ac: add emit_vertex to the abi

Reviewed-by: Marek Olšák <[email protected]>

3 files changed, 31 insertions, 19 deletions

diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index f922b32bf7b..3d9f613d93b 100644
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -3934,14 +3934,13 @@ static LLVMValueRef visit_interp(struct nir_to_llvm_context *ctx,
 }
 
 static void
-visit_emit_vertex(struct nir_to_llvm_context *ctx,
-		  const nir_intrinsic_instr *instr)
+visit_emit_vertex(struct ac_shader_abi *abi, unsigned stream, LLVMValueRef *addrs)
 {
 	LLVMValueRef gs_next_vertex;
 	LLVMValueRef can_emit;
 	int idx;
+	struct nir_to_llvm_context *ctx = nir_to_llvm_context_from_abi(abi);
 
-	assert(instr->const_index[0] == 0);
 	/* Write vertex attribute values to GSVS ring */
 	gs_next_vertex = LLVMBuildLoad(ctx->builder,
 				       ctx->gs_next_vertex,
@@ -3959,7 +3958,7 @@ visit_emit_vertex(struct nir_to_llvm_context *ctx,
 	/* loop num outputs */
 	idx = 0;
 	for (unsigned i = 0; i < RADEON_LLVM_MAX_OUTPUTS; ++i) {
-		LLVMValueRef *out_ptr = &ctx->nir->outputs[i * 4];
+		LLVMValueRef *out_ptr = &addrs[i * 4];
 		int length = 4;
 		int slot = idx;
 		int slot_inc = 1;
@@ -4197,7 +4196,8 @@ static void visit_intrinsic(struct ac_nir_context *ctx,
 		result = visit_interp(ctx->nctx, instr);
 		break;
 	case nir_intrinsic_emit_vertex:
-		visit_emit_vertex(ctx->nctx, instr);
+		assert(instr->const_index[0] == 0);
+		ctx->abi->emit_vertex(ctx->abi, 0, ctx->outputs);
 		break;
 	case nir_intrinsic_end_primitive:
 		visit_end_primitive(ctx->nctx, instr);
@@ -6527,6 +6527,7 @@ LLVMModuleRef ac_translate_nir_to_llvm(LLVMTargetMachineRef tm,
 
 	ctx.abi.inputs = &ctx.inputs[0];
 	ctx.abi.emit_outputs = handle_shader_outputs_post;
+	ctx.abi.emit_vertex = visit_emit_vertex;
 	ctx.abi.load_ssbo = radv_load_ssbo;
 	ctx.abi.load_sampler_desc = radv_get_sampler_desc;
 	ctx.abi.clamp_shadow_reference = false;
diff --git a/src/amd/common/ac_shader_abi.h b/src/amd/common/ac_shader_abi.h
index 14517d55700..27586d0212f 100644
--- a/src/amd/common/ac_shader_abi.h
+++ b/src/amd/common/ac_shader_abi.h
@@ -58,6 +58,10 @@ struct ac_shader_abi {
 			     unsigned max_outputs,
 			     LLVMValueRef *addrs);
 
+	void (*emit_vertex)(struct ac_shader_abi *abi,
+			    unsigned stream,
+			    LLVMValueRef *addrs);
+
 	LLVMValueRef (*load_ubo)(struct ac_shader_abi *abi, LLVMValueRef index);
 
 	/**
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index afb9c10a79d..3293dd44c63 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -4038,15 +4038,14 @@ static unsigned si_llvm_get_stream(struct lp_build_tgsi_context *bld_base,
 }
 
 /* Emit one vertex from the geometry shader */
-static void si_llvm_emit_vertex(
-	const struct lp_build_tgsi_action *action,
-	struct lp_build_tgsi_context *bld_base,
-	struct lp_build_emit_data *emit_data)
+static void si_llvm_emit_vertex(struct ac_shader_abi *abi,
+				unsigned stream,
+				LLVMValueRef *addrs)
 {
-	struct si_shader_context *ctx = si_shader_context(bld_base);
-	struct lp_build_context *uint = &bld_base->uint_bld;
+	struct si_shader_context *ctx = si_shader_context_from_abi(abi);
+	struct tgsi_shader_info *info = &ctx->shader->selector->info;
+	struct lp_build_context *uint = &ctx->bld_base.uint_bld;
 	struct si_shader *shader = ctx->shader;
-	struct tgsi_shader_info *info = &shader->selector->info;
 	struct lp_build_if_state if_state;
 	LLVMValueRef soffset = LLVMGetParam(ctx->main_fn,
 					    ctx->param_gs2vs_offset);
@@ -4054,9 +4053,6 @@ static void si_llvm_emit_vertex(
 	LLVMValueRef can_emit;
 	unsigned chan, offset;
 	int i;
-	unsigned stream;
-
-	stream = si_llvm_get_stream(bld_base, emit_data);
 
 	/* Write vertex attribute values to GSVS ring */
 	gs_next_vertex = LLVMBuildLoad(ctx->ac.builder,
@@ -4084,14 +4080,12 @@ static void si_llvm_emit_vertex(
 
 	offset = 0;
 	for (i = 0; i < info->num_outputs; i++) {
-		LLVMValueRef *out_ptr = ctx->outputs[i];
-
 		for (chan = 0; chan < 4; chan++) {
 			if (!(info->output_usagemask[i] & (1 << chan)) ||
 			    ((info->output_streams[i] >> (2 * chan)) & 3) != stream)
 				continue;
 
-			LLVMValueRef out_val = LLVMBuildLoad(ctx->ac.builder, out_ptr[chan], "");
+			LLVMValueRef out_val = LLVMBuildLoad(ctx->ac.builder, addrs[4 * i + chan], "");
 			LLVMValueRef voffset =
 				LLVMConstInt(ctx->i32, offset *
 					     shader->selector->gs_max_out_vertices, 0);
@@ -4122,6 +4116,18 @@ static void si_llvm_emit_vertex(
 		lp_build_endif(&if_state);
 }
 
+/* Emit one vertex from the geometry shader */
+static void si_tgsi_emit_vertex(
+	const struct lp_build_tgsi_action *action,
+	struct lp_build_tgsi_context *bld_base,
+	struct lp_build_emit_data *emit_data)
+{
+	struct si_shader_context *ctx = si_shader_context(bld_base);
+	unsigned stream = si_llvm_get_stream(bld_base, emit_data);
+
+	si_llvm_emit_vertex(&ctx->abi, stream, ctx->outputs[0]);
+}
+
 /* Cut one primitive from the geometry shader */
 static void si_llvm_emit_primitive(
 	const struct lp_build_tgsi_action *action,
@@ -5644,7 +5650,7 @@ static void si_init_shader_ctx(struct si_shader_context *ctx,
 	bld_base->op_actions[TGSI_OPCODE_READ_INVOC].fetch_args = read_invoc_fetch_args;
 	bld_base->op_actions[TGSI_OPCODE_READ_INVOC].emit = read_lane_emit;
 
-	bld_base->op_actions[TGSI_OPCODE_EMIT].emit = si_llvm_emit_vertex;
+	bld_base->op_actions[TGSI_OPCODE_EMIT].emit = si_tgsi_emit_vertex;
 	bld_base->op_actions[TGSI_OPCODE_ENDPRIM].emit = si_llvm_emit_primitive;
 	bld_base->op_actions[TGSI_OPCODE_BARRIER].emit = si_llvm_emit_barrier;
 }
@@ -5758,6 +5764,7 @@ static bool si_compile_tgsi_main(struct si_shader_context *ctx,
 		break;
 	case PIPE_SHADER_GEOMETRY:
 		bld_base->emit_fetch_funcs[TGSI_FILE_INPUT] = fetch_input_gs;
+		ctx->abi.emit_vertex = si_llvm_emit_vertex;
 		bld_base->emit_epilogue = si_llvm_emit_gs_epilogue;
 		break;
 	case PIPE_SHADER_FRAGMENT: