[Mesa-dev] [PATCH v2 02/20] radeonsi: add shared memory

[Bas Nieuwenhuizen]

Declares the shared memory as a global variable so that
LLVM is aware of it and it does not conflict with passes
like AMDGPUPromoteAlloca.

v2: - Use ctx->i8.
    - Dropped null-check for declare_memory_region.
    - Changed memory region array to single region.

Signed-off-by: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
---
 src/gallium/drivers/radeon/radeon_llvm.h           |  3 +++
 .../drivers/radeon/radeon_setup_tgsi_llvm.c        |  4 ++++
 src/gallium/drivers/radeonsi/si_shader.c           | 27 ++++++++++++++++++++++
 src/gallium/drivers/radeonsi/si_shader.h           |  3 +++
 4 files changed, 37 insertions(+)

diff --git a/src/gallium/drivers/radeon/radeon_llvm.h b/src/gallium/drivers/radeon/radeon_llvm.h
index 0a164bb..3e11b36 100644
--- a/src/gallium/drivers/radeon/radeon_llvm.h
+++ b/src/gallium/drivers/radeon/radeon_llvm.h
@@ -68,6 +68,9 @@ struct radeon_llvm_context {
 			unsigned index,
 			const struct tgsi_full_declaration *decl);
 
+	void (*declare_memory_region)(struct radeon_llvm_context *,
+			const struct tgsi_full_declaration *decl);
+
 	/** This array contains the input values for the shader.  Typically these
 	  * values will be in the form of a target intrinsic that will inform the
 	  * backend how to load the actual inputs to the shader. 
diff --git a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
index fb883cb..0828197 100644
--- a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
+++ b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
@@ -366,6 +366,10 @@ static void emit_declaration(
 		break;
 	}
 
+	case TGSI_FILE_MEMORY:
+		ctx->declare_memory_region(ctx, decl);
+		break;
+
 	default:
 		break;
 	}
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 1ccdcac..5a76435 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -121,6 +121,8 @@ struct si_shader_context
 	LLVMTypeRef v4i32;
 	LLVMTypeRef v4f32;
 	LLVMTypeRef v8i32;
+
+	LLVMValueRef shared_memory;
 };
 
 static struct si_shader_context *si_shader_context(
@@ -1320,6 +1322,30 @@ static void declare_system_value(
 	radeon_bld->system_values[index] = value;
 }
 
+static void declare_compute_memory(struct radeon_llvm_context *radeon_bld,
+                                   const struct tgsi_full_declaration *decl)
+{
+	struct si_shader_context *ctx =
+		si_shader_context(&radeon_bld->soa.bld_base);
+	struct si_shader_selector *sel = ctx->shader->selector;
+	struct gallivm_state *gallivm = &radeon_bld->gallivm;
+
+	LLVMTypeRef i8p = LLVMPointerType(ctx->i8, LOCAL_ADDR_SPACE);
+	LLVMValueRef var;
+
+	assert(decl->Declaration.MemType == TGSI_MEMORY_TYPE_SHARED);
+	assert(decl->Range.First == decl->Range.Last);
+	assert(!ctx->shared_memory);
+
+	var = LLVMAddGlobalInAddressSpace(gallivm->module,
+	                                  LLVMArrayType(ctx->i8, sel->local_size),
+	                                  "compute_lds",
+	                                  LOCAL_ADDR_SPACE);
+	LLVMSetAlignment(var, 4);
+
+	ctx->shared_memory = LLVMBuildBitCast(gallivm->builder, var, i8p, "");
+}
+
 static LLVMValueRef fetch_constant(
 	struct lp_build_tgsi_context *bld_base,
 	const struct tgsi_full_src_register *reg,
@@ -5824,6 +5850,7 @@ int si_compile_tgsi_shader(struct si_screen *sscreen,
 			bld_base->emit_epilogue = si_llvm_return_fs_outputs;
 		break;
 	case TGSI_PROCESSOR_COMPUTE:
+		ctx.radeon_bld.declare_memory_region = declare_compute_memory;
 		break;
 	default:
 		assert(!"Unsupported shader type");
diff --git a/src/gallium/drivers/radeonsi/si_shader.h b/src/gallium/drivers/radeonsi/si_shader.h
index 5043d43..70ae46f 100644
--- a/src/gallium/drivers/radeonsi/si_shader.h
+++ b/src/gallium/drivers/radeonsi/si_shader.h
@@ -222,6 +222,9 @@ struct si_shader_selector {
 	 */
 	unsigned	colors_written_4bit;
 
+	/* CS parameters */
+	unsigned local_size;
+
 	/* masks of "get_unique_index" bits */
 	uint64_t	outputs_written;
 	uint32_t	patch_outputs_written;
-- 
2.8.0