radv/ac: cleanup ddxy emission

This cleans up the ddxy emission along the same lines as
radeonsi. It also means we don't use LDS on VI chips we
use the dspermute interface, it also removes some duplicated
code.

Reviewed-by: Bas Nieuwenhuizen <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>

1 files changed, 43 insertions, 93 deletions

diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index 1d71795426c..17cefea34a9 100644
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -130,6 +130,8 @@ struct nir_to_llvm_context {
 	bool has_ddxy;
 	unsigned num_clips;
 	unsigned num_culls;
+
+	bool has_ds_bpermute;
 };
 
 struct ac_tex_info {
@@ -377,14 +379,23 @@ static LLVMValueRef to_float(struct nir_to_llvm_context *ctx, LLVMValueRef v)
 	return v;
 }
 
+static LLVMValueRef build_gep0(struct nir_to_llvm_context *ctx,
+			       LLVMValueRef base_ptr, LLVMValueRef index)
+{
+	LLVMValueRef indices[2] = {
+		ctx->i32zero,
+		index,
+	};
+	return LLVMBuildGEP(ctx->builder, base_ptr,
+			    indices, 2, "");
+}
+
 static LLVMValueRef build_indexed_load(struct nir_to_llvm_context *ctx,
 				       LLVMValueRef base_ptr, LLVMValueRef index,
 				       bool uniform)
 {
 	LLVMValueRef pointer;
-	LLVMValueRef indices[] = {ctx->i32zero, index};
-
-	pointer = LLVMBuildGEP(ctx->builder, base_ptr, indices, 2, "");
+	pointer = build_gep0(ctx, base_ptr, index);
 	if (uniform)
 		LLVMSetMetadata(pointer, ctx->uniform_md_kind, ctx->empty_md);
 	return LLVMBuildLoad(ctx->builder, pointer, "");
@@ -1132,55 +1143,44 @@ static LLVMValueRef get_thread_id(struct nir_to_llvm_context *ctx)
 #define TID_MASK_TOP      0xfffffffd
 #define TID_MASK_LEFT     0xfffffffe
 static LLVMValueRef emit_ddxy(struct nir_to_llvm_context *ctx,
-			      nir_alu_instr *instr,
+			      nir_op op,
 			      LLVMValueRef src0)
 {
-	LLVMValueRef indices[2];
-	LLVMValueRef store_ptr, load_ptr0, load_ptr1;
 	LLVMValueRef tl, trbl, result;
 	LLVMValueRef tl_tid, trbl_tid;
 	LLVMValueRef args[2];
+	LLVMValueRef thread_id;
 	unsigned mask;
 	int idx;
 	ctx->has_ddxy = true;
-	if (!ctx->lds)
+
+	if (!ctx->lds && !ctx->has_ds_bpermute)
 		ctx->lds = LLVMAddGlobalInAddressSpace(ctx->module,
 						       LLVMArrayType(ctx->i32, 64),
 						       "ddxy_lds", LOCAL_ADDR_SPACE);
 
-	indices[0] = ctx->i32zero;
-	indices[1] = get_thread_id(ctx);
-	store_ptr = LLVMBuildGEP(ctx->builder, ctx->lds,
-				 indices, 2, "");
-
-	if (instr->op == nir_op_fddx_fine || instr->op == nir_op_fddx)
+	thread_id = get_thread_id(ctx);
+	if (op == nir_op_fddx_fine || op == nir_op_fddx)
 		mask = TID_MASK_LEFT;
-	else if (instr->op == nir_op_fddy_fine || instr->op == nir_op_fddy)
+	else if (op == nir_op_fddy_fine || op == nir_op_fddy)
 		mask = TID_MASK_TOP;
 	else
 		mask = TID_MASK_TOP_LEFT;
 
-	tl_tid = LLVMBuildAnd(ctx->builder, indices[1],
+	tl_tid = LLVMBuildAnd(ctx->builder, thread_id,
 			      LLVMConstInt(ctx->i32, mask, false), "");
-	indices[1] = tl_tid;
-	load_ptr0 = LLVMBuildGEP(ctx->builder, ctx->lds,
-				 indices, 2, "");
-
 	/* for DDX we want to next X pixel, DDY next Y pixel. */
-	if (instr->op == nir_op_fddx_fine ||
-	    instr->op == nir_op_fddx_coarse ||
-	    instr->op == nir_op_fddx)
+	if (op == nir_op_fddx_fine ||
+	    op == nir_op_fddx_coarse ||
+	    op == nir_op_fddx)
 		idx = 1;
 	else
 		idx = 2;
 
-	trbl_tid = LLVMBuildAdd(ctx->builder, indices[1],
+	trbl_tid = LLVMBuildAdd(ctx->builder, tl_tid,
 				LLVMConstInt(ctx->i32, idx, false), "");
-	indices[1] = trbl_tid;
-	load_ptr1 = LLVMBuildGEP(ctx->builder, ctx->lds,
-				 indices, 2, "");
 
-	if (ctx->options->family >= CHIP_TONGA) {
+	if (ctx->has_ds_bpermute) {
 		args[0] = LLVMBuildMul(ctx->builder, tl_tid,
 				       LLVMConstInt(ctx->i32, 4, false), "");
 		args[1] = src0;
@@ -1194,8 +1194,13 @@ static LLVMValueRef emit_ddxy(struct nir_to_llvm_context *ctx,
 					   ctx->i32, args, 2,
 					   AC_FUNC_ATTR_READNONE);
 	} else {
-		LLVMBuildStore(ctx->builder, src0, store_ptr);
+		LLVMValueRef store_ptr, load_ptr0, load_ptr1;
+
+		store_ptr = build_gep0(ctx, ctx->lds, thread_id);
+		load_ptr0 = build_gep0(ctx, ctx->lds, tl_tid);
+		load_ptr1 = build_gep0(ctx, ctx->lds, trbl_tid);
 
