[GlobalISel] Use-Vector-Truncate Opt Needs Elt Type Check #146003
Conversation
In the pre-legalizer combiner, there exists a bug with UseVectorTruncate match-apply optimization. When the destinations' types do not match the vector element type of the G_UNMERGE_VALUES instruction, the resulting collapsed truncate does not preserve original functional behavior. This commit introduces a simple type check to ensure that the destination types match the vector element type.
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be notified. If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers. If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
|
@llvm/pr-subscribers-backend-aarch64 Author: Daniel Man (daniel-man) ChangesIn the pre-legalizer combiner, there exists a bug with UseVectorTruncate Full diff: https://github.com/llvm/llvm-project/pull/146003.diff 2 Files Affected:
diff --git a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
index b1e851183de0d..05dd269d48921 100644
--- a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
@@ -3487,6 +3487,13 @@ bool CombinerHelper::matchUseVectorTruncate(MachineInstr &MI,
if (!DstTy.getElementCount().isKnownMultipleOf(UnmergeSrcTy.getNumElements()))
return false;
+ // Check the unmerge source and destination element types match
+ LLT UnmergeSrcEltTy = UnmergeSrcTy.getElementType();
+ Register UnmergeDstReg = UnmergeMI->getOperand(0).getReg();
+ LLT UnmergeDstEltTy = MRI.getType(UnmergeDstReg);
+ if (UnmergeSrcEltTy != UnmergeDstEltTy)
+ return false;
+
// Only generate legal instructions post-legalizer
if (!IsPreLegalize) {
LLT MidTy = DstTy.changeElementType(UnmergeSrcTy.getScalarType());
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-use-vector-truncate.mir b/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-use-vector-truncate.mir
new file mode 100644
index 0000000000000..278f7e71f3d8b
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-use-vector-truncate.mir
@@ -0,0 +1,50 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=aarch64 -run-pass=aarch64-prelegalizer-combiner -global-isel -verify-machineinstrs %s -o - | FileCheck %s
+---
+name: test_foldable_s32
+body: |
+ bb.0:
+ liveins: $x0, $x1
+
+ ; CHECK-LABEL: name: test_foldable_s32
+ ; CHECK: liveins: $x0, $x1
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[SRC:%src]]:_(<2 x s32>) = COPY $x1
+ ; CHECK-NEXT: [[UNDEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
+ ; CHECK-NEXT: [[MID:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[SRC]](<2 x s32>), [[UNDEF]](<2 x s32>)
+ ; CHECK-NEXT: [[DST:%dst]]:_(<4 x s16>) = G_TRUNC [[MID]]
+ ; CHECK-NEXT: $x0 = COPY [[DST]]
+
+ %src:_(<2 x s32>) = COPY $x1
+ %a:_(s32), %b:_(s32) = G_UNMERGE_VALUES %src:_(<2 x s32>)
+ %T_a:_(s16) = G_TRUNC %a:_(s32)
+ %T_b:_(s16) = G_TRUNC %b:_(s32)
+ %Undef:_(s16) = G_IMPLICIT_DEF
+ %dst:_(<4 x s16>) = G_BUILD_VECTOR %T_a:_(s16), %T_b:_(s16), %Undef:_(s16), %Undef:_(s16)
+ $x0 = COPY %dst
+...
+---
+name: test_unfoldable_s32
+body: |
+ bb.0:
+ liveins: $w0, $x1
+
+ ; CHECK-LABEL: name: test_unfoldable_s32
+ ; CHECK: liveins: $w0, $x1
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[SRC:%src]]:_(<2 x s32>) = COPY $x1
+ ; CHECK-NEXT: [[A:%a]]:_(s16), [[B:%b]]:_(s16), [[C:%c]]:_(s16), %d:_(s16) = G_UNMERGE_VALUES [[SRC]](<2 x s32>)
+ ; CHECK-NEXT: [[T_A:%T_a]]:_(s8) = G_TRUNC [[A]](s16)
+ ; CHECK-NEXT: [[T_B:%T_b]]:_(s8) = G_TRUNC [[B]](s16)
+ ; CHECK-NEXT: [[T_C:%T_c]]:_(s8) = G_TRUNC [[C]](s16)
+ ; CHECK-NEXT: [[UNDEF:%Undef]]:_(s8) = G_IMPLICIT_DEF
+ ; CHECK-NEXT: [[DST:%dst]]:_(<4 x s8>) = G_BUILD_VECTOR [[T_A]](s8), [[T_B]](s8), [[T_C]](s8), [[UNDEF]](s8)
+ ; CHECK-NEXT: $w0 = COPY [[DST]](<4 x s8>)
+ %src:_(<2 x s32>) = COPY $x1
+ %a:_(s16), %b:_(s16), %c:_(s16), %d:_(s16) = G_UNMERGE_VALUES %src:_(<2 x s32>)
+ %T_a:_(s8) = G_TRUNC %a:_(s16)
+ %T_b:_(s8) = G_TRUNC %b:_(s16)
+ %T_c:_(s8) = G_TRUNC %c:_(s16)
+ %Undef:_(s8) = G_IMPLICIT_DEF
+ %dst:_(<4 x s8>) = G_BUILD_VECTOR %T_a:_(s8), %T_b:_(s8), %T_c:_(s8), %Undef:_(s8)
+ $w0 = COPY %dst
|
|
@llvm/pr-subscribers-llvm-globalisel Author: Daniel Man (daniel-man) ChangesIn the pre-legalizer combiner, there exists a bug with UseVectorTruncate Full diff: https://github.com/llvm/llvm-project/pull/146003.diff 2 Files Affected:
diff --git a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
index b1e851183de0d..05dd269d48921 100644
--- a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
@@ -3487,6 +3487,13 @@ bool CombinerHelper::matchUseVectorTruncate(MachineInstr &MI,
if (!DstTy.getElementCount().isKnownMultipleOf(UnmergeSrcTy.getNumElements()))
return false;
+ // Check the unmerge source and destination element types match
+ LLT UnmergeSrcEltTy = UnmergeSrcTy.getElementType();
+ Register UnmergeDstReg = UnmergeMI->getOperand(0).getReg();
+ LLT UnmergeDstEltTy = MRI.getType(UnmergeDstReg);
+ if (UnmergeSrcEltTy != UnmergeDstEltTy)
+ return false;
+
// Only generate legal instructions post-legalizer
if (!IsPreLegalize) {
LLT MidTy = DstTy.changeElementType(UnmergeSrcTy.getScalarType());
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-use-vector-truncate.mir b/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-use-vector-truncate.mir
new file mode 100644
index 0000000000000..278f7e71f3d8b
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-use-vector-truncate.mir
@@ -0,0 +1,50 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=aarch64 -run-pass=aarch64-prelegalizer-combiner -global-isel -verify-machineinstrs %s -o - | FileCheck %s
+---
+name: test_foldable_s32
+body: |
+ bb.0:
+ liveins: $x0, $x1
+
+ ; CHECK-LABEL: name: test_foldable_s32
+ ; CHECK: liveins: $x0, $x1
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[SRC:%src]]:_(<2 x s32>) = COPY $x1
+ ; CHECK-NEXT: [[UNDEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
+ ; CHECK-NEXT: [[MID:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[SRC]](<2 x s32>), [[UNDEF]](<2 x s32>)
+ ; CHECK-NEXT: [[DST:%dst]]:_(<4 x s16>) = G_TRUNC [[MID]]
+ ; CHECK-NEXT: $x0 = COPY [[DST]]
+
+ %src:_(<2 x s32>) = COPY $x1
+ %a:_(s32), %b:_(s32) = G_UNMERGE_VALUES %src:_(<2 x s32>)
+ %T_a:_(s16) = G_TRUNC %a:_(s32)
+ %T_b:_(s16) = G_TRUNC %b:_(s32)
+ %Undef:_(s16) = G_IMPLICIT_DEF
+ %dst:_(<4 x s16>) = G_BUILD_VECTOR %T_a:_(s16), %T_b:_(s16), %Undef:_(s16), %Undef:_(s16)
+ $x0 = COPY %dst
+...
