/*
Copyright (C) Intel Corp. 2006. All Rights Reserved.
- Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ Intel funded Tungsten Graphics to
develop this 3D driver.
-
+
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
distribute, sublicense, 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 NONINFRINGEMENT.
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.
-
+
**********************************************************************/
/*
* Authors:
- * Keith Whitwell <keith@tungstengraphics.com>
+ * Keith Whitwell <keithw@vmware.com>
*/
-
+
#include <pthread.h>
#include "main/imports.h"
#include "main/enums.h"
#include "main/shaderobj.h"
#include "program/prog_parameter.h"
+#include "program/prog_print.h"
#include "program/program.h"
#include "program/programopt.h"
#include "tnl/tnl.h"
-#include "glsl/ralloc.h"
+#include "util/ralloc.h"
+#include "glsl/ir.h"
#include "brw_context.h"
+#include "brw_shader.h"
+#include "brw_nir.h"
#include "brw_wm.h"
+#include "intel_batchbuffer.h"
static unsigned
get_new_program_id(struct intel_screen *screen)
return id;
}
-static void brwBindProgram( struct gl_context *ctx,
- GLenum target,
- struct gl_program *prog )
-{
- struct brw_context *brw = brw_context(ctx);
-
- switch (target) {
- case GL_VERTEX_PROGRAM_ARB:
- brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
- break;
- case GL_FRAGMENT_PROGRAM_ARB:
- brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM;
- break;
- }
-}
-
static struct gl_program *brwNewProgram( struct gl_context *ctx,
- GLenum target,
+ GLenum target,
GLuint id )
{
struct brw_context *brw = brw_context(ctx);
if (prog) {
prog->id = get_new_program_id(brw->intelScreen);
- return _mesa_init_vertex_program( ctx, &prog->program,
- target, id );
+ return _mesa_init_gl_program(&prog->program.Base, target, id);
}
else
return NULL;
if (prog) {
prog->id = get_new_program_id(brw->intelScreen);
- return _mesa_init_fragment_program( ctx, &prog->program,
- target, id );
+ return _mesa_init_gl_program(&prog->program.Base, target, id);
}
else
return NULL;
}
+ case GL_GEOMETRY_PROGRAM_NV: {
+ struct brw_geometry_program *prog = CALLOC_STRUCT(brw_geometry_program);
+ if (prog) {
+ prog->id = get_new_program_id(brw->intelScreen);
+
+ return _mesa_init_gl_program(&prog->program, target, id);
+ } else {
+ return NULL;
+ }
+ }
+
+ case GL_COMPUTE_PROGRAM_NV: {
+ struct brw_compute_program *prog = CALLOC_STRUCT(brw_compute_program);
+ if (prog) {
+ prog->id = get_new_program_id(brw->intelScreen);
+
+ return _mesa_init_gl_program(&prog->program.Base, target, id);
+ } else {
+ return NULL;
+ }
+ }
+
default:
- return _mesa_new_program(ctx, target, id);
+ unreachable("Unsupported target in brwNewProgram()");
}
}
}
-static GLboolean
-brwIsProgramNative(struct gl_context *ctx,
- GLenum target,
- struct gl_program *prog)
-{
- return true;
-}
-
static GLboolean
brwProgramStringNotify(struct gl_context *ctx,
GLenum target,
brw_fragment_program_const(brw->fragment_program);
if (newFP == curFP)
- brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM;
+ brw->ctx.NewDriverState |= BRW_NEW_FRAGMENT_PROGRAM;
newFP->id = get_new_program_id(brw->intelScreen);
+
+ brw_add_texrect_params(prog);
+
+ prog->nir = brw_create_nir(brw, NULL, prog, MESA_SHADER_FRAGMENT, true);
+
+ brw_fs_precompile(ctx, NULL, prog);
break;
}
case GL_VERTEX_PROGRAM_ARB: {
brw_vertex_program_const(brw->vertex_program);
if (newVP == curVP)
- brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
+ brw->ctx.NewDriverState |= BRW_NEW_VERTEX_PROGRAM;
if (newVP->program.IsPositionInvariant) {
_mesa_insert_mvp_code(ctx, &newVP->program);
}
/* Also tell tnl about it:
*/
_tnl_program_string(ctx, target, prog);
+
+ brw_add_texrect_params(prog);
+
+ prog->nir = brw_create_nir(brw, NULL, prog, MESA_SHADER_VERTEX,
+ brw->intelScreen->compiler->scalar_vs);
+
+ brw_vs_precompile(ctx, NULL, prog);
break;
}
default:
* this function should only ever be called with a target of
* GL_VERTEX_PROGRAM_ARB or GL_FRAGMENT_PROGRAM_ARB.
