[Mesa-dev] [PATCH 09/11] ac/radeonsi: add num_work_groups to the abi

[Timothy Arceri]

---
 src/amd/common/ac_nir_to_llvm.c                   | 5 ++---
 src/amd/common/ac_shader_abi.h                    | 1 +
 src/gallium/drivers/radeonsi/si_shader.c          | 4 ++--
 src/gallium/drivers/radeonsi/si_shader_internal.h | 1 -
 4 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index bf36c7ef84..28f73a30b3 100644
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -90,7 +90,6 @@ struct nir_to_llvm_context {
 	LLVMValueRef ring_offsets;
 	LLVMValueRef push_constants;
 	LLVMValueRef view_index;
-	LLVMValueRef num_work_groups;
 	LLVMValueRef tg_size;
 
 	LLVMValueRef vertex_buffers;
@@ -780,7 +779,7 @@ static void create_function(struct nir_to_llvm_context *ctx,
 
 		if (ctx->shader_info->info.cs.uses_grid_size) {
 			add_arg(&args, ARG_SGPR, ctx->ac.v3i32,
-				&ctx->num_work_groups);
+				&ctx->abi.num_work_groups);
 		}
 
 		for (int i = 0; i < 3; i++) {
@@ -4385,7 +4384,7 @@ static void visit_intrinsic(struct ac_nir_context *ctx,
 		result = ctx->abi->instance_id;
 		break;
 	case nir_intrinsic_load_num_work_groups:
-		result = ctx->nctx->num_work_groups;
+		result = ctx->abi->num_work_groups;
 		break;
 	case nir_intrinsic_load_local_invocation_index:
 		result = visit_load_local_invocation_index(ctx->nctx);
diff --git a/src/amd/common/ac_shader_abi.h b/src/amd/common/ac_shader_abi.h
index 55842ea342..9499c1a15f 100644
--- a/src/amd/common/ac_shader_abi.h
+++ b/src/amd/common/ac_shader_abi.h
@@ -56,6 +56,7 @@ struct ac_shader_abi {
 	LLVMValueRef prim_mask;
 	/* CS */
 	LLVMValueRef local_invocation_ids;
+	LLVMValueRef num_work_groups;
 	LLVMValueRef workgroup_ids[3];
 
 	/* For VS and PS: pre-loaded shader inputs.
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 08088450d7..797450845b 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -2139,7 +2139,7 @@ void si_load_system_value(struct si_shader_context *ctx,
 		break;
 
 	case TGSI_SEMANTIC_GRID_SIZE:
-		value = LLVMGetParam(ctx->main_fn, ctx->param_grid_size);
+		value = ctx->abi.num_work_groups;
 		break;
 
 	case TGSI_SEMANTIC_BLOCK_SIZE:
@@ -4946,7 +4946,7 @@ static void create_function(struct si_shader_context *ctx)
 		declare_global_desc_pointers(ctx, &fninfo);
 		declare_per_stage_desc_pointers(ctx, &fninfo, true);
 		if (shader->selector->info.uses_grid_size)
-			ctx->param_grid_size = add_arg(&fninfo, ARG_SGPR, v3i32);
+			add_arg_assign(&fninfo, ARG_SGPR, v3i32, &ctx->abi.num_work_groups);
 		if (shader->selector->info.uses_block_size)
 			ctx->param_block_size = add_arg(&fninfo, ARG_SGPR, v3i32);
 
diff --git a/src/gallium/drivers/radeonsi/si_shader_internal.h b/src/gallium/drivers/radeonsi/si_shader_internal.h
index c0d83fb93e..aec0fe3102 100644
--- a/src/gallium/drivers/radeonsi/si_shader_internal.h
+++ b/src/gallium/drivers/radeonsi/si_shader_internal.h
@@ -184,7 +184,6 @@ struct si_shader_context {
 	int param_gs_vtx23_offset; /* in dwords (GFX9) */
 	int param_gs_vtx45_offset; /* in dwords (GFX9) */
 	/* CS */
-	int param_grid_size;
 	int param_block_size;
 
 	LLVMTargetMachineRef tm;
-- 
2.14.3