diff --git a/libc/cmake/modules/CheckCompilerFeatures.cmake b/libc/cmake/modules/CheckCompilerFeatures.cmake
index d84c07b35d2d7..a6d793d495c45 100644
--- a/libc/cmake/modules/CheckCompilerFeatures.cmake
+++ b/libc/cmake/modules/CheckCompilerFeatures.cmake
@@ -2,7 +2,15 @@
 # Compiler features definition and flags
 # ------------------------------------------------------------------------------
 
-set(ALL_COMPILER_FEATURES "float16" "float128" "fixed_point")
+set(
+  ALL_COMPILER_FEATURES
+    "builtin_ceil_floor_rint_trunc"
+    "builtin_round"
+    "builtin_roundeven"
+    "float16"
+    "float128"
+    "fixed_point"
+)
 
 # Making sure ALL_COMPILER_FEATURES is sorted.
 list(SORT ALL_COMPILER_FEATURES)
@@ -39,11 +47,22 @@ endfunction()
 set(AVAILABLE_COMPILER_FEATURES "")
 
 # Try compile a C file to check if flag is supported.
-set(CMAKE_TRY_COMPILE_TARGET_TYPE STATIC_LIBRARY)
 foreach(feature IN LISTS ALL_COMPILER_FEATURES)
+  set(CMAKE_TRY_COMPILE_TARGET_TYPE STATIC_LIBRARY)
   set(compile_options ${LIBC_COMPILE_OPTIONS_NATIVE})
+  set(link_options "")
   if(${feature} STREQUAL "fixed_point")
     list(APPEND compile_options "-ffixed-point")
+  elseif(${feature} MATCHES "^builtin_")
+    set(compile_options ${LIBC_COMPILE_OPTIONS_DEFAULT})
+    set(link_options -nostdlib)
+    # The compiler might handle calls to rounding builtins by generating calls
+    # to the respective libc math functions, in which case we cannot use these
+    # builtins in our implementations of these functions. We check that this is
+    # not the case by trying to link an executable, since linking would fail due
+    # to unresolved references with -nostdlib if calls to libc functions were
+    # generated.
+    set(CMAKE_TRY_COMPILE_TARGET_TYPE EXECUTABLE)
   endif()
 
   try_compile(
@@ -51,6 +70,7 @@ foreach(feature IN LISTS ALL_COMPILER_FEATURES)
     ${CMAKE_CURRENT_BINARY_DIR}/compiler_features
     SOURCES ${LIBC_SOURCE_DIR}/cmake/modules/compiler_features/check_${feature}.cpp
     COMPILE_DEFINITIONS -I${LIBC_SOURCE_DIR} ${compile_options}
+    LINK_OPTIONS ${link_options}
   )
   if(has_feature)
     list(APPEND AVAILABLE_COMPILER_FEATURES ${feature})
@@ -60,6 +80,12 @@ foreach(feature IN LISTS ALL_COMPILER_FEATURES)
       set(LIBC_TYPES_HAS_FLOAT128 TRUE)
     elseif(${feature} STREQUAL "fixed_point")
       set(LIBC_COMPILER_HAS_FIXED_POINT TRUE)
+    elseif(${feature} STREQUAL "builtin_ceil_floor_rint_trunc")
+      set(LIBC_COMPILER_HAS_BUILTIN_CEIL_FLOOR_RINT_TRUNC TRUE)
+    elseif(${feature} STREQUAL "builtin_round")
+      set(LIBC_COMPILER_HAS_BUILTIN_ROUND TRUE)
+    elseif(${feature} STREQUAL "builtin_roundeven")
+      set(LIBC_COMPILER_HAS_BUILTIN_ROUNDEVEN TRUE)
     endif()
   endif()
 endforeach()
diff --git a/libc/cmake/modules/LLVMLibCCompileOptionRules.cmake b/libc/cmake/modules/LLVMLibCCompileOptionRules.cmake
index c5e7dfe8abd0f..253da4ae890e5 100644
--- a/libc/cmake/modules/LLVMLibCCompileOptionRules.cmake
+++ b/libc/cmake/modules/LLVMLibCCompileOptionRules.cmake
@@ -4,7 +4,7 @@ function(_get_compile_options_from_flags output_var)
   if(LIBC_TARGET_ARCHITECTURE_IS_RISCV64 OR(LIBC_CPU_FEATURES MATCHES "FMA"))
     check_flag(ADD_FMA_FLAG ${FMA_OPT_FLAG} ${ARGN})
   endif()
-  check_flag(ADD_SSE4_2_FLAG ${ROUND_OPT_FLAG} ${ARGN})
+  check_flag(ADD_ROUND_OPT_FLAG ${ROUND_OPT_FLAG} ${ARGN})
   check_flag(ADD_EXPLICIT_SIMD_OPT_FLAG ${EXPLICIT_SIMD_OPT_FLAG} ${ARGN})
 
