[PowerPC] Drop invalid range metadata when lowering i64 load to fp in INT_TO_FP#198705
Merged
Conversation
|
@llvm/pr-subscribers-backend-powerpc Author: Amy Kwan (amy-kwan) ChangesWhen lowering an i64 load in LowerINT_TO_FP, we were forwarding the original !range metadata to a new f64 load. This is invalid because the metadata no longer matches the value type/semantics, and can trigger assertions when lowering i64 to fp (double or float) conversions. This patch fixes this by passing a nullptr for the Ranges operand when calling getLoad() and adds extra test cases to cover signed/unsigned i64 to f32/f64 conversions and to ensure they do not assert when the !range metadata is present. The assertion this patch attempts to fix is: This assert was originally seen when building Rust on AIX. Full diff: https://github.com/llvm/llvm-project/pull/198705.diff 2 Files Affected:
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index d8ed51ec9b26f..d121e8e2cc578 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -8865,7 +8865,7 @@ SDValue PPCTargetLowering::LowerINT_TO_FP(SDValue Op,
MachineFunction &MF = DAG.getMachineFunction();
if (canReuseLoadAddress(SINT, MVT::i64, RLI, DAG)) {
Bits = DAG.getLoad(MVT::f64, dl, RLI.Chain, RLI.Ptr, RLI.MPI,
- RLI.Alignment, RLI.MMOFlags(), RLI.AAInfo, RLI.Ranges);
+ RLI.Alignment, RLI.MMOFlags(), RLI.AAInfo, nullptr);
if (RLI.ResChain)
DAG.makeEquivalentMemoryOrdering(RLI.ResChain, Bits.getValue(1));
} else if (Subtarget.hasLFIWAX() &&
diff --git a/llvm/test/CodeGen/PowerPC/ppc-i64-to-fp.ll b/llvm/test/CodeGen/PowerPC/ppc-i64-to-fp.ll
new file mode 100644
index 0000000000000..420163f6ad101
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/ppc-i64-to-fp.ll
@@ -0,0 +1,124 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mcpu=pwr8 -ppc-asm-full-reg-names -mtriple=powerpc64-ibm-aix < %s | \
+; RUN: FileCheck %s --check-prefix=CHECK-AIX64
+; RUN: llc -mcpu=pwr8 -ppc-asm-full-reg-names -mtriple=powerpc-ibm-aix < %s | \
+; RUN: FileCheck %s --check-prefix=CHECK-AIX32
+; RUN: llc -mcpu=pwr8 -ppc-asm-full-reg-names -mtriple=powerpc64le-unknown-linux-gnu < %s | \
+; RUN: FileCheck %s --check-prefix=CHECK-LINUX
+
+; Test that assertions related to the range metadata when converting from i64
+; to i32 does not occur.
+
+define float @case_i64_uitofp_f32(ptr %p) {
+; CHECK-AIX64-LABEL: case_i64_uitofp_f32:
+; CHECK-AIX64: # %bb.0: # %entry
+; CHECK-AIX64-NEXT: lfd f0, 0(r3)
+; CHECK-AIX64-NEXT: xscvuxdsp f1, f0
+; CHECK-AIX64-NEXT: blr
+;
+; CHECK-AIX32-LABEL: case_i64_uitofp_f32:
+; CHECK-AIX32: # %bb.0: # %entry
+; CHECK-AIX32-NEXT: lwz r4, 4(r3)
+; CHECK-AIX32-NEXT: stw r4, -4(r1)
+; CHECK-AIX32-NEXT: lwz r3, 0(r3)
+; CHECK-AIX32-NEXT: stw r3, -8(r1)
+; CHECK-AIX32-NEXT: lfd f0, -8(r1)
+; CHECK-AIX32-NEXT: xscvuxdsp f1, f0
+; CHECK-AIX32-NEXT: blr
+;
+; CHECK-LINUX-LABEL: case_i64_uitofp_f32:
+; CHECK-LINUX: # %bb.0: # %entry
+; CHECK-LINUX-NEXT: lfd f0, 0(r3)
+; CHECK-LINUX-NEXT: xscvuxdsp f1, f0
+; CHECK-LINUX-NEXT: blr
+entry:
