/**************************************************************************
*
* Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2010 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* 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
+ * IN NO EVENT SHALL THE AUTHORS 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.
**************************************************************************/
/**
- * \brief Quad depth testing
+ * \brief Quad depth / stencil testing
*/
#include "pipe/p_defines.h"
+#include "util/u_format.h"
#include "util/u_memory.h"
#include "tgsi/tgsi_scan.h"
#include "sp_context.h"
#include "sp_quad.h"
-#include "sp_surface.h"
#include "sp_quad_pipe.h"
#include "sp_tile_cache.h"
#include "sp_state.h" /* for sp_fragment_shader */
data->bzzzz[j] = tile->data.depth32[y][x];
}
break;
- case PIPE_FORMAT_X8Z24_UNORM:
- case PIPE_FORMAT_S8Z24_UNORM:
+ case PIPE_FORMAT_Z24X8_UNORM:
+ case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
for (j = 0; j < QUAD_SIZE; j++) {
int x = quad->input.x0 % TILE_SIZE + (j & 1);
int y = quad->input.y0 % TILE_SIZE + (j >> 1);
data->bzzzz[j] = tile->data.depth32[y][x] & 0xffffff;
data->stencilVals[j] = tile->data.depth32[y][x] >> 24;
}
- break;
- case PIPE_FORMAT_Z24X8_UNORM:
- case PIPE_FORMAT_Z24S8_UNORM:
+ break;
+ case PIPE_FORMAT_X8Z24_UNORM:
+ case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
for (j = 0; j < QUAD_SIZE; j++) {
int x = quad->input.x0 % TILE_SIZE + (j & 1);
int y = quad->input.y0 % TILE_SIZE + (j >> 1);
data->stencilVals[j] = tile->data.depth32[y][x] & 0xff;
}
break;
+ case PIPE_FORMAT_S8_USCALED:
+ for (j = 0; j < QUAD_SIZE; j++) {
+ int x = quad->input.x0 % TILE_SIZE + (j & 1);
+ int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+ data->bzzzz[j] = 0;
+ data->stencilVals[j] = tile->data.stencil8[y][x];
+ }
+ break;
default:
assert(0);
}
}
-/* If the shader has not been run, interpolate the depth values
+
+/**
+ * If the shader has not been run, interpolate the depth values
* ourselves.
*/
static void
}
+/**
+ * Compute the depth_data::qzzzz[] values from the float fragment Z values.
+ */
static void
convert_quad_depth( struct depth_data *data,
const struct quad_header *quad )
}
}
break;
- case PIPE_FORMAT_X8Z24_UNORM:
- case PIPE_FORMAT_S8Z24_UNORM:
+ case PIPE_FORMAT_Z24X8_UNORM:
+ case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
{
float scale = (float) ((1 << 24) - 1);
}
}
break;
- case PIPE_FORMAT_Z24X8_UNORM:
- case PIPE_FORMAT_Z24S8_UNORM:
+ case PIPE_FORMAT_X8Z24_UNORM:
+ case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
{
float scale = (float) ((1 << 24) - 1);
+/**
+ * Write data->bzzzz[] values and data->stencilVals into the Z/stencil buffer.
+ */
static void
write_depth_stencil_values( struct depth_data *data,
struct quad_header *quad )
tile->data.depth16[y][x] = (ushort) data->bzzzz[j];
}
break;
- case PIPE_FORMAT_X8Z24_UNORM:
+ case PIPE_FORMAT_Z24X8_UNORM:
case PIPE_FORMAT_Z32_UNORM:
for (j = 0; j < QUAD_SIZE; j++) {
int x = quad->input.x0 % TILE_SIZE + (j & 1);
tile->data.depth32[y][x] = data->bzzzz[j];
}
break;
- case PIPE_FORMAT_S8Z24_UNORM:
+ case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
for (j = 0; j < QUAD_SIZE; j++) {
int x = quad->input.x0 % TILE_SIZE + (j & 1);
int y = quad->input.y0 % TILE_SIZE + (j >> 1);
tile->data.depth32[y][x] = (data->stencilVals[j] << 24) | data->bzzzz[j];
}
break;
- case PIPE_FORMAT_Z24S8_UNORM:
+ case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
for (j = 0; j < QUAD_SIZE; j++) {
int x = quad->input.x0 % TILE_SIZE + (j & 1);
int y = quad->input.y0 % TILE_SIZE + (j >> 1);
tile->data.depth32[y][x] = (data->bzzzz[j] << 8) | data->stencilVals[j];
}
break;
- case PIPE_FORMAT_Z24X8_UNORM:
+ case PIPE_FORMAT_X8Z24_UNORM:
for (j = 0; j < QUAD_SIZE; j++) {
int x = quad->input.x0 % TILE_SIZE + (j & 1);
int y = quad->input.y0 % TILE_SIZE + (j >> 1);
tile->data.depth32[y][x] = data->bzzzz[j] << 8;
}
break;
+ case PIPE_FORMAT_S8_USCALED:
+ for (j = 0; j < QUAD_SIZE; j++) {
+ int x = quad->input.x0 % TILE_SIZE + (j & 1);
+ int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+ tile->data.stencil8[y][x] = data->stencilVals[j];
+ }
+ break;
+
default:
assert(0);
}
-
/** Only 8-bit stencil supported */
#define STENCIL_MAX 0xff
-/*
+/**
* To increase efficiency, we should probably have multiple versions
* of this function that are specifically for Z16, Z32 and FP Z buffers.
* Try to effectively do that with codegen...
*/
-
static boolean
depth_test_quad(struct quad_stage *qs,
struct depth_data *data,
* Do stencil (and depth) testing. Stenciling depends on the outcome of
* depth testing.
*/
-static boolean
+static void
depth_stencil_test_quad(struct quad_stage *qs,
struct depth_data *data,
struct quad_header *quad)
face = 0;
}
+ /* 0 = front-face, 1 = back-face */
+ assert(face == 0 || face == 1);
+
/* choose front or back face function, operator, etc */
/* XXX we could do these initializations once per primitive */
func = softpipe->depth_stencil->stencil[face].func;
failOp = softpipe->depth_stencil->stencil[face].fail_op;
zFailOp = softpipe->depth_stencil->stencil[face].zfail_op;
zPassOp = softpipe->depth_stencil->stencil[face].zpass_op;
- ref = softpipe->depth_stencil->stencil[face].ref_value;
+ ref = softpipe->stencil_ref.ref_value[face];
wrtMask = softpipe->depth_stencil->stencil[face].writemask;
valMask = softpipe->depth_stencil->stencil[face].valuemask;
-
/* do the stencil test first */
{
unsigned passMask, failMask;
/* update stencil buffer values according to z pass/fail result */
if (zFailOp != PIPE_STENCIL_OP_KEEP) {
- const unsigned failMask = origMask & ~quad->inout.mask;
- apply_stencil_op(data, failMask, zFailOp, ref, wrtMask);
+ const unsigned zFailMask = origMask & ~quad->inout.mask;
+ apply_stencil_op(data, zFailMask, zFailOp, ref, wrtMask);
}
if (zPassOp != PIPE_STENCIL_OP_KEEP) {
- const unsigned passMask = origMask & quad->inout.mask;
- apply_stencil_op(data, passMask, zPassOp, ref, wrtMask);
+ const unsigned zPassMask = origMask & quad->inout.mask;
+ apply_stencil_op(data, zPassMask, zPassOp, ref, wrtMask);
}
}
else {
apply_stencil_op(data, quad->inout.mask, zPassOp, ref, wrtMask);
}
}
-
- return quad->inout.mask != 0;
}
#define ALPHATEST( FUNC, COMP ) \
- static int \
+ static int \
alpha_test_quads_##FUNC( struct quad_stage *qs, \
struct quad_header *quads[], \
unsigned nr ) \
}
}
+
static unsigned mask_count[16] =
{
0, /* 0x0 */
+/**
+ * General depth/stencil test function. Used when there's no fast-path.
+ */
static void
depth_test_quads_fallback(struct quad_stage *qs,
struct quad_header *quads[],
nr = alpha_test_quads(qs, quads, nr);
}
- if (qs->softpipe->framebuffer.zsbuf &&
- (qs->softpipe->depth_stencil->depth.enabled ||
- qs->softpipe->depth_stencil->stencil[0].enabled)) {
+ if (qs->softpipe->depth_stencil->depth.enabled ||
+ qs->softpipe->depth_stencil->stencil[0].enabled) {
data.ps = qs->softpipe->framebuffer.zsbuf;
data.format = data.ps->format;
}
if (qs->softpipe->depth_stencil->stencil[0].enabled) {
- if (!depth_stencil_test_quad(qs, &data, quads[i]))
- continue;
+ depth_stencil_test_quad(qs, &data, quads[i]);
+ write_depth_stencil_values(&data, quads[i]);
}
else {
if (!depth_test_quad(qs, &data, quads[i]))
continue;
- }
- if (qs->softpipe->depth_stencil->stencil[0].enabled ||
- qs->softpipe->depth_stencil->depth.writemask)
- write_depth_stencil_values(&data, quads[i]);
+ if (qs->softpipe->depth_stencil->depth.writemask)
+ write_depth_stencil_values(&data, quads[i]);
+ }
quads[pass++] = quads[i];
}
qs->next->run(qs->next, quads, nr);
}
-/* XXX: this function assumes setup function actually emits linear
- * spans of quads. It seems a lot more natural to do (early)
- * depth-testing on spans rather than quads.
