71 lines · plain
1; RUN: opt -S -passes='dxil-data-scalarization' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s --check-prefixes=SCHECK,CHECK2; RUN: opt -S -passes='dxil-data-scalarization,dxil-flatten-arrays' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s --check-prefixes=FCHECK,CHECK3 4@"arrayofVecData" = local_unnamed_addr addrspace(3) global [8 x <4 x i32>] zeroinitializer, align 165@"vecData" = external addrspace(3) global <4 x i32>, align 46 7; SCHECK: [[arrayofVecData:@arrayofVecData.*]] = local_unnamed_addr addrspace(3) global [8 x [4 x i32]] zeroinitializer, align 168; FCHECK: [[arrayofVecData:@arrayofVecData.*]] = local_unnamed_addr addrspace(3) global [32 x i32] zeroinitializer, align 169; CHECK: [[vecData:@vecData.*]] = external addrspace(3) global [4 x i32], align 410 11; CHECK-LABEL: subtype_array_test12define <4 x i32> @subtype_array_test() {13 ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)14 ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [4 x i32], ptr addrspace(3) [[arrayofVecData]], i32 [[tid]]15 ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 416 ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]17 ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]18 ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 419 ; CHECK: ret <4 x i32> [[x]]20 %i = tail call i32 @llvm.dx.thread.id(i32 0)21 %gep = getelementptr inbounds nuw [4 x i32], ptr addrspace(3) @"arrayofVecData", i32 %i22 %x = load <4 x i32>, ptr addrspace(3) %gep, align 423 ret <4 x i32> %x24}25 26; CHECK-LABEL: subtype_vector_test27define <4 x i32> @subtype_vector_test() {28 ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)29 ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [4 x i32], ptr addrspace(3) [[arrayofVecData]], i32 [[tid]]30 ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 431 ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]32 ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]33 ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 434 ; CHECK: ret <4 x i32> [[x]]35 %i = tail call i32 @llvm.dx.thread.id(i32 0)36 %gep = getelementptr inbounds nuw <4 x i32>, ptr addrspace(3) @"arrayofVecData", i32 %i37 %x = load <4 x i32>, ptr addrspace(3) %gep, align 438 ret <4 x i32> %x39}40 41; CHECK-LABEL: subtype_scalar_test42define <4 x i32> @subtype_scalar_test() {43 ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)44 ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw i32, ptr addrspace(3) [[arrayofVecData]], i32 [[tid]]45 ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 146 ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]47 ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]48 ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 449 ; CHECK: ret <4 x i32> [[x]]50 %i = tail call i32 @llvm.dx.thread.id(i32 0)51 %gep = getelementptr inbounds nuw i32, ptr addrspace(3) @"arrayofVecData", i32 %i52 %x = load <4 x i32>, ptr addrspace(3) %gep, align 453 ret <4 x i32> %x54}55 56; CHECK-LABEL: subtype_i8_test57define <4 x i32> @subtype_i8_test() {58 ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)59 ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw i8, ptr addrspace(3) [[arrayofVecData]], i32 [[tid]]60 ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 161 ; FCHECK: [[flatidx_lshr:%.*]] = lshr i32 [[flatidx_mul]], 262 ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_lshr]]63 ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]64 ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 465 ; CHECK: ret <4 x i32> [[x]]66 %i = tail call i32 @llvm.dx.thread.id(i32 0)67 %gep = getelementptr inbounds nuw i8, ptr addrspace(3) @"arrayofVecData", i32 %i68 %x = load <4 x i32>, ptr addrspace(3) %gep, align 469 ret <4 x i32> %x70}71