+		LLVMBuildStore(ctx->builder, src0, store_ptr);
 		tl = LLVMBuildLoad(ctx->builder, load_ptr0, "");
 		trbl = LLVMBuildLoad(ctx->builder, load_ptr1, "");
 	}
@@ -1214,72 +1219,15 @@ static LLVMValueRef emit_ddxy_interp(
 	struct nir_to_llvm_context *ctx,
 	LLVMValueRef interp_ij)
 {
-	LLVMValueRef indices[2];
-	LLVMValueRef store_ptr, load_ptr_x, load_ptr_y, load_ptr_ddx, load_ptr_ddy, temp, temp2;
-	LLVMValueRef tl, tr, bl, result[4];
-	unsigned c;
-
-	if (!ctx->lds)
-		ctx->lds = LLVMAddGlobalInAddressSpace(ctx->module,
-						       LLVMArrayType(ctx->i32, 64),
-						       "ddxy_lds", LOCAL_ADDR_SPACE);
-
-	indices[0] = ctx->i32zero;
-	indices[1] = get_thread_id(ctx);
-	store_ptr = LLVMBuildGEP(ctx->builder, ctx->lds,
-				 indices, 2, "");
-
-	temp = LLVMBuildAnd(ctx->builder, indices[1],
-			    LLVMConstInt(ctx->i32, TID_MASK_LEFT, false), "");
-
-	temp2 = LLVMBuildAnd(ctx->builder, indices[1],
-			     LLVMConstInt(ctx->i32, TID_MASK_TOP, false), "");
-
-	indices[1] = temp;
-	load_ptr_x = LLVMBuildGEP(ctx->builder, ctx->lds,
-				  indices, 2, "");
-
-	indices[1] = temp2;
-	load_ptr_y = LLVMBuildGEP(ctx->builder, ctx->lds,
-				  indices, 2, "");
-
-	indices[1] = LLVMBuildAdd(ctx->builder, temp,
-				  LLVMConstInt(ctx->i32, 1, false), "");
-	load_ptr_ddx = LLVMBuildGEP(ctx->builder, ctx->lds,
-				   indices, 2, "");
-
-	indices[1] = LLVMBuildAdd(ctx->builder, temp2,
-				  LLVMConstInt(ctx->i32, 2, false), "");
-	load_ptr_ddy = LLVMBuildGEP(ctx->builder, ctx->lds,
-				   indices, 2, "");
-
-	for (c = 0; c < 2; ++c) {
-		LLVMValueRef store_val;
-		LLVMValueRef c_ll = LLVMConstInt(ctx->i32, c, false);
-
-		store_val = LLVMBuildExtractElement(ctx->builder,
-						    interp_ij, c_ll, "");
-		LLVMBuildStore(ctx->builder,
-			       store_val,
-			       store_ptr);
-
-		tl = LLVMBuildLoad(ctx->builder, load_ptr_x, "");
-		tl = LLVMBuildBitCast(ctx->builder, tl, ctx->f32, "");
-
-		tr = LLVMBuildLoad(ctx->builder, load_ptr_ddx, "");
-		tr = LLVMBuildBitCast(ctx->builder, tr, ctx->f32, "");
-
-		result[c] = LLVMBuildFSub(ctx->builder, tr, tl, "");
-
-		tl = LLVMBuildLoad(ctx->builder, load_ptr_y, "");
-		tl = LLVMBuildBitCast(ctx->builder, tl, ctx->f32, "");
-
-		bl = LLVMBuildLoad(ctx->builder, load_ptr_ddy, "");
-		bl = LLVMBuildBitCast(ctx->builder, bl, ctx->f32, "");
+	LLVMValueRef result[4], a;
+	unsigned i;
 
-		result[c + 2] = LLVMBuildFSub(ctx->builder, bl, tl, "");
+	for (i = 0; i < 2; i++) {
+		a = LLVMBuildExtractElement(ctx->builder, interp_ij,
+					    LLVMConstInt(ctx->i32, i, false), "");
+		result[i] = emit_ddxy(ctx, nir_op_fddx, a);
+		result[2+i] = emit_ddxy(ctx, nir_op_fddy, a);
 	}
-
 	return build_gather_values(ctx, result, 4);
 }
 
@@ -1593,7 +1541,7 @@ static void visit_alu(struct nir_to_llvm_context *ctx, nir_alu_instr *instr)
 	case nir_op_fddy_fine:
 	case nir_op_fddx_coarse:
 	case nir_op_fddy_coarse:
-		result = emit_ddxy(ctx, instr, src[0]);
+		result = emit_ddxy(ctx, instr->op, src[0]);
 		break;
 	default:
 		fprintf(stderr, "Unknown NIR alu instr: ");
@@ -4558,6 +4506,8 @@ LLVMModuleRef ac_translate_nir_to_llvm(LLVMTargetMachineRef tm,
 	ctx.context = LLVMContextCreate();
 	ctx.module = LLVMModuleCreateWithNameInContext("shader", ctx.context);
 
+	ctx.has_ds_bpermute = ctx.options->chip_class >= VI;
+
 	memset(shader_info, 0, sizeof(*shader_info));
 
 	LLVMSetTarget(ctx.module, "amdgcn--");