From 909184ac9cf59f23803915773f5659f05c161394 Mon Sep 17 00:00:00 2001 From: Alex Smith Date: Mon, 26 Jun 2017 17:17:32 +0100 Subject: [PATCH] ac/nir: Use correct LLVM intrinsics for atomic ops on imageBuffers The buffer intrinsics should be used instead of the image ones. Signed-off-by: Alex Smith Cc: Reviewed-by: Bas Nieuwenhuizen --- src/amd/common/ac_nir_to_llvm.c | 63 ++++++++++++++++++++++------------------- 1 file changed, 34 insertions(+), 29 deletions(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index a02968e..68f0218 100644 --- a/src/amd/common/ac_nir_to_llvm.c +++ b/src/amd/common/ac_nir_to_llvm.c @@ -3400,37 +3400,14 @@ static LLVMValueRef visit_image_atomic(struct nir_to_llvm_context *ctx, int param_count = 0; const nir_variable *var = instr->variables[0]->var; - const char *base_name = "llvm.amdgcn.image.atomic"; const char *atomic_name; - LLVMValueRef coords; - char intrinsic_name[41], coords_type[8]; + char intrinsic_name[41]; const struct glsl_type *type = glsl_without_array(var->type); + MAYBE_UNUSED int length; if (ctx->stage == MESA_SHADER_FRAGMENT) ctx->shader_info->fs.writes_memory = true; - params[param_count++] = get_src(ctx, instr->src[2]); - if (instr->intrinsic == nir_intrinsic_image_atomic_comp_swap) - params[param_count++] = get_src(ctx, instr->src[3]); - - if (glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_BUF) { - params[param_count++] = get_sampler_desc(ctx, instr->variables[0], DESC_BUFFER); - coords = params[param_count++] = LLVMBuildExtractElement(ctx->builder, get_src(ctx, instr->src[0]), - LLVMConstInt(ctx->i32, 0, false), ""); /* vindex */ - params[param_count++] = ctx->i32zero; /* voffset */ - params[param_count++] = ctx->i1false; /* glc */ - params[param_count++] = ctx->i1false; /* slc */ - } else { - bool da = glsl_sampler_type_is_array(type) || - glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_CUBE; - - coords = params[param_count++] = get_image_coords(ctx, instr); - params[param_count++] = get_sampler_desc(ctx, instr->variables[0], DESC_IMAGE); - params[param_count++] = ctx->i1false; /* r128 */ - params[param_count++] = da ? ctx->i1true : ctx->i1false; /* da */ - params[param_count++] = ctx->i1false; /* slc */ - } - switch (instr->intrinsic) { case nir_intrinsic_image_atomic_add: atomic_name = "add"; @@ -3459,11 +3436,39 @@ static LLVMValueRef visit_image_atomic(struct nir_to_llvm_context *ctx, default: abort(); } - build_int_type_name(LLVMTypeOf(coords), - coords_type, sizeof(coords_type)); - MAYBE_UNUSED const int length = snprintf(intrinsic_name, sizeof(intrinsic_name), - "%s.%s.%s", base_name, atomic_name, coords_type); + params[param_count++] = get_src(ctx, instr->src[2]); + if (instr->intrinsic == nir_intrinsic_image_atomic_comp_swap) + params[param_count++] = get_src(ctx, instr->src[3]); + + if (glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_BUF) { + params[param_count++] = get_sampler_desc(ctx, instr->variables[0], DESC_BUFFER); + params[param_count++] = LLVMBuildExtractElement(ctx->builder, get_src(ctx, instr->src[0]), + LLVMConstInt(ctx->i32, 0, false), ""); /* vindex */ + params[param_count++] = ctx->i32zero; /* voffset */ + params[param_count++] = ctx->i1false; /* slc */ + + length = snprintf(intrinsic_name, sizeof(intrinsic_name), + "llvm.amdgcn.buffer.atomic.%s", atomic_name); + } else { + char coords_type[8]; + + bool da = glsl_sampler_type_is_array(type) || + glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_CUBE; + + LLVMValueRef coords = params[param_count++] = get_image_coords(ctx, instr); + params[param_count++] = get_sampler_desc(ctx, instr->variables[0], DESC_IMAGE); + params[param_count++] = ctx->i1false; /* r128 */ + params[param_count++] = da ? ctx->i1true : ctx->i1false; /* da */ + params[param_count++] = ctx->i1false; /* slc */ + + build_int_type_name(LLVMTypeOf(coords), + coords_type, sizeof(coords_type)); + + length = snprintf(intrinsic_name, sizeof(intrinsic_name), + "llvm.amdgcn.image.atomic.%s.%s", atomic_name, coords_type); + } + assert(length < sizeof(intrinsic_name)); return ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->i32, params, param_count, 0); } -- 2.7.4