[libc] Add nearest integer instructions to fputil.

[lntue]

Summary

Add round to nearest integer instructions to fputil. This will be
used in sinf implementation https://reviews.llvm.org/D123154

[michaelrj]

LGTM

[tschuett]

I am probably wrong, but ...
The Arm® Architecture Reference Manual defines a feature called FEAT_FRINTTS. It was added with Armv8.5.
FEAT_FRINTTS, Floating-point to integer instructions.

frintn is probably only available on modern ARM cores ala iPhone 12 from 2020.

You include arm_acle.h in a few places. There are intrinsics for frint with a feature flag, see https://gcc.gnu.org/legacy-ml/gcc-patches/2019-09/msg00053.html.

[lntue]

In D129776#3689493, @tschuett wrote:

I am probably wrong, but ...
The Arm® Architecture Reference Manual defines a feature called FEAT_FRINTTS. It was added with Armv8.5.
FEAT_FRINTTS, Floating-point to integer instructions.

frintn is probably only available on modern ARM cores ala iPhone 12 from 2020.

You include arm_acle.h in a few places. There are intrinsics for frint with a feature flag, see https://gcc.gnu.org/legacy-ml/gcc-patches/2019-09/msg00053.html.

I do think frintn is available on aarch64. I got it from https://developer.arm.com/documentation/ddi0487/ga section C7.2.162 (page 1873). I also tested it on a Raspberry Pi 400,

Also thanks for the suggestion about the intrinsics. I would definitely replacing these arm assembly with compiler intrinsics if they are available.

[tschuett]

I can confirm that I found frintn on page 1873. What worries me is A2.8.1. There is a feature FEAT_FRINTTS added by Armv8.5.

This uncertainty can be easily solved by using the intrinsics. Thanks.

libc/src/__support/FPUtil/CMakeLists.txt

 add_header_library(
   polyeval
   HDRS
     PolyEval.h
   DEPENDS
     .multiply_add
 )
 
+add_header_library(
+  nearest_integer
+  HDRS
+    nearest_integer.h
+  DEPENDS
+    libc.src.__support.common
+)
+
 add_subdirectory(generic)

libc/src/__support/FPUtil/aarch64/nearest_integer.h

+//===--- Round floating point to nearest integer on aarch64 -----*- 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_FPUTIL_AARCH64_NEAREST_INTEGER_H
+#define LLVM_LIBC_SRC_SUPPORT_FPUTIL_AARCH64_NEAREST_INTEGER_H
+
+#include "src/__support/architectures.h"
+
+#if !defined(LLVM_LIBC_ARCH_AARCH64)
+#error "Invalid include"
+#endif
+
+namespace __llvm_libc {
+namespace fputil {
+
+static inline double nearest_integer(double x) {
+  double result;
+  __asm__ __volatile__("frintn %d0, %d1\n\t" : "=w"(result) : "w"(x));
+  return result;
+}
+
+} // namespace fputil
+} // namespace __llvm_libc
+
+#endif // LLVM_LIBC_SRC_SUPPORT_FPUTIL_AARCH64_NEAREST_INTEGER_H

libc/src/__support/FPUtil/nearest_integer.h

+//===-- Fast rounding to nearest integer for floating point -----*- 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_FPUTIL_NEAREST_INTEGER_H
+#define LLVM_LIBC_SRC_SUPPORT_FPUTIL_NEAREST_INTEGER_H
+
+#include "src/__support/architectures.h"
+#include "src/__support/common.h"
+
+#if (defined(LLVM_LIBC_ARCH_X86_64) && defined(__SSE4_2__))
+#include "x86_64/nearest_integer.h"
+#elif defined(LLVM_LIBC_ARCH_AARCH64)
+#include "aarch64/nearest_integer.h"
+#else
+
+namespace __llvm_libc {
+namespace fputil {
+
+// This is a fast implementation for rounding to a nearest integer that, in case
+// of a tie, might pick a random one among 2 closest integers when the rounding
+// mode is not FE_TONEAREST.
+//
+// Notice that for AARCH64 and x86-64 with SSE4.2 support, we will use their
+// corresponding rounding instruction instead.  And in those cases, the results
+// are rounded to the nearest integer, tie-to-even.
+static inline double nearest_integer(double x) {
+  if (x < 0x1p53 && x > -0x1p53) {
+    double r = x < 0 ? (x - 0x1.0p52) + 0x1.0p52 : (x + 0x1.0p52) - 0x1.0p52;
+    double diff = x - r;
+    // The expression above is correct for the default rounding mode, round-to-
+    // nearest, tie-to-even.  For other rounding modes, it might be off by 1,
+    // which is corrected below.
+    if (unlikely(diff > 0.5))
+      return r + 1.0;
+    if (unlikely(diff < -0.5))
+      return r - 1.0;
+    return r;
+  }
+  return x;
+}
+
+} // namespace fputil
+} // namespace __llvm_libc
+
+#endif
+#endif // LLVM_LIBC_SRC_SUPPORT_FPUTIL_NEAREST_INTEGER_H

libc/src/__support/FPUtil/x86_64/nearest_integer.h

+//===--- Round floating point to nearest integer on x86-64 ------*- 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_FPUTIL_X86_64_NEAREST_INTEGER_H
+#define LLVM_LIBC_SRC_SUPPORT_FPUTIL_X86_64_NEAREST_INTEGER_H
+
+#include "src/__support/architectures.h"
+
+#if !defined(LLVM_LIBC_ARCH_X86_64)
+#error "Invalid include"
+#endif
+
+#if !defined(__SSE4_2__)
+#error "SSE4.2 instruction set is not supported"
+#endif
+
+#include <immintrin.h>
+
+namespace __llvm_libc {
+namespace fputil {
+
+static inline double nearest_integer(double x) {
+  __m128d xmm = _mm_set_sd(x); // NOLINT
+  __m128d ymm =
+      _mm_round_sd(xmm, xmm, _MM_ROUND_NEAREST | _MM_FROUND_NO_EXC); // NOLINT
+  return ymm[0];
+}
+
+} // namespace fputil
+} // namespace __llvm_libc
+
+#endif // LLVM_LIBC_SRC_SUPPORT_FPUTIL_X86_64_NEAREST_INTEGER_H

utils/bazel/llvm-project-overlay/libc/BUILD.bazel

cc_library(
     hdrs = [
         "src/__support/FPUtil/PolyEval.h",
     ],
     deps = [
         ":__support_fputil_multiply_add",
     ],
 )
 
+nearest_integer_common_hdrs = [
+    "src/__support/FPUtil/nearest_integer.h",
+]
+
+nearest_integer_platform_hdrs = [
+    "src/__support/FPUtil/x86_64/nearest_integer.h",
+    "src/__support/FPUtil/aarch64/nearest_integer.h",
+]
+
+cc_library(
+    name = "__support_fputil_nearest_integer",
+    hdrs = nearest_integer_common_hdrs,
+    # These are conditionally included and will #error out if the platform
+    # doesn't support rounding instructions, so they can't be compiled on their
+    # own.
+    textual_hdrs = nearest_integer_platform_hdrs,
+    deps = [
+        ":__support_common",
+        ":libc_root",
+    ],
+)
+
 ################################ fenv targets ################################
 
 libc_function(
     name = "fetestexcept",
     srcs = ["src/fenv/fetestexcept.cpp"],
     hdrs = ["src/fenv/fetestexcept.h"],
     deps = [
         ":__support_common",