Revert "Temporarily Revert "Add basic loop fusion pass.""

The reversion apparently deleted the test/Transforms directory.

Will be re-reverting again.

llvm-svn: 358552

3 files changed, 62 insertions, 0 deletions

diff --git a/llvm/test/Transforms/InferAddressSpaces/NVPTX/bug31948.ll b/llvm/test/Transforms/InferAddressSpaces/NVPTX/bug31948.ll
new file mode 100644
index 00000000000..b4e05b2e429
--- /dev/null
+++ b/llvm/test/Transforms/InferAddressSpaces/NVPTX/bug31948.ll
@@ -0,0 +1,24 @@
+; RUN: opt -S -mtriple=nvptx64-nvidia-cuda -infer-address-spaces %s | FileCheck %s
+
+target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64"
+
+%struct.bar = type { float, float* }
+
+@var1 = local_unnamed_addr addrspace(3) externally_initialized global %struct.bar undef, align 8
+
+; CHECK-LABEL: @bug31948(
+; CHECK: %tmp = load float*, float* addrspace(3)* getelementptr inbounds (%struct.bar, %struct.bar addrspace(3)* @var1, i64 0, i32 1), align 8
+; CHECK: %tmp1 = load float, float* %tmp, align 4
+; CHECK: store float %conv1, float* %tmp, align 4
+; CHECK: store i32 32, i32 addrspace(3)* bitcast (float* addrspace(3)* getelementptr inbounds (%struct.bar, %struct.bar addrspace(3)* @var1, i64 0, i32 1) to i32 addrspace(3)*), align 4
+define void @bug31948(float %a, float* nocapture readnone %x, float* nocapture readnone %y) local_unnamed_addr #0 {
+entry:
+  %tmp = load float*, float** getelementptr (%struct.bar, %struct.bar* addrspacecast (%struct.bar addrspace(3)* @var1 to %struct.bar*), i64 0, i32 1), align 8
+  %tmp1 = load float, float* %tmp, align 4
+  %conv1 = fadd float %tmp1, 1.000000e+00
+  store float %conv1, float* %tmp, align 4
+  store i32 32, i32* bitcast (float** getelementptr (%struct.bar, %struct.bar* addrspacecast (%struct.bar addrspace(3)* @var1 to %struct.bar*), i64 0, i32 1) to i32*), align 4
+  ret void
+}
+
+attributes #0 = { norecurse nounwind }
diff --git a/llvm/test/Transforms/InferAddressSpaces/NVPTX/clone_constexpr.ll b/llvm/test/Transforms/InferAddressSpaces/NVPTX/clone_constexpr.ll
new file mode 100644
index 00000000000..1b324062057
--- /dev/null
+++ b/llvm/test/Transforms/InferAddressSpaces/NVPTX/clone_constexpr.ll
@@ -0,0 +1,36 @@
+; RUN: opt -S -mtriple=nvptx64-nvidia-cuda -infer-address-spaces %s | FileCheck %s
+
+%struct.S = type { [5 x i32] }
+
+$g1 = comdat any
+
+@g1 = linkonce_odr addrspace(3) global %struct.S zeroinitializer, comdat, align 4
+
+; CHECK-LABEL: @foo(
+; CHECK:  %x0 = tail call i32 @llvm.nvvm.read.ptx.sreg.tid.x() #2
+; CHECK:  %idxprom.i = zext i32 %x0 to i64
+; CHECK:  %arrayidx.i = getelementptr %struct.S, %struct.S* addrspacecast (%struct.S addrspace(3)* @g1 to %struct.S*), i64 0, i32 0, i64 %idxprom.i
+; CHECK:  tail call void @f1(i32* %arrayidx.i, i32 undef) #0
+; CHECK:  %x1 = load i32, i32* getelementptr (%struct.S, %struct.S* addrspacecast (%struct.S addrspace(3)* @g1 to %struct.S*), i64 0, i32 0, i64 0), align 4
+; CHECK:  %L.sroa.0.0.insert.ext.i = zext i32 %x1 to i64
+; CHECK:  tail call void @f2(i64* null, i64 %L.sroa.0.0.insert.ext.i) #0
+; CHECK:  ret void
+define void @foo() local_unnamed_addr #0 {
+entry:
+  %x0 = tail call i32 @llvm.nvvm.read.ptx.sreg.tid.x() #2
+  %idxprom.i = zext i32 %x0 to i64
+  %arrayidx.i = getelementptr %struct.S, %struct.S* addrspacecast (%struct.S addrspace(3)* @g1 to %struct.S*), i64 0, i32 0, i64 %idxprom.i
+  tail call void @f1(i32* %arrayidx.i, i32 undef) #0
+  %x1 = load i32, i32* getelementptr (%struct.S, %struct.S* addrspacecast (%struct.S addrspace(3)* @g1 to %struct.S*), i64 0, i32 0, i64 0), align 4
+  %L.sroa.0.0.insert.ext.i = zext i32 %x1 to i64
+  tail call void @f2(i64* null, i64 %L.sroa.0.0.insert.ext.i) #0
+  ret void
+}
+
+declare void @f1(i32*, i32) local_unnamed_addr #0
+declare void @f2(i64*, i64) local_unnamed_addr #0
+declare i32 @llvm.nvvm.read.ptx.sreg.tid.x() #1
+
+attributes #0 = { convergent nounwind }
+attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind }
diff --git a/llvm/test/Transforms/InferAddressSpaces/NVPTX/lit.local.cfg b/llvm/test/Transforms/InferAddressSpaces/NVPTX/lit.local.cfg
new file mode 100644
index 00000000000..2cb98eb371b
--- /dev/null
+++ b/llvm/test/Transforms/InferAddressSpaces/NVPTX/lit.local.cfg
@@ -0,0 +1,2 @@
+if not 'NVPTX' in config.root.targets:
+    config.unsupported = True