[clang] [CIR] X86 vector fcmp-sse vector builtins (PR #167125)

[Amr Hesham via cfe-commits]

================
@@ -0,0 +1,102 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -O2 -fclangir -emit-cir %s -o -  | FileCheck %s --check-prefix=CIR
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -O2 -fclangir -emit-llvm %s -o - | FileCheck %s -check-prefix=LLVM
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -O2 -emit-llvm %s -o - | FileCheck %s -check-prefix=OGCG
+
+typedef float __m128 __attribute__((__vector_size__(16), __aligned__(16)));
+typedef double __m128d __attribute__((__vector_size__(16), __aligned__(16)));
+
+__m128 test_cmpnleps(__m128 A, __m128 B) {
+  // CIR-LABEL: @test_cmpnleps
+  // CIR: [[CMP:%.*]] = cir.vec.cmp(le, [[A:%.*]], [[B:%.*]]) : !cir.vector<!cir.float x 4>, !cir.vector<!s32i x 4>
+  // CIR: [[NOTCMP:%.*]] = cir.unary(not, [[CMP]]) : !cir.vector<!s32i x 4>, !cir.vector<!s32i x 4>
+  // CIR-NEXT: [[CAST:%.*]] = cir.cast bitcast [[NOTCMP:%.*]] : !cir.vector<!s32i x 4> -> !cir.vector<!cir.float x 4>
+  // CIR-NEXT: cir.store [[CAST]], [[ALLOCA:%.*]] :  !cir.vector<!cir.float x 4>, !cir.ptr<!cir.vector<!cir.float x 4>>
+  // CIR-NEXT: [[LD:%.*]] = cir.load [[ALLOCA]] :
+  // CIR-NEXT: cir.return [[LD]] : !cir.vector<!cir.float x 4>
+
+  // LLVM-LABEL: test_cmpnleps
+  // LLVM: [[CMP:%.*]] = fcmp ugt <4 x float> {{.*}}, {{.*}}
+  // LLVM-NEXT: [[SEXT:%.*]] = sext <4 x i1> [[CMP]] to <4 x i32>
+  // LLVM-NEXT: [[CAST:%.*]] = bitcast <4 x i32> [[SEXT]] to <4 x float>
+  // LLVM-NEXT: ret <4 x float> [[CAST]]
+
+  // OGCG-LABEL: test_cmpnleps
+  // OGCG: [[CMP:%.*]] = fcmp ugt <4 x float> {{.*}}, {{.*}}
+  // OGCG-NEXT: [[SEXT:%.*]] = sext <4 x i1> [[CMP]] to <4 x i32>
+  // OGCG-NEXT: [[CAST:%.*]] = bitcast <4 x i32> [[SEXT]] to <4 x float>
+  // OGCG-NEXT: ret <4 x float> [[CAST]]
+  return __builtin_ia32_cmpnleps(A, B); //done!
----------------
AmrDeveloper wrote:

```suggestion
  return __builtin_ia32_cmpnleps(A, B);
```

https://github.com/llvm/llvm-project/pull/167125