[llvm] [GlobalISel] Adding support for handling G_ASSERT_{SEXT,ZEXT,ALIGN} i… (PR #74196)
via llvm-commits
llvm-commits at lists.llvm.org
Sat Dec 2 06:44:55 PST 2023
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-globalisel
Author: Dávid Ferenc Szabó (dfszabo)
<details>
<summary>Changes</summary>
…n artifact combiner
These instructions are hint for optimizations and can be treated as copies and are handled as such with this change. Without it is possible to run into an assertion, since tryCombineUnmergeValues rightfully use getDefIgnoringCopies to get the source MI, which already handle these hint instructions and treat them as copies. The problem is that markDefDead only considers COPYs, which will lead to crash with assertion for cases like testUnmergeHintOfTrunc.
---
Full diff: https://github.com/llvm/llvm-project/pull/74196.diff
3 Files Affected:
- (modified) llvm/include/llvm/CodeGen/GlobalISel/LegalizationArtifactCombiner.h (+10-11)
- (modified) llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-hoisted-constants.ll (+1-4)
- (modified) llvm/test/CodeGen/AArch64/GlobalISel/legalize-simple.mir (+64)
``````````diff
diff --git a/llvm/include/llvm/CodeGen/GlobalISel/LegalizationArtifactCombiner.h b/llvm/include/llvm/CodeGen/GlobalISel/LegalizationArtifactCombiner.h
index 851353042cc26..da330b517c280 100644
--- a/llvm/include/llvm/CodeGen/GlobalISel/LegalizationArtifactCombiner.h
+++ b/llvm/include/llvm/CodeGen/GlobalISel/LegalizationArtifactCombiner.h
@@ -1366,6 +1366,9 @@ class LegalizationArtifactCombiner {
// Adding Use to ArtifactList.
WrapperObserver.changedInstr(Use);
break;
+ case TargetOpcode::G_ASSERT_SEXT:
+ case TargetOpcode::G_ASSERT_ZEXT:
+ case TargetOpcode::G_ASSERT_ALIGN:
case TargetOpcode::COPY: {
Register Copy = Use.getOperand(0).getReg();
if (Copy.isVirtual())
@@ -1392,6 +1395,9 @@ class LegalizationArtifactCombiner {
case TargetOpcode::G_ANYEXT:
case TargetOpcode::G_SEXT:
case TargetOpcode::G_EXTRACT:
+ case TargetOpcode::G_ASSERT_SEXT:
+ case TargetOpcode::G_ASSERT_ZEXT:
+ case TargetOpcode::G_ASSERT_ALIGN:
return MI.getOperand(1).getReg();
case TargetOpcode::G_UNMERGE_VALUES:
return MI.getOperand(MI.getNumOperands() - 1).getReg();
@@ -1425,7 +1431,8 @@ class LegalizationArtifactCombiner {
if (MRI.hasOneUse(PrevRegSrc)) {
if (TmpDef != &DefMI) {
assert((TmpDef->getOpcode() == TargetOpcode::COPY ||
- isArtifactCast(TmpDef->getOpcode())) &&
+ isArtifactCast(TmpDef->getOpcode()) ||
+ isPreISelGenericOptimizationHint(TmpDef->getOpcode())) &&
"Expecting copy or artifact cast here");
DeadInsts.push_back(TmpDef);
@@ -1509,16 +1516,8 @@ class LegalizationArtifactCombiner {
/// Looks through copy instructions and returns the actual
/// source register.
