[llvm] [GlobalISel] Import extract/insert subvector (PR #110287)

via llvm-commits llvm-commits at lists.llvm.org
Fri Sep 27 08:49:44 PDT 2024


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-aarch64

Author: Thorsten Schütt (tschuett)

<details>
<summary>Changes</summary>

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

---
Full diff: https://github.com/llvm/llvm-project/pull/110287.diff


2 Files Affected:

- (modified) llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp (+14) 
- (added) llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-subvector.ll (+78) 


``````````diff
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
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/110287


More information about the llvm-commits mailing list