[libc-commits] [libc] [libc][math][c23] Add cosf16 function (PR #118785)

[via libc-commits]

================
@@ -0,0 +1,84 @@
+//===-- Half-precision cos(x) function ------------------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/math/cosf16.h"
+#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 {
+
+constexpr size_t N_EXCEPTS = 4;
+
+constexpr fputil::ExceptValues<float16, N_EXCEPTS> COSF16_EXCEPTS{{
+    // (input, RZ output, RU offset, RD offset, RN offset)
+    {0x2b7c, 0x3bfc, 1, 0, 1},
+    {0x4ac1, 0x38b5, 1, 0, 0},
+    {0x5c49, 0xb8c6, 0, 1, 0},
+    {0x7acc, 0xa474, 0, 1, 0},
+}};
+
+LLVM_LIBC_FUNCTION(float16, cosf16, (float16 x)) {
+  using FPBits = fputil::FPBits<float16>;
+  FPBits xbits(x);
+
+  uint16_t x_u = xbits.uintval();
+  uint16_t x_abs = x_u & 0x7fff;
+  float xf = x;
+
+  // Range reduction:
+  // For |x| > pi/32, we perform range reduction as follows:
+  // Find k and y such that:
+  //   x = (k + y) * pi/32
+  //   k is an integer, |y| < 0.5
+  //
+  // This is done by performing:
+  //   k = round(x * 32/pi)
+  //   y = x * 32/pi - k
+  //
+  // Once k and y are computed, we then deduce the answer by the cosine of sum
+  // formula:
+  //   cos(x) = cos((k + y) * pi/32)
+  //          = cos(k * pi/32) * cos(y * pi/32) -
+  //            sin(k * pi/32) * sin(y * pi/32)
+
+  // Handle exceptional values
+
+  if (LIBC_UNLIKELY(x_abs == 0x2b7c || x_abs == 0x4ac1 || x_abs == 0x5c49 ||
+                    x_abs == 0x7acc)) {
+    if (auto r = COSF16_EXCEPTS.lookup(x_abs); LIBC_UNLIKELY(r.has_value()))
+      return r.value();
+  }
+
+  if (LIBC_UNLIKELY(x_abs == 0U))
+    return fputil::cast<float16>(1.0f);
+
+  if (xbits.is_inf_or_nan()) {
----------------
overmighty wrote:

More of a nit, but I would still add comments here.

```suggestion
  // cos(+/-0) = 1
  if (LIBC_UNLIKELY(x_abs == 0U))
    return fputil::cast<float16>(1.0f);

  // cos(+/-inf) = NaN, and cos(NaN) = NaN.
  if (xbits.is_inf_or_nan()) {
```

https://github.com/llvm/llvm-project/pull/118785