   if(LLVM_COMPILER_IS_GCC_COMPATIBLE)
@@ -16,8 +16,23 @@ function(_get_compile_options_from_flags output_var)
         list(APPEND compile_options "-D__LIBC_RISCV_USE_FMA")
       endif()
     endif()
-    if(ADD_SSE4_2_FLAG)
-      list(APPEND compile_options "-msse4.2")
+    if(ADD_ROUND_OPT_FLAG)
+      if(LIBC_TARGET_ARCHITECTURE_IS_X86)
+        # ROUND_OPT_FLAG is only enabled if SSE4.2 is detected, not just SSE4.1,
+        # because there was code to check for SSE4.2 already, and few CPUs only
+        # have SSE4.1.
+        list(APPEND compile_options "-msse4.2")
+      endif()
+      if(LIBC_COMPILER_HAS_BUILTIN_CEIL_FLOOR_RINT_TRUNC)
+        list(APPEND compile_options
+             "-D__LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC")
+      endif()
+      if(LIBC_COMPILER_HAS_BUILTIN_ROUND)
+        list(APPEND compile_options "-D__LIBC_USE_BUILTIN_ROUND")
+      endif()
+      if(LIBC_COMPILER_HAS_BUILTIN_ROUNDEVEN)
+        list(APPEND compile_options "-D__LIBC_USE_BUILTIN_ROUNDEVEN")
+      endif()
     endif()
     if(ADD_EXPLICIT_SIMD_OPT_FLAG)
       list(APPEND compile_options "-D__LIBC_EXPLICIT_SIMD_OPT")
diff --git a/libc/cmake/modules/LLVMLibCFlagRules.cmake b/libc/cmake/modules/LLVMLibCFlagRules.cmake
index 18e36dfde5cc1..eca7ba8d183e6 100644
--- a/libc/cmake/modules/LLVMLibCFlagRules.cmake
+++ b/libc/cmake/modules/LLVMLibCFlagRules.cmake
@@ -277,6 +277,7 @@ if(NOT(LIBC_TARGET_ARCHITECTURE_IS_X86 AND (LIBC_CPU_FEATURES MATCHES "SSE2")))
 endif()
 
