Merge branch 'main' into users/meinersbur/flang_runtime_split-headersusers/meinersbur/flang_runtime_split-headers

7 files changed, 133 insertions, 11 deletions

diff --git a/libc/src/math/CMakeLists.txt b/libc/src/math/CMakeLists.txt
index 390a59d07a28..e4e2c49642f2 100644
--- a/libc/src/math/CMakeLists.txt
+++ b/libc/src/math/CMakeLists.txt
@@ -89,6 +89,7 @@ add_math_entrypoint_object(copysignf128)
 
 add_math_entrypoint_object(cos)
 add_math_entrypoint_object(cosf)
+add_math_entrypoint_object(cosf16)
 
 add_math_entrypoint_object(cosh)
 add_math_entrypoint_object(coshf)
diff --git a/libc/src/math/cosf16.h b/libc/src/math/cosf16.h
new file mode 100644
index 000000000000..cc179a6e16c6
--- /dev/null
+++ b/libc/src/math/cosf16.h
@@ -0,0 +1,21 @@
+//===-- Implementation header for cosf16 ------------------------*- 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_MATH_COSF16_H
+#define LLVM_LIBC_SRC_MATH_COSF16_H
+
+#include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/types.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+float16 cosf16(float16 x);
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC_MATH_COSF16_H
diff --git a/libc/src/math/docs/add_math_function.md b/libc/src/math/docs/add_math_function.md
index f02d502399e2..daaf1a3ec563 100644
--- a/libc/src/math/docs/add_math_function.md
+++ b/libc/src/math/docs/add_math_function.md
@@ -18,7 +18,7 @@ together with its specifications:
 ```
 - Add function specs to the file:
 ```
-  libc/hdrgen/yaml/math.yaml
+  libc/include/math.yaml
 ```
 
 ## Implementation
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index aeb758d4a092..b3d461291519 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -419,6 +419,25 @@ add_entrypoint_object(
 )
 
 add_entrypoint_object(
+  cosf16
+  SRCS
+    cosf16.cpp
+  HDRS
+    ../cosf16.h
+  DEPENDS
+    .sincosf16_utils
+    libc.hdr.errno_macros
+    libc.hdr.fenv_macros
+    libc.src.__support.FPUtil.cast
+    libc.src.__support.FPUtil.fenv_impl
+    libc.src.__support.FPUtil.fp_bits
+    libc.src.__support.FPUtil.except_value_utils
+    libc.src.__support.FPUtil.multiply_add
+    libc.src.__support.macros.optimization
+    libc.src.__support.macros.properties.types
+)
+
+add_entrypoint_object(
   cospif
   SRCS
     cospif.cpp
diff --git a/libc/src/math/generic/cosf16.cpp b/libc/src/math/generic/cosf16.cpp
new file mode 100644
index 000000000000..534e07854474
--- /dev/null
+++ b/libc/src/math/generic/cosf16.cpp
@@ -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 (auto r = COSF16_EXCEPTS.lookup(x_abs); LIBC_UNLIKELY(r.has_value()))
+    return r.value();
+
+  // 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()) {
+    if (xbits.is_inf()) {
+      fputil::set_errno_if_required(EDOM);
+      fputil::raise_except_if_required(FE_INVALID);
+    }
+
+    return x + FPBits::quiet_nan().get_val();
+  }
+
+  float sin_k, cos_k, sin_y, cosm1_y;
+  sincosf16_eval(xf, sin_k, cos_k, sin_y, cosm1_y);
+  // Since, cosm1_y = cos_y - 1, therefore:
+  //   cos(x) = cos_k * cos_y - sin_k * sin_y
+  //          = cos_k * (cos_y - 1 + 1) - sin_k * sin_y
+  //          = cos_k * cosm1_y - sin_k * sin_y + cos_k
+  return fputil::cast<float16>(fputil::multiply_add(
+      cos_k, cosm1_y, fputil::multiply_add(-sin_k, sin_y, cos_k)));
+}
+
+} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/cospif16.cpp b/libc/src/math/generic/cospif16.cpp
index 384b39ff8e2c..ee74bdb4a369 100644
--- a/libc/src/math/generic/cospif16.cpp
+++ b/libc/src/math/generic/cospif16.cpp
@@ -73,7 +73,7 @@ LLVM_LIBC_FUNCTION(float16, cospif16, (float16 x)) {
     return fputil::cast<float16>(0.0f);
 
   // Since, cosm1_y = cos_y - 1, therefore:
-  // 	cos(x * pi) = cos_k(cosm1_y) + cos_k - sin_k * sin_y
+  //   cos(x * pi) = cos_k(cosm1_y) + cos_k - sin_k * sin_y
   return fputil::cast<float16>(fputil::multiply_add(
       cos_k, cosm1_y, fputil::multiply_add(-sin_k, sin_y, cos_k)));
 }
diff --git a/libc/src/math/generic/sinf16.cpp b/libc/src/math/generic/sinf16.cpp
index 86546348ba73..7c15951cccec 100644
--- a/libc/src/math/generic/sinf16.cpp
+++ b/libc/src/math/generic/sinf16.cpp
@@ -54,13 +54,10 @@ LLVM_LIBC_FUNCTION(float16, sinf16, (float16 x)) {
   //   	        sin(y * pi/32) * cos(k * pi/32)
 
   // Handle exceptional values
-  if (LIBC_UNLIKELY(x_abs == 0x585c || x_abs == 0x5cb0 || x_abs == 0x51f5 ||
-                    x_abs == 0x2b45)) {
-    bool x_sign = x_u >> 15;
-    if (auto r = SINF16_EXCEPTS.lookup_odd(x_abs, x_sign);
-        LIBC_UNLIKELY(r.has_value()))
-      return r.value();
-  }
+  bool x_sign = x_u >> 15;
+  if (auto r = SINF16_EXCEPTS.lookup_odd(x_abs, x_sign);
+      LIBC_UNLIKELY(r.has_value()))
+    return r.value();
 
   int rounding = fputil::quick_get_round();
 
@@ -99,8 +96,8 @@ LLVM_LIBC_FUNCTION(float16, sinf16, (float16 x)) {
   if (LIBC_UNLIKELY(sin_y == 0 && sin_k == 0))
     return FPBits::zero(xbits.sign()).get_val();
 
-  // Since, cosm1_y = cos_y - 1, therfore:
-  //    sin(x) = cos_k * sin_y + sin_k + (cosm1_y * sin_k)
+  // Since, cosm1_y = cos_y - 1, therefore:
+  //   sin(x) = cos_k * sin_y + sin_k + (cosm1_y * sin_k)
   return fputil::cast<float16>(fputil::multiply_add(
       sin_y, cos_k, fputil::multiply_add(cosm1_y, sin_k, sin_k)));
 }