aboutsummaryrefslogtreecommitdiff
path: root/llvm/test/CodeGen/DirectX
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/DirectX')
-rw-r--r--llvm/test/CodeGen/DirectX/llvm_assume.ll9
-rw-r--r--llvm/test/CodeGen/DirectX/scalarize-alloca.ll65
-rw-r--r--llvm/test/CodeGen/DirectX/scalarize-global.ll70
3 files changed, 144 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/DirectX/llvm_assume.ll b/llvm/test/CodeGen/DirectX/llvm_assume.ll
new file mode 100644
index 0000000..d739592
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/llvm_assume.ll
@@ -0,0 +1,9 @@
+; RUN: opt -S -dxil-intrinsic-expansion -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
+
+define void @test_llvm_assume(i1 %0) {
+; CHECK-LABEL: test_llvm_assume
+; CHECK-NEXT: ret void
+tail call void @llvm.assume(i1 %0)
+ret void
+}
+
diff --git a/llvm/test/CodeGen/DirectX/scalarize-alloca.ll b/llvm/test/CodeGen/DirectX/scalarize-alloca.ll
index a8557e4..475935d 100644
--- a/llvm/test/CodeGen/DirectX/scalarize-alloca.ll
+++ b/llvm/test/CodeGen/DirectX/scalarize-alloca.ll
@@ -42,3 +42,68 @@ define void @alloca_2d_gep_test() {
%3 = getelementptr inbounds nuw [2 x <2 x i32>], ptr %1, i32 0, i32 %2
ret void
}
+
+; CHECK-LABEL: subtype_array_test
+define void @subtype_array_test() {
+ ; SCHECK: [[alloca_val:%.*]] = alloca [8 x [4 x i32]], align 4
+ ; FCHECK: [[alloca_val:%.*]] = alloca [32 x i32], align 4
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [8 x [4 x i32]], ptr [[alloca_val]], i32 0, i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 4
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr [[alloca_val]], i32 0, i32 [[flatidx]]
+ ; CHECK: ret void
+ %arr = alloca [8 x [4 x i32]], align 4
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw [4 x i32], ptr %arr, i32 %i
+ ret void
+}
+
+; CHECK-LABEL: subtype_vector_test
+define void @subtype_vector_test() {
+ ; SCHECK: [[alloca_val:%.*]] = alloca [8 x [4 x i32]], align 4
+ ; FCHECK: [[alloca_val:%.*]] = alloca [32 x i32], align 4
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [8 x [4 x i32]], ptr [[alloca_val]], i32 0, i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 4
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr [[alloca_val]], i32 0, i32 [[flatidx]]
+ ; CHECK: ret void
+ %arr = alloca [8 x <4 x i32>], align 4
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw <4 x i32>, ptr %arr, i32 %i
+ ret void
+}
+
+; CHECK-LABEL: subtype_scalar_test
+define void @subtype_scalar_test() {
+ ; SCHECK: [[alloca_val:%.*]] = alloca [8 x [4 x i32]], align 4
+ ; FCHECK: [[alloca_val:%.*]] = alloca [32 x i32], align 4
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [8 x [4 x i32]], ptr [[alloca_val]], i32 0, i32 0, i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 1
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr [[alloca_val]], i32 0, i32 [[flatidx]]
+ ; CHECK: ret void
+ %arr = alloca [8 x [4 x i32]], align 4
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw i32, ptr %arr, i32 %i
+ ret void
+}
+
+; CHECK-LABEL: subtype_i8_test
+define void @subtype_i8_test() {
+ ; SCHECK: [[alloca_val:%.*]] = alloca [8 x [4 x i32]], align 4
+ ; FCHECK: [[alloca_val:%.*]] = alloca [32 x i32], align 4
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw i8, ptr [[alloca_val]], i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 1
+ ; FCHECK: [[flatidx_lshr:%.*]] = lshr i32 [[flatidx_mul]], 2
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_lshr]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr [[alloca_val]], i32 0, i32 [[flatidx]]
+ ; CHECK: ret void
+ %arr = alloca [8 x [4 x i32]], align 4
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw i8, ptr %arr, i32 %i