+---
+name: test_unfoldable_s32
+body: |
+ bb.0:
+ liveins: $w0, $x1
+
+ ; CHECK-LABEL: name: test_unfoldable_s32
+ ; CHECK: liveins: $w0, $x1
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[SRC:%src]]:_(<2 x s32>) = COPY $x1
+ ; CHECK-NEXT: [[A:%a]]:_(s16), [[B:%b]]:_(s16), [[C:%c]]:_(s16), %d:_(s16) = G_UNMERGE_VALUES [[SRC]](<2 x s32>)
+ ; CHECK-NEXT: [[T_A:%T_a]]:_(s8) = G_TRUNC [[A]](s16)
+ ; CHECK-NEXT: [[T_B:%T_b]]:_(s8) = G_TRUNC [[B]](s16)
+ ; CHECK-NEXT: [[T_C:%T_c]]:_(s8) = G_TRUNC [[C]](s16)
+ ; CHECK-NEXT: [[UNDEF:%Undef]]:_(s8) = G_IMPLICIT_DEF
+ ; CHECK-NEXT: [[DST:%dst]]:_(<4 x s8>) = G_BUILD_VECTOR [[T_A]](s8), [[T_B]](s8), [[T_C]](s8), [[UNDEF]](s8)
+ ; CHECK-NEXT: $w0 = COPY [[DST]](<4 x s8>)
+ %src:_(<2 x s32>) = COPY $x1
+ %a:_(s16), %b:_(s16), %c:_(s16), %d:_(s16) = G_UNMERGE_VALUES %src:_(<2 x s32>)
+ %T_a:_(s8) = G_TRUNC %a:_(s16)
+ %T_b:_(s8) = G_TRUNC %b:_(s16)
+ %T_c:_(s8) = G_TRUNC %c:_(s16)
+ %Undef:_(s8) = G_IMPLICIT_DEF
+ %dst:_(<4 x s8>) = G_BUILD_VECTOR %T_a:_(s8), %T_b:_(s8), %T_c:_(s8), %Undef:_(s8)
+ $w0 = COPY %dst
|
llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-use-vector-truncate.mir
Show resolved
Hide resolved
| @@ -0,0 +1,50 @@ | |||
| # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py | |||
| # RUN: llc -mtriple=aarch64 -run-pass=aarch64-prelegalizer-combiner -global-isel -verify-machineinstrs %s -o - | FileCheck %s | |||
There was a problem hiding this comment.
| # RUN: llc -mtriple=aarch64 -run-pass=aarch64-prelegalizer-combiner -global-isel -verify-machineinstrs %s -o - | FileCheck %s | |
| # RUN: llc -mtriple=aarch64 -run-pass=aarch64-prelegalizer-combiner %s -o - | FileCheck %s |
-global-isel is never needed when directly running passes, and llc ought to verify any mir run before and after
|
Thanks @arsenm for the review, could you also approve the 2 workflows that require approval? |
arsenm
changed the title
daniel-man
changed the title
|
@daniel-man Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR. Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues. How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
In the pre-legalizer combiner, there exists a bug with UseVectorTruncate match-apply optimization. When the destinations' types do not match the vector element type of the G_UNMERGE_VALUES instruction, the resulting collapsed truncate does not preserve original functional behavior. This commit introduces a simple type check to ensure that the destination types match the vector element type.
In the pre-legalizer combiner, there exists a bug with UseVectorTruncate
match-apply optimization. When the destinations' types do not match the
vector element type of the G_UNMERGE_VALUES instruction, the resulting
collapsed truncate does not preserve original functional behavior. This
commit introduces a simple type check to ensure that the destination
types match the vector element type.