summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorNicolai Hähnle <nicolai.haehnle@amd.com>2017-05-10 13:26:39 +0200
committerNicolai Hähnle <nicolai.haehnle@amd.com>2017-05-12 10:46:05 +0200
commitcfe6e30f1b74cc5617791aad9188f6cd4a07b78d (patch)
treeadf727f4324dff3223b36c758d55c435ba2324c0
parent7091fe887b3b676fe70c44ff10e5027f4b44513d (diff)
radeonsi: skip generic out/in indices without a shader IO index
OpenGL uses at most 32 generic outputs/inputs in any stage, and they always have a shader IO index and therefore fit into the outputs_written/ inputs_read/kill_outputs fields. However, Nine uses semantic indices more liberally. We support that in VS-PS pipelines, except that the optimization of killing outputs must be skipped. Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de> Reviewed-by: Marek Olšák <marek.olsak@amd.com>
-rw-r--r--src/gallium/drivers/radeonsi/si_shader.c4
-rw-r--r--src/gallium/drivers/radeonsi/si_state_shaders.c6
2 files changed, 9 insertions, 1 deletions
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index bdc9cc10fa..98292867c8 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -2293,6 +2293,10 @@ static void si_llvm_export_vs(struct lp_build_tgsi_context *bld_base,
case TGSI_SEMANTIC_EDGEFLAG:
break;
case TGSI_SEMANTIC_GENERIC:
+ /* don't process indices the function can't handle */
+ if (semantic_index >= SI_MAX_IO_GENERIC)
+ break;
+ /* fall through */
case TGSI_SEMANTIC_CLIPDIST:
if (shader->key.opt.hw_vs.kill_outputs &
(1ull << si_shader_io_get_unique_index(semantic_name, semantic_index)))
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index 8193fe041d..6020bec9c0 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -2032,8 +2032,12 @@ static void *si_create_shader_selector(struct pipe_context *ctx,
unsigned index = sel->info.input_semantic_index[i];
switch (name) {
- case TGSI_SEMANTIC_CLIPDIST:
case TGSI_SEMANTIC_GENERIC:
+ /* don't process indices the function can't handle */
+ if (index >= SI_MAX_IO_GENERIC)
+ break;
+ /* fall through */
+ case TGSI_SEMANTIC_CLIPDIST:
sel->inputs_read |=
1llu << si_shader_io_get_unique_index(name, index);
break;