8 ***/
static void
a8_l8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
const ushort *src
= ((const ushort *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_A8_L8);
assert(x < (int) sps->surface.width - 1);
assert(y < (int) sps->surface.height - 1);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
const ushort p = src[j];
rrrr[0][i * 2 + j] =
rrrr[1][i * 2 + j] =
rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT(p >> 8);
rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(p & 0xff);
}
src += sps->surface.region->pitch;
}
}
static void
a8_l8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y,
float (*rrrr)[QUAD_SIZE])
{
ushort *dst
= ((ushort *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
unsigned i, j;
assert(sps->surface.format == PIPE_FORMAT_U_A8_L8);
for (i = 0; i < 2; i++) { /* loop over pixel row */
for (j = 0; j < 2; j++) { /* loop over pixel column */
ubyte l, a;
UNCLAMPED_FLOAT_TO_UBYTE(l, rrrr[0][i * 2 + j]); /*R*/
UNCLAMPED_FLOAT_TO_UBYTE(a, rrrr[3][i * 2 + j]); /*A*/
dst[j] = (l << 8) | a;
}
dst += sps->surface.region->pitch;
}
}
static void
a8_l8_get_tile(struct pipe_surface *ps,
unsigned x, unsigned y, unsigned w, unsigned h, float *p)
{
const ushort *src
= ((const ushort *) (ps->region->map + ps->offset))
+ y * ps->region->pitch + x;
unsigned i, j;
unsigned w0 = w;
assert(ps->format == PIPE_FORMAT_U_A8_L8);
#if 0
assert(x + w <= ps->width);
assert(y + h <= ps->height);
#else
/* temp clipping hack */
if (x + w > ps->width)
w = ps->width - x;
if (y + h > ps->height)
h = ps->height -y;
#endif
for (i = 0; i < h; i++) {
float *pRow = p;
for (j = 0; j < w; j++) {
const ushort p = src[j];
pRow[0] =
pRow[1] =
pRow[2] = UBYTE_TO_FLOAT(p & 0xff);
pRow[3] = UBYTE_TO_FLOAT(p >> 8);
pRow += 4;
}
src += ps->region->pitch;
p += w0 * 4;
}
}
/*** PIPE_FORMAT_U_Z32 ***/
static void
z32_read_quad_z(struct softpipe_surface *sps,
int x, int y, unsigned zzzz[QUAD_SIZE])
{
const unsigned *src
= ((unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_U_Z32);
zzzz[0] = src[0];
zzzz[1] = src[1];
src += sps->surface.region->pitch;
zzzz[2] = src[0];
zzzz[3] = src[1];
}
static void
z32_write_quad_z(struct softpipe_surface *sps,
int x, int y, const unsigned zzzz[QUAD_SIZE])
{
unsigned *dst
= ((unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_U_Z32);
dst[0] = zzzz[0];
dst[1] = zzzz[1];
dst += sps->surface.region->pitch;
dst[0] = zzzz[2];
dst[1] = zzzz[3];
}
/*** PIPE_FORMAT_S8_Z24 ***/
static void
s8z24_read_quad_z(struct softpipe_surface *sps,
int x, int y, unsigned zzzz[QUAD_SIZE])
{
static const unsigned mask = 0x00ffffff;
const unsigned *src
= ((unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_S8_Z24);
/* extract lower three bytes */
zzzz[0] = src[0] & mask;
zzzz[1] = src[1] & mask;
src += sps->surface.region->pitch;
zzzz[2] = src[0] & mask;
zzzz[3] = src[1] & mask;
}
static void
s8z24_write_quad_z(struct softpipe_surface *sps,
int x, int y, const unsigned zzzz[QUAD_SIZE])
{
static const unsigned mask = 0xff000000;
unsigned *dst
= ((unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_S8_Z24);
assert(zzzz[0] <= 0xffffff);
dst[0] = (dst[0] & mask) | zzzz[0];
dst[1] = (dst[1] & mask) | zzzz[1];
dst += sps->surface.region->pitch;
dst[0] = (dst[0] & mask) | zzzz[2];
dst[1] = (dst[1] & mask) | zzzz[3];
}
static void
s8z24_read_quad_stencil(struct softpipe_surface *sps,
int x, int y, ubyte ssss[QUAD_SIZE])
{
const unsigned *src
= ((unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_S8_Z24);
ssss[0] = src[0] >> 24;
ssss[1] = src[1] >> 24;
src += sps->surface.region->pitch;
ssss[2] = src[0] >> 24;
ssss[3] = src[1] >> 24;
}
static void
s8z24_write_quad_stencil(struct softpipe_surface *sps,
int x, int y, const ubyte ssss[QUAD_SIZE])
{
static const unsigned mask = 0x00ffffff;
unsigned *dst
= ((unsigned *) (sps->surface.region->map + sps->surface.offset))
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_S8_Z24);
dst[0] = (dst[0] & mask) | (ssss[0] << 24);
dst[1] = (dst[1] & mask) | (ssss[1] << 24);
dst += sps->surface.region->pitch;
dst[0] = (dst[0] & mask) | (ssss[2] << 24);
dst[1] = (dst[1] & mask) | (ssss[3] << 24);
}
/*** PIPE_FORMAT_U_S8 ***/
static void
s8_read_quad_stencil(struct softpipe_surface *sps,
int x, int y, ubyte ssss[QUAD_SIZE])
{
const ubyte *src
= sps->surface.region->map + sps->surface.offset
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_U_S8);
ssss[0] = src[0];
ssss[1] = src[1];
src += sps->surface.region->pitch;
ssss[2] = src[0];
ssss[3] = src[1];
}
static void
s8_write_quad_stencil(struct softpipe_surface *sps,
int x, int y, const ubyte ssss[QUAD_SIZE])
{
ubyte *dst
= sps->surface.region->map + sps->surface.offset
+ y * sps->surface.region->pitch + x;
assert(sps->surface.format == PIPE_FORMAT_U_S8);
dst[0] = ssss[0];
dst[1] = ssss[1];
dst += sps->surface.region->pitch;
dst[0] = ssss[2];
dst[1] = ssss[3];
}
/**
* Initialize the quad_read/write and get/put_tile() methods.
