[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-17 Thread via cfe-commits 


@@ -1964,11 +1964,25 @@ Instruction *InstCombinerImpl::visitFPToSI(FPToSIInst 
) {
 }
 
 Instruction *InstCombinerImpl::visitUIToFP(CastInst ) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+return R;
+  if (!CI.hasNonNeg() && isKnownNonNegative(CI.getOperand(0), SQ)) {
+CI.setNonNeg();
+return 
+  }
+  return nullptr;
 }
 
 Instruction *InstCombinerImpl::visitSIToFP(CastInst ) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+return R;
+  if (isKnownNonNegative(CI.getOperand(0), SQ)) {
+auto UI =

goldsteinn wrote:

My mistake, ill push a fix.

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-17 Thread Yingwei Zheng via cfe-commits 


@@ -1964,11 +1964,25 @@ Instruction *InstCombinerImpl::visitFPToSI(FPToSIInst 
) {
 }
 
 Instruction *InstCombinerImpl::visitUIToFP(CastInst ) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+return R;
+  if (!CI.hasNonNeg() && isKnownNonNegative(CI.getOperand(0), SQ)) {
+CI.setNonNeg();
+return 
+  }
+  return nullptr;
 }
 
 Instruction *InstCombinerImpl::visitSIToFP(CastInst ) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+return R;
+  if (isKnownNonNegative(CI.getOperand(0), SQ)) {
+auto UI =

dtcxzyw wrote:

@goldsteinn We always use `auto *` for pointers.

https://llvm.org/docs/CodingStandards.html#beware-unnecessary-copies-with-auto

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-16 Thread via cfe-commits 

https://github.com/goldsteinn closed 
https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-11 Thread Nikita Popov via cfe-commits 

nikic wrote:

> > Can the implementation of foldFBinOpOfIntCastsFromSign be simplified to use 
> > nneg instead of KnownBits after this change?
> 
> yeah we could. Should I do a survey of existing folds first to ensure we 
> don't incorrectly keep flags (like with `trunc nuw/nsw`) before integrating 
> too deeply?

I don't think this is important for uitofp nneg, mainly because we only have 
very few folds for it, so there's just not a great deal of potential for things 
to go wrong...

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-11 Thread Yingwei Zheng via cfe-commits 

https://github.com/dtcxzyw approved this pull request.

LGTM.

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-11 Thread Yingwei Zheng via cfe-commits 


@@ -1964,11 +1964,25 @@ Instruction *InstCombinerImpl::visitFPToSI(FPToSIInst 
) {
 }
 
 Instruction *InstCombinerImpl::visitUIToFP(CastInst ) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+return R;
+  if (!CI.hasNonNeg() && isKnownNonNegative(CI.getOperand(0), SQ)) {
+CI.setNonNeg();
+return 
+  }
+  return nullptr;
 }
 
 Instruction *InstCombinerImpl::visitSIToFP(CastInst ) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+return R;
+  if (isKnownNonNegative(CI.getOperand(0), SQ)) {
+auto UI =

dtcxzyw wrote:

```suggestion
auto *UI =
```

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-10 Thread via cfe-commits 

goldsteinn wrote:

> Can the implementation of foldFBinOpOfIntCastsFromSign be simplified to use 
> nneg instead of KnownBits after this change?

yeah we could. Should I do a survey of existing folds first to ensure we don't 
incorrectly keep flags (like with `trunc nuw/nsw`) before integrating too 
deeply?

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-10 Thread Nikita Popov via cfe-commits 

nikic wrote:

Can the implementation of foldFBinOpOfIntCastsFromSign be simplified to use 
nneg instead of KnownBits after this change?

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-10 Thread Nikita Popov via cfe-commits 

https://github.com/nikic approved this pull request.

LGTM

https://github.com/llvm/llvm-project/pull/88299
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-10 Thread via cfe-commits 

llvmbot wrote:




@llvm/pr-subscribers-clang

Author: None (goldsteinn)


Changes

This is essentially the same as #82404 but has the `nneg` flag which
allows the backend to reliably undo the transform.


---

Patch is 61.71 KiB, truncated to 20.00 KiB below, full version: 
https://github.com/llvm/llvm-project/pull/88299.diff


12 Files Affected:

- (modified) clang/test/Headers/__clang_hip_math.hip (+12-12) 
- (modified) llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp (+16-2) 
- (modified) llvm/test/Transforms/InstCombine/add-sitofp.ll (+10-10) 
- (modified) llvm/test/Transforms/InstCombine/binop-itofp.ll (+33-33) 
- (modified) llvm/test/Transforms/InstCombine/clamp-to-minmax.ll (+5-5) 
- (modified) llvm/test/Transforms/InstCombine/fpcast.ll (+12-12) 
- (modified) llvm/test/Transforms/InstCombine/minmax-fold.ll (+5-5) 
- (modified) llvm/test/Transforms/InstCombine/minmax-fp.ll (+1-1) 
- (modified) llvm/test/Transforms/InstCombine/pr27236.ll (+1-1) 
- (modified) llvm/test/Transforms/InstCombine/sitofp.ll (+1-1) 
- (modified) llvm/test/Transforms/LoopVectorize/X86/float-induction-x86.ll 
(+3-3) 
- (modified) llvm/test/Transforms/LoopVectorize/float-induction.ll (+28-28) 


``diff
diff --git a/clang/test/Headers/__clang_hip_math.hip 
b/clang/test/Headers/__clang_hip_math.hip
index 37099de74fb8ec..bff1708120604b 100644
--- a/clang/test/Headers/__clang_hip_math.hip
+++ b/clang/test/Headers/__clang_hip_math.hip
@@ -1685,7 +1685,7 @@ extern "C" __device__ double test_j1(double x) {
 // DEFAULT-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// DEFAULT-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // DEFAULT-NEXT:[[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:[[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], 
[[DIV_I]]
 // DEFAULT-NEXT:[[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1718,7 +1718,7 @@ extern "C" __device__ double test_j1(double x) {
 // FINITEONLY-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// FINITEONLY-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // FINITEONLY-NEXT:[[DIV_I:%.*]] = fdiv nnan ninf contract float 
[[CONV_I]], [[Y]]
 // FINITEONLY-NEXT:[[MUL8_I:%.*]] = fmul nnan ninf contract float 
[[__X1_0_I3]], [[DIV_I]]
 // FINITEONLY-NEXT:[[SUB_I]] = fsub nnan ninf contract float [[MUL8_I]], 
[[__X0_0_I2]]
@@ -1751,7 +1751,7 @@ extern "C" __device__ double test_j1(double x) {
 // APPROX-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// APPROX-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// APPROX-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // APPROX-NEXT:[[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // APPROX-NEXT:[[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], 
[[DIV_I]]
 // APPROX-NEXT:[[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1788,7 +1788,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // DEFAULT-NEXT:[[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// DEFAULT-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // DEFAULT-NEXT:[[DIV_I:%.*]] = fdiv contract double [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:[[MUL8_I:%.*]] = fmul contract double [[__X1_0_I3]], 
[[DIV_I]]
 // DEFAULT-NEXT:[[SUB_I]] = fsub contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -1821,7 +1821,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // FINITEONLY-NEXT:[[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:[[CONV_I:%.*]] = sitofp i32

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-10 Thread via cfe-commits 

llvmbot wrote:




@llvm/pr-subscribers-llvm-transforms

Author: None (goldsteinn)


Changes

This is essentially the same as #82404 but has the `nneg` flag which
allows the backend to reliably undo the transform.


---

Patch is 61.71 KiB, truncated to 20.00 KiB below, full version: 
https://github.com/llvm/llvm-project/pull/88299.diff


12 Files Affected:

- (modified) clang/test/Headers/__clang_hip_math.hip (+12-12) 
- (modified) llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp (+16-2) 
- (modified) llvm/test/Transforms/InstCombine/add-sitofp.ll (+10-10) 
- (modified) llvm/test/Transforms/InstCombine/binop-itofp.ll (+33-33) 
- (modified) llvm/test/Transforms/InstCombine/clamp-to-minmax.ll (+5-5) 
- (modified) llvm/test/Transforms/InstCombine/fpcast.ll (+12-12) 
- (modified) llvm/test/Transforms/InstCombine/minmax-fold.ll (+5-5) 
- (modified) llvm/test/Transforms/InstCombine/minmax-fp.ll (+1-1) 
- (modified) llvm/test/Transforms/InstCombine/pr27236.ll (+1-1) 
- (modified) llvm/test/Transforms/InstCombine/sitofp.ll (+1-1) 
- (modified) llvm/test/Transforms/LoopVectorize/X86/float-induction-x86.ll 
(+3-3) 
- (modified) llvm/test/Transforms/LoopVectorize/float-induction.ll (+28-28) 


``diff
diff --git a/clang/test/Headers/__clang_hip_math.hip 
b/clang/test/Headers/__clang_hip_math.hip
index 37099de74fb8ec..bff1708120604b 100644
--- a/clang/test/Headers/__clang_hip_math.hip
+++ b/clang/test/Headers/__clang_hip_math.hip
@@ -1685,7 +1685,7 @@ extern "C" __device__ double test_j1(double x) {
 // DEFAULT-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// DEFAULT-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // DEFAULT-NEXT:[[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:[[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], 
[[DIV_I]]
 // DEFAULT-NEXT:[[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1718,7 +1718,7 @@ extern "C" __device__ double test_j1(double x) {
 // FINITEONLY-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// FINITEONLY-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // FINITEONLY-NEXT:[[DIV_I:%.*]] = fdiv nnan ninf contract float 
[[CONV_I]], [[Y]]
 // FINITEONLY-NEXT:[[MUL8_I:%.*]] = fmul nnan ninf contract float 
[[__X1_0_I3]], [[DIV_I]]
 // FINITEONLY-NEXT:[[SUB_I]] = fsub nnan ninf contract float [[MUL8_I]], 
[[__X0_0_I2]]
@@ -1751,7 +1751,7 @@ extern "C" __device__ double test_j1(double x) {
 // APPROX-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// APPROX-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// APPROX-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // APPROX-NEXT:[[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // APPROX-NEXT:[[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], 
[[DIV_I]]
 // APPROX-NEXT:[[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1788,7 +1788,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // DEFAULT-NEXT:[[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// DEFAULT-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // DEFAULT-NEXT:[[DIV_I:%.*]] = fdiv contract double [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:[[MUL8_I:%.*]] = fmul contract double [[__X1_0_I3]], 
[[DIV_I]]
 // DEFAULT-NEXT:[[SUB_I]] = fsub contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -1821,7 +1821,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // FINITEONLY-NEXT:[[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:[[CONV_I:%.*]] = sitofp

[clang] [llvm] [InstCombine] Add canonicalization of `sitofp` -> `uitofp nneg` (PR #88299)

2024-04-10 Thread via cfe-commits 

https://github.com/goldsteinn created 
https://github.com/llvm/llvm-project/pull/88299

This is essentially the same as #82404 but has the `nneg` flag which
allows the backend to reliably undo the transform.


>From 46f76652548eed85a0947a93af701115875dbf6d Mon Sep 17 00:00:00 2001
From: Noah Goldstein 
Date: Thu, 21 Mar 2024 11:01:21 -0500
Subject: [PATCH] [InstCombine] Add canonicalization of `sitofp` -> `uitofp
 nneg`

This is essentially the same as #82404 but has the `nneg` flag which
allows the backend to reliably undo the transform.
---
 clang/test/Headers/__clang_hip_math.hip   | 24 +++
 .../InstCombine/InstCombineCasts.cpp  | 18 -
 .../test/Transforms/InstCombine/add-sitofp.ll | 20 +++---
 .../Transforms/InstCombine/binop-itofp.ll | 66 +--
 .../Transforms/InstCombine/clamp-to-minmax.ll | 10 +--
 llvm/test/Transforms/InstCombine/fpcast.ll| 24 +++
 .../Transforms/InstCombine/minmax-fold.ll | 10 +--
 llvm/test/Transforms/InstCombine/minmax-fp.ll |  2 +-
 llvm/test/Transforms/InstCombine/pr27236.ll   |  2 +-
 llvm/test/Transforms/InstCombine/sitofp.ll|  2 +-
 .../LoopVectorize/X86/float-induction-x86.ll  |  6 +-
 .../LoopVectorize/float-induction.ll  | 56 
 12 files changed, 127 insertions(+), 113 deletions(-)

diff --git a/clang/test/Headers/__clang_hip_math.hip 
b/clang/test/Headers/__clang_hip_math.hip
index 37099de74fb8ec..bff1708120604b 100644
--- a/clang/test/Headers/__clang_hip_math.hip
+++ b/clang/test/Headers/__clang_hip_math.hip
@@ -1685,7 +1685,7 @@ extern "C" __device__ double test_j1(double x) {
 // DEFAULT-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// DEFAULT-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // DEFAULT-NEXT:[[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:[[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], 
[[DIV_I]]
 // DEFAULT-NEXT:[[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1718,7 +1718,7 @@ extern "C" __device__ double test_j1(double x) {
 // FINITEONLY-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// FINITEONLY-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // FINITEONLY-NEXT:[[DIV_I:%.*]] = fdiv nnan ninf contract float 
[[CONV_I]], [[Y]]
 // FINITEONLY-NEXT:[[MUL8_I:%.*]] = fmul nnan ninf contract float 
[[__X1_0_I3]], [[DIV_I]]
 // FINITEONLY-NEXT:[[SUB_I]] = fsub nnan ninf contract float [[MUL8_I]], 
[[__X0_0_I2]]
@@ -1751,7 +1751,7 @@ extern "C" __device__ double test_j1(double x) {
 // APPROX-NEXT:[[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:[[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// APPROX-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// APPROX-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // APPROX-NEXT:[[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // APPROX-NEXT:[[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], 
[[DIV_I]]
 // APPROX-NEXT:[[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1788,7 +1788,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // DEFAULT-NEXT:[[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:[[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:[[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// DEFAULT-NEXT:[[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // DEFAULT-NEXT:[[DIV_I:%.*]] = fdiv contract double [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:[[MUL8_I:%.*]] = fmul contract double [[__X1_0_I3]], 
[[DIV_I]]
 // DEFAULT-NEXT:[[SUB_I]] = fsub contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -1821,7 +1821,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // FINITEONLY-NEXT:[[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], 
[[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:[[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], 
[[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT: