GlobalISel: fewerElementsVector for a few more trivial ops

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@352165 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index e9d55ea889bcba4920b36ab8863617d510aed70d..5be14018fb9f70c2f93664e6f88bcf08e037c6ec 100644 (file)
--- a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -1407,6 +1407,12 @@ LegalizerHelper::fewerElementsVector(MachineInstr &MI, unsigned TypeIdx,
   case TargetOpcode::G_FDIV:
   case TargetOpcode::G_FREM:
   case TargetOpcode::G_FMA:
+  case TargetOpcode::G_FPOW:
+  case TargetOpcode::G_FEXP:
+  case TargetOpcode::G_FEXP2:
+  case TargetOpcode::G_FLOG:
+  case TargetOpcode::G_FLOG2:
+  case TargetOpcode::G_FLOG10:
   case TargetOpcode::G_FCEIL: {
     unsigned NarrowSize = NarrowTy.getSizeInBits();
     unsigned DstReg = MI.getOperand(0).getReg();

diff --git a/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 6bf378293bf15e462e11e0832918838e441c3d5d..5ad9a43bb960802f61f66171725d4cbf31b40307 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -176,10 +176,6 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST,
   getActionDefinitionsBuilder({G_FPTOSI, G_FPTOUI})
     .legalFor({{S32, S32}, {S32, S64}});
 
-  setAction({G_FPOW, S32}, Legal);
-  setAction({G_FEXP2, S32}, Legal);
-  setAction({G_FLOG2, S32}, Legal);
-
   getActionDefinitionsBuilder({G_INTRINSIC_TRUNC, G_INTRINSIC_ROUND})
     .legalFor({S32, S64});
 
@@ -198,7 +194,11 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST,
     .clampMaxNumElements(0, S1, 1)
     .clampMaxNumElements(1, S32, 1);
 
-
+  // FIXME: fexp, flog2, flog10 needs to be custom lowered.
+  getActionDefinitionsBuilder({G_FPOW, G_FEXP, G_FEXP2,
+                               G_FLOG, G_FLOG2, G_FLOG10})
+    .legalFor({S32})
+    .scalarize(0);
 
   setAction({G_CTLZ, S32}, Legal);
   setAction({G_CTLZ_ZERO_UNDEF, S32}, Legal);

diff --git a/test/CodeGen/AMDGPU/GlobalISel/legalize-fexp.mir b/test/CodeGen/AMDGPU/GlobalISel/legalize-fexp.mir
new file mode 100644 (file)
index 0000000..d780957
--- /dev/null
+++ b/test/CodeGen/AMDGPU/GlobalISel/legalize-fexp.mir
@@ -0,0 +1,54 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: test_fexp_s32
+body: |
+  bb.0:
+    liveins: $vgpr0
+
+    ; CHECK-LABEL: name: test_fexp_s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[FEXP_:%[0-9]+]]:_(s32) = G_FEXP [[COPY]]
+    ; CHECK: $vgpr0 = COPY [[FEXP_]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = G_FEXP %0
+    $vgpr0 = COPY %1
+...
+
+---
+name: test_fexp_v2s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+
+    ; CHECK-LABEL: name: test_fexp_v2s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[FEXP_:%[0-9]+]]:_(s32) = G_FEXP [[UV]]
+    ; CHECK: [[FEXP_1:%[0-9]+]]:_(s32) = G_FEXP [[UV1]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FEXP_]](s32), [[FEXP_1]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(<2 x s32>) = G_FEXP %0
+    $vgpr0_vgpr1 = COPY %1
+...
+
+---
+name: test_fexp_v3s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1_vgpr2
+
+    ; CHECK-LABEL: name: test_fexp_v3s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK: [[FEXP_:%[0-9]+]]:_(s32) = G_FEXP [[UV]]
+    ; CHECK: [[FEXP_1:%[0-9]+]]:_(s32) = G_FEXP [[UV1]]
+    ; CHECK: [[FEXP_2:%[0-9]+]]:_(s32) = G_FEXP [[UV2]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FEXP_]](s32), [[FEXP_1]](s32), [[FEXP_2]](s32)
+    ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+    %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    %1:_(<3 x  s32>) = G_FEXP %0
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+...

diff --git a/test/CodeGen/AMDGPU/GlobalISel/legalize-fexp2.mir b/test/CodeGen/AMDGPU/GlobalISel/legalize-fexp2.mir
new file mode 100644 (file)
index 0000000..36ee554
--- /dev/null
+++ b/test/CodeGen/AMDGPU/GlobalISel/legalize-fexp2.mir
@@ -0,0 +1,54 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: test_fexp2_s32
+body: |
+  bb.0:
+    liveins: $vgpr0
+
+    ; CHECK-LABEL: name: test_fexp2_s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[FEXP2_:%[0-9]+]]:_(s32) = G_FEXP2 [[COPY]]
+    ; CHECK: $vgpr0 = COPY [[FEXP2_]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = G_FEXP2 %0
+    $vgpr0 = COPY %1
+...
+
+---
+name: test_fexp2_v2s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+
+    ; CHECK-LABEL: name: test_fexp2_v2s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[FEXP2_:%[0-9]+]]:_(s32) = G_FEXP2 [[UV]]
+    ; CHECK: [[FEXP2_1:%[0-9]+]]:_(s32) = G_FEXP2 [[UV1]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FEXP2_]](s32), [[FEXP2_1]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(<2 x s32>) = G_FEXP2 %0
+    $vgpr0_vgpr1 = COPY %1
+...
+
+---
+name: test_fexp2_v3s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1_vgpr2
+
+    ; CHECK-LABEL: name: test_fexp2_v3s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK: [[FEXP2_:%[0-9]+]]:_(s32) = G_FEXP2 [[UV]]
+    ; CHECK: [[FEXP2_1:%[0-9]+]]:_(s32) = G_FEXP2 [[UV1]]
+    ; CHECK: [[FEXP2_2:%[0-9]+]]:_(s32) = G_FEXP2 [[UV2]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FEXP2_]](s32), [[FEXP2_1]](s32), [[FEXP2_2]](s32)
+    ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+    %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    %1:_(<3 x  s32>) = G_FEXP2 %0
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+...

diff --git a/test/CodeGen/AMDGPU/GlobalISel/legalize-flog.mir b/test/CodeGen/AMDGPU/GlobalISel/legalize-flog.mir
new file mode 100644 (file)
index 0000000..08dd07b
--- /dev/null
+++ b/test/CodeGen/AMDGPU/GlobalISel/legalize-flog.mir
@@ -0,0 +1,54 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: test_flog_s32
+body: |
+  bb.0:
+    liveins: $vgpr0
+
+    ; CHECK-LABEL: name: test_flog_s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[FLOG:%[0-9]+]]:_(s32) = G_FLOG [[COPY]]
+    ; CHECK: $vgpr0 = COPY [[FLOG]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = G_FLOG %0
+    $vgpr0 = COPY %1
+...
+
+---
+name: test_flog_v2s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+
+    ; CHECK-LABEL: name: test_flog_v2s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[FLOG:%[0-9]+]]:_(s32) = G_FLOG [[UV]]
+    ; CHECK: [[FLOG1:%[0-9]+]]:_(s32) = G_FLOG [[UV1]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FLOG]](s32), [[FLOG1]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(<2 x s32>) = G_FLOG %0
+    $vgpr0_vgpr1 = COPY %1
+...
+
+---
+name: test_flog_v3s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1_vgpr2
+
+    ; CHECK-LABEL: name: test_flog_v3s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK: [[FLOG:%[0-9]+]]:_(s32) = G_FLOG [[UV]]
+    ; CHECK: [[FLOG1:%[0-9]+]]:_(s32) = G_FLOG [[UV1]]
+    ; CHECK: [[FLOG2:%[0-9]+]]:_(s32) = G_FLOG [[UV2]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FLOG]](s32), [[FLOG1]](s32), [[FLOG2]](s32)
+    ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+    %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    %1:_(<3 x  s32>) = G_FLOG %0
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+...

diff --git a/test/CodeGen/AMDGPU/GlobalISel/legalize-flog10.mir b/test/CodeGen/AMDGPU/GlobalISel/legalize-flog10.mir
new file mode 100644 (file)
index 0000000..7bcfa6b
--- /dev/null
+++ b/test/CodeGen/AMDGPU/GlobalISel/legalize-flog10.mir
@@ -0,0 +1,54 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: test_flog10_s32
+body: |
+  bb.0:
+    liveins: $vgpr0
+
+    ; CHECK-LABEL: name: test_flog10_s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[FLOG10_:%[0-9]+]]:_(s32) = G_FLOG10 [[COPY]]
+    ; CHECK: $vgpr0 = COPY [[FLOG10_]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = G_FLOG10 %0
+    $vgpr0 = COPY %1
+...
+
+---
+name: test_flog10_v2s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+
+    ; CHECK-LABEL: name: test_flog10_v2s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[FLOG10_:%[0-9]+]]:_(s32) = G_FLOG10 [[UV]]
+    ; CHECK: [[FLOG10_1:%[0-9]+]]:_(s32) = G_FLOG10 [[UV1]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FLOG10_]](s32), [[FLOG10_1]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(<2 x s32>) = G_FLOG10 %0
+    $vgpr0_vgpr1 = COPY %1
+...
+
+---
+name: test_flog10_v3s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1_vgpr2
+
+    ; CHECK-LABEL: name: test_flog10_v3s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK: [[FLOG10_:%[0-9]+]]:_(s32) = G_FLOG10 [[UV]]
+    ; CHECK: [[FLOG10_1:%[0-9]+]]:_(s32) = G_FLOG10 [[UV1]]
+    ; CHECK: [[FLOG10_2:%[0-9]+]]:_(s32) = G_FLOG10 [[UV2]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FLOG10_]](s32), [[FLOG10_1]](s32), [[FLOG10_2]](s32)
+    ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+    %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    %1:_(<3 x  s32>) = G_FLOG10 %0
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+...

