ac: add load_sample_mask_in() to the ABI

Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Bas Nieuwenhuizen <[email protected]>

3 files changed, 17 insertions, 8 deletions

diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index f8103cf1e9b..406f55d8c40 100644
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -4022,9 +4022,12 @@ static LLVMValueRef load_sample_pos(struct ac_nir_context *ctx)
 	return ac_build_gather_values(&ctx->ac, values, 2);
 }
 
-static LLVMValueRef load_sample_mask_in(struct ac_nir_context *ctx)
+static LLVMValueRef load_sample_mask_in(struct ac_shader_abi *abi)
 {
-	uint8_t log2_ps_iter_samples = ctx->nctx->shader_info->info.ps.force_persample ? ctx->nctx->options->key.fs.log2_num_samples : ctx->nctx->options->key.fs.log2_ps_iter_samples;
+	struct nir_to_llvm_context *ctx = nir_to_llvm_context_from_abi(abi);
+	uint8_t log2_ps_iter_samples = ctx->shader_info->info.ps.force_persample ?
+		ctx->options->key.fs.log2_num_samples :
+		ctx->options->key.fs.log2_ps_iter_samples;
 
 	/* The bit pattern matches that used by fixed function fragment
 	 * processing. */
@@ -4040,9 +4043,9 @@ static LLVMValueRef load_sample_mask_in(struct ac_nir_context *ctx)
 	uint32_t ps_iter_mask = ps_iter_masks[log2_ps_iter_samples];
 
 	LLVMValueRef result, sample_id;
-	sample_id = unpack_param(&ctx->ac, ctx->abi->ancillary, 8, 4);
+	sample_id = unpack_param(&ctx->ac, abi->ancillary, 8, 4);
 	sample_id = LLVMBuildShl(ctx->ac.builder, LLVMConstInt(ctx->ac.i32, ps_iter_mask, false), sample_id, "");
-	result = LLVMBuildAnd(ctx->ac.builder, sample_id, ctx->abi->sample_coverage, "");
+	result = LLVMBuildAnd(ctx->ac.builder, sample_id, abi->sample_coverage, "");
 	return result;
 }
 
@@ -4353,10 +4356,7 @@ static void visit_intrinsic(struct ac_nir_context *ctx,
 		result = load_sample_pos(ctx);
 		break;
 	case nir_intrinsic_load_sample_mask_in:
-		if (ctx->nctx)
-			result = load_sample_mask_in(ctx);
-		else
-			result = ctx->abi->sample_coverage;
+		result = ctx->abi->load_sample_mask_in(ctx->abi);
 		break;
 	case nir_intrinsic_load_frag_coord: {
 		LLVMValueRef values[4] = {
@@ -6827,6 +6827,7 @@ LLVMModuleRef ac_translate_nir_to_llvm(LLVMTargetMachineRef tm,
 			shader_info->fs.can_discard = shaders[i]->info.fs.uses_discard;
 			ctx.abi.lookup_interp_param = lookup_interp_param;
 			ctx.abi.load_sample_position = load_sample_position;
+			ctx.abi.load_sample_mask_in = load_sample_mask_in;
 		}
 
 		if (i)
diff --git a/src/amd/common/ac_shader_abi.h b/src/amd/common/ac_shader_abi.h
index d8fea18b351..c5f7fcd2f87 100644
--- a/src/amd/common/ac_shader_abi.h
+++ b/src/amd/common/ac_shader_abi.h
@@ -165,6 +165,8 @@ struct ac_shader_abi {
 
 	LLVMValueRef (*load_local_group_size)(struct ac_shader_abi *abi);
 
+	LLVMValueRef (*load_sample_mask_in)(struct ac_shader_abi *abi);
+
 	/* Whether to clamp the shadow reference value to [0,1]on VI. Radeonsi currently
 	 * uses it due to promoting D16 to D32, but radv needs it off. */
 	bool clamp_shadow_reference;
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 4ff00db55f3..ca43530d924 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -1934,6 +1934,11 @@ static LLVMValueRef load_sample_position(struct ac_shader_abi *abi, LLVMValueRef
 	return lp_build_gather_values(&ctx->gallivm, pos, 4);
 }
 
+static LLVMValueRef load_sample_mask_in(struct ac_shader_abi *abi)
+{
+	return abi->sample_coverage;
+}
+
 static LLVMValueRef si_load_tess_coord(struct ac_shader_abi *abi,
 				       LLVMTypeRef type,
 				       unsigned num_components)
@@ -5960,6 +5965,7 @@ static bool si_compile_tgsi_main(struct si_shader_context *ctx,
 		bld_base->emit_epilogue = si_tgsi_emit_epilogue;
 		ctx->abi.lookup_interp_param = si_nir_lookup_interp_param;
 		ctx->abi.load_sample_position = load_sample_position;
+		ctx->abi.load_sample_mask_in = load_sample_mask_in;
 		break;
 	case PIPE_SHADER_COMPUTE:
 		ctx->abi.load_local_group_size = get_block_size;