[clang] Emit constraint intrinsics for arc and hyperbolic trig clang builtins #98949

farzonl · 2024-07-15T19:48:24Z

Change(s)

Builtins.td - Add f16 support for libm arc and hyperbolic trig functions
CGBuiltin.cpp - Emit constraint intrinsics for trig clang builtins

History

This change is part of an implementation of #87367 investigation on supporting IEEE math operations as intrinsics. Which was discussed in this RFC:
https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294

This change adds wasm lowering cases for acos, asin, atan, cosh, sinh, and tanh.

#70079
#70080
#70081
#70083
#70084
#95966

Precursor PR(s)

Note this PR needs Merge after:

…builtins - `Builtins.td` - Add f16 support for libm arc and hyperbolic trig functions - `CGBuiltin.cpp` - Emit constraint intrinsics for trig clang builtins This change is part of an implementation of llvm#87367 investigation on supporting IEEE math operations as intrinsics. Which was discussed in this RFC: https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294 This change adds wasm lowering cases for `acos`, `asin`, `atan`, `cosh`, `sinh`, and `tanh`. llvm#70079 llvm#70080 llvm#70081 llvm#70083 llvm#70084 llvm#95966 Note this PR needs Merge after: - llvm#98937 - llvm#98755

llvmbot · 2024-07-15T19:48:54Z

@llvm/pr-subscribers-backend-x86

@llvm/pr-subscribers-clang

Author: Farzon Lotfi (farzonl)

Changes

Change(s)

Builtins.td - Add f16 support for libm arc and hyperbolic trig functions
CGBuiltin.cpp - Emit constraint intrinsics for trig clang builtins

History

This change is part of an implementation of #87367 investigation on supporting IEEE math operations as intrinsics. Which was discussed in this RFC:
https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294

This change adds wasm lowering cases for acos, asin, atan, cosh, sinh, and tanh.

#70079
#70080
#70081
#70083
#70084
#95966

Precursor PR(s)

Note this PR needs Merge after:

#98937
#98755

Patch is 28.53 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/98949.diff

6 Files Affected:

(modified) clang/include/clang/Basic/Builtins.td (+18-18)
(modified) clang/lib/CodeGen/CGBuiltin.cpp (+66)
(modified) clang/test/CodeGen/X86/math-builtins.c (+24-24)
(modified) clang/test/CodeGen/constrained-math-builtins.c (+41)
(modified) clang/test/CodeGen/math-libcalls.c (+37-36)
(modified) clang/test/CodeGenOpenCL/builtins-f16.cl (+18)

diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index f5b15cf90d1f8..4133f6ff40cf3 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -83,11 +83,11 @@ class BitInt_Long_LongLongTemplate :
 // - _Constant: Argument has to constant-fold to an integer constant expression
 
 // __fp16 and __float128 builtin variants of libc/libm functions.
-def AcosF128 : Builtin {
-  let Spellings = ["__builtin_acosf128"];
+def AcosF16F128 : Builtin, F16F128MathTemplate {
+  let Spellings = ["__builtin_acos"];
   let Attributes = [FunctionWithBuiltinPrefix, NoThrow,
                     ConstIgnoringErrnoAndExceptions];
-  let Prototype = "__float128(__float128)";
+  let Prototype = "T(T)";
 }
 
 def AcoshF128 : Builtin {
@@ -97,11 +97,11 @@ def AcoshF128 : Builtin {
   let Prototype = "__float128(__float128)";
 }
 
-def AsinF128 : Builtin {
-  let Spellings = ["__builtin_asinf128"];
+def AsinF16F128 : Builtin, F16F128MathTemplate {
+  let Spellings = ["__builtin_asin"];
   let Attributes = [FunctionWithBuiltinPrefix, NoThrow,
                     ConstIgnoringErrnoAndExceptions];
-  let Prototype = "__float128(__float128)";
+  let Prototype = "T(T)";
 }
 
 def AsinhF128 : Builtin {
@@ -111,11 +111,11 @@ def AsinhF128 : Builtin {
   let Prototype = "__float128(__float128)";
 }
 
