[clang] [llvm] [clang] Reland Add tanf16 builtin and support for tan constrained intrinsic (PR #94559)
Farzon Lotfi via llvm-commits
llvm-commits at lists.llvm.org
Thu Jun 6 10:08:06 PDT 2024
https://github.com/farzonl updated https://github.com/llvm/llvm-project/pull/94559
>From 3c461b62ae530ddc965ddeea78246d782c8bb9c6 Mon Sep 17 00:00:00 2001
From: Farzon Lotfi <farzonlotfi at microsoft.com>
Date: Fri, 24 May 2024 10:01:52 -0400
Subject: [PATCH 1/2] [clang] Add tanf16 builtin and support for tan
constrained intrinsic
---
clang/include/clang/Basic/Builtins.td | 6 ++--
clang/lib/CodeGen/CGBuiltin.cpp | 12 +++++++
clang/test/CodeGen/X86/math-builtins.c | 8 ++---
.../test/CodeGen/constrained-math-builtins.c | 13 +++++++
clang/test/CodeGen/math-libcalls.c | 12 +++----
clang/test/CodeGenOpenCL/builtins-f16.cl | 3 ++
llvm/docs/LangRef.rst | 36 +++++++++++++++++++
llvm/include/llvm/IR/ConstrainedOps.def | 1 +
llvm/include/llvm/IR/Intrinsics.td | 4 +++
llvm/test/Assembler/fp-intrinsics-attr.ll | 8 +++++
llvm/test/Feature/fp-intrinsics.ll | 11 ++++++
11 files changed, 101 insertions(+), 13 deletions(-)
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 11982af3fa609..7bef5fd7ad40f 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -482,11 +482,11 @@ def SqrtF16F128 : Builtin, F16F128MathTemplate {
let Prototype = "T(T)";
}
-def TanF128 : Builtin {
- let Spellings = ["__builtin_tanf128"];
+def TanF16F128 : Builtin, F16F128MathTemplate {
+ let Spellings = ["__builtin_tan"];
let Attributes = [FunctionWithBuiltinPrefix, NoThrow,
ConstIgnoringErrnoAndExceptions];
- let Prototype = "__float128(__float128)";
+ let Prototype = "T(T)";
}
def TanhF128 : Builtin {
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 37d0c478e0330..a76ce82830d8f 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -2923,6 +2923,18 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
SetSqrtFPAccuracy(Call);
return RValue::get(Call);
}
+
+ case Builtin::BItan:
+ case Builtin::BItanf:
+ case Builtin::BItanl:
+ case Builtin::BI__builtin_tan:
+ case Builtin::BI__builtin_tanf:
+ case Builtin::BI__builtin_tanf16:
+ case Builtin::BI__builtin_tanl:
+ case Builtin::BI__builtin_tanf128:
+ return RValue::get(emitUnaryMaybeConstrainedFPBuiltin(
+ *this, E, Intrinsic::tan, Intrinsic::experimental_constrained_tan));
+
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 093239b448260..1e0f129b98610 100644
--- a/clang/test/CodeGen/X86/math-builtins.c
+++ b/clang/test/CodeGen/X86/math-builtins.c
@@ -674,10 +674,10 @@ __builtin_sqrt(f); __builtin_sqrtf(f); __builtin_sqrtl(f); __builtin_
__builtin_tan(f); __builtin_tanf(f); __builtin_tanl(f); __builtin_tanf128(f);
-// NO__ERRNO: declare double @tan(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @tanf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @tanl(x86_fp80 noundef) [[READNONE]]
-// NO__ERRNO: declare fp128 @tanf128(fp128 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.tan.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.tan.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.tan.f80(x86_fp80) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare fp128 @llvm.tan.f128(fp128) [[READNONE_INTRINSIC]]
// HAS_ERRNO: declare double @tan(double noundef) [[NOT_READNONE]]
