radeonsi: make si_declare_compute_memory() more generic and call for nir

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 9340360376d029b68b5f0851a18507cfc612d5de..825cb9dd0ea2874a5d8e85d58148d90caae7959d 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -2231,16 +2231,13 @@ void si_load_system_value(struct si_shader_context *ctx,
        ctx->system_values[index] = value;
 }
 
-void si_declare_compute_memory(struct si_shader_context *ctx,
-                              const struct tgsi_full_declaration *decl)
+void si_declare_compute_memory(struct si_shader_context *ctx)
 {
        struct si_shader_selector *sel = ctx->shader->selector;
 
        LLVMTypeRef i8p = LLVMPointerType(ctx->i8, AC_LOCAL_ADDR_SPACE);
        LLVMValueRef var;
 
-       assert(decl->Declaration.MemType == TGSI_MEMORY_TYPE_SHARED);
-       assert(decl->Range.First == decl->Range.Last);
        assert(!ctx->ac.lds);
 
        var = LLVMAddGlobalInAddressSpace(ctx->ac.module,
@@ -2252,6 +2249,15 @@ void si_declare_compute_memory(struct si_shader_context *ctx,
        ctx->ac.lds = LLVMBuildBitCast(ctx->ac.builder, var, i8p, "");
 }
 
+void si_tgsi_declare_compute_memory(struct si_shader_context *ctx,
+                                   const struct tgsi_full_declaration *decl)
+{
+       assert(decl->Declaration.MemType == TGSI_MEMORY_TYPE_SHARED);
+       assert(decl->Range.First == decl->Range.Last);
+
+       si_declare_compute_memory(ctx);
+}
+
 static LLVMValueRef load_const_buffer_desc(struct si_shader_context *ctx, int i)
 {
        LLVMValueRef list_ptr = LLVMGetParam(ctx->main_fn,

diff --git a/src/gallium/drivers/radeonsi/si_shader_internal.h b/src/gallium/drivers/radeonsi/si_shader_internal.h
index 02d5fc8e379c4af1f34f984433512b001419e434..571df5597709ce6c4be2978f76389d85c73d386d 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader_internal.h
+++ b/src/gallium/drivers/radeonsi/si_shader_internal.h
@@ -320,8 +320,9 @@ LLVMValueRef si_load_image_desc(struct si_shader_context *ctx,
 void si_load_system_value(struct si_shader_context *ctx,
                          unsigned index,
                          const struct tgsi_full_declaration *decl);
-void si_declare_compute_memory(struct si_shader_context *ctx,
-                              const struct tgsi_full_declaration *decl);
+void si_declare_compute_memory(struct si_shader_context *ctx);
+void si_tgsi_declare_compute_memory(struct si_shader_context *ctx,
+                                   const struct tgsi_full_declaration *decl);
 
 void si_llvm_load_input_vs(
        struct si_shader_context *ctx,

diff --git a/src/gallium/drivers/radeonsi/si_shader_nir.c b/src/gallium/drivers/radeonsi/si_shader_nir.c
index f467ce282af8a4aa5cecf973d73f1487273f234e..f46e5c6aac41a135562675148463ca4cd2cbbbb8 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader_nir.c
+++ b/src/gallium/drivers/radeonsi/si_shader_nir.c
@@ -857,6 +857,10 @@ bool si_nir_build_llvm(struct si_shader_context *ctx, struct nir_shader *nir)
        ctx->num_samplers = util_last_bit(info->samplers_declared);
        ctx->num_images = util_last_bit(info->images_declared);
 
+       if (ctx->shader->selector->local_size) {
+               assert(nir->info.stage == MESA_SHADER_COMPUTE);
+               si_declare_compute_memory(ctx);
+       }
        ac_nir_translate(&ctx->ac, &ctx->abi, nir, NULL);
 
        return true;

diff --git a/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c b/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
index 44294f28d167312b33ed4d83309cff484c2cab1e..8707be504e563587367de8171e26f790e9353722 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
+++ b/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
@@ -851,7 +851,7 @@ static void emit_declaration(struct lp_build_tgsi_context *bld_base,
        }
 
        case TGSI_FILE_MEMORY:
-               si_declare_compute_memory(ctx, decl);
+               si_tgsi_declare_compute_memory(ctx, decl);
                break;
 
        default: