[AArch64 NEON] Fix a bug about vcles_f32 and vcled_f64.

As vcles_f32() and vcled_f64 are implemented by FCMGE, operands
should make a swap.

llvm-svn: 199866

2 files changed, 29 insertions, 28 deletions

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 267dea069e7..103fe3f5406 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -2305,6 +2305,7 @@ static Value *EmitAArch64ScalarBuiltinExpr(CodeGenFunction &CGF,
   case AArch64::BI__builtin_neon_vcles_f32:
   case AArch64::BI__builtin_neon_vcled_f64:
     Int = Intrinsic::aarch64_neon_fcge; s = "vcge";
+    std::swap(Ops[0], Ops[1]);
     IntTypes = VectorRet | ScalarArg0 | ScalarArg1; break;
   // Scalar Floating-point Compare Less Than Or Equal To Zero
   case AArch64::BI__builtin_neon_vclezs_f32:
diff --git a/clang/test/CodeGen/aarch64-neon-intrinsics.c b/clang/test/CodeGen/aarch64-neon-intrinsics.c
index 35dd06ecc76..7c15e88ae09 100644
--- a/clang/test/CodeGen/aarch64-neon-intrinsics.c
+++ b/clang/test/CodeGen/aarch64-neon-intrinsics.c
@@ -8598,169 +8598,169 @@ int32_t test_vqmovnd_u64(int64_t a) {
 
 uint32_t test_vceqs_f32(float32_t a, float32_t b) {
 // CHECK: test_vceqs_f32
-// CHECK: fcmeq {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: fcmeq s0, s0, s1
   return (uint32_t)vceqs_f32(a, b);
 }
 
 uint64_t test_vceqd_f64(float64_t a, float64_t b) {
 // CHECK: test_vceqd_f64
-// CHECK: fcmeq {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: fcmeq d0, d0, d1
   return (uint64_t)vceqd_f64(a, b);
 }
 
 uint32_t test_vceqzs_f32(float32_t a) {
 // CHECK: test_vceqzs_f32
-// CHECK: fcmeq {{s[0-9]+}}, {{s[0-9]+}}, #0.0
+// CHECK: fcmeq s0, s0, #0.0
   return (uint32_t)vceqzs_f32(a);
 }
 
 uint64_t test_vceqzd_f64(float64_t a) {
 // CHECK: test_vceqzd_f64
-// CHECK: fcmeq {{d[0-9]+}}, {{d[0-9]+}}, #0.0
+// CHECK: fcmeq d0, d0, #0.0
   return (uint64_t)vceqzd_f64(a);
 }
 
 uint32_t test_vcges_f32(float32_t a, float32_t b) {
 // CHECK: test_vcges_f32
-// CHECK: fcmge {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: fcmge s0, s0, s1
   return (uint32_t)vcges_f32(a, b);
 }
 
 uint64_t test_vcged_f64(float64_t a, float64_t b) {
 // CHECK: test_vcged_f64
-// CHECK: fcmge {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: fcmge d0, d0, d1
   return (uint64_t)vcged_f64(a, b);
 }
 
 uint32_t test_vcgezs_f32(float32_t a) {
 // CHECK: test_vcgezs_f32
-// CHECK: fcmge {{s[0-9]+}}, {{s[0-9]+}}, #0.0
+// CHECK: fcmge s0, s0, #0.0
   return (uint32_t)vcgezs_f32(a);
 }
 
 uint64_t test_vcgezd_f64(float64_t a) {
 // CHECK: test_vcgezd_f64
-// CHECK: fcmge {{d[0-9]+}}, {{d[0-9]+}}, #0.0
+// CHECK: fcmge d0, d0, #0.0
   return (uint64_t)vcgezd_f64(a);
 }
 
 uint32_t test_vcgts_f32(float32_t a, float32_t b) {
 // CHECK: test_vcgts_f32
-// CHECK: fcmgt {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: fcmgt s0, s0, s1
   return (uint32_t)vcgts_f32(a, b);
 }
 
 uint64_t test_vcgtd_f64(float64_t a, float64_t b) {
 // CHECK: test_vcgtd_f64
-// CHECK: fcmgt {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: fcmgt d0, d0, d1
   return (uint64_t)vcgtd_f64(a, b);
 }
 
 uint32_t test_vcgtzs_f32(float32_t a) {
 // CHECK: test_vcgtzs_f32
-// CHECK: fcmgt {{s[0-9]+}}, {{s[0-9]+}}, #0.0
+// CHECK: fcmgt s0, s0, #0.0
   return (uint32_t)vcgtzs_f32(a);
 }
 