// HAS_ERRNO: declare float @tanf(float noundef) [[NOT_READNONE]]
// HAS_ERRNO: declare x86_fp80 @tanl(x86_fp80 noundef) [[NOT_READNONE]]
diff --git a/clang/test/CodeGen/constrained-math-builtins.c b/clang/test/CodeGen/constrained-math-builtins.c
index 2de832dd2b6ca..6cc3a10a1e794 100644
--- a/clang/test/CodeGen/constrained-math-builtins.c
+++ b/clang/test/CodeGen/constrained-math-builtins.c
@@ -183,6 +183,14 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c, _
// CHECK: call x86_fp80 @llvm.experimental.constrained.sqrt.f80(x86_fp80 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
// CHECK: call fp128 @llvm.experimental.constrained.sqrt.f128(fp128 %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+ __builtin_tan(f); __builtin_tanf(f); __builtin_tanl(f); __builtin_tanf128(f);
+
+// CHECK: call double @llvm.experimental.constrained.tan.f64(double %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// CHECK: call float @llvm.experimental.constrained.tan.f32(float %{{.*}}, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// 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_trunc(f); __builtin_truncf(f); __builtin_truncl(f); __builtin_truncf128(f);
// CHECK: call double @llvm.experimental.constrained.trunc.f64(double %{{.*}}, metadata !"fpexcept.strict")
@@ -315,6 +323,11 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c, _
// CHECK: declare x86_fp80 @llvm.experimental.constrained.sqrt.f80(x86_fp80, metadata, metadata)
// CHECK: declare fp128 @llvm.experimental.constrained.sqrt.f128(fp128, metadata, metadata)
+// CHECK: declare double @llvm.experimental.constrained.tan.f64(double, metadata, metadata)
+// CHECK: declare float @llvm.experimental.constrained.tan.f32(float, metadata, metadata)
+// CHECK: declare x86_fp80 @llvm.experimental.constrained.tan.f80(x86_fp80, metadata, metadata)
+// CHECK: declare fp128 @llvm.experimental.constrained.tan.f128(fp128, metadata, metadata)
+
// CHECK: declare double @llvm.experimental.constrained.trunc.f64(double, metadata)
// CHECK: declare float @llvm.experimental.constrained.trunc.f32(float, metadata)
// CHECK: declare x86_fp80 @llvm.experimental.constrained.trunc.f80(x86_fp80, metadata)
diff --git a/clang/test/CodeGen/math-libcalls.c b/clang/test/CodeGen/math-libcalls.c
index 29c312ba0ecac..a249182692762 100644
--- a/clang/test/CodeGen/math-libcalls.c
+++ b/clang/test/CodeGen/math-libcalls.c
@@ -662,15 +662,15 @@ void foo(double *d, float f, float *fp, long double *l, int *i, const char *c) {
tan(f); tanf(f); tanl(f);
-// NO__ERRNO: declare double @tan(double noundef) [[READNONE]]
-// NO__ERRNO: declare float @tanf(float noundef) [[READNONE]]
-// NO__ERRNO: declare x86_fp80 @tanl(x86_fp80 noundef) [[READNONE]]
+// NO__ERRNO: declare double @llvm.tan.f64(double) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare float @llvm.tan.f32(float) [[READNONE_INTRINSIC]]
+// NO__ERRNO: declare x86_fp80 @llvm.tan.f80(x86_fp80) [[READNONE_INTRINSIC]]
// HAS_ERRNO: declare double @tan(double noundef) [[NOT_READNONE]]
// HAS_ERRNO: declare float @tanf(float noundef) [[NOT_READNONE]]
// HAS_ERRNO: declare x86_fp80 @tanl(x86_fp80 noundef) [[NOT_READNONE]]
-// HAS_MAYTRAP: declare double @tan(double noundef) [[NOT_READNONE]]