*/
void
softpipe_init_surface_funcs(struct softpipe_surface *sps)
{
assert(sps->surface.format);
switch (sps->surface.format) {
case PIPE_FORMAT_U_A8_R8_G8_B8:
sps->read_quad_f_swz = a8r8g8b8_read_quad_f_swz;
sps->write_quad_f_swz = a8r8g8b8_write_quad_f_swz;
sps->surface.get_tile = a8r8g8b8_get_tile;
break;
case PIPE_FORMAT_U_A1_R5_G5_B5:
sps->surface.get_tile = a1r5g5b5_get_tile;
break;
case PIPE_FORMAT_U_L8:
sps->read_quad_f_swz = l8_read_quad_f_swz;
sps->write_quad_f_swz = l8_write_quad_f_swz;
sps->surface.get_tile = l8_get_tile;
break;
case PIPE_FORMAT_U_A8:
sps->read_quad_f_swz = a8_read_quad_f_swz;
sps->write_quad_f_swz = a8_write_quad_f_swz;
sps->surface.get_tile = a8_get_tile;
break;
case PIPE_FORMAT_U_I8:
sps->read_quad_f_swz = i8_read_quad_f_swz;
sps->write_quad_f_swz = i8_write_quad_f_swz;
sps->surface.get_tile = i8_get_tile;
break;
case PIPE_FORMAT_U_A8_L8:
sps->read_quad_f_swz = a8_l8_read_quad_f_swz;
sps->write_quad_f_swz = a8_l8_write_quad_f_swz;
sps->surface.get_tile = a8_l8_get_tile;
break;
case PIPE_FORMAT_U_Z16:
sps->read_quad_z = z16_read_quad_z;
sps->write_quad_z = z16_write_quad_z;
break;
case PIPE_FORMAT_U_Z32:
sps->read_quad_z = z32_read_quad_z;
sps->write_quad_z = z32_write_quad_z;
break;
case PIPE_FORMAT_S8_Z24:
sps->read_quad_z = s8z24_read_quad_z;
sps->write_quad_z = s8z24_write_quad_z;
sps->read_quad_stencil = s8z24_read_quad_stencil;
sps->write_quad_stencil = s8z24_write_quad_stencil;
break;
case PIPE_FORMAT_U_S8:
sps->read_quad_stencil = s8_read_quad_stencil;
sps->write_quad_stencil = s8_write_quad_stencil;
break;
default:
assert(0);
}
}
static struct pipe_surface *
softpipe_surface_alloc(struct pipe_context *pipe, unsigned pipeFormat)
{
struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface);
if (!sps)
return NULL;
assert(pipeFormat < PIPE_FORMAT_COUNT);
sps->surface.format = pipeFormat;
sps->surface.refcount = 1;
softpipe_init_surface_funcs(sps);
return &sps->surface;
}
/**
* Called via pipe->get_tex_surface()
* XXX is this in the right place?
*/
struct pipe_surface *
softpipe_get_tex_surface(struct pipe_context *pipe,
struct pipe_mipmap_tree *mt,
unsigned face, unsigned level, unsigned zslice)
{
struct pipe_surface *ps;
unsigned offset; /* in bytes */
offset = mt->level[level].level_offset;
if (mt->target == PIPE_TEXTURE_CUBE) {
offset += mt->level[level].image_offset[face] * mt->cpp;
}
else if (mt->target == PIPE_TEXTURE_3D) {
offset += mt->level[level].image_offset[zslice] * mt->cpp;
}
else {
assert(face == 0);
assert(zslice == 0);
}
ps = pipe->surface_alloc(pipe, mt->format);
if (ps) {
assert(ps->format);
assert(ps->refcount);
ps->region = mt->region;
ps->width = mt->level[level].width;
ps->height = mt->level[level].height;
ps->offset = offset;
}
return ps;
}
void
sp_init_surface_functions(struct softpipe_context *sp)
{
sp->pipe.surface_alloc = softpipe_surface_alloc;
}