[X86] For lzcnt/tzcnt intrinsics use cttz/ctlz intrinsics with zero_undef flag set to false.

Previously we used a select and the zero_undef=true intrinsic. In -O2 this pattern will get optimized to zero_undef=false. But in -O0 this optimization won't happen. This results in a compare and cmov being wrapped around a tzcnt/lzcnt instruction.

By using the zero_undef=false intrinsic directly without the select, we can improve the -O0 codegen to just an lzcnt/tzcnt instruction.

Differential Revision: https://reviews.llvm.org/D52392

llvm-svn: 343126

7 files changed, 43 insertions, 33 deletions

diff --git a/clang/include/clang/Basic/BuiltinsX86.def b/clang/include/clang/Basic/BuiltinsX86.def
index db81f58e082..45d366341c0 100644
--- a/clang/include/clang/Basic/BuiltinsX86.def
+++ b/clang/include/clang/Basic/BuiltinsX86.def
@@ -727,8 +727,14 @@ TARGET_BUILTIN(__builtin_ia32_subborrow_u32, "UcUcUiUiUi*", "n", "")
 TARGET_BUILTIN(__builtin_ia32_rdseed16_step, "UiUs*", "n", "rdseed")
 TARGET_BUILTIN(__builtin_ia32_rdseed32_step, "UiUi*", "n", "rdseed")
 
+// LZCNT
+TARGET_BUILTIN(__builtin_ia32_lzcnt_u16, "UsUs", "nc", "lzcnt")
+TARGET_BUILTIN(__builtin_ia32_lzcnt_u32, "UiUi", "nc", "lzcnt")
+
 // BMI
 TARGET_BUILTIN(__builtin_ia32_bextr_u32, "UiUiUi", "nc", "bmi")
+TARGET_BUILTIN(__builtin_ia32_tzcnt_u16, "UsUs", "nc", "")
+TARGET_BUILTIN(__builtin_ia32_tzcnt_u32, "UiUi", "nc", "")
 
 // BMI2
 TARGET_BUILTIN(__builtin_ia32_bzhi_si, "UiUiUi", "nc", "bmi2")
diff --git a/clang/include/clang/Basic/BuiltinsX86_64.def b/clang/include/clang/Basic/BuiltinsX86_64.def
index a9aaadca848..b59f82d8c9e 100644
--- a/clang/include/clang/Basic/BuiltinsX86_64.def
+++ b/clang/include/clang/Basic/BuiltinsX86_64.def
@@ -81,7 +81,9 @@ TARGET_BUILTIN(__builtin_ia32_addcarry_u64, "UcUcULLiULLiULLi*", "n", "")
 TARGET_BUILTIN(__builtin_ia32_subborrow_u64, "UcUcULLiULLiULLi*", "n", "")
 TARGET_BUILTIN(__builtin_ia32_rdrand64_step, "UiULLi*", "n", "rdrnd")
 TARGET_BUILTIN(__builtin_ia32_rdseed64_step, "UiULLi*", "n", "rdseed")
+TARGET_BUILTIN(__builtin_ia32_lzcnt_u64, "ULLiULLi", "nc", "lzcnt")
 TARGET_BUILTIN(__builtin_ia32_bextr_u64, "ULLiULLiULLi", "nc", "bmi")
+TARGET_BUILTIN(__builtin_ia32_tzcnt_u64, "ULLiULLi", "nc", "")
 TARGET_BUILTIN(__builtin_ia32_bzhi_di, "ULLiULLiULLi", "nc", "bmi2")
 TARGET_BUILTIN(__builtin_ia32_pdep_di, "ULLiULLiULLi", "nc", "bmi2")
 TARGET_BUILTIN(__builtin_ia32_pext_di, "ULLiULLiULLi", "nc", "bmi2")
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 23c760207b8..df71dbb4b4a 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -9164,6 +9164,18 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
                                       Ops[0]);
     return Builder.CreateExtractValue(Call, 0);
   }
+  case X86::BI__builtin_ia32_lzcnt_u16:
+  case X86::BI__builtin_ia32_lzcnt_u32:
+  case X86::BI__builtin_ia32_lzcnt_u64: {
+    Value *F = CGM.getIntrinsic(Intrinsic::ctlz, Ops[0]->getType());
+    return Builder.CreateCall(F, {Ops[0], Builder.getInt1(false)});
+  }
+  case X86::BI__builtin_ia32_tzcnt_u16:
+  case X86::BI__builtin_ia32_tzcnt_u32:
+  case X86::BI__builtin_ia32_tzcnt_u64: {
+    Value *F = CGM.getIntrinsic(Intrinsic::cttz, Ops[0]->getType());
+    return Builder.CreateCall(F, {Ops[0], Builder.getInt1(false)});
+  }
   case X86::BI__builtin_ia32_undef128:
   case X86::BI__builtin_ia32_undef256:
   case X86::BI__builtin_ia32_undef512:
diff --git a/clang/lib/Headers/bmiintrin.h b/clang/lib/Headers/bmiintrin.h
index d03bef442a2..56c20b78d34 100644
--- a/clang/lib/Headers/bmiintrin.h
+++ b/clang/lib/Headers/bmiintrin.h
@@ -62,7 +62,7 @@
 static __inline__ unsigned short __RELAXED_FN_ATTRS
 __tzcnt_u16(unsigned short __X)
 {
-  return __X ? __builtin_ctzs(__X) : 16;
+  return __builtin_ia32_tzcnt_u16(__X);
 }
 