- */
-static void
-depth_interp_z16_less_write(struct quad_stage *qs,
- struct quad_header *quads[],
- unsigned nr)
-{
- unsigned i, pass = 0;
- const unsigned ix = quads[0]->input.x0;
- const unsigned iy = quads[0]->input.y0;
- const float fx = (float) ix;
- const float fy = (float) iy;
- const float dzdx = quads[0]->posCoef->dadx[2];
- const float dzdy = quads[0]->posCoef->dady[2];
- const float z0 = quads[0]->posCoef->a0[2] + dzdx * fx + dzdy * fy;
- struct softpipe_cached_tile *tile;
- ushort (*depth16)[TILE_SIZE];
- ushort idepth[4], depth_step;
- const float scale = 65535.0;
-
- idepth[0] = (ushort)((z0) * scale);
- idepth[1] = (ushort)((z0 + dzdx) * scale);
- idepth[2] = (ushort)((z0 + dzdy) * scale);
- idepth[3] = (ushort)((z0 + dzdx + dzdy) * scale);
-
- depth_step = (ushort)(dzdx * 2 * scale);
-
- tile = sp_get_cached_tile(qs->softpipe->zsbuf_cache, ix, iy);
-
- depth16 = (ushort (*)[TILE_SIZE])
- &tile->data.depth16[iy % TILE_SIZE][ix % TILE_SIZE];
- for (i = 0; i < nr; i++) {
- unsigned outmask = quads[i]->inout.mask;
- unsigned mask = 0;
-
- if ((outmask & 1) && idepth[0] < depth16[0][0]) {
- depth16[0][0] = idepth[0];
- mask |= (1 << 0);
- }
-
- if ((outmask & 2) && idepth[1] < depth16[0][1]) {
- depth16[0][1] = idepth[1];
- mask |= (1 << 1);
- }
-
- if ((outmask & 4) && idepth[2] < depth16[1][0]) {
- depth16[1][0] = idepth[2];
- mask |= (1 << 2);
- }
-
- if ((outmask & 8) && idepth[3] < depth16[1][1]) {
- depth16[1][1] = idepth[3];
- mask |= (1 << 3);
- }
-
- idepth[0] += depth_step;
- idepth[1] += depth_step;
- idepth[2] += depth_step;
- idepth[3] += depth_step;
-
- depth16 = (ushort (*)[TILE_SIZE]) &depth16[0][2];
-
- quads[i]->inout.mask = mask;
- if (quads[i]->inout.mask)
- quads[pass++] = quads[i];
- }
-
- if (pass)
- qs->next->run(qs->next, quads, pass);
-
-}
-
-
-static void
-depth_interp_z16_lequal_write(struct quad_stage *qs,
- struct quad_header *quads[],
- unsigned nr)
-{
- unsigned i, pass = 0;
- const unsigned ix = quads[0]->input.x0;
- const unsigned iy = quads[0]->input.y0;
- const float fx = (float) ix;
- const float fy = (float) iy;
- const float dzdx = quads[0]->posCoef->dadx[2];
- const float dzdy = quads[0]->posCoef->dady[2];
- const float z0 = quads[0]->posCoef->a0[2] + dzdx * fx + dzdy * fy;
- struct softpipe_cached_tile *tile;
- ushort (*depth16)[TILE_SIZE];
- ushort idepth[4], depth_step;
- const float scale = 65535.0;
-
- idepth[0] = (ushort)((z0) * scale);
- idepth[1] = (ushort)((z0 + dzdx) * scale);
- idepth[2] = (ushort)((z0 + dzdy) * scale);
- idepth[3] = (ushort)((z0 + dzdx + dzdy) * scale);
-
- depth_step = (ushort)(dzdx * 2 * scale);
-
- tile = sp_get_cached_tile(qs->softpipe->zsbuf_cache, ix, iy);
-
- depth16 = (ushort (*)[TILE_SIZE])
- &tile->data.depth16[iy % TILE_SIZE][ix % TILE_SIZE];
-
- for (i = 0; i < nr; i++) {
- unsigned outmask = quads[i]->inout.mask;
- unsigned mask = 0;
-
- if ((outmask & 1) && idepth[0] <= depth16[0][0]) {
- depth16[0][0] = idepth[0];
- mask |= (1 << 0);
- }
-
- if ((outmask & 2) && idepth[1] <= depth16[0][1]) {
- depth16[0][1] = idepth[1];
- mask |= (1 << 1);
- }
-
- if ((outmask & 4) && idepth[2] <= depth16[1][0]) {
- depth16[1][0] = idepth[2];
- mask |= (1 << 2);
- }
-
- if ((outmask & 8) && idepth[3] <= depth16[1][1]) {
- depth16[1][1] = idepth[3];
- mask |= (1 << 3);
- }
+/**
+ * Special-case Z testing for 16-bit Zbuffer and Z buffer writes enabled.