 # Skip ROUND_OPT flag for targets that don't support SSE 4.2.
-if(NOT(LIBC_TARGET_ARCHITECTURE_IS_X86 AND (LIBC_CPU_FEATURES MATCHES "SSE4_2")))
+if(NOT((LIBC_TARGET_ARCHITECTURE_IS_X86 AND (LIBC_CPU_FEATURES MATCHES "SSE4_2")) OR
+       LIBC_TARGET_ARCHITECTURE_IS_AARCH64))
   set(SKIP_FLAG_EXPANSION_ROUND_OPT TRUE)
 endif()
diff --git a/libc/cmake/modules/compiler_features/check_builtin_ceil_floor_rint_trunc.cpp b/libc/cmake/modules/compiler_features/check_builtin_ceil_floor_rint_trunc.cpp
new file mode 100644
index 0000000000000..946200001d69f
--- /dev/null
+++ b/libc/cmake/modules/compiler_features/check_builtin_ceil_floor_rint_trunc.cpp
@@ -0,0 +1,11 @@
+float try_builtin_ceilf(float x) { return __builtin_ceilf(x); }
+float try_builtin_floorf(float x) { return __builtin_floorf(x); }
+float try_builtin_rintf(float x) { return __builtin_rintf(x); }
+float try_builtin_truncf(float x) { return __builtin_truncf(x); }
+
+double try_builtin_ceil(double x) { return __builtin_ceil(x); }
+double try_builtin_floor(double x) { return __builtin_floor(x); }
+double try_builtin_rint(double x) { return __builtin_rint(x); }
+double try_builtin_trunc(double x) { return __builtin_trunc(x); }
+
+extern "C" void _start() {}
diff --git a/libc/cmake/modules/compiler_features/check_builtin_round.cpp b/libc/cmake/modules/compiler_features/check_builtin_round.cpp
new file mode 100644
index 0000000000000..79a347ada8b60
--- /dev/null
+++ b/libc/cmake/modules/compiler_features/check_builtin_round.cpp
@@ -0,0 +1,5 @@
+float try_builtin_roundf(float x) { return __builtin_roundf(x); }
+
+double try_builtin_round(double x) { return __builtin_round(x); }
+
+extern "C" void _start() {}
diff --git a/libc/cmake/modules/compiler_features/check_builtin_roundeven.cpp b/libc/cmake/modules/compiler_features/check_builtin_roundeven.cpp
new file mode 100644
index 0000000000000..0aa40dc7f4b7a
--- /dev/null
+++ b/libc/cmake/modules/compiler_features/check_builtin_roundeven.cpp
@@ -0,0 +1,5 @@
+float try_builtin_roundevenf(float x) { return __builtin_roundevenf(x); }
+
+double try_builtin_roundeven(double x) { return __builtin_roundeven(x); }
+
+extern "C" void _start() {}
diff --git a/libc/src/math/aarch64/CMakeLists.txt b/libc/src/math/aarch64/CMakeLists.txt
deleted file mode 100644
index bbe927a1c7c88..0000000000000
--- a/libc/src/math/aarch64/CMakeLists.txt
+++ /dev/null
@@ -1,79 +0,0 @@
-add_entrypoint_object(
-  ceil
-  SRCS
-    ceil.cpp
-  HDRS
-    ../ceil.h
-  COMPILE_OPTIONS
-    -O2
-)
-
-add_entrypoint_object(
-  ceilf
-  SRCS
-    ceilf.cpp
-  HDRS
-    ../ceilf.h
-  COMPILE_OPTIONS
-    -O2
-)
-
-add_entrypoint_object(
-  floor
-  SRCS
-    floor.cpp
-  HDRS
-    ../floor.h
-  COMPILE_OPTIONS
-    -O2
-)
-
-add_entrypoint_object(
-  floorf
-  SRCS
-    floorf.cpp
-  HDRS
-    ../floorf.h
-  COMPILE_OPTIONS
-    -O2
-)
-
-add_entrypoint_object(
-  trunc
-  SRCS
-    trunc.cpp
-  HDRS
-    ../trunc.h
-  COMPILE_OPTIONS
-    -O2
-)
-
-add_entrypoint_object(
-  truncf
-  SRCS
-    truncf.cpp
-  HDRS
-    ../truncf.h
-  COMPILE_OPTIONS
-    -O2
-)
-
-add_entrypoint_object(
-  round
-  SRCS
-    round.cpp
-  HDRS
-    ../round.h
-  COMPILE_OPTIONS
-    -O2
-)
-
-add_entrypoint_object(
-  roundf
-  SRCS
-    roundf.cpp
-  HDRS
-    ../roundf.h
-  COMPILE_OPTIONS
-    -O2
-)
diff --git a/libc/src/math/aarch64/ceil.cpp b/libc/src/math/aarch64/ceil.cpp
deleted file mode 100644
index 5bfd053d603de..0000000000000
--- a/libc/src/math/aarch64/ceil.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the ceil function for aarch64 -------------------===//
-//
-// 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/ceil.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(double, ceil, (double x)) {
-  double y;
-  __asm__ __volatile__("frintp %d0, %d1\n\t" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/aarch64/ceilf.cpp b/libc/src/math/aarch64/ceilf.cpp
deleted file mode 100644
index 2352245bc8303..0000000000000
--- a/libc/src/math/aarch64/ceilf.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the ceilf function for aarch64 ------------------===//
-//
-// 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/ceilf.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(float, ceilf, (float x)) {
-  float y;
-  __asm__ __volatile__("frintp %s0, %s1\n\t" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/aarch64/floor.cpp b/libc/src/math/aarch64/floor.cpp
deleted file mode 100644
index f9da52bdd45a8..0000000000000
--- a/libc/src/math/aarch64/floor.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the floor function for aarch64 ------------------===//
-//
-// 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/floor.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(double, floor, (double x)) {
-  double y;
-  __asm__ __volatile__("frintm %d0, %d1\n\t" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/aarch64/floorf.cpp b/libc/src/math/aarch64/floorf.cpp
deleted file mode 100644
index 980b3c52a00b5..0000000000000
--- a/libc/src/math/aarch64/floorf.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the floorf function for aarch64 -----------------===//
-//
-// 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/floorf.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(float, floorf, (float x)) {
-  float y;
-  __asm__ __volatile__("frintm %s0, %s1\n\t" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/aarch64/round.cpp b/libc/src/math/aarch64/round.cpp
