[libc-commits] [libc] [llvm] [libc][math] Refactor tanpif16 to header-only (PR #181240)

[Weiwen He via libc-commits]

================
@@ -0,0 +1,126 @@
+//===-- Implementation header for tanpif16 ----------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_TANPIF16_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_TANPIF16_H
+#include "include/llvm-libc-macros/float16-macros.h"
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "hdr/errno_macros.h"
+#include "hdr/fenv_macros.h"
+#include "sincosf16_utils.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/cast.h"
+#include "src/__support/FPUtil/except_value_utils.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/macros/optimization.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+namespace math {
+
+namespace tanpif16_internal {
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+LIBC_INLINE_VAR constexpr size_t N_EXCEPTS = 21;
+
+LIBC_INLINE_VAR constexpr fputil::ExceptValues<float16, N_EXCEPTS>
+    TANPIF16_EXCEPTS{{
+        // (input, RZ output, RU offset, RD offset, RN offset)
+        {0x07f2, 0x0e3d, 1, 0, 0}, {0x086a, 0x0eee, 1, 0, 1},
+        {0x08db, 0x0fa0, 1, 0, 0}, {0x094c, 0x1029, 1, 0, 0},
+        {0x0b10, 0x118c, 1, 0, 0}, {0x1ce0, 0x23a8, 1, 0, 1},
+        {0x1235, 0x18e0, 1, 0, 0}, {0x2579, 0x2c4e, 1, 0, 0},
+        {0x28b2, 0x2f68, 1, 0, 1}, {0x2a43, 0x30f4, 1, 0, 1},
+        {0x31b7, 0x3907, 1, 0, 0}, {0x329d, 0x3a12, 1, 0, 1},
+        {0x34f1, 0x3dd7, 1, 0, 0}, {0x3658, 0x41ee, 1, 0, 0},
+        {0x38d4, 0xc1ee, 0, 1, 0}, {0x3d96, 0x41ee, 1, 0, 0},
+        {0x3e6a, 0xc1ee, 0, 1, 0}, {0x40cb, 0x41ee, 1, 0, 0},
+        {0x4135, 0xc1ee, 0, 1, 0}, {0x42cb, 0x41ee, 1, 0, 0},
+        {0x4335, 0xc1ee, 0, 1, 0},
+    }};
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+} // namespace tanpif16_internal
+
+LIBC_INLINE constexpr float16 tanpif16(float16 x) {
----------------
he-weiwen wrote:

Thanks for these suggestions. Only question: I thought the purpose of the original RFC is to make functions like this `constexpr`, so presumably the reason this cannot yet be `constexpr` is that some of its dependencies still need some work, and the intention is that the entire call chain could be 'constexpr' at some point? Let me know if my understanding is correct and if there's any plan for further work on this.

https://github.com/llvm/llvm-project/pull/181240