+ ret void
+}
diff --git a/llvm/test/CodeGen/DirectX/scalarize-global.ll b/llvm/test/CodeGen/DirectX/scalarize-global.ll
new file mode 100644
index 0000000..ca10f6e
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/scalarize-global.ll
@@ -0,0 +1,70 @@
+; RUN: opt -S -passes='dxil-data-scalarization' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s --check-prefixes=SCHECK,CHECK
+; RUN: opt -S -passes='dxil-data-scalarization,dxil-flatten-arrays' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s --check-prefixes=FCHECK,CHECK
+
+@"arrayofVecData" = local_unnamed_addr addrspace(3) global [8 x <4 x i32>] zeroinitializer, align 16
+@"vecData" = external addrspace(3) global <4 x i32>, align 4
+
+; SCHECK: [[arrayofVecData:@arrayofVecData.*]] = local_unnamed_addr addrspace(3) global [8 x [4 x i32]] zeroinitializer, align 16
+; FCHECK: [[arrayofVecData:@arrayofVecData.*]] = local_unnamed_addr addrspace(3) global [32 x i32] zeroinitializer, align 16
+; CHECK: [[vecData:@vecData.*]] = external addrspace(3) global [4 x i32], align 4
+
+; CHECK-LABEL: subtype_array_test
+define <4 x i32> @subtype_array_test() {
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [8 x [4 x i32]], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 4
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]
+ ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 4
+ ; CHECK: ret <4 x i32> [[x]]
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw [4 x i32], ptr addrspace(3) @"arrayofVecData", i32 %i
+ %x = load <4 x i32>, ptr addrspace(3) %gep, align 4
+ ret <4 x i32> %x
+}
+
+; CHECK-LABEL: subtype_vector_test
+define <4 x i32> @subtype_vector_test() {
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [8 x [4 x i32]], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 4
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]
+ ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 4
+ ; CHECK: ret <4 x i32> [[x]]
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw <4 x i32>, ptr addrspace(3) @"arrayofVecData", i32 %i
+ %x = load <4 x i32>, ptr addrspace(3) %gep, align 4
+ ret <4 x i32> %x
+}
+
+; CHECK-LABEL: subtype_scalar_test
+define <4 x i32> @subtype_scalar_test() {
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw [8 x [4 x i32]], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 0, i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 1
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_mul]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]
+ ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 4
+ ; CHECK: ret <4 x i32> [[x]]
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw i32, ptr addrspace(3) @"arrayofVecData", i32 %i
+ %x = load <4 x i32>, ptr addrspace(3) %gep, align 4
+ ret <4 x i32> %x
+}
+
+; CHECK-LABEL: subtype_i8_test
+define <4 x i32> @subtype_i8_test() {
+ ; CHECK: [[tid:%.*]] = tail call i32 @llvm.dx.thread.id(i32 0)
+ ; SCHECK: [[gep:%.*]] = getelementptr inbounds nuw i8, ptr addrspace(3) [[arrayofVecData]], i32 [[tid]]
+ ; FCHECK: [[flatidx_mul:%.*]] = mul i32 [[tid]], 1
+ ; FCHECK: [[flatidx_lshr:%.*]] = lshr i32 [[flatidx_mul]], 2
+ ; FCHECK: [[flatidx:%.*]] = add i32 0, [[flatidx_lshr]]
+ ; FCHECK: [[gep:%.*]] = getelementptr inbounds nuw [32 x i32], ptr addrspace(3) [[arrayofVecData]], i32 0, i32 [[flatidx]]
+ ; CHECK: [[x:%.*]] = load <4 x i32>, ptr addrspace(3) [[gep]], align 4
+ ; CHECK: ret <4 x i32> [[x]]
+ %i = tail call i32 @llvm.dx.thread.id(i32 0)
+ %gep = getelementptr inbounds nuw i8, ptr addrspace(3) @"arrayofVecData", i32 %i
+ %x = load <4 x i32>, ptr addrspace(3) %gep, align 4
+ ret <4 x i32> %x
+}
generated by cgit v1.1 at 2025-11-08 04:14:56 +0000