-def AtanF128 : Builtin {
-  let Spellings = ["__builtin_atanf128"];
+def AtanF16F128 : Builtin, F16F128MathTemplate {
+  let Spellings = ["__builtin_atan"];
   let Attributes = [FunctionWithBuiltinPrefix, NoThrow,
                     ConstIgnoringErrnoAndExceptions];
-  let Prototype = "__float128(__float128)";
+  let Prototype = "T(T)";
 }
 
 def AtanhF128 : Builtin {
@@ -143,10 +143,10 @@ def CosF16F128 : Builtin, F16F128MathTemplate {
   let Prototype = "T(T)";
 }
 
-def CoshF128 : Builtin {
-  let Spellings = ["__builtin_coshf128"];
+def CoshF16F128 : Builtin, F16F128MathTemplate {
+  let Spellings = ["__builtin_cosh"];
   let Attributes = [FunctionWithBuiltinPrefix, NoThrow, ConstIgnoringErrnoAndExceptions];
-  let Prototype = "__float128(__float128)";
+  let Prototype = "T(T)";
 }
 
 def ErfF128 : Builtin {
@@ -468,11 +468,11 @@ def SinF16F128 : Builtin, F16F128MathTemplate {
   let Prototype = "T(T)";
 }
 
-def SinhF128 : Builtin {
-  let Spellings = ["__builtin_sinhf128"];
+def SinhF16F128 : Builtin, F16F128MathTemplate {
+  let Spellings = ["__builtin_sinh"];
   let Attributes = [FunctionWithBuiltinPrefix, NoThrow,
                     ConstIgnoringErrnoAndExceptions];
-  let Prototype = "__float128(__float128)";
+  let Prototype = "T(T)";
 }
 
 def SqrtF16F128 : Builtin, F16F128MathTemplate {
@@ -489,11 +489,11 @@ def TanF16F128 : Builtin, F16F128MathTemplate {
   let Prototype = "T(T)";
 }
 
-def TanhF128 : Builtin {
-  let Spellings = ["__builtin_tanhf128"];
+def TanhF16F128 : Builtin, F16F128MathTemplate {
+  let Spellings = ["__builtin_tanh"];
   let Attributes = [FunctionWithBuiltinPrefix, NoThrow,
                     ConstIgnoringErrnoAndExceptions];
-  let Prototype = "__float128(__float128)";
+  let Prototype = "T(T)";
 }
 
 def TgammaF128 : Builtin {
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index a54fa7bf87aad..d78efb2217931 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -2640,6 +2640,39 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
   }
   if (GenerateIntrinsics) {
     switch (BuiltinIDIfNoAsmLabel) {
+    case Builtin::BIacos:
+    case Builtin::BIacosf:
+    case Builtin::BIacosl:
+    case Builtin::BI__builtin_acos:
+    case Builtin::BI__builtin_acosf:
+    case Builtin::BI__builtin_acosf16:
+    case Builtin::BI__builtin_acosl:
+    case Builtin::BI__builtin_acosf128:
+      return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
+          *this, E, Intrinsic::acos, Intrinsic::experimental_constrained_acos));
+
+    case Builtin::BIasin:
+    case Builtin::BIasinf:
+    case Builtin::BIasinl:
+    case Builtin::BI__builtin_asin:
+    case Builtin::BI__builtin_asinf:
+    case Builtin::BI__builtin_asinf16:
+    case Builtin::BI__builtin_asinl:
+    case Builtin::BI__builtin_asinf128:
+      return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
+          *this, E, Intrinsic::asin, Intrinsic::experimental_constrained_asin));
+
+    case Builtin::BIatan:
+    case Builtin::BIatanf:
+    case Builtin::BIatanl:
+    case Builtin::BI__builtin_atan:
+    case Builtin::BI__builtin_atanf:
+    case Builtin::BI__builtin_atanf16:
+    case Builtin::BI__builtin_atanl:
+    case Builtin::BI__builtin_atanf128:
+      return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
+          *this, E, Intrinsic::atan, Intrinsic::experimental_constrained_atan));
+
     case Builtin::BIceil:
     case Builtin::BIceilf:
     case Builtin::BIceill:
@@ -2675,6 +2708,17 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
                                    Intrinsic::cos,
                                    Intrinsic::experimental_constrained_cos));
 