+ %x = load i64, ptr %p, align 8, !range !0
+ %y = uitofp i64 %x to float
+ ret float %y
+}
+
+define double @case_i64_sitofp_f64(ptr %p) {
+; CHECK-AIX64-LABEL: case_i64_sitofp_f64:
+; CHECK-AIX64: # %bb.0: # %entry
+; CHECK-AIX64-NEXT: lfd f0, 0(r3)
+; CHECK-AIX64-NEXT: xscvsxddp f1, f0
+; CHECK-AIX64-NEXT: blr
+;
+; CHECK-AIX32-LABEL: case_i64_sitofp_f64:
+; CHECK-AIX32: # %bb.0: # %entry
+; CHECK-AIX32-NEXT: lwz r4, 4(r3)
+; CHECK-AIX32-NEXT: stw r4, -4(r1)
+; CHECK-AIX32-NEXT: lwz r3, 0(r3)
+; CHECK-AIX32-NEXT: stw r3, -8(r1)
+; CHECK-AIX32-NEXT: lfd f0, -8(r1)
+; CHECK-AIX32-NEXT: xscvsxddp f1, f0
+; CHECK-AIX32-NEXT: blr
+;
+; CHECK-LINUX-LABEL: case_i64_sitofp_f64:
+; CHECK-LINUX: # %bb.0: # %entry
+; CHECK-LINUX-NEXT: lfd f0, 0(r3)
+; CHECK-LINUX-NEXT: xscvsxddp f1, f0
+; CHECK-LINUX-NEXT: blr
+entry:
+ %x = load i64, ptr %p, align 8, !range !0
+ %y = sitofp i64 %x to double
+ ret double %y
+}
+
+define double @case_i64_uitofp_f64(ptr %p) {
+; CHECK-AIX64-LABEL: case_i64_uitofp_f64:
+; CHECK-AIX64: # %bb.0: # %entry
+; CHECK-AIX64-NEXT: lfd f0, 0(r3)
+; CHECK-AIX64-NEXT: xscvuxddp f1, f0
+; CHECK-AIX64-NEXT: blr
+;
+; CHECK-AIX32-LABEL: case_i64_uitofp_f64:
+; CHECK-AIX32: # %bb.0: # %entry
+; CHECK-AIX32-NEXT: lwz r4, 4(r3)
+; CHECK-AIX32-NEXT: stw r4, -4(r1)
+; CHECK-AIX32-NEXT: lwz r3, 0(r3)
+; CHECK-AIX32-NEXT: stw r3, -8(r1)
+; CHECK-AIX32-NEXT: lfd f0, -8(r1)
+; CHECK-AIX32-NEXT: xscvuxddp f1, f0
+; CHECK-AIX32-NEXT: blr
+;
+; CHECK-LINUX-LABEL: case_i64_uitofp_f64:
+; CHECK-LINUX: # %bb.0: # %entry
+; CHECK-LINUX-NEXT: lfd f0, 0(r3)
+; CHECK-LINUX-NEXT: xscvuxddp f1, f0
+; CHECK-LINUX-NEXT: blr
+entry:
+ %x = load i64, ptr %p, align 8, !range !0
+ %y = uitofp i64 %x to double
+ ret double %y
+}
+
+define float @case_i64_sitofp_f32(ptr %p) {
+; CHECK-AIX64-LABEL: case_i64_sitofp_f32:
+; CHECK-AIX64: # %bb.0: # %entry
+; CHECK-AIX64-NEXT: lfd f0, 0(r3)
+; CHECK-AIX64-NEXT: xscvsxdsp f1, f0
+; CHECK-AIX64-NEXT: blr
+;
+; CHECK-AIX32-LABEL: case_i64_sitofp_f32:
+; CHECK-AIX32: # %bb.0: # %entry
+; CHECK-AIX32-NEXT: lwz r4, 4(r3)
+; CHECK-AIX32-NEXT: stw r4, -4(r1)
+; CHECK-AIX32-NEXT: lwz r3, 0(r3)
+; CHECK-AIX32-NEXT: stw r3, -8(r1)
+; CHECK-AIX32-NEXT: lfd f0, -8(r1)
+; CHECK-AIX32-NEXT: xscvsxdsp f1, f0
+; CHECK-AIX32-NEXT: blr
+;
+; CHECK-LINUX-LABEL: case_i64_sitofp_f32:
+; CHECK-LINUX: # %bb.0: # %entry
+; CHECK-LINUX-NEXT: lfd f0, 0(r3)
+; CHECK-LINUX-NEXT: xscvsxdsp f1, f0
+; CHECK-LINUX-NEXT: blr
+entry:
+ %x = load i64, ptr %p, align 8, !range !0
+ %y = sitofp i64 %x to float
+ ret float %y
+}
+
+!0 = !{i64 1, i64 0}
|
🐧 Linux x64 Test Results
✅ The build succeeded and all tests passed. |
lei137
approved these changes
May 21, 2026
| ; RUN: llc -mcpu=pwr8 -ppc-asm-full-reg-names -mtriple=powerpc-ibm-aix < %s | \ | ||
| ; RUN: FileCheck %s --check-prefix=CHECK-AIX32 | ||
| ; RUN: llc -mcpu=pwr8 -ppc-asm-full-reg-names -mtriple=powerpc64le-unknown-linux-gnu < %s | \ | ||
| ; RUN: FileCheck %s --check-prefix=CHECK-LINUX |
Contributor
There was a problem hiding this comment.
