[CIR] Maximum across vector (IEEE754)

[Ko496-glitch]

Part of #185382

Move the test cases to
intrinsics.c
Removed the test cases from
neon-intrinsics.c

Removed neon-across.c

[llvmorg-github-actions]

@llvm/pr-subscribers-clangir

@llvm/pr-subscribers-clang

Author: Kartik Ohlan (Ko496-glitch)

Changes

Part of #185382

Move the test cases to
intrinsics.c
Removed the test cases from
neon-intrinsics.c

Removed neon-across.c

---

Full diff: https://github.com/llvm/llvm-project/pull/199779.diff

3 Files Affected:

• (modified) clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp (+3)
• (removed) clang/test/CodeGen/AArch64/neon-across.c (-18)
• (modified) clang/test/CodeGen/AArch64/neon/intrinsics.c (+36)

diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
index ca5a4680ee79f..a956a4dddb57f 100644
--- a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
@@ -372,6 +372,9 @@ static mlir::Value emitCommonNeonSISDBuiltinExpr(
   case NEON::BI__builtin_neon_vqrshrund_n_s64:
   case NEON::BI__builtin_neon_vqrshrnd_n_s64:
   case NEON::BI__builtin_neon_vqrshrnd_n_u64:
+  case NEON::BI__builtin_neon_vmaxnmv_f32:
+  case NEON::BI__builtin_neon_vmaxnmvq_f32:
+  case NEON::BI__builtin_neon_vmaxnmvq_f64:
     return emitNeonCall(cgf.cgm, cgf.getBuilder(),
                         {cgf.convertType(expr->getArg(0)->getType())}, ops,
                         llvmIntrName, cgf.convertType(expr->getType()), loc);
diff --git a/clang/test/CodeGen/AArch64/neon-across.c b/clang/test/CodeGen/AArch64/neon-across.c
deleted file mode 100644
index f248e31b374ba..0000000000000
--- a/clang/test/CodeGen/AArch64/neon-across.c
+++ /dev/null
@@ -1,18 +0,0 @@
-// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature
-// RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
-// RUN:  -disable-O0-optnone -emit-llvm -o - %s | opt -S -passes=mem2reg | FileCheck %s
-
-// REQUIRES: aarch64-registered-target || arm-registered-target
-
-#include <arm_neon.h>
-
-
-// CHECK-LABEL: define {{[^@]+}}@test_vmaxnmvq_f32
-// CHECK-SAME: (<4 x float> noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VMAXNMVQ_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v4f32(<4 x float> [[A]])
-// CHECK-NEXT:    ret float [[VMAXNMVQ_F32_I]]
-//
-float32_t test_vmaxnmvq_f32(float32x4_t a) {
-  return vmaxnmvq_f32(a);
-}
diff --git a/clang/test/CodeGen/AArch64/neon/intrinsics.c b/clang/test/CodeGen/AArch64/neon/intrinsics.c
index b07f61a712f90..8b5c495d9cc2a 100644
--- a/clang/test/CodeGen/AArch64/neon/intrinsics.c
+++ b/clang/test/CodeGen/AArch64/neon/intrinsics.c
@@ -24,6 +24,42 @@
 
 #include <arm_neon.h>
 
+//===------------------------------------------------------===//
+// 2.1.1.13.5  Maximum across vector (IEEE754)
+
+//===------------------------------------------------------===//
+
+
+//ALL-LABEL: @test_vmaxnmv_f32(
+float32_t test_vmaxnmv_f32(float32x2_t a) {
+    //CIR: cir.call_llvm_intrinsic "aarch64.neon.fmaxnmv"
+    // 
+    // LLVM-SAME:<2 x float> {{.*}} [[A:%.*]])
+    // LLVM:  [[VMAXNMV_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v2f32(<2 x float> [[A]])
+    // LLVM: ret float [[VMAXNMV_F32_I]]
+  return vmaxnmv_f32(a);
+}
+
+//ALL-LABEL: @test_vmaxnmvq_f32(
+float32_t test_vmaxnmvq_f32(float32x4_t a) {
+    //CIR: cir.call_llvm_intrinsic "aarch64.neon.fmaxnmv"
+    // 
+    // LLVM-SAME: (<4 x float> {{.*}}[[A:%.*]])
+    // LLVM:  [[VMAXNMVQ_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v4f32(<4 x float> [[A]])
+    // LLVM: ret float [[VMAXNMVQ_F32_I]]
+  return vmaxnmvq_f32(a);
+}
+
+//ALL-LABEL: @test_vmaxnmvq_f64(
+float64_t test_vmaxnmvq_f64(float64x2_t a) {
+    //CIR: cir.call_llvm_intrinsic "aarch64.neon.fmaxnmv"
+    // 
+    // LLVM-SAME:  <2 x double> {{.*}} [[A:%.*]])
+    // LLVM:  [[VMAXNMVQ_F64_I:%.*]] = call double @llvm.aarch64.neon.fmaxnmv.f64.v2f64(<2 x double> [[A]])
+    // LLVM:  ret double [[VMAXNMVQ_F64_I]]
+  return vmaxnmvq_f64(a);
+}
+
 //===------------------------------------------------------===//
 // 2.1.3.2.7  Vector saturating rounding shift right and narrow
 // TODO: Implement SISD variants

[Ko496-glitch]

ping @banach-space

[github-actions]

🪟 Windows x64 Test Results

• 55687 tests passed
• 3387 tests skipped

✅ The build succeeded and all tests passed.

[github-actions]

🐧 Linux x64 Test Results

• 119684 tests passed
• 3992 tests skipped

✅ The build succeeded and all tests passed.

[Ko496-glitch]

@banach-space looks like during conflict my change is spitting

return emitNeonCall(cgf.cgm, cgf.getBuilder(),
                        {cgf.convertType(expr->getArg(0)->getType())}, ops,
                        llvmIntrName, cgf.convertType(expr->getType()), loc);

which should be removed after ##196776 landed ?

[banach-space]

Thanks!

There should be 2 more tests that are removed.

[banach-space]

[nit]

Suggested change

	//CIR: cir.call @vmaxnmv_f32
	//
	// LLVM-SAME: <2 x float> {{.*}} [[A:%.*]])
	// LLVM: [[VMAXNMV_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v2f32(<2 x float> [[A]])
	// LLVM: ret float [[VMAXNMV_F32_I]]
	return vmaxnmv_f32(a);
	// CIR: cir.call @vmaxnmv_f32
	// LLVM-SAME: <2 x float> {{.*}} [[A:%.*]])
	// LLVM: [[VMAXNMV_F32_I:%.*]] = call float @llvm.aarch64.neon.fmaxnmv.f32.v2f32(<2 x float> [[A]])
	// LLVM: ret float [[VMAXNMV_F32_I]]
	return vmaxnmv_f32(a);