From: Marek Olšák <[email protected]>
also, screen wasn't initialized for compute shaders
---
src/gallium/drivers/radeonsi/si_compute.c | 1 +
src/gallium/drivers/radeonsi/si_shader.c | 13 ++++++++++++-
2 files changed, 13 insertions(+), 1 deletion(-)
diff --git a/src/gallium/drivers/radeonsi/si_compute.c
b/src/gallium/drivers/radeonsi/si_compute.c
index 33ebe2e..2b2efae 100644
--- a/src/gallium/drivers/radeonsi/si_compute.c
+++ b/src/gallium/drivers/radeonsi/si_compute.c
@@ -106,20 +106,21 @@ static void si_create_compute_state_async(void *job, int
thread_index)
assert(thread_index < ARRAY_SIZE(program->screen->tm));
tm = program->screen->tm[thread_index];
if (!debug->async)
debug = NULL;
} else {
tm = program->compiler_ctx_state.tm;
}
memset(&sel, 0, sizeof(sel));
+ sel.screen = program->screen;
tgsi_scan_shader(program->tokens, &sel.info);
sel.tokens = program->tokens;
sel.type = PIPE_SHADER_COMPUTE;
sel.local_size = program->local_size;
program->shader.selector = &sel;
program->shader.is_monolithic = true;
if (si_shader_create(program->screen, tm, &program->shader, debug)) {
program->shader.compilation_failed = true;
diff --git a/src/gallium/drivers/radeonsi/si_shader.c
b/src/gallium/drivers/radeonsi/si_shader.c
index 8b21ff7..dab95e2 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -92,20 +92,31 @@ static void si_build_ps_epilog_function(struct
si_shader_context *ctx,
/* The VS location of the PrimitiveID input is the same in the epilog,
* so that the main shader part doesn't have to move it.
*/
#define VS_EPILOG_PRIMID_LOC 2
enum {
CONST_ADDR_SPACE = 2,
LOCAL_ADDR_SPACE = 3,
};
+static bool is_merged_shader(struct si_shader *shader)
+{
+ if (shader->selector->screen->b.chip_class <= VI)
+ return false;
+
+ return shader->key.as_ls ||
+ shader->key.as_es ||
+ shader->selector->type == PIPE_SHADER_TESS_CTRL ||
+ shader->selector->type == PIPE_SHADER_GEOMETRY;
+}
+
/**
* Returns a unique index for a semantic name and index. The index must be
* less than 64, so that a 64-bit bitmask of used inputs or outputs can be
* calculated.
*/
unsigned si_shader_io_get_unique_index(unsigned semantic_name, unsigned index)
{
switch (semantic_name) {
case TGSI_SEMANTIC_POSITION:
return 0;
@@ -7818,21 +7829,21 @@ int si_compile_tgsi_shader(struct si_screen *sscreen,
/* Just terminate the process, because dependent
* shaders can hang due to bad input data, but use
* the env var to allow shader-db to work.
*/
if (!debug_get_bool_option("SI_PASS_BAD_SHADERS",
false))
abort();
}
}
/* Add the scratch offset to input SGPRs. */
- if (shader->config.scratch_bytes_per_wave)
+ if (shader->config.scratch_bytes_per_wave && !is_merged_shader(shader))
shader->info.num_input_sgprs += 1; /* scratch byte offset */
/* Calculate the number of fragment input VGPRs. */
if (ctx.type == PIPE_SHADER_FRAGMENT) {
shader->info.num_input_vgprs = 0;
shader->info.face_vgpr_index = -1;
if (G_0286CC_PERSP_SAMPLE_ENA(shader->config.spi_ps_input_addr))
shader->info.num_input_vgprs += 2;
if (G_0286CC_PERSP_CENTER_ENA(shader->config.spi_ps_input_addr))
--
2.7.4
_______________________________________________
mesa-dev mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