+    case Builtin::BIcosh:
+    case Builtin::BIcoshf:
+    case Builtin::BIcoshl:
+    case Builtin::BI__builtin_cosh:
+    case Builtin::BI__builtin_coshf:
+    case Builtin::BI__builtin_coshf16:
+    case Builtin::BI__builtin_coshl:
+    case Builtin::BI__builtin_coshf128:
+      return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
+          *this, E, Intrinsic::cosh, Intrinsic::experimental_constrained_cosh));
+
     case Builtin::BIexp:
     case Builtin::BIexpf:
     case Builtin::BIexpl:
@@ -2891,6 +2935,17 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
                                    Intrinsic::sin,
                                    Intrinsic::experimental_constrained_sin));
 
+    case Builtin::BIsinh:
+    case Builtin::BIsinhf:
+    case Builtin::BIsinhl:
+    case Builtin::BI__builtin_sinh:
+    case Builtin::BI__builtin_sinhf:
+    case Builtin::BI__builtin_sinhf16:
+    case Builtin::BI__builtin_sinhl:
+    case Builtin::BI__builtin_sinhf128:
+      return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
+          *this, E, Intrinsic::sinh, Intrinsic::experimental_constrained_sinh));
+
     case Builtin::BIsqrt:
     case Builtin::BIsqrtf:
     case Builtin::BIsqrtl:
@@ -2917,6 +2972,17 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
       return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
           *this, E, Intrinsic::tan, Intrinsic::experimental_constrained_tan));
 
+    case Builtin::BItanh:
+    case Builtin::BItanhf:
+    case Builtin::BItanhl:
+    case Builtin::BI__builtin_tanh:
+    case Builtin::BI__builtin_tanhf:
+    case Builtin::BI__builtin_tanhf16:
+    case Builtin::BI__builtin_tanhl:
+    case Builtin::BI__builtin_tanhf128:
+      return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
+          *this, E, Intrinsic::tanh, Intrinsic::experimental_constrained_tanh));
+
     case Builtin::BItrunc:
     case Builtin::BItruncf:
     case Builtin::BItruncl:
diff --git a/clang/test/CodeGen/X86/math-builtins.c b/clang/test/CodeGen/X86/math-builtins.c
index d26db19574051..48465df21cca1 100644
--- a/clang/test/CodeGen/X86/math-builtins.c
+++ b/clang/test/CodeGen/X86/math-builtins.c
@@ -168,10 +168,10 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c) {
   /* math */
   __builtin_acos(f);       __builtin_acosf(f);      __builtin_acosl(f); __builtin_acosf128(f);
 
-// NO__ERRNO: declare double @acos(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @acosf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @acosl(x86_fp80 noundef) [[READNONE]]
-// NO__ERRNO: declare fp128 @acosf128(fp128 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.acos.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.acos.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.acos.f80(x86_fp80) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare fp128 @llvm.acos.f128(fp128) [[READNONE_INTRINSIC]]
 // HAS_ERRNO: declare double @acos(double noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare float @acosf(float noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare x86_fp80 @acosl(x86_fp80 noundef) [[NOT_READNONE]]
@@ -190,10 +190,10 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c) {
 
   __builtin_asin(f);       __builtin_asinf(f);      __builtin_asinl(f); __builtin_asinf128(f);
 
-// NO__ERRNO: declare double @asin(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @asinf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @asinl(x86_fp80 noundef) [[READNONE]]
-// NO__ERRNO: declare fp128 @asinf128(fp128 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.asin.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.asin.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.asin.f80(x86_fp80) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare fp128 @llvm.asin.f128(fp128) [[READNONE_INTRINSIC]]
 // HAS_ERRNO: declare double @asin(double noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare float @asinf(float noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare x86_fp80 @asinl(x86_fp80 noundef) [[NOT_READNONE]]
@@ -212,10 +212,10 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c) {
 
   __builtin_atan(f);       __builtin_atanf(f);      __builtin_atanl(f); __builtin_atanf128(f);
 
-// NO__ERRNO: declare double @atan(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @atanf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @atanl(x86_fp80 noundef) [[READNONE]]
-// NO__ERRNO: declare fp128 @atanf128(fp128 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.atan.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.atan.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.atan.f80(x86_fp80) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare fp128 @llvm.atan.f128(fp128) [[READNONE_INTRINSIC]]
 // HAS_ERRNO: declare double @atan(double noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare float @atanf(float noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare x86_fp80 @atanl(x86_fp80 noundef) [[NOT_READNONE]]
@@ -267,10 +267,10 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c) {
 
   __builtin_cosh(f);       __builtin_coshf(f);      __builtin_coshl(f); __builtin_coshf128(f);
 
-// NO__ERRNO: declare double @cosh(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @coshf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @coshl(x86_fp80 noundef) [[READNONE]]
-// NO__ERRNO: declare fp128 @coshf128(fp128 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.cosh.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.cosh.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.cosh.f80(x86_fp80) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare fp128 @llvm.cosh.f128(fp128) [[READNONE_INTRINSIC]]
 // HAS_ERRNO: declare double @cosh(double noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare float @coshf(float noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare x86_fp80 @coshl(x86_fp80 noundef) [[NOT_READNONE]]
@@ -656,10 +656,10 @@ __builtin_sin(f);        __builtin_sinf(f);       __builtin_sinl(f); __builtin_s
 
 __builtin_sinh(f);       __builtin_sinhf(f);      __builtin_sinhl(f); __builtin_sinhf128(f);
 
-// NO__ERRNO: declare double @sinh(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @sinhf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @sinhl(x86_fp80 noundef) [[READNONE]]
-// NO__ERRNO: declare fp128 @sinhf128(fp128 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.sinh.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.sinh.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.sinh.f80(x86_fp80) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare fp128 @llvm.sinh.f128(fp128) [[READNONE_INTRINSIC]]
 // HAS_ERRNO: declare double @sinh(double noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare float @sinhf(float noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare x86_fp80 @sinhl(x86_fp80 noundef) [[NOT_READNONE]]
@@ -689,10 +689,10 @@ __builtin_tan(f);        __builtin_tanf(f);       __builtin_tanl(f); __builtin_t
 