diff --git a/test/CodeGen/AMDGPU/GlobalISel/legalize-flog2.mir b/test/CodeGen/AMDGPU/GlobalISel/legalize-flog2.mir
new file mode 100644 (file)
index 0000000..91f3dc5
--- /dev/null
+++ b/test/CodeGen/AMDGPU/GlobalISel/legalize-flog2.mir
@@ -0,0 +1,54 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: test_flog2_s32
+body: |
+  bb.0:
+    liveins: $vgpr0
+
+    ; CHECK-LABEL: name: test_flog2_s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[FLOG2_:%[0-9]+]]:_(s32) = G_FLOG2 [[COPY]]
+    ; CHECK: $vgpr0 = COPY [[FLOG2_]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = G_FLOG2 %0
+    $vgpr0 = COPY %1
+...
+
+---
+name: test_flog2_v2s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+
+    ; CHECK-LABEL: name: test_flog2_v2s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[FLOG2_:%[0-9]+]]:_(s32) = G_FLOG2 [[UV]]
+    ; CHECK: [[FLOG2_1:%[0-9]+]]:_(s32) = G_FLOG2 [[UV1]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FLOG2_]](s32), [[FLOG2_1]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(<2 x s32>) = G_FLOG2 %0
+    $vgpr0_vgpr1 = COPY %1
+...
+
+---
+name: test_flog2_v3s32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1_vgpr2
+
+    ; CHECK-LABEL: name: test_flog2_v3s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK: [[FLOG2_:%[0-9]+]]:_(s32) = G_FLOG2 [[UV]]
+    ; CHECK: [[FLOG2_1:%[0-9]+]]:_(s32) = G_FLOG2 [[UV1]]
+    ; CHECK: [[FLOG2_2:%[0-9]+]]:_(s32) = G_FLOG2 [[UV2]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FLOG2_]](s32), [[FLOG2_1]](s32), [[FLOG2_2]](s32)
+    ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+    %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    %1:_(<3 x  s32>) = G_FLOG2 %0
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+...

diff --git a/test/CodeGen/AMDGPU/GlobalISel/legalize-fpow.mir b/test/CodeGen/AMDGPU/GlobalISel/legalize-fpow.mir
new file mode 100644 (file)
index 0000000..a03b6e2
--- /dev/null
+++ b/test/CodeGen/AMDGPU/GlobalISel/legalize-fpow.mir
@@ -0,0 +1,62 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: test_fpow_s32
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    ; CHECK-LABEL: name: test_fpow_s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
+    ; CHECK: [[FPOW:%[0-9]+]]:_(s32) = G_FPOW [[COPY]], [[COPY1]]
+    ; CHECK: $vgpr0 = COPY [[FPOW]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = COPY $vgpr1
+    %2:_(s32) = G_FPOW %0, %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_fpow_v2s32
+body: |
+  bb.0.entry:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; CHECK-LABEL: name: test_fpow_v2s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>)
+    ; CHECK: [[FPOW:%[0-9]+]]:_(s32) = G_FPOW [[UV]], [[UV2]]
+    ; CHECK: [[FPOW1:%[0-9]+]]:_(s32) = G_FPOW [[UV1]], [[UV3]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[FPOW]](s32), [[FPOW1]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(<2 x s32>) = COPY $vgpr2_vgpr3
+    %2:_(<2 x s32>) = G_FPOW %0, %1
+    $vgpr0_vgpr1 = COPY %2
+...
+
+---
+name: test_fpow_v3s32
+body: |
+  bb.0.entry:
+    liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
+
+    ; CHECK-LABEL: name: test_fpow_v3s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CHECK: [[FPOW:%[0-9]+]]:_(s32) = G_FPOW [[UV]], [[UV3]]
+    ; CHECK: [[FPOW1:%[0-9]+]]:_(s32) = G_FPOW [[UV1]], [[UV4]]
+    ; CHECK: [[FPOW2:%[0-9]+]]:_(s32) = G_FPOW [[UV2]], [[UV5]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FPOW]](s32), [[FPOW1]](s32), [[FPOW2]](s32)
+    ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+    %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
+    %1:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
+    %2:_(<3 x s32>) = G_FPOW %0, %1
+    $vgpr0_vgpr1_vgpr2 = COPY %2
+...