[GlobalISel] Import extract/insert subvector #110287
Merged
Conversation
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
Member
|
@llvm/pr-subscribers-backend-aarch64 Author: Thorsten Schütt (tschuett) ChangesTests are limited to fixed-length vectors. Test: AArch64/GlobalISel/irtranslator-subvector.ll Reference: https://llvm.org/docs/LangRef.html#llvm-vector-extract-intrinsic Full diff: https://github.com/llvm/llvm-project/pull/110287.diff 2 Files Affected:
diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
index 7ff8d2446eec5d..a0649f712bd642 100644
--- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
@@ -2588,6 +2588,20 @@ bool IRTranslator::translateKnownIntrinsic(const CallInst &CI, Intrinsic::ID ID,
getOrCreateVReg(*CI.getOperand(0)),
getOrCreateVReg(*CI.getOperand(1)));
return true;
+ case Intrinsic::vector_extract: {
+ ConstantInt *Index = cast<ConstantInt>(CI.getOperand(1));
+ MIRBuilder.buildExtractSubvector(getOrCreateVReg(CI),
+ getOrCreateVReg(*CI.getOperand(0)),
+ Index->getZExtValue());
+ return true;
+ }
+ case Intrinsic::vector_insert: {
+ ConstantInt *Index = cast<ConstantInt>(CI.getOperand(2));
+ MIRBuilder.buildInsertSubvector(
+ getOrCreateVReg(CI), getOrCreateVReg(*CI.getOperand(0)),
+ getOrCreateVReg(*CI.getOperand(1)), Index->getZExtValue());
+ return true;
+ }
case Intrinsic::prefetch: {
Value *Addr = CI.getOperand(0);
unsigned RW = cast<ConstantInt>(CI.getOperand(1))->getZExtValue();
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-subvector.ll b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-subvector.ll
new file mode 100644
index 00000000000000..bdcd8e3d99af87
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-subvector.ll
@@ -0,0 +1,78 @@
+; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+; RUN: llc -O0 -mtriple=aarch64-linux-gnu -global-isel -stop-after=irtranslator %s -o - | FileCheck %s
+
+define i32 @extract_v4i32_vector_insert_const(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_insert_const
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; CHECK-NEXT: [[INSERT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_INSERT_SUBVECTOR [[COPY]], [[COPY1]](<2 x s32>), 0
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[INSERT_SUBVECTOR]](<4 x s32>), [[C]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.insert.v4i32.v2i32(<4 x i32> %a, <2 x i32> %b, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 1
+ ret i32 %d
+}
+
+define i32 @extract_v4i32_vector_insert(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_insert
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[INSERT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_INSERT_SUBVECTOR [[COPY]], [[COPY1]](<2 x s32>), 0
+ ; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[INSERT_SUBVECTOR]](<4 x s32>), [[ZEXT]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.insert.v4i32.v2i32(<4 x i32> %a, <2 x i32> %b, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 %c
+ ret i32 %d
+}
+
+define i32 @extract_v4i32_vector_extract(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_extract
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[EXTRACT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT_SUBVECTOR [[COPY]](<4 x s32>), 0
+ ; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[EXTRACT_SUBVECTOR]](<4 x s32>), [[ZEXT]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.extract.v2i32.v4i32(<4 x i32> %a, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 %c
+ ret i32 %d
+}
+
+define i32 @extract_v4i32_vector_extract_const(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_extract_const
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
+ ; CHECK-NEXT: [[EXTRACT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT_SUBVECTOR [[COPY]](<4 x s32>), 0
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[EXTRACT_SUBVECTOR]](<4 x s32>), [[C]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.extract.v2i32.v4i32(<4 x i32> %a, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 0
+ ret i32 %d
+}
|
Member
|
@llvm/pr-subscribers-llvm-globalisel Author: Thorsten Schütt (tschuett) ChangesTests are limited to fixed-length vectors. Test: AArch64/GlobalISel/irtranslator-subvector.ll Reference: https://llvm.org/docs/LangRef.html#llvm-vector-extract-intrinsic Full diff: https://github.com/llvm/llvm-project/pull/110287.diff 2 Files Affected:
diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
index 7ff8d2446eec5d..a0649f712bd642 100644
--- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
@@ -2588,6 +2588,20 @@ bool IRTranslator::translateKnownIntrinsic(const CallInst &CI, Intrinsic::ID ID,
getOrCreateVReg(*CI.getOperand(0)),
getOrCreateVReg(*CI.getOperand(1)));
return true;
+ case Intrinsic::vector_extract: {
+ ConstantInt *Index = cast<ConstantInt>(CI.getOperand(1));
+ MIRBuilder.buildExtractSubvector(getOrCreateVReg(CI),
+ getOrCreateVReg(*CI.getOperand(0)),
+ Index->getZExtValue());
+ return true;
+ }
+ case Intrinsic::vector_insert: {
+ ConstantInt *Index = cast<ConstantInt>(CI.getOperand(2));
+ MIRBuilder.buildInsertSubvector(
+ getOrCreateVReg(CI), getOrCreateVReg(*CI.getOperand(0)),
+ getOrCreateVReg(*CI.getOperand(1)), Index->getZExtValue());
+ return true;
+ }
case Intrinsic::prefetch: {
Value *Addr = CI.getOperand(0);
unsigned RW = cast<ConstantInt>(CI.getOperand(1))->getZExtValue();
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-subvector.ll b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-subvector.ll
new file mode 100644
index 00000000000000..bdcd8e3d99af87
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-subvector.ll
@@ -0,0 +1,78 @@
+; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+; RUN: llc -O0 -mtriple=aarch64-linux-gnu -global-isel -stop-after=irtranslator %s -o - | FileCheck %s
+
+define i32 @extract_v4i32_vector_insert_const(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_insert_const
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; CHECK-NEXT: [[INSERT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_INSERT_SUBVECTOR [[COPY]], [[COPY1]](<2 x s32>), 0
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[INSERT_SUBVECTOR]](<4 x s32>), [[C]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.insert.v4i32.v2i32(<4 x i32> %a, <2 x i32> %b, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 1
+ ret i32 %d
+}
+
+define i32 @extract_v4i32_vector_insert(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_insert
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[INSERT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_INSERT_SUBVECTOR [[COPY]], [[COPY1]](<2 x s32>), 0
+ ; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[INSERT_SUBVECTOR]](<4 x s32>), [[ZEXT]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.insert.v4i32.v2i32(<4 x i32> %a, <2 x i32> %b, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 %c
+ ret i32 %d
+}
+
+define i32 @extract_v4i32_vector_extract(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_extract
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[EXTRACT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT_SUBVECTOR [[COPY]](<4 x s32>), 0
+ ; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[EXTRACT_SUBVECTOR]](<4 x s32>), [[ZEXT]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.extract.v2i32.v4i32(<4 x i32> %a, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 %c
+ ret i32 %d
+}
+
+define i32 @extract_v4i32_vector_extract_const(<4 x i32> %a, <2 x i32> %b, i32 %c) {
+ ; CHECK-LABEL: name: extract_v4i32_vector_extract_const
+ ; CHECK: bb.1.entry:
+ ; CHECK-NEXT: liveins: $d1, $q0, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $w0
+ ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
+ ; CHECK-NEXT: [[EXTRACT_SUBVECTOR:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT_SUBVECTOR [[COPY]](<4 x s32>), 0
+ ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[EXTRACT_SUBVECTOR]](<4 x s32>), [[C]](s64)
+ ; CHECK-NEXT: $w0 = COPY [[EVEC]](s32)
+ ; CHECK-NEXT: RET_ReallyLR implicit $w0
+entry:
+ %vector = call <4 x i32> @llvm.vector.extract.v2i32.v4i32(<4 x i32> %a, i64 0)
+ %d = extractelement <4 x i32> %vector, i32 0
+ ret i32 %d
+}
|
qcolombet
reviewed
Sep 28, 2024
added 4 commits
September 29, 2024 13:30
Tests are limited to fixed-length vectors. Test: AArch64/GlobalISel/irtranslator-subvector.ll Reference: https://llvm.org/docs/LangRef.html#llvm-vector-extract-intrinsic https://llvm.org/docs/LangRef.html#llvm-vector-insert-intrinsic
arsenm
reviewed
Sep 30, 2024
| } | ||
| if (auto *InputType = | ||
| dyn_cast<FixedVectorType>(U.getOperand(0)->getType())) { | ||
| // We are inserting an illegal fixed vector into a fixed vector, use the |
Contributor
There was a problem hiding this comment.
InputType not used, is this missing the InputType && InputType->getNumElements() == 1 like above?
Author
There was a problem hiding this comment.
Nope. It is meant to say InputType && InputType->getNumElements() != 1.
Author
|
The unused |
arsenm
reviewed
Sep 30, 2024
Contributor
arsenm
left a comment
arsenm
left a comment
There was a problem hiding this comment.
LGTM, but could use some test coverage for different element classes
| %vector = call <1 x i32> @llvm.vector.insert.v1i32.v4i32(<1 x i32> %a, <1 x i32> %b, i64 0) | ||
| store <1 x i32> %vector, ptr %p, align 16 | ||
| ret i32 1 | ||
| } |
Contributor
There was a problem hiding this comment.
Can you add some tests with FP types, and vectors of pointers?
Author
There was a problem hiding this comment.
Sure. There are some limitations with scalable vectors.
added 2 commits
September 30, 2024 17:48
arsenm
approved these changes
Sep 30, 2024
| ; CHECK-NEXT: $d0 = COPY [[EVEC]](s64) | ||
| ; CHECK-NEXT: RET_ReallyLR implicit $d0 | ||
| entry: | ||
| %vector = call <4 x double> @llvm.vector.insert.v4double.v2double(<4 x double> %a, <2 x double> %b, i64 0) |
Contributor
There was a problem hiding this comment.
Use canonical mangling, f64 not double
arsenm
approved these changes
Sep 30, 2024
Author
|
Thanks for review. |
Sterling-Augustine
pushed a commit
to Sterling-Augustine/llvm-project
that referenced
this pull request
Oct 3, 2024
Test: AArch64/GlobalISel/irtranslator-subvector.ll Reference: https://llvm.org/docs/LangRef.html#llvm-vector-extract-intrinsic https://llvm.org/docs/LangRef.html#llvm-vector-insert-intrinsic
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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.
Test: AArch64/GlobalISel/irtranslator-subvector.ll
Reference:
https://llvm.org/docs/LangRef.html#llvm-vector-extract-intrinsic
https://llvm.org/docs/LangRef.html#llvm-vector-insert-intrinsic