[flang] optimize tand precision
#155629
[flang] optimize tand precision
#155629
Conversation
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Connector Switch (c8ef) ChangesCloses #150452. Full diff: https://github.com/llvm/llvm-project/pull/155629.diff 2 Files Affected:
diff --git a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
index f6d71a9d1d933..70a3734798379 100644
--- a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
+++ b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
@@ -8290,10 +8290,11 @@ mlir::Value IntrinsicLibrary::genTand(mlir::Type resultType,
mlir::MLIRContext *context = builder.getContext();
mlir::FunctionType ftype =
mlir::FunctionType::get(context, {resultType}, {args[0].getType()});
- llvm::APFloat pi = llvm::APFloat(llvm::numbers::pi);
- mlir::Value dfactor = builder.createRealConstant(
- loc, mlir::Float64Type::get(context), pi / llvm::APFloat(180.0));
- mlir::Value factor = builder.createConvert(loc, args[0].getType(), dfactor);
+ const llvm::fltSemantics &fltSem =
+ llvm::cast<mlir::FloatType>(resultType).getFloatSemantics();
+ llvm::APFloat pi = llvm::APFloat(fltSem, llvm::numbers::pis);
+ mlir::Value factor = builder.createRealConstant(
+ loc, resultType, pi / llvm::APFloat(fltSem, "180.0"));
mlir::Value arg = mlir::arith::MulFOp::create(builder, loc, args[0], factor);
return getRuntimeCallGenerator("tan", ftype)(builder, loc, {arg});
}
diff --git a/flang/test/Lower/Intrinsics/tand.f90 b/flang/test/Lower/Intrinsics/tand.f90
index b0f0c52400f78..8c8927e6f5a99 100644
--- a/flang/test/Lower/Intrinsics/tand.f90
+++ b/flang/test/Lower/Intrinsics/tand.f90
@@ -1,3 +1,4 @@
+! REQUIRES: flang-supports-f128-math
! RUN: bbc -emit-fir %s -o - | FileCheck %s --check-prefixes="CHECK,CHECK-FAST"
! RUN: bbc --math-runtime=precise -emit-fir %s -o - | FileCheck %s --check-prefixes="CHECK,CHECK-PRECISE"
! RUN: %flang_fc1 -emit-fir %s -o - | FileCheck %s --check-prefixes="CHECK,CHECK-FAST"
@@ -8,8 +9,7 @@ function test_real4(x)
end function
! CHECK-LABEL: @_QPtest_real4
-! CHECK: %[[dfactor:.*]] = arith.constant 0.017453292519943295 : f64
-! CHECK: %[[factor:.*]] = fir.convert %[[dfactor]] : (f64) -> f32
+! CHECK: %[[factor:.*]] = arith.constant 0.0174532924 : f32
! CHECK: %[[arg:.*]] = arith.mulf %{{[A-Za-z0-9._]+}}, %[[factor]] fastmath<contract> : f32
! CHECK-PRECISE: %{{.*}} = fir.call @tanf(%[[arg]]) fastmath<contract> : (f32) -> f32
! CHECK-FAST: %{{.*}} = math.tan %[[arg]] fastmath<contract> : f32
@@ -24,3 +24,13 @@ function test_real8(x)
! CHECK: %[[arg:.*]] = arith.mulf %{{[A-Za-z0-9._]+}}, %[[factor]] fastmath<contract> : f64
! CHECK-PRECISE: %{{.*}} = fir.call @tan(%[[arg]]) fastmath<contract> : (f64) -> f64
! CHECK-FAST: %{{.*}} = math.tan %[[arg]] fastmath<contract> : f64
+
+function test_real16(x)
+ real(16) :: x, test_real16
+ test_real16 = tand(x)
+end function
+
+! CHECK-LABEL: @_QPtest_real16
+! CHECK: %[[factor:.*]] = arith.constant 0.0174532925199432957692369076848861{{.*}} : f128
+! CHECK: %[[arg:.*]] = arith.mulf %{{[A-Za-z0-9._]+}}, %[[factor]] fastmath<contract> : f128
+! CHECK: %[[result:.*]] = fir.call @_FortranATanF128({{.*}}) fastmath<contract> : (f128) -> f128
|
|
BTW, I have a question about the f128 provider in Flang. In the current main branch, |
|
We support f128 if either libquadmath is available or if libm defines macros indicating that it supports f128: |
Thanks for the pointer, I'll look into it later. On my Ubuntu 22.04 development box, the glibc version is 2.35. My thought is to have f128 usable by default on both x86_64 and aarch64, but on x86, it's currently unsupported by default. |
|
@c8ef, the support for newer versions of glibc needs to be implemented and tested. There is no fundamental reasons not to support it, its just has not been done yet. If you would like to work on that, I will be glad to review it. |
Will do! |
Closes #150452.