-// HAS_MAYTRAP: declare float @tanf(float noundef) [[NOT_READNONE]]
-// HAS_MAYTRAP: declare x86_fp80 @tanl(x86_fp80 noundef) [[NOT_READNONE]]
+// HAS_MAYTRAP: declare double @llvm.experimental.constrained.tan.f64(
+// HAS_MAYTRAP: declare float @llvm.experimental.constrained.tan.f32(
+// HAS_MAYTRAP: declare x86_fp80 @llvm.experimental.constrained.tan.f80(
tanh(f); tanhf(f); tanhl(f);
diff --git a/clang/test/CodeGenOpenCL/builtins-f16.cl b/clang/test/CodeGenOpenCL/builtins-f16.cl
index adf7cdde154f5..d7bffdad5c548 100644
--- a/clang/test/CodeGenOpenCL/builtins-f16.cl
+++ b/clang/test/CodeGenOpenCL/builtins-f16.cl
@@ -66,6 +66,9 @@ void test_half_builtins(half h0, half h1, half h2, int i0) {
// CHECK: call half @llvm.sqrt.f16(half %h0)
res = __builtin_sqrtf16(h0);
+ // CHECK: call half @llvm.tan.f16(half %h0)
+ res = __builtin_tanf16(h0);
+
// CHECK: call half @llvm.trunc.f16(half %h0)
res = __builtin_truncf16(h0);
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index 9d7ade8eb523b..08a41170d9b9b 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -26231,6 +26231,42 @@ same values as the libm ``cos`` functions would, and handles error
conditions in the same way.
+'``llvm.experimental.constrained.tan``' Intrinsic
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax:
+"""""""
+
+::
+
+ declare <type>
+ @llvm.experimental.constrained.tan(<type> <op1>,
+ metadata <rounding mode>,
+ metadata <exception behavior>)
+
+Overview:
+"""""""""
+
+The '``llvm.experimental.constrained.tan``' intrinsic returns the tangent of the
+first operand.
+
+Arguments:
+""""""""""
+
+The first argument and the return type are floating-point numbers of the same
+type.
+
+The second and third arguments specify the rounding mode and exception
+behavior as described above.
+
+Semantics:
+""""""""""
+
+This function returns the tangent of the specified operand, returning the
+same values as the libm ``tan`` functions would, and handles error
+conditions in the same way.
+
+
'``llvm.experimental.constrained.exp``' Intrinsic
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/llvm/include/llvm/IR/ConstrainedOps.def b/llvm/include/llvm/IR/ConstrainedOps.def
index 41aa44de957f9..a7b37c5cb204d 100644
--- a/llvm/include/llvm/IR/ConstrainedOps.def
+++ b/llvm/include/llvm/IR/ConstrainedOps.def
@@ -95,6 +95,7 @@ DAG_FUNCTION(round, 1, 0, experimental_constrained_round, FROUND)
DAG_FUNCTION(roundeven, 1, 0, experimental_constrained_roundeven, FROUNDEVEN)
DAG_FUNCTION(sin, 1, 1, experimental_constrained_sin, FSIN)
DAG_FUNCTION(sqrt, 1, 1, experimental_constrained_sqrt, FSQRT)
+DAG_FUNCTION(tan, 1, 1, experimental_constrained_tan, FTAN)
DAG_FUNCTION(trunc, 1, 0, experimental_constrained_trunc, FTRUNC)
// This is definition for fmuladd intrinsic function, that is converted into
diff --git a/llvm/include/llvm/IR/Intrinsics.td b/llvm/include/llvm/IR/Intrinsics.td
index 107442623ab7b..4c506a6ace23e 100644
--- a/llvm/include/llvm/IR/Intrinsics.td
+++ b/llvm/include/llvm/IR/Intrinsics.td
@@ -1218,6 +1218,10 @@ let IntrProperties = [IntrInaccessibleMemOnly, IntrWillReturn, IntrStrictFP] in
[ LLVMMatchType<0>,
llvm_metadata_ty,
llvm_metadata_ty ]>;
+ def int_experimental_constrained_tan : DefaultAttrsIntrinsic<[ llvm_anyfloat_ty ],
+ [ LLVMMatchType<0>,
+ llvm_metadata_ty,
+ llvm_metadata_ty ]>;
def int_experimental_constrained_pow : DefaultAttrsIntrinsic<[ llvm_anyfloat_ty ],
[ LLVMMatchType<0>,
LLVMMatchType<0>,
diff --git a/llvm/test/Assembler/fp-intrinsics-attr.ll b/llvm/test/Assembler/fp-intrinsics-attr.ll
index 6546d1a275c99..613630e1a2b4d 100644
--- a/llvm/test/Assembler/fp-intrinsics-attr.ll
+++ b/llvm/test/Assembler/fp-intrinsics-attr.ll
@@ -85,6 +85,11 @@ define void @func(double %a, double %b, double %c, i32 %i) strictfp {
metadata !"round.dynamic",
metadata !"fpexcept.strict")
+ %tan = call double @llvm.experimental.constrained.tan.f64(
+ double %a,
+ metadata !"round.dynamic",
+ metadata !"fpexcept.strict")
+
%pow = call double @llvm.experimental.constrained.pow.f64(
double %a, double %b,
metadata !"round.dynamic",
@@ -244,6 +249,9 @@ declare double @llvm.experimental.constrained.sin.f64(double, metadata, metadata
declare double @llvm.experimental.constrained.cos.f64(double, metadata, metadata)
; CHECK: @llvm.experimental.constrained.cos.f64({{.*}}) #[[ATTR1]]
+declare double @llvm.experimental.constrained.tan.f64(double, metadata, metadata)
+; CHECK: @llvm.experimental.constrained.tan.f64({{.*}}) #[[ATTR1]]
+
declare double @llvm.experimental.constrained.pow.f64(double, double, metadata, metadata)
; CHECK: @llvm.experimental.constrained.pow.f64({{.*}}) #[[ATTR1]]
diff --git a/llvm/test/Feature/fp-intrinsics.ll b/llvm/test/Feature/fp-intrinsics.ll
index b92408a1bf1cd..7759813dc2e11 100644
--- a/llvm/test/Feature/fp-intrinsics.ll
+++ b/llvm/test/Feature/fp-intrinsics.ll
@@ -151,6 +151,17 @@ entry:
ret double %result
}
+; Verify that tan(42.0) isn't simplified when the rounding mode is unknown.
+; CHECK-LABEL: ftan
+; CHECK: call double @llvm.experimental.constrained.tan
+define double @ftan() #0 {
+entry:
+ %result = call double @llvm.experimental.constrained.tan.f64(double 42.0,
+ metadata !"round.dynamic",
+ metadata !"fpexcept.strict") #0
+ ret double %result
+}
+
; Verify that exp(42.0) isn't simplified when the rounding mode is unknown.
; CHECK-LABEL: f10
; CHECK: call double @llvm.experimental.constrained.exp
>From ce4e2a09e59127ec3a7107d0cc299d5ef423bb62 Mon Sep 17 00:00:00 2001
From: Farzon Lotfi <farzonlotfi at microsoft.com>
Date: Thu, 6 Jun 2024 13:07:48 -0400
Subject: [PATCH 2/2] add arm64 llvm.experimental.constrained.tan tests
---
llvm/test/CodeGen/AArch64/fp-intrinsics.ll | 31 ++++++++++++++++++++++
1 file changed, 31 insertions(+)
diff --git a/llvm/test/CodeGen/AArch64/fp-intrinsics.ll b/llvm/test/CodeGen/AArch64/fp-intrinsics.ll
index 685efbb7cad43..67d0b63f4076f 100644
--- a/llvm/test/CodeGen/AArch64/fp-intrinsics.ll
+++ b/llvm/test/CodeGen/AArch64/fp-intrinsics.ll
@@ -146,6 +146,13 @@ define float @cos_f32(float %x) #0 {
ret float %val
}
+; CHECK-LABEL: tan_f32:
+; CHECK: bl tanf
+define float @tan_f32(float %x) #0 {
+ %val = call float @llvm.experimental.constrained.tan.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+ ret float %val
+}
+
; CHECK-LABEL: pow_f32:
; CHECK: bl powf
define float @pow_f32(float %x, float %y) #0 {
@@ -630,6 +637,13 @@ define double @cos_f64(double %x) #0 {
ret double %val
}
+; CHECK-LABEL: tan_f64:
+; CHECK: bl tan
+define double @tan_f64(double %x) #0 {
+ %val = call double @llvm.experimental.constrained.tan.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+ ret double %val
+}
+
; CHECK-LABEL: pow_f64:
; CHECK: bl pow
define double @pow_f64(double %x, double %y) #0 {
@@ -1114,6 +1128,13 @@ define fp128 @cos_f128(fp128 %x) #0 {
ret fp128 %val
}
+; CHECK-LABEL: tan_f128:
+; CHECK: bl tanl
+define fp128 @tan_f128(fp128 %x) #0 {
+ %val = call fp128 @llvm.experimental.constrained.tan.f128(fp128 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+ ret fp128 %val
+}
+
; CHECK-LABEL: pow_f128:
; CHECK: bl powl
define fp128 @pow_f128(fp128 %x, fp128 %y) #0 {
@@ -1491,6 +1512,13 @@ define <1 x double> @cos_v1f64(<1 x double> %x, <1 x double> %y) #0 {
ret <1 x double> %val
}
+; CHECK-LABEL: tan_v1f64:
+; CHECK: bl tan
+define <1 x double> @tan_v1f64(<1 x double> %x, <1 x double> %y) #0 {
+ %val = call <1 x double> @llvm.experimental.constrained.tan.v1f64(<1 x double> %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+ ret <1 x double> %val
+}
+
; CHECK-LABEL: pow_v1f64:
; CHECK: bl pow
define <1 x double> @pow_v1f64(<1 x double> %x, <1 x double> %y) #0 {
@@ -1555,6 +1583,7 @@ declare float @llvm.experimental.constrained.sqrt.f32(float, metadata, metadata)
declare float @llvm.experimental.constrained.powi.f32(float, i32, metadata, metadata)
declare float @llvm.experimental.constrained.sin.f32(float, metadata, metadata)
declare float @llvm.experimental.constrained.cos.f32(float, metadata, metadata)
+declare float @llvm.experimental.constrained.tan.f32(float, metadata, metadata)
declare float @llvm.experimental.constrained.pow.f32(float, float, metadata, metadata)
declare float @llvm.experimental.constrained.log.f32(float, metadata, metadata)
declare float @llvm.experimental.constrained.log10.f32(float, metadata, metadata)
@@ -1599,6 +1628,7 @@ declare double @llvm.experimental.constrained.sqrt.f64(double, metadata, metadat
declare double @llvm.experimental.constrained.powi.f64(double, i32, metadata, metadata)
declare double @llvm.experimental.constrained.sin.f64(double, metadata, metadata)
declare double @llvm.experimental.constrained.cos.f64(double, metadata, metadata)
+declare double @llvm.experimental.constrained.tan.f64(double, metadata, metadata)
declare double @llvm.experimental.constrained.pow.f64(double, double, metadata, metadata)
declare double @llvm.experimental.constrained.log.f64(double, metadata, metadata)
declare double @llvm.experimental.constrained.log10.f64(double, metadata, metadata)
@@ -1643,6 +1673,7 @@ declare fp128 @llvm.experimental.constrained.sqrt.f128(fp128, metadata, metadata
declare fp128 @llvm.experimental.constrained.powi.f128(fp128, i32, metadata, metadata)
declare fp128 @llvm.experimental.constrained.sin.f128(fp128, metadata, metadata)
declare fp128 @llvm.experimental.constrained.cos.f128(fp128, metadata, metadata)
+declare fp128 @llvm.experimental.constrained.tan.f128(fp128, metadata, metadata)
declare fp128 @llvm.experimental.constrained.pow.f128(fp128, fp128, metadata, metadata)
declare fp128 @llvm.experimental.constrained.log.f128(fp128, metadata, metadata)
declare fp128 @llvm.experimental.constrained.log10.f128(fp128, metadata, metadata)
More information about the llvm-commits
mailing list