+ */
- idepth[0] += depth_step;
- idepth[1] += depth_step;
- idepth[2] += depth_step;
- idepth[3] += depth_step;
+#define NAME depth_interp_z16_less_write
+#define OPERATOR <
+#include "sp_quad_depth_test_tmp.h"
- depth16 = (ushort (*)[TILE_SIZE]) &depth16[0][2];
+#define NAME depth_interp_z16_equal_write
+#define OPERATOR ==
+#include "sp_quad_depth_test_tmp.h"
- quads[i]->inout.mask = mask;
- if (quads[i]->inout.mask)
- quads[pass++] = quads[i];
- }
+#define NAME depth_interp_z16_lequal_write
+#define OPERATOR <=
+#include "sp_quad_depth_test_tmp.h"
- if (pass)
- qs->next->run(qs->next, quads, pass);
+#define NAME depth_interp_z16_greater_write
+#define OPERATOR >
+#include "sp_quad_depth_test_tmp.h"
-}
+#define NAME depth_interp_z16_notequal_write
+#define OPERATOR !=
+#include "sp_quad_depth_test_tmp.h"
+#define NAME depth_interp_z16_gequal_write
+#define OPERATOR >=
+#include "sp_quad_depth_test_tmp.h"
+#define NAME depth_interp_z16_always_write
+#define ALWAYS 1
+#include "sp_quad_depth_test_tmp.h"
boolean alpha = qs->softpipe->depth_stencil->alpha.enabled;
- boolean depth = (qs->softpipe->framebuffer.zsbuf &&
- qs->softpipe->depth_stencil->depth.enabled);
+ boolean depth = qs->softpipe->depth_stencil->depth.enabled;
unsigned depthfunc = qs->softpipe->depth_stencil->depth.func;
boolean occlusion = qs->softpipe->active_query_count;
+ /* default */
+ qs->run = depth_test_quads_fallback;
+ /* look for special cases */
if (!alpha &&
!depth &&
!stencil) {
!occlusion &&
!stencil)
{
- switch (depthfunc) {
- case PIPE_FUNC_LESS:
- switch (qs->softpipe->framebuffer.zsbuf->format) {
- case PIPE_FORMAT_Z16_UNORM:
+ if (qs->softpipe->framebuffer.zsbuf->format == PIPE_FORMAT_Z16_UNORM) {
+ switch (depthfunc) {
+ case PIPE_FUNC_NEVER:
+ qs->run = depth_test_quads_fallback;
+ break;
+ case PIPE_FUNC_LESS:
qs->run = depth_interp_z16_less_write;
break;
- default:
- qs->run = depth_test_quads_fallback;
+ case PIPE_FUNC_EQUAL:
+ qs->run = depth_interp_z16_equal_write;
break;
- }
- break;
- case PIPE_FUNC_LEQUAL:
- switch (qs->softpipe->framebuffer.zsbuf->format) {
- case PIPE_FORMAT_Z16_UNORM:
+ case PIPE_FUNC_LEQUAL:
qs->run = depth_interp_z16_lequal_write;
break;
+ case PIPE_FUNC_GREATER:
+ qs->run = depth_interp_z16_greater_write;
+ break;
+ case PIPE_FUNC_NOTEQUAL:
+ qs->run = depth_interp_z16_notequal_write;
+ break;
+ case PIPE_FUNC_GEQUAL:
+ qs->run = depth_interp_z16_gequal_write;
+ break;
+ case PIPE_FUNC_ALWAYS:
+ qs->run = depth_interp_z16_always_write;
+ break;
default:
qs->run = depth_test_quads_fallback;
break;
}
- break;
- default:
- qs->run = depth_test_quads_fallback;
}
}
- else {
- qs->run = depth_test_quads_fallback;
- }
-
+ /* next quad/fragment stage */
qs->run( qs, quads, nr );
}
-
-
-static void depth_test_begin(struct quad_stage *qs)
+static void
+depth_test_begin(struct quad_stage *qs)
{
qs->run = choose_depth_test;
qs->next->begin(qs->next);
}
-static void depth_test_destroy(struct quad_stage *qs)
+static void
+depth_test_destroy(struct quad_stage *qs)
{
FREE( qs );
}
-struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe )
+struct quad_stage *
+sp_quad_depth_test_stage(struct softpipe_context *softpipe)
{
struct quad_stage *stage = CALLOC_STRUCT(quad_stage);