radeonsi: move gfx10_ngg_export_vertex to si_shader_llvm.c

It's now also used by non-ngg pipeline and older GPUs.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19489>

diff --git a/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c b/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c
index 2392e73..c053d43 100644 (file)
--- a/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c
+++ b/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c
@@ -95,35 +95,6 @@ bool gfx10_ngg_export_prim_early(struct si_shader *shader)
           !gfx10_ngg_writes_user_edgeflags(shader);
 }
 
-void gfx10_ngg_export_vertex(struct ac_shader_abi *abi)
-{
-   struct si_shader_context *ctx = si_shader_context_from_abi(abi);
-   struct si_shader_info *info = &ctx->shader->selector->info;
-   struct si_shader_output_values outputs[PIPE_MAX_SHADER_OUTPUTS];
-   LLVMValueRef *addrs = ctx->abi.outputs;
-
-   unsigned num_outputs = info->num_outputs;
-   /* if needed, nir ngg lower will append primitive id export at last */
-   if (ctx->shader->key.ge.mono.u.vs_export_prim_id)
-      num_outputs++;
-
-   for (unsigned i = 0; i < num_outputs; i++) {
-      if (i < info->num_outputs) {
-         outputs[i].semantic = info->output_semantic[i];
-         outputs[i].vertex_streams = info->output_streams[i];
-      } else {
-         outputs[i].semantic = VARYING_SLOT_PRIMITIVE_ID;
-         outputs[i].vertex_streams = 0;
-      }
-
-      for (unsigned j = 0; j < 4; j++)
-         outputs[i].values[j] =
-            LLVMBuildLoad2(ctx->ac.builder, ctx->ac.f32, addrs[4 * i + j], "");
-   }
-
-   si_llvm_build_vs_exports(ctx, outputs, num_outputs);
-}
-
 static void clamp_gsprims_to_esverts(unsigned *max_gsprims, unsigned max_esverts,
                                      unsigned min_verts_per_prim, bool use_adjacency)
 {

diff --git a/src/gallium/drivers/radeonsi/si_shader_internal.h b/src/gallium/drivers/radeonsi/si_shader_internal.h
index 3207c52..d506647 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader_internal.h
+++ b/src/gallium/drivers/radeonsi/si_shader_internal.h
@@ -183,7 +183,6 @@ void si_fix_resource_usage(struct si_screen *sscreen, struct si_shader *shader);
 LLVMValueRef gfx10_get_thread_id_in_tg(struct si_shader_context *ctx);
 unsigned gfx10_ngg_get_vertices_per_prim(struct si_shader *shader);
 bool gfx10_ngg_export_prim_early(struct si_shader *shader);
-void gfx10_ngg_export_vertex(struct ac_shader_abi *abi);
 unsigned gfx10_ngg_get_scratch_dw_size(struct si_shader *shader);
 bool gfx10_ngg_calculate_subgroup_info(struct si_shader *shader);
 

diff --git a/src/gallium/drivers/radeonsi/si_shader_llvm.c b/src/gallium/drivers/radeonsi/si_shader_llvm.c
index aa9cf56..d809d21 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader_llvm.c
+++ b/src/gallium/drivers/radeonsi/si_shader_llvm.c
@@ -774,6 +774,35 @@ static LLVMValueRef si_llvm_load_sampler_desc(struct ac_shader_abi *abi, LLVMVal
    return index;
 }
 
+static void si_llvm_export_vertex(struct ac_shader_abi *abi)
+{
+   struct si_shader_context *ctx = si_shader_context_from_abi(abi);
+   struct si_shader_info *info = &ctx->shader->selector->info;
+   struct si_shader_output_values outputs[PIPE_MAX_SHADER_OUTPUTS];
+   LLVMValueRef *addrs = ctx->abi.outputs;
+
+   unsigned num_outputs = info->num_outputs;
+   /* if needed, nir lower will append primitive id export at last */
+   if (ctx->shader->key.ge.mono.u.vs_export_prim_id)
+      num_outputs++;
+
+   for (unsigned i = 0; i < num_outputs; i++) {
+      if (i < info->num_outputs) {
+         outputs[i].semantic = info->output_semantic[i];
+         outputs[i].vertex_streams = info->output_streams[i];
+      } else {
+         outputs[i].semantic = VARYING_SLOT_PRIMITIVE_ID;
+         outputs[i].vertex_streams = 0;
+      }
+
+      for (unsigned j = 0; j < 4; j++)
+         outputs[i].values[j] =
+            LLVMBuildLoad2(ctx->ac.builder, ctx->ac.f32, addrs[4 * i + j], "");
+   }
+
+   si_llvm_build_vs_exports(ctx, outputs, num_outputs);
+}
+
 bool si_llvm_translate_nir(struct si_shader_context *ctx, struct si_shader *shader,
                            struct nir_shader *nir, bool free_nir)
 {
@@ -790,7 +819,7 @@ bool si_llvm_translate_nir(struct si_shader_context *ctx, struct si_shader *shad
    ctx->num_images = info->base.num_images;
 
    ctx->abi.intrinsic_load = si_llvm_load_intrinsic;
-   ctx->abi.export_vertex = gfx10_ngg_export_vertex;
+   ctx->abi.export_vertex = si_llvm_export_vertex;
    ctx->abi.load_sampler_desc = si_llvm_load_sampler_desc;
 
    si_llvm_create_main_func(ctx);