 uint64_t test_vcgtzd_f64(float64_t a) {
 // CHECK: test_vcgtzd_f64
-// CHECK: fcmgt {{d[0-9]+}}, {{d[0-9]+}}, #0.0
+// CHECK: fcmgt d0, d0, #0.0
   return (uint64_t)vcgtzd_f64(a);
 }
 
 uint32_t test_vcles_f32(float32_t a, float32_t b) {
 // CHECK: test_vcles_f32
-// CHECK: fcmge {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: fcmge s0, s1, s0
   return (uint32_t)vcles_f32(a, b);
 }
 
 uint64_t test_vcled_f64(float64_t a, float64_t b) {
 // CHECK: test_vcled_f64
-// CHECK: fcmge {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: fcmge d0, d1, d0
   return (uint64_t)vcled_f64(a, b);
 }
 
 uint32_t test_vclezs_f32(float32_t a) {
 // CHECK: test_vclezs_f32
-// CHECK: fcmle {{s[0-9]+}}, {{s[0-9]+}}, #0.0
+// CHECK: fcmle s0, s0, #0.0
   return (uint32_t)vclezs_f32(a);
 }
 
 uint64_t test_vclezd_f64(float64_t a) {
 // CHECK: test_vclezd_f64
-// CHECK: fcmle {{d[0-9]+}}, {{d[0-9]+}}, #0.0
+// CHECK: fcmle d0, d0, #0.0
   return (uint64_t)vclezd_f64(a);
 }
 
 uint32_t test_vclts_f32(float32_t a, float32_t b) {
 // CHECK: test_vclts_f32
-// CHECK: fcmgt {{s[0-9]+}}, s1, s0
+// CHECK: fcmgt s0, s1, s0
   return (uint32_t)vclts_f32(a, b);
 }
 
 uint64_t test_vcltd_f64(float64_t a, float64_t b) {
 // CHECK: test_vcltd_f64
-// CHECK: fcmgt {{d[0-9]+}}, d1, d0
+// CHECK: fcmgt d0, d1, d0
   return (uint64_t)vcltd_f64(a, b);
 }
 
 uint32_t test_vcltzs_f32(float32_t a) {
 // CHECK: test_vcltzs_f32
-// CHECK: fcmlt {{s[0-9]+}}, {{s[0-9]+}}, #0.0
+// CHECK: fcmlt s0, s0, #0.0
   return (uint32_t)vcltzs_f32(a);
 }
 
 uint64_t test_vcltzd_f64(float64_t a) {
 // CHECK: test_vcltzd_f64
-// CHECK: fcmlt {{d[0-9]+}}, {{d[0-9]+}}, #0.0
+// CHECK: fcmlt d0, d0, #0.0
   return (uint64_t)vcltzd_f64(a);
 }
 
 uint32_t test_vcages_f32(float32_t a, float32_t b) {
 // CHECK: test_vcages_f32
-// CHECK: facge {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: facge s0, s0, s1
   return (uint32_t)vcages_f32(a, b);
 }
 
 uint64_t test_vcaged_f64(float64_t a, float64_t b) {
 // CHECK: test_vcaged_f64
-// CHECK: facge {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: facge d0, d0, d1
   return (uint64_t)vcaged_f64(a, b);
 }
 
 uint32_t test_vcagts_f32(float32_t a, float32_t b) {
 // CHECK: test_vcagts_f32
-// CHECK: facgt {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: facgt s0, s0, s1
   return (uint32_t)vcagts_f32(a, b);
 }
 
 uint64_t test_vcagtd_f64(float64_t a, float64_t b) {
 // CHECK: test_vcagtd_f64
-// CHECK: facgt {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: facgt d0, d0, d1
   return (uint64_t)vcagtd_f64(a, b);
 }
 
 uint32_t test_vcales_f32(float32_t a, float32_t b) {
 // CHECK: test_vcales_f32
-// CHECK: facge {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: facge s0, s1, s0
   return (uint32_t)vcales_f32(a, b);
 }
 
 uint64_t test_vcaled_f64(float64_t a, float64_t b) {
 // CHECK: test_vcaled_f64
-// CHECK: facge {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: facge d0, d1, d0
   return (uint64_t)vcaled_f64(a, b);
 }
 
 uint32_t test_vcalts_f32(float32_t a, float32_t b) {
 // CHECK: test_vcalts_f32
-// CHECK: facgt {{s[0-9]+}}, {{s[0-9]+}}, {{s[0-9]+}}
+// CHECK: facgt s0, s1, s0
   return (uint32_t)vcalts_f32(a, b);
 }
 
 uint64_t test_vcaltd_f64(float64_t a, float64_t b) {
 // CHECK: test_vcaltd_f64
-// CHECK: facgt {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK: facgt d0, d1, d0
   return (uint64_t)vcaltd_f64(a, b);
 }