nit: maybe we can just use default CHECK for 64bit runs since I don't see a diff in the asm generated.
amy-kwan
force-pushed
the
amyk/fix-i64-to-f64-ranges
branch
from
May 21, 2026 20:19
513a0a9 to
d4e6eb7
Compare
🪟 Windows x64 Test Results
✅ The build succeeded and all tests passed. |
… INT_TO_FP When lowering an i64 load in LowerINT_TO_FP, we were forwarding the original !range metadata to a new f64 load. This is invalid because the metadata no longer matches the value type/semantics, and can trigger assertions when lowering i64 to fp (double or float) conversions. This patch fixes this by passing a nullptr for the Ranges operand when calling getLoad() and adds extra test cases to cover signed/unsigned i64 to f32/f64 conversions and to ensure they do not assert when the !range metadata is present. The assertion this patch attempts to fix is: ``` Assertion failed: (!MMO->getRanges() || (mdconst::extract<ConstantInt>(MMO->getRanges()->getOperand(0)) ->getBitWidth() == MemVT.getScalarSizeInBits() && MemVT.isInteger())) && "Range metadata and load type must match!" ``` This assert was originally seen when building Rust on AIX.
amy-kwan
force-pushed
the
amyk/fix-i64-to-f64-ranges
branch
from
May 26, 2026 19:05
d4e6eb7 to
a397685
Compare
github-project-automation
Bot
moved this from Needs Triage
to Done
in LLVM Release Status
Member
Author
|
/cherry-pick 78f5f77 |
Member
|
/pull-request #199985 |
c-rhodes
pushed a commit
to llvmbot/llvm-project
that referenced
this pull request
Jun 1, 2026
… INT_TO_FP (llvm#198705) When lowering an i64 load in LowerINT_TO_FP, we were forwarding the original !range metadata to a new f64 load. This is invalid because the metadata no longer matches the value type/semantics, and can trigger assertions when lowering i64 to fp (double or float) conversions. This patch fixes this by passing a nullptr for the Ranges operand when calling getLoad() and adds extra test cases to cover signed/unsigned i64 to f32/f64 conversions and to ensure they do not assert when the !range metadata is present. The assertion this patch attempts to fix is: ``` Assertion failed: (!MMO->getRanges() || (mdconst::extract<ConstantInt>(MMO->getRanges()->getOperand(0)) ->getBitWidth() == MemVT.getScalarSizeInBits() && MemVT.isInteger())) && "Range metadata and load type must match!" ``` This assert was originally seen when building Rust on AIX. (cherry picked from commit 78f5f77)
daunabomba
pushed a commit
to daunabomba/llvm-project
that referenced
this pull request
Jun 2, 2026
… INT_TO_FP (llvm#198705) When lowering an i64 load in LowerINT_TO_FP, we were forwarding the original !range metadata to a new f64 load. This is invalid because the metadata no longer matches the value type/semantics, and can trigger assertions when lowering i64 to fp (double or float) conversions. This patch fixes this by passing a nullptr for the Ranges operand when calling getLoad() and adds extra test cases to cover signed/unsigned i64 to f32/f64 conversions and to ensure they do not assert when the !range metadata is present. The assertion this patch attempts to fix is: ``` Assertion failed: (!MMO->getRanges() || (mdconst::extract<ConstantInt>(MMO->getRanges()->getOperand(0)) ->getBitWidth() == MemVT.getScalarSizeInBits() && MemVT.isInteger())) && "Range metadata and load type must match!" ``` This assert was originally seen when building Rust on AIX. (cherry picked from commit 78f5f77)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
When lowering an i64 load in LowerINT_TO_FP, we were forwarding the original !range metadata to a new f64 load. This is invalid because the metadata no longer matches the value type/semantics, and can trigger assertions when lowering i64 to fp (double or float) conversions.
This patch fixes this by passing a nullptr for the Ranges operand when calling getLoad() and adds extra test cases to cover signed/unsigned i64 to f32/f64 conversions and to ensure they do not assert when the !range metadata is present.
The assertion this patch attempts to fix is:
This assert was originally seen when building Rust on AIX.