Register lookThroughCopyInstrs(Register Reg) {
- using namespace llvm::MIPatternMatch;
-
- Register TmpReg;
- while (mi_match(Reg, MRI, m_Copy(m_Reg(TmpReg)))) {
- if (MRI.getType(TmpReg).isValid())
- Reg = TmpReg;
- else
- break;
- }
- return Reg;
+ Register TmpReg = getSrcRegIgnoringCopies(Reg, MRI);
+ return TmpReg.isValid() ? TmpReg : Reg;
}
};
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-hoisted-constants.ll b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-hoisted-constants.ll
index a7d7a1e81617e..5867326c18aa6 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-hoisted-constants.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-hoisted-constants.ll
@@ -46,14 +46,11 @@ define i32 @test(i32 %a, i1 %c) {
; PRESELECTION-NEXT: {{ $}}
; PRESELECTION-NEXT: [[COPY:%[0-9]+]]:gpr(s32) = COPY $w0
; PRESELECTION-NEXT: [[COPY1:%[0-9]+]]:gpr(s32) = COPY $w1
- ; PRESELECTION-NEXT: [[TRUNC:%[0-9]+]]:gpr(s8) = G_TRUNC [[COPY1]](s32)
- ; PRESELECTION-NEXT: [[ASSERT_ZEXT:%[0-9]+]]:gpr(s8) = G_ASSERT_ZEXT [[TRUNC]], 1
; PRESELECTION-NEXT: [[C:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 0
; PRESELECTION-NEXT: [[C1:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 100000
; PRESELECTION-NEXT: [[CONSTANT_FOLD_BARRIER:%[0-9]+]]:gpr(s32) = G_CONSTANT_FOLD_BARRIER [[C1]]
- ; PRESELECTION-NEXT: [[ANYEXT:%[0-9]+]]:gpr(s32) = G_ANYEXT [[ASSERT_ZEXT]](s8)
; PRESELECTION-NEXT: [[C2:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 1
- ; PRESELECTION-NEXT: [[AND:%[0-9]+]]:gpr(s32) = G_AND [[ANYEXT]], [[C2]]
+ ; PRESELECTION-NEXT: [[AND:%[0-9]+]]:gpr(s32) = G_AND [[COPY1]], [[C2]]
; PRESELECTION-NEXT: G_BRCOND [[AND]](s32), %bb.3
; PRESELECTION-NEXT: G_BR %bb.2
; PRESELECTION-NEXT: {{ $}}
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/legalize-simple.mir b/llvm/test/CodeGen/AArch64/GlobalISel/legalize-simple.mir
index 62865bcfdf081..ce02aa380efe1 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/legalize-simple.mir
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/legalize-simple.mir
@@ -136,6 +136,47 @@ body: |
$x0 = COPY %3(s64)
RET_ReallyLR implicit $x0
+...
+---
+name: testExtOfHintOfTrunc
+body: |
+ bb.0:
+ liveins: $x0
+
+ ; CHECK-LABEL: name: testExtOfHintOfTrunc
+ ; CHECK: liveins: $x0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
+ ; CHECK-NEXT: $x0 = COPY [[COPY]](s64)
+ ; CHECK-NEXT: RET_ReallyLR implicit $x0
+ %0:_(s64) = COPY $x0
+ %1:_(s8) = G_TRUNC %0(s64)
+ %2:_(s8) = G_ASSERT_ZEXT %1(s8), 1
+ %3:_(s64) = G_ANYEXT %2(s8)
+ $x0 = COPY %3(s64)
+ RET_ReallyLR implicit $x0
+
+...
+---
+name: testExtOfCopyAndHintOfTrunc
+body: |
+ bb.0:
+ liveins: $x0
+
+ ; CHECK-LABEL: name: testExtOfCopyAndHintOfTrunc
+ ; CHECK: liveins: $x0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
+ ; CHECK-NEXT: $x0 = COPY [[COPY]](s64)
+ ; CHECK-NEXT: RET_ReallyLR implicit $x0
+ %0:_(s64) = COPY $x0
+ %1:_(s8) = G_TRUNC %0(s64)
+ %2:_(s8) = COPY %1(s8)
+ %3:_(s8) = G_ASSERT_ZEXT %2(s8), 1
+ %4:_(s64) = G_ANYEXT %3(s8)
+ $x0 = COPY %4(s64)
+ RET_ReallyLR implicit $x0
+
...
---
name: testExtOf2CopyOfTrunc
@@ -158,3 +199,26 @@ body: |
RET_ReallyLR implicit $x0
...
+---
+name: testUnmergeHintOfTrunc
+body: |
+ bb.0:
+ liveins: $x0
+
+ ; CHECK-LABEL: name: testUnmergeHintOfTrunc
+ ; CHECK: liveins: $x0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
+ ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8), [[UV3:%[0-9]+]]:_(s8), [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[COPY]](s64)
+ ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[UV1]](s8)
+ ; CHECK-NEXT: $x0 = COPY [[ANYEXT]](s64)
+ ; CHECK-NEXT: RET_ReallyLR implicit $x0
+ %0:_(s64) = COPY $x0
+ %1:_(s16) = G_TRUNC %0(s64)
+ %2:_(s16) = G_ASSERT_ZEXT %1(s16), 1
+ %3:_(s8), %4:_(s8) = G_UNMERGE_VALUES %2(s16)
+ %5:_(s64) = G_ANYEXT %4(s8)
+ $x0 = COPY %5(s64)
+ RET_ReallyLR implicit $x0
+
+...
``````````
</details>
https://github.com/llvm/llvm-project/pull/74196
More information about the llvm-commits
mailing list