 /// Performs a bitwise AND of the second operand with the one's
@@ -196,7 +196,7 @@ __blsr_u32(unsigned int __X)
 static __inline__ unsigned int __RELAXED_FN_ATTRS
 __tzcnt_u32(unsigned int __X)
 {
-  return __X ? __builtin_ctz(__X) : 32;
+  return __builtin_ia32_tzcnt_u32(__X);
 }
 
 /// Counts the number of trailing zero bits in the operand.
@@ -212,7 +212,7 @@ __tzcnt_u32(unsigned int __X)
 static __inline__ int __RELAXED_FN_ATTRS
 _mm_tzcnt_32(unsigned int __X)
 {
-  return __X ? __builtin_ctz(__X) : 32;
+  return __builtin_ia32_tzcnt_u32(__X);
 }
 
 #ifdef __x86_64__
@@ -359,7 +359,7 @@ __blsr_u64(unsigned long long __X)
 static __inline__ unsigned long long __RELAXED_FN_ATTRS
 __tzcnt_u64(unsigned long long __X)
 {
-  return __X ? __builtin_ctzll(__X) : 64;
+  return __builtin_ia32_tzcnt_u64(__X);
 }
 
 /// Counts the number of trailing zero bits in the operand.
@@ -375,7 +375,7 @@ __tzcnt_u64(unsigned long long __X)
 static __inline__ long long __RELAXED_FN_ATTRS
 _mm_tzcnt_64(unsigned long long __X)
 {
-  return __X ? __builtin_ctzll(__X) : 64;
+  return __builtin_ia32_tzcnt_u64(__X);
 }
 
 #endif /* __x86_64__ */
diff --git a/clang/lib/Headers/lzcntintrin.h b/clang/lib/Headers/lzcntintrin.h
index 558f1828f0e..f2674d2a5ab 100644
--- a/clang/lib/Headers/lzcntintrin.h
+++ b/clang/lib/Headers/lzcntintrin.h
@@ -44,7 +44,7 @@
 static __inline__ unsigned short __DEFAULT_FN_ATTRS
 __lzcnt16(unsigned short __X)
 {
-  return __X ? __builtin_clzs(__X) : 16;
+  return __builtin_ia32_lzcnt_u16(__X);
 }
 