 __builtin_tanh(f);       __builtin_tanhf(f);      __builtin_tanhl(f); __builtin_tanhf128(f);
 
-// NO__ERRNO: declare double @tanh(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @tanhf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @tanhl(x86_fp80 noundef) [[READNONE]]
-// NO__ERRNO: declare fp128 @tanhf128(fp128 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.tanh.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.tanh.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.tanh.f80(x86_fp80) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare fp128 @llvm.tanh.f128(fp128) [[READNONE_INTRINSIC]]
 // HAS_ERRNO: declare double @tanh(double noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare float @tanhf(float noundef) [[NOT_READNONE]]
 // HAS_ERRNO: declare x86_fp80 @tanhl(x86_fp80 noundef) [[NOT_READNONE]]
diff --git a/clang/test/CodeGen/constrained-math-builtins.c b/clang/test/CodeGen/constrained-math-builtins.c
index 42c9e3c5008a3..aa77620b44535 100644
--- a/clang/test/CodeGen/constrained-math-builtins.c
+++ b/clang/test/CodeGen/constrained-math-builtins.c
@@ -36,6 +36,27 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c, _
 // CHECK: call float @llvm.experimental.constrained.ldexp.f32.i32(float %{{.*}}, i32 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 // CHECK: call x86_fp80 @llvm.experimental.constrained.ldexp.f80.i32(x86_fp80 %{{.*}}, i32 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 
+  __builtin_acos(f);        __builtin_acosf(f);       __builtin_acosl(f); __builtin_acosf128(f);
+
+// CHECK: call double @llvm.experimental.constrained.acos.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call float @llvm.experimental.constrained.acos.f32(float %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call x86_fp80 @llvm.experimental.constrained.acos.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call fp128 @llvm.experimental.constrained.acos.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+
+__builtin_asin(f);        __builtin_asinf(f);       __builtin_asinl(f); __builtin_asinf128(f);
+
+// CHECK: call double @llvm.experimental.constrained.asin.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call float @llvm.experimental.constrained.asin.f32(float %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call x86_fp80 @llvm.experimental.constrained.asin.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call fp128 @llvm.experimental.constrained.asin.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+
+__builtin_atan(f);        __builtin_atanf(f);       __builtin_atanl(f); __builtin_atanf128(f);
+
+// CHECK: call double @llvm.experimental.constrained.atan.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call float @llvm.experimental.constrained.atan.f32(float %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call x86_fp80 @llvm.experimental.constrained.atan.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call fp128 @llvm.experimental.constrained.atan.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+
   __builtin_ceil(f);       __builtin_ceilf(f);      __builtin_ceill(f); __builtin_ceilf128(f);
 
 // CHECK: call double @llvm.experimental.constrained.ceil.f64(double %{{.*}}, metadata !"fpexcept.strict")
@@ -50,6 +71,13 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c, _
 // CHECK: call x86_fp80 @llvm.experimental.constrained.cos.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 // CHECK: call fp128 @llvm.experimental.constrained.cos.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 
+  __builtin_cosh(f);        __builtin_coshf(f);       __builtin_coshl(f); __builtin_coshf128(f);
+
+// CHECK: call double @llvm.experimental.constrained.cosh.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call float @llvm.experimental.constrained.cosh.f32(float %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call x86_fp80 @llvm.experimental.constrained.cosh.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call fp128 @llvm.experimental.constrained.cosh.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+
   __builtin_exp(f);        __builtin_expf(f);       __builtin_expl(f); __builtin_expf128(f);
 
 // CHECK: call double @llvm.experimental.constrained.exp.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
@@ -177,6 +205,13 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c, _
 // CHECK: call x86_fp80 @llvm.experimental.constrained.sin.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 // CHECK: call fp128 @llvm.experimental.constrained.sin.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 
+  __builtin_sinh(f);        __builtin_sinhf(f);       __builtin_sinhl(f); __builtin_sinhf128(f);
+
+// CHECK: call double @llvm.experimental.constrained.sinh.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call float @llvm.experimental.constrained.sinh.f32(float %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call x86_fp80 @llvm.experimental.constrained.sinh.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call fp128 @llvm.experimental.constrained.sinh.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+
   __builtin_sqrt(f);       __builtin_sqrtf(f);      __builtin_sqrtl(f); __builtin_sqrtf128(f);
 
 // CHECK: call double @llvm.experimental.constrained.sqrt.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
@@ -191,6 +226,12 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c, _
 // CHECK: call x86_fp80 @llvm.experimental.constrained.tan.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 // CHECK: call fp128 @llvm.experimental.constrained.tan.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 
+  __builtin_tanh(f);        __builtin_tanhf(f);       __builtin_tanhl(f); __builtin_tanhf128(f);
+
+// CHECK: call double @llvm.experimental.constrained.tanh.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call float @llvm.experimental.constrained.tanh.f32(float %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call x86_fp80 @llvm.experimental.constrained.tanh.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call fp128 @llvm.experimental.constrained.tanh.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
 
