radv/gfx10: fix storing/loading NGG stream outputs for VS and TES

The LDS storage allocated for stream outputs is 4 * N, where N
is the number of outputs. So, we have to store/load with N as index
and not with the output location as index.

This doesn't fix anything known but it should fix out-of-bounds
access and it also reduces the number of outputs written to the
LDS storage.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

diff --git a/src/amd/vulkan/radv_nir_to_llvm.c b/src/amd/vulkan/radv_nir_to_llvm.c
index 313a5a6..8339bc0 100644 (file)
--- a/src/amd/vulkan/radv_nir_to_llvm.c
+++ b/src/amd/vulkan/radv_nir_to_llvm.c
@@ -3202,12 +3202,11 @@ static void build_streamout_vertex(struct radv_shader_context *ctx,
                if (stream != output->stream)
                        continue;
 
-               unsigned loc = output->location;
                struct radv_shader_output_values out = {};
 
                for (unsigned comp = 0; comp < 4; comp++) {
                        tmp = ac_build_gep0(&ctx->ac, vertexptr,
-                                           LLVMConstInt(ctx->ac.i32, 4 * loc + comp, false));
+                                           LLVMConstInt(ctx->ac.i32, 4 * i + comp, false));
                        out.values[comp] = LLVMBuildLoad(builder, tmp, "");
                }
 
@@ -3576,6 +3575,7 @@ static LLVMValueRef ngg_nogs_vertex_ptr(struct radv_shader_context *ctx,
 static void
 handle_ngg_outputs_post_1(struct radv_shader_context *ctx)
 {
+       struct radv_streamout_info *so = &ctx->shader_info->so;
        LLVMBuilderRef builder = ctx->ac.builder;
        LLVMValueRef vertex_ptr = NULL;
        LLVMValueRef tmp, tmp2;
@@ -3588,15 +3588,17 @@ handle_ngg_outputs_post_1(struct radv_shader_context *ctx)
 
        vertex_ptr = ngg_nogs_vertex_ptr(ctx, get_thread_id_in_tg(ctx));
 
-       for (unsigned i = 0; i < AC_LLVM_MAX_OUTPUTS; ++i) {
-               if (!(ctx->output_mask & (1ull << i)))
-                       continue;
+       for (unsigned i = 0; i < so->num_outputs; ++i) {
+               struct radv_stream_output *output =
+                       &ctx->shader_info->so.outputs[i];
 
-               for (unsigned j = 0; j < 4; j++) {
+               unsigned loc = output->location;
+
+               for (unsigned comp = 0; comp < 4; comp++) {
                        tmp = ac_build_gep0(&ctx->ac, vertex_ptr,
-                                           LLVMConstInt(ctx->ac.i32, 4 * i + j, false));
+                                           LLVMConstInt(ctx->ac.i32, 4 * i + comp, false));
                        tmp2 = LLVMBuildLoad(builder,
-                                            ctx->abi.outputs[4 * i + j], "");
+                                            ctx->abi.outputs[4 * loc + comp], "");
                        tmp2 = ac_to_integer(&ctx->ac, tmp2);
                        LLVMBuildStore(builder, tmp2, tmp);
                }