return NULL;
}
- case MESA_GEOMETRY_PROGRAM: {
+ 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);
brw_add_texrect_params(prog);
if (ctx->Const.ShaderCompilerOptions[MESA_SHADER_FRAGMENT].NirOptions) {
- prog->nir = brw_create_nir(brw, NULL, prog, MESA_SHADER_FRAGMENT);
+ prog->nir = brw_create_nir(brw, NULL, prog, MESA_SHADER_FRAGMENT, true);
}
brw_fs_precompile(ctx, NULL, prog);
brw_add_texrect_params(prog);
if (ctx->Const.ShaderCompilerOptions[MESA_SHADER_VERTEX].NirOptions) {
- prog->nir = brw_create_nir(brw, NULL, prog, MESA_SHADER_VERTEX);
+ prog->nir = brw_create_nir(brw, NULL, prog, MESA_SHADER_VERTEX,
+ brw->intelScreen->compiler->scalar_vs);
}
brw_vs_precompile(ctx, NULL, prog);
unsigned bits = (PIPE_CONTROL_DATA_CACHE_INVALIDATE |
PIPE_CONTROL_NO_WRITE |
PIPE_CONTROL_CS_STALL);
- assert(brw->gen >= 7 && brw->gen <= 8);
+ assert(brw->gen >= 7 && brw->gen <= 9);
if (barriers & (GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT |
GL_ELEMENT_ARRAY_BARRIER_BIT |
void brwInitFragProgFuncs( struct dd_function_table *functions )
{
- assert(functions->ProgramStringNotify == _tnl_program_string);
+ /* assert(functions->ProgramStringNotify == _tnl_program_string); */
functions->NewProgram = brwNewProgram;
functions->DeleteProgram = brwDeleteProgram;
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_GS || 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)
{
fprintf(stderr, "%-6s%-18s", stage, name);
- if (shader_num != -1)
+ if (shader_num != 0)
fprintf(stderr, "%4d: ", shader_num);
else
fprintf(stderr, " : ");
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_GS_WRITTEN:
- case ST_GS_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 {
case ST_GS:
case ST_FS8:
case ST_FS16:
+ case ST_CS:
total_by_type[type] += scaled[i];
break;
default:
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;
- struct gl_shader_program *prog = brw->shader_time.shader_programs[i];
if (scaled[i] == 0)
continue;
- int shader_num = -1;
- if (prog) {
- shader_num = prog->Name;
-
- /* The fixed function fragment shader generates GLSL IR with a Name
- * of 0, and nothing else does.
- */
- if (prog->Label) {
- shader_name = prog->Label;
- } else 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:
case ST_FS16:
stage = "fs16";
break;
+ case ST_CS:
+ stage = "cs";
+ break;
default:
stage = "other";
break;
}
fprintf(stderr, "\n");
- print_shader_time_line("total", "vs", -1, total_by_type[ST_VS], total);
- print_shader_time_line("total", "gs", -1, total_by_type[ST_GS], 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);
+ 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;
}
struct gl_shader *shader, struct gl_program *prog)
{
if (shader_prog) {
- 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");
+ 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);