   __builtin_trunc(f);      __builtin_truncf(f);     __builtin_truncl(f); __builtin_truncf128(f);
 
diff --git a/clang/test/CodeGen/math-libcalls.c b/clang/test/CodeGen/math-libcalls.c
index a249182692762..5b23a4a3faef3 100644
--- a/clang/test/CodeGen/math-libcalls.c
+++ b/clang/test/CodeGen/math-libcalls.c
@@ -121,15 +121,15 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c) {
   /* math */
   acos(f);       acosf(f);      acosl(f);
 
-// NO__ERRNO: declare double @acos(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @acosf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @acosl(x86_fp80 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.acos.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.acos.f32(float) [[READNONE_INTRINSIC]...
[truncated]

efriedma-quic

LGTM

…builtins (llvm#98949) ## Change(s) - `Builtins.td` - Add f16 support for libm arc and hyperbolic trig functions - `CGBuiltin.cpp` - Emit constraint intrinsics for trig clang builtins ## History This change is part of an implementation of llvm#87367 investigation on supporting IEEE math operations as intrinsics. Which was discussed in this RFC: https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294 This change adds wasm lowering cases for `acos`, `asin`, `atan`, `cosh`, `sinh`, and `tanh`. llvm#70079 llvm#70080 llvm#70081 llvm#70083 llvm#70084 llvm#95966 ## Precursor PR(s) Note this PR needs Merge after: - llvm#98937 - llvm#98755

…builtins (#98949) Summary: ## Change(s) - `Builtins.td` - Add f16 support for libm arc and hyperbolic trig functions - `CGBuiltin.cpp` - Emit constraint intrinsics for trig clang builtins ## History This change is part of an implementation of #87367 investigation on supporting IEEE math operations as intrinsics. Which was discussed in this RFC: https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294 This change adds wasm lowering cases for `acos`, `asin`, `atan`, `cosh`, `sinh`, and `tanh`. #70079 #70080 #70081 #70083 #70084 #95966 ## Precursor PR(s) Note this PR needs Merge after: - #98937 - #98755 Test Plan: Reviewers: Subscribers: Tasks: Tags: Differential Revision: https://phabricator.intern.facebook.com/D60251574

alanzhao1 · 2024-07-30T22:57:26Z

We (Chrome) are seeing linker failures on our x86 Windows builds due to this PR: https://crbug.com/355455678

[12702/86218] LINK glmark2_wgl.exe glmark2_wgl.exe.pdb
..\..\third_party\llvm-build\Release+Asserts\bin\lld-link.exe /OUT:./glmark2_wgl.exe /nologo -libpath:..\..\third_party\llvm-build\Release+Asserts\lib\clang\20\lib\windows /winsysroot:../../third_party/depot_tools/win_toolchain/vs_files/7393122652 /MACHINE:X86 /PDB:./glmark2_wgl.exe.pdb @./glmark2_wgl.exe.rsp
lld-link: error: undefined symbol: _asinf
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\model.cpp:266
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/model.obj:(public: void __thiscall Model::calculate_texcoords(void))
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\model.cpp:267
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/model.obj:(public: void __thiscall Model::calculate_texcoords(void))

lld-link: error: undefined symbol: _atanf
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\scene-build.cpp:179
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/scene-build.obj:(public: virtual bool __thiscall SceneBuild::setup(void))
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\scene-refract.cpp:382
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/scene-refract.obj:(public: bool __thiscall RefractPrivate::setup(class std::__Cr::map<class std::__Cr::basic_string<char, struct std::__Cr::char_traits<char>, class std::__Cr::allocator<char>>, struct Scene::Option, struct std::__Cr::less<class std::__Cr::basic_string<char, struct std::__Cr::char_traits<char>, class std::__Cr::allocator<char>>>, class std::__Cr::allocator<struct std::__Cr::pair<class std::__Cr::basic_string<char, struct std::__Cr::char_traits<char>, class std::__Cr::allocator<char>> const, struct Scene::Option>>> &))
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\scene-shading.cpp:250
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/scene-shading.obj:(public: virtual bool __thiscall SceneShading::setup(void))
>>> referenced 2 more times
exit=1
error: 1 steps failed

efriedma-quic · 2024-07-30T23:04:51Z

Right, asinf doesn't exist on Windows (see llvm/lib/Analysis/TargetLibraryInfo.cpp). We probably need code to promote calls to llvm.asin.f32 to 64-bit. Not sure how hard it would be to revert this... probably easier to fix forward if we can do that quickly.