*/
- assert(!"Unexpected target in brwProgramStringNotify");
- break;
+ unreachable("Unexpected target in brwProgramStringNotify");
}
- brw_add_texrect_params(prog);
-
return true;
}
+static void
+brw_memory_barrier(struct gl_context *ctx, GLbitfield barriers)
+{
+ struct brw_context *brw = brw_context(ctx);
+ unsigned bits = (PIPE_CONTROL_DATA_CACHE_INVALIDATE |
+ PIPE_CONTROL_NO_WRITE |
+ PIPE_CONTROL_CS_STALL);
+ assert(brw->gen >= 7 && brw->gen <= 9);
+
+ if (barriers & (GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT |
+ GL_ELEMENT_ARRAY_BARRIER_BIT |
+ GL_COMMAND_BARRIER_BIT))
+ bits |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
+
+ if (barriers & GL_UNIFORM_BARRIER_BIT)
+ bits |= (PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE |
+ PIPE_CONTROL_CONST_CACHE_INVALIDATE);
+
+ if (barriers & GL_TEXTURE_FETCH_BARRIER_BIT)
+ bits |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
+
+ if (barriers & GL_TEXTURE_UPDATE_BARRIER_BIT)
+ bits |= PIPE_CONTROL_RENDER_TARGET_FLUSH;
+
+ if (barriers & GL_FRAMEBUFFER_BARRIER_BIT)
+ bits |= (PIPE_CONTROL_DEPTH_CACHE_FLUSH |
+ PIPE_CONTROL_RENDER_TARGET_FLUSH);
+
+ /* Typed surface messages are handled by the render cache on IVB, so we
+ * need to flush it too.
+ */
+ if (brw->gen == 7 && !brw->is_haswell)
+ bits |= PIPE_CONTROL_RENDER_TARGET_FLUSH;
+
+ brw_emit_pipe_control_flush(brw, bits);
+}
+
void
brw_add_texrect_params(struct gl_program *prog)
{
}
}
-/* Per-thread scratch space is a power-of-two multiple of 1KB. */
-int
-brw_get_scratch_size(int size)
-{
- int i;
-
- for (i = 1024; i < size; i *= 2)
- ;
-
- return i;
-}
-
void
brw_get_scratch_bo(struct brw_context *brw,
drm_intel_bo **scratch_bo, int size)
void brwInitFragProgFuncs( struct dd_function_table *functions )
{
- assert(functions->ProgramStringNotify == _tnl_program_string);
+ /* assert(functions->ProgramStringNotify == _tnl_program_string); */
- functions->BindProgram = brwBindProgram;
functions->NewProgram = brwNewProgram;
functions->DeleteProgram = brwDeleteProgram;
- functions->IsProgramNative = brwIsProgramNative;
functions->ProgramStringNotify = brwProgramStringNotify;
functions->NewShader = brw_new_shader;
- functions->NewShaderProgram = brw_new_shader_program;
functions->LinkShader = brw_link_shader;
+
+ functions->MemoryBarrier = brw_memory_barrier;
}
+struct shader_times {
+ uint64_t time;
+ uint64_t written;
+ uint64_t reset;
+};
+
void
brw_init_shader_time(struct brw_context *brw)
{
- const int max_entries = 4096;
- brw->shader_time.bo = drm_intel_bo_alloc(brw->bufmgr, "shader time",
- max_entries * SHADER_TIME_STRIDE,
- 4096);
- brw->shader_time.shader_programs = rzalloc_array(brw, struct gl_shader_program *,
- max_entries);
- brw->shader_time.programs = rzalloc_array(brw, struct gl_program *,
- max_entries);
+ const int max_entries = 2048;
+ brw->shader_time.bo =
+ drm_intel_bo_alloc(brw->bufmgr, "shader time",
+ max_entries * SHADER_TIME_STRIDE * 3, 4096);
+ brw->shader_time.names = rzalloc_array(brw, const char *, max_entries);
+ brw->shader_time.ids = rzalloc_array(brw, int, max_entries);
brw->shader_time.types = rzalloc_array(brw, enum shader_time_shader_type,
max_entries);
- brw->shader_time.cumulative = rzalloc_array(brw, uint64_t,
+ brw->shader_time.cumulative = rzalloc_array(brw, struct shader_times,
max_entries);
brw->shader_time.max_entries = max_entries;
}
return 1;
}
-static void
-get_written_and_reset(struct brw_context *brw, int i,
- uint64_t *written, uint64_t *reset)
-{
- enum shader_time_shader_type type = brw->shader_time.types[i];
- assert(type == ST_VS || type == ST_FS8 || type == ST_FS16);
-
- /* Find where we recorded written and reset. */
- int wi, ri;
-
- for (wi = i; brw->shader_time.types[wi] != type + 1; wi++)
- ;
-
- for (ri = i; brw->shader_time.types[ri] != type + 2; ri++)
- ;
-
- *written = brw->shader_time.cumulative[wi];
- *reset = brw->shader_time.cumulative[ri];
-}
-
static void
print_shader_time_line(const char *stage, const char *name,
int shader_num, uint64_t time, uint64_t total)
{
- printf("%-6s%-6s", stage, name);
+ fprintf(stderr, "%-6s%-18s", stage, name);
- if (shader_num != -1)
- printf("%4d: ", shader_num);
+ if (shader_num != 0)
+ fprintf(stderr, "%4d: ", shader_num);
else
- printf(" : ");
+ fprintf(stderr, " : ");
- printf("%16lld (%7.2f Gcycles) %4.1f%%\n",
- (long long)time,
- (double)time / 1000000000.0,
- (double)time / total * 100.0);
+ fprintf(stderr, "%16lld (%7.2f Gcycles) %4.1f%%\n",
+ (long long)time,
+ (double)time / 1000000000.0,
+ (double)time / total * 100.0);
}
static void
uint64_t scaled[brw->shader_time.num_entries];
uint64_t *sorted[brw->shader_time.num_entries];
- uint64_t total_by_type[ST_FS16 + 1];
+ uint64_t total_by_type[ST_CS + 1];
memset(total_by_type, 0, sizeof(total_by_type));
double total = 0;
for (int i = 0; i < brw->shader_time.num_entries; i++) {
sorted[i] = &scaled[i];
switch (type) {
- case ST_VS_WRITTEN:
- case ST_VS_RESET:
- case ST_FS8_WRITTEN:
- case ST_FS8_RESET:
- case ST_FS16_WRITTEN:
- case ST_FS16_RESET:
- /* We'll handle these when along with the time. */
- scaled[i] = 0;
- continue;
-
case ST_VS:
+ case ST_GS:
case ST_FS8:
case ST_FS16:
- get_written_and_reset(brw, i, &written, &reset);
+ case ST_CS:
+ written = brw->shader_time.cumulative[i].written;
+ reset = brw->shader_time.cumulative[i].reset;
break;
default:
break;
}
- uint64_t time = brw->shader_time.cumulative[i];
+ uint64_t time = brw->shader_time.cumulative[i].time;
if (written) {
scaled[i] = time / written * (written + reset);
} else {
switch (type) {
case ST_VS:
+ case ST_GS:
case ST_FS8:
case ST_FS16:
+ case ST_CS:
total_by_type[type] += scaled[i];
break;
default:
}
if (total == 0) {
- printf("No shader time collected yet\n");
+ fprintf(stderr, "No shader time collected yet\n");
return;
}
qsort(sorted, brw->shader_time.num_entries, sizeof(sorted[0]), compare_time);
- printf("\n");
- printf("type ID cycles spent %% of total\n");
+ fprintf(stderr, "\n");
+ fprintf(stderr, "type ID cycles spent %% of total\n");
for (int s = 0; s < brw->shader_time.num_entries; s++) {
- const char *shader_name;
const char *stage;
/* Work back from the sorted pointers times to a time to print. */
int i = sorted[s] - scaled;
if (scaled[i] == 0)
continue;
- int shader_num = -1;
- if (brw->shader_time.shader_programs[i]) {
- shader_num = brw->shader_time.shader_programs[i]->Name;
-
- /* The fixed function fragment shader generates GLSL IR with a Name
- * of 0, and nothing else does.
- */
- if (shader_num == 0 &&
- (brw->shader_time.types[i] == ST_FS8 ||
- brw->shader_time.types[i] == ST_FS16)) {
- shader_name = "ff";
- shader_num = -1;
- } else {
- shader_name = "glsl";
- }
- } else if (brw->shader_time.programs[i]) {
- shader_num = brw->shader_time.programs[i]->Id;
- if (shader_num == 0) {
- shader_name = "ff";
- shader_num = -1;
- } else {
- shader_name = "prog";
- }
- } else {
- shader_name = "other";
- }
+ int shader_num = brw->shader_time.ids[i];
+ const char *shader_name = brw->shader_time.names[i];
switch (brw->shader_time.types[i]) {
case ST_VS:
stage = "vs";
break;
+ case ST_GS:
+ stage = "gs";
+ break;
case ST_FS8:
stage = "fs8";
break;
case ST_FS16:
stage = "fs16";
break;
+ case ST_CS:
+ stage = "cs";
+ break;
default:
stage = "other";
break;
scaled[i], total);
}
- printf("\n");
- print_shader_time_line("total", "vs", -1, total_by_type[ST_VS], total);
- print_shader_time_line("total", "fs8", -1, total_by_type[ST_FS8], total);
- print_shader_time_line("total", "fs16", -1, total_by_type[ST_FS16], total);
+ fprintf(stderr, "\n");
+ print_shader_time_line("total", "vs", 0, total_by_type[ST_VS], total);
+ print_shader_time_line("total", "gs", 0, total_by_type[ST_GS], total);
+ print_shader_time_line("total", "fs8", 0, total_by_type[ST_FS8], total);
+ print_shader_time_line("total", "fs16", 0, total_by_type[ST_FS16], total);
+ print_shader_time_line("total", "cs", 0, total_by_type[ST_CS], total);
}
static void
* overhead compared to the cost of tracking the time in the first place.
*/
drm_intel_bo_map(brw->shader_time.bo, true);
-
- uint32_t *times = brw->shader_time.bo->virtual;
+ void *bo_map = brw->shader_time.bo->virtual;
for (int i = 0; i < brw->shader_time.num_entries; i++) {
- brw->shader_time.cumulative[i] += times[i * SHADER_TIME_STRIDE / 4];
+ uint32_t *times = bo_map + i * 3 * SHADER_TIME_STRIDE;
+
+ brw->shader_time.cumulative[i].time += times[SHADER_TIME_STRIDE * 0 / 4];
+ brw->shader_time.cumulative[i].written += times[SHADER_TIME_STRIDE * 1 / 4];
+ brw->shader_time.cumulative[i].reset += times[SHADER_TIME_STRIDE * 2 / 4];
}
/* Zero the BO out to clear it out for our next collection.
*/
- memset(times, 0, brw->shader_time.bo->size);
+ memset(bo_map, 0, brw->shader_time.bo->size);
drm_intel_bo_unmap(brw->shader_time.bo);
}
struct gl_program *prog,
enum shader_time_shader_type type)
{
- struct gl_context *ctx = &brw->ctx;
-
int shader_time_index = brw->shader_time.num_entries++;
assert(shader_time_index < brw->shader_time.max_entries);
brw->shader_time.types[shader_time_index] = type;
- _mesa_reference_shader_program(ctx,
- &brw->shader_time.shader_programs[shader_time_index],
- shader_prog);
+ int id = shader_prog ? shader_prog->Name : prog->Id;
+ const char *name;
+ if (id == 0) {
+ name = "ff";
+ } else if (!shader_prog) {
+ name = "prog";
+ } else if (shader_prog->Label) {
+ name = ralloc_strdup(brw->shader_time.names, shader_prog->Label);
+ } else {
+ name = "glsl";
+ }
- _mesa_reference_program(ctx,
- &brw->shader_time.programs[shader_time_index],
- prog);
+ brw->shader_time.names[shader_time_index] = name;
+ brw->shader_time.ids[shader_time_index] = id;
return shader_time_index;
}
drm_intel_bo_unreference(brw->shader_time.bo);
brw->shader_time.bo = NULL;
}
+
+void
+brw_stage_prog_data_free(const void *p)
+{
+ struct brw_stage_prog_data *prog_data = (struct brw_stage_prog_data *)p;
+
+ ralloc_free(prog_data->param);
+ ralloc_free(prog_data->pull_param);
+ ralloc_free(prog_data->image_param);
+}
+
+void
+brw_dump_ir(const char *stage, struct gl_shader_program *shader_prog,
+ struct gl_shader *shader, struct gl_program *prog)
+{
+ if (shader_prog) {
+ if (shader->ir) {
+ fprintf(stderr,
+ "GLSL IR for native %s shader %d:\n",
+ stage, shader_prog->Name);
+ _mesa_print_ir(stderr, shader->ir, NULL);
+ fprintf(stderr, "\n\n");
+ }
+ } else {
+ fprintf(stderr, "ARB_%s_program %d ir for native %s shader\n",
+ stage, prog->Id, stage);
+ _mesa_print_program(prog);
+ }
+}
+
+void
+brw_setup_tex_for_precompile(struct brw_context *brw,
+ struct brw_sampler_prog_key_data *tex,
+ struct gl_program *prog)
+{
+ const bool has_shader_channel_select = brw->is_haswell || brw->gen >= 8;
+ unsigned sampler_count = _mesa_fls(prog->SamplersUsed);
+ for (unsigned i = 0; i < sampler_count; i++) {
+ if (!has_shader_channel_select && (prog->ShadowSamplers & (1 << i))) {
+ /* Assume DEPTH_TEXTURE_MODE is the default: X, X, X, 1 */
+ tex->swizzles[i] =
+ MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_ONE);
+ } else {
+ /* Color sampler: assume no swizzling. */
+ tex->swizzles[i] = SWIZZLE_XYZW;
+ }
+ }
+}
projects / mesa.git / blobdiff