 /// Counts the number of leading zero bits in the operand.
@@ -61,7 +61,7 @@ __lzcnt16(unsigned short __X)
 static __inline__ unsigned int __DEFAULT_FN_ATTRS
 __lzcnt32(unsigned int __X)
 {
-  return __X ? __builtin_clz(__X) : 32;
+  return __builtin_ia32_lzcnt_u32(__X);
 }
 
 /// Counts the number of leading zero bits in the operand.
@@ -78,7 +78,7 @@ __lzcnt32(unsigned int __X)
 static __inline__ unsigned int __DEFAULT_FN_ATTRS
 _lzcnt_u32(unsigned int __X)
 {
-  return __X ? __builtin_clz(__X) : 32;
+  return __builtin_ia32_lzcnt_u32(__X);
 }
 
 #ifdef __x86_64__
@@ -96,7 +96,7 @@ _lzcnt_u32(unsigned int __X)
 static __inline__ unsigned long long __DEFAULT_FN_ATTRS
 __lzcnt64(unsigned long long __X)
 {
-  return __X ? __builtin_clzll(__X) : 64;
+  return __builtin_ia32_lzcnt_u64(__X);
 }
 
 /// Counts the number of leading zero bits in the operand.
@@ -113,7 +113,7 @@ __lzcnt64(unsigned long long __X)
 static __inline__ unsigned long long __DEFAULT_FN_ATTRS
 _lzcnt_u64(unsigned long long __X)
 {
-  return __X ? __builtin_clzll(__X) : 64;
+  return __builtin_ia32_lzcnt_u64(__X);
 }
 #endif
 
diff --git a/clang/test/CodeGen/bmi-builtins.c b/clang/test/CodeGen/bmi-builtins.c
index 91271f0b359..69eb2bdb807 100644
--- a/clang/test/CodeGen/bmi-builtins.c
+++ b/clang/test/CodeGen/bmi-builtins.c
@@ -15,9 +15,7 @@
 
 unsigned short test__tzcnt_u16(unsigned short __X) {
   // CHECK-LABEL: test__tzcnt_u16
-  // CHECK: zext i16 %{{.*}} to i32
-  // CHECK: icmp ne i32 %{{.*}}, 0
-  // CHECK: i16 @llvm.cttz.i16(i16 %{{.*}}, i1 true)
+  // CHECK: i16 @llvm.cttz.i16(i16 %{{.*}}, i1 false)
   return __tzcnt_u16(__X);
 }
 
@@ -57,15 +55,13 @@ unsigned int test__blsr_u32(unsigned int __X) {
 
 unsigned int test__tzcnt_u32(unsigned int __X) {
   // CHECK-LABEL: test__tzcnt_u32
-  // CHECK: icmp ne i32 %{{.*}}, 0
-  // CHECK: i32 @llvm.cttz.i32(i32 %{{.*}}, i1 true)
+  // CHECK: i32 @llvm.cttz.i32(i32 %{{.*}}, i1 false)
   return __tzcnt_u32(__X);
 }
 
 int test_mm_tzcnt_32(unsigned int __X) {
   // CHECK-LABEL: test_mm_tzcnt_32
-  // CHECK: icmp ne i32 %{{.*}}, 0
-  // CHECK: i32 @llvm.cttz.i32(i32 %{{.*}}, i1 true)
+  // CHECK: i32 @llvm.cttz.i32(i32 %{{.*}}, i1 false)
   return _mm_tzcnt_32(__X);
 }
 
@@ -105,15 +101,13 @@ unsigned long long test__blsr_u64(unsigned long long __X) {
 
 unsigned long long test__tzcnt_u64(unsigned long long __X) {
   // CHECK-LABEL: test__tzcnt_u64
-  // CHECK: icmp ne i64 %{{.*}}, 0
-  // CHECK: i64 @llvm.cttz.i64(i64 %{{.*}}, i1 true)
+  // CHECK: i64 @llvm.cttz.i64(i64 %{{.*}}, i1 false)
   return __tzcnt_u64(__X);
 }
 
 long long test_mm_tzcnt_64(unsigned long long __X) {
   // CHECK-LABEL: test_mm_tzcnt_64
-  // CHECK: icmp ne i64 %{{.*}}, 0
-  // CHECK: i64 @llvm.cttz.i64(i64 %{{.*}}, i1 true)
+  // CHECK: i64 @llvm.cttz.i64(i64 %{{.*}}, i1 false)
   return _mm_tzcnt_64(__X);
 }
 
@@ -121,9 +115,7 @@ long long test_mm_tzcnt_64(unsigned long long __X) {
 
 unsigned short test_tzcnt_u16(unsigned short __X) {
   // CHECK-LABEL: test_tzcnt_u16
-  // CHECK: zext i16 %{{.*}} to i32
-  // CHECK: icmp ne i32 %{{.*}}, 0
-  // CHECK: i16 @llvm.cttz.i16(i16 %{{.*}}, i1 true)
+  // CHECK: i16 @llvm.cttz.i16(i16 %{{.*}}, i1 false)
   return _tzcnt_u16(__X);
 }
 
@@ -168,8 +160,7 @@ unsigned int test_blsr_u32(unsigned int __X) {
 
 unsigned int test_tzcnt_u32(unsigned int __X) {
   // CHECK-LABEL: test_tzcnt_u32
-  // CHECK: icmp ne i32 %{{.*}}, 0
-  // CHECK: i32 @llvm.cttz.i32(i32 %{{.*}}, i1 true)
+  // CHECK: i32 @llvm.cttz.i32(i32 %{{.*}}, i1 false)
   return _tzcnt_u32(__X);
 }
 
@@ -215,7 +206,6 @@ unsigned long long test_blsr_u64(unsigned long long __X) {
 
 unsigned long long test_tzcnt_u64(unsigned long long __X) {
   // CHECK-LABEL: test_tzcnt_u64
-  // CHECK: icmp ne i64 %{{.*}}, 0
-  // CHECK: i64 @llvm.cttz.i64(i64 %{{.*}}, i1 true)
+  // CHECK: i64 @llvm.cttz.i64(i64 %{{.*}}, i1 false)
   return _tzcnt_u64(__X);
 }
diff --git a/clang/test/CodeGen/lzcnt-builtins.c b/clang/test/CodeGen/lzcnt-builtins.c
index b1414a3b56f..9255207ffae 100644
--- a/clang/test/CodeGen/lzcnt-builtins.c
+++ b/clang/test/CodeGen/lzcnt-builtins.c
@@ -5,30 +5,30 @@
 
 unsigned short test__lzcnt16(unsigned short __X)
 {
-  // CHECK: @llvm.ctlz.i16
+  // CHECK: @llvm.ctlz.i16(i16 %{{.*}}, i1 false)
   return __lzcnt16(__X);
 }
 
 unsigned int test_lzcnt32(unsigned int __X)
 {
-  // CHECK: @llvm.ctlz.i32
+  // CHECK: @llvm.ctlz.i32(i32 %{{.*}}, i1 false)
   return __lzcnt32(__X);
 }
 
 unsigned long long test__lzcnt64(unsigned long long __X)
 {
-  // CHECK: @llvm.ctlz.i64
+  // CHECK: @llvm.ctlz.i64(i64 %{{.*}}, i1 false)
   return __lzcnt64(__X);
 }
 
 unsigned int test_lzcnt_u32(unsigned int __X)
 {
-  // CHECK: @llvm.ctlz.i32
+  // CHECK: @llvm.ctlz.i32(i32 %{{.*}}, i1 false)
   return _lzcnt_u32(__X);
 }
 
 unsigned long long test__lzcnt_u64(unsigned long long __X)
 {
-  // CHECK: @llvm.ctlz.i64
+  // CHECK: @llvm.ctlz.i64(i64 %{{.*}}, i1 false)
   return _lzcnt_u64(__X);
 }