deleted file mode 100644
index c85445aa19ab5..0000000000000
--- a/libc/src/math/aarch64/round.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the round function for aarch64 ------------------===//
-//
-// 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/round.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(double, round, (double x)) {
-  double y;
-  __asm__ __volatile__("frinta %d0, %d1\n\t" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/aarch64/roundf.cpp b/libc/src/math/aarch64/roundf.cpp
deleted file mode 100644
index 0c7f7640e8278..0000000000000
--- a/libc/src/math/aarch64/roundf.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the roundf function for aarch64 -----------------===//
-//
-// 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/roundf.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(float, roundf, (float x)) {
-  float y;
-  __asm__ __volatile__("frinta %s0, %s1\n\t" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/aarch64/trunc.cpp b/libc/src/math/aarch64/trunc.cpp
deleted file mode 100644
index 1ef26f44fd234..0000000000000
--- a/libc/src/math/aarch64/trunc.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the trunc function for aarch64 ------------------===//
-//
-// 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/trunc.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(double, trunc, (double x)) {
-  double y;
-  __asm__ __volatile__("frintz %d0, %d1\n" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/aarch64/truncf.cpp b/libc/src/math/aarch64/truncf.cpp
deleted file mode 100644
index 0c64ef60e91c4..0000000000000
--- a/libc/src/math/aarch64/truncf.cpp
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- Implementation of the truncf function for aarch64 -----------------===//
-//
-// 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/truncf.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-
-namespace LIBC_NAMESPACE_DECL {
-
-LLVM_LIBC_FUNCTION(float, truncf, (float x)) {
-  float y;
-  __asm__ __volatile__("frintz %s0, %s1\n\t" : "=w"(y) : "w"(x));
-  return y;
-}
-
-} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index 5e920307d39de..c2f58fb1a4f71 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -70,6 +70,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -82,6 +84,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -107,6 +111,9 @@ add_entrypoint_object(
   DEPENDS
     libc.src.__support.macros.properties.types
     libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.macros.properties.architectures
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -455,6 +462,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -467,6 +476,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -492,6 +503,9 @@ add_entrypoint_object(
   DEPENDS
     libc.src.__support.macros.properties.types
     libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.macros.properties.architectures
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -517,6 +531,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -529,6 +545,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -554,6 +572,9 @@ add_entrypoint_object(
   DEPENDS
     libc.src.__support.macros.properties.types
     libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.macros.properties.architectures
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -579,6 +600,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -591,6 +614,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -616,6 +641,9 @@ add_entrypoint_object(
   DEPENDS
     libc.src.__support.macros.properties.types
     libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.macros.properties.architectures
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -641,6 +669,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -653,6 +683,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -678,6 +710,9 @@ add_entrypoint_object(
   DEPENDS
     libc.src.__support.macros.properties.types
     libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.macros.properties.architectures
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -827,6 +862,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -839,6 +876,8 @@ add_entrypoint_object(
     -O3
   DEPENDS
     libc.src.__support.FPUtil.nearest_integer_operations
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
@@ -864,6 +903,9 @@ add_entrypoint_object(
   DEPENDS
     libc.src.__support.macros.properties.types
     libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.macros.properties.architectures
+  FLAGS
+    ROUND_OPT
 )
 
 add_entrypoint_object(
diff --git a/libc/src/math/generic/ceil.cpp b/libc/src/math/generic/ceil.cpp
index a5ac1348834d8..72c6e990fcc71 100644
--- a/libc/src/math/generic/ceil.cpp
+++ b/libc/src/math/generic/ceil.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(double, ceil, (double x)) { return fputil::ceil(x); }
+LLVM_LIBC_FUNCTION(double, ceil, (double x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_ceil(x);
+#else
+  return fputil::ceil(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/ceilf.cpp b/libc/src/math/generic/ceilf.cpp
index 0fd54361648dc..dfd0dc62bc51b 100644
--- a/libc/src/math/generic/ceilf.cpp
+++ b/libc/src/math/generic/ceilf.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float, ceilf, (float x)) { return fputil::ceil(x); }
+LLVM_LIBC_FUNCTION(float, ceilf, (