We'll need a fix for the regression for the 19 branch.

farzonl · 2024-07-30T23:26:18Z

Right, asinf doesn't exist on Windows (see llvm/lib/Analysis/TargetLibraryInfo.cpp). We probably need code to promote calls to llvm.asin.f32 to 64-bit. Not sure how hard it would be to revert this... probably easier to fix forward if we can do that quickly.

We'll need a fix for the regression for the 19 branch.

I see that promoting seems like it could work, but it is also disabled for f80\f128:
https://github.com/llvm/llvm-project/blob/39e192b379362e9e645427631c35450d55ed517d/llvm/lib/Analysis/TargetLibraryInfo.cpp#L322C5-L324C39

How should we handle that case?

farzonl · 2024-07-30T23:39:02Z

We (Chrome) are seeing linker failures on our x86 Windows builds due to this PR: https://crbug.com/355455678

[12702/86218] LINK glmark2_wgl.exe glmark2_wgl.exe.pdb
..\..\third_party\llvm-build\Release+Asserts\bin\lld-link.exe /OUT:./glmark2_wgl.exe /nologo -libpath:..\..\third_party\llvm-build\Release+Asserts\lib\clang\20\lib\windows /winsysroot:../../third_party/depot_tools/win_toolchain/vs_files/7393122652 /MACHINE:X86 /PDB:./glmark2_wgl.exe.pdb @./glmark2_wgl.exe.rsp
lld-link: error: undefined symbol: _asinf
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\model.cpp:266
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/model.obj:(public: void __thiscall Model::calculate_texcoords(void))
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\model.cpp:267
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/model.obj:(public: void __thiscall Model::calculate_texcoords(void))

lld-link: error: undefined symbol: _atanf
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\scene-build.cpp:179
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/scene-build.obj:(public: virtual bool __thiscall SceneBuild::setup(void))
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\scene-refract.cpp:382
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/scene-refract.obj:(public: bool __thiscall RefractPrivate::setup(class std::__Cr::map<class std::__Cr::basic_string<char, struct std::__Cr::char_traits<char>, class std::__Cr::allocator<char>>, struct Scene::Option, struct std::__Cr::less<class std::__Cr::basic_string<char, struct std::__Cr::char_traits<char>, class std::__Cr::allocator<char>>>, class std::__Cr::allocator<struct std::__Cr::pair<class std::__Cr::basic_string<char, struct std::__Cr::char_traits<char>, class std::__Cr::allocator<char>> const, struct Scene::Option>>> &))
>>> referenced by .\..\..\third_party\angle\third_party\glmark2\src\src\scene-shading.cpp:250
>>>               obj/third_party/angle/third_party/glmark2/glmark2_common_gl/scene-shading.obj:(public: virtual bool __thiscall SceneShading::setup(void))
>>> referenced 2 more times
exit=1
error: 1 steps failed

I'm preparing a fix, building it now should be up in the next 20-30 minutes. Thanks for reporting this issue!

Windows does not support float C89 math functions like: - acosf - asinf - atanf - coshf - sinhf - tanhf These 6 libfuncs need to be type promoted. This PR fixes the bug introduced by llvm#98949

Windows does not support float C89 math functions like: - acosf - asinf - atanf - coshf - sinhf - tanhf All 6 are disabled in [TargetLibraryInfo.cpp:293](https://github.com/llvm/llvm-project/blob/39e192b379362e9e645427631c35450d55ed517d/llvm/lib/Analysis/TargetLibraryInfo.cpp#L293C4-L296C41) These 6 libfuncs need to be type promoted. This PR fixes the bug introduced by #98949

farzonl · 2024-07-31T00:49:11Z

@alanzhao1 @efriedma-quic It should be fixed now by #101268 if you role the clang build to latest.

alanzhao1 · 2024-07-31T01:59:46Z

@alanzhao1 @efriedma-quic It should be fixed now by #101268 if you role the clang build to latest.

Confirmed. Thanks!

Windows does not support float C89 math functions like: - acosf - asinf - atanf - coshf - sinhf - tanhf All 6 are disabled in [TargetLibraryInfo.cpp:293](https://github.com/llvm/llvm-project/blob/39e192b379362e9e645427631c35450d55ed517d/llvm/lib/Analysis/TargetLibraryInfo.cpp#L293C4-L296C41) These 6 libfuncs need to be type promoted. This PR fixes the bug introduced by llvm#98949

This change due to llvm/llvm-project#98949.

Windows does not support float C89 math functions like: - acosf - asinf - atanf - coshf - sinhf - tanhf These 6 libfuncs need to be type promoted. This PR fixes the bug introduced by llvm#98949

farzonl requested review from davemgreen and efriedma-quic July 15, 2024 19:48

llvmbot added clang Clang issues not falling into any other category backend:X86 clang:frontend Language frontend issues, e.g. anything involving "Sema" clang:codegen labels Jul 15, 2024

fix test failure

db0ee81

efriedma-quic approved these changes Jul 15, 2024

View reviewed changes

farzonl self-assigned this Jul 16, 2024

farzonl merged commit a14baec into llvm:main Jul 19, 2024
7 checks passed

farzonl mentioned this pull request Jul 31, 2024

[x86][Windows] Fix chromium build break #101268

Merged

farzonl deleted the arc-hyp-trig-constraint-intrinsics branch July 31, 2024 00:47

zahiraam mentioned this pull request Aug 12, 2024

Add missing fpbuiltin math functions. intel/llvm#15039

Merged

jsji pushed a commit to intel/llvm that referenced this pull request Aug 12, 2024

Add missing fpbuiltin math functions. (#15039)

746f3d9

This change due to llvm/llvm-project#98949.

jsji pushed a commit to intel/llvm that referenced this pull request Aug 16, 2024

Add missing fpbuiltin math functions. (#15039)

22f9f89

This change due to llvm/llvm-project#98949.

farzonl mentioned this pull request Oct 4, 2024

[x86][Windows] Fix chromium build break #111218

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[clang] Emit constraint intrinsics for arc and hyperbolic trig clang builtins #98949

[clang] Emit constraint intrinsics for arc and hyperbolic trig clang builtins #98949

farzonl commented Jul 15, 2024

llvmbot commented Jul 15, 2024 •

edited

Loading

Change(s)

History

Precursor PR(s)

efriedma-quic left a comment

alanzhao1 commented Jul 30, 2024

efriedma-quic commented Jul 30, 2024

farzonl commented Jul 30, 2024

farzonl commented Jul 30, 2024

farzonl commented Jul 31, 2024

alanzhao1 commented Jul 31, 2024

[clang] Emit constraint intrinsics for arc and hyperbolic trig clang builtins #98949

[clang] Emit constraint intrinsics for arc and hyperbolic trig clang builtins #98949

Conversation

farzonl commented Jul 15, 2024

Change(s)

History

Precursor PR(s)

llvmbot commented Jul 15, 2024 • edited Loading

Change(s)

History

Precursor PR(s)

efriedma-quic left a comment

Choose a reason for hiding this comment

alanzhao1 commented Jul 30, 2024

efriedma-quic commented Jul 30, 2024

farzonl commented Jul 30, 2024

farzonl commented Jul 30, 2024

farzonl commented Jul 31, 2024

alanzhao1 commented Jul 31, 2024

llvmbot commented Jul 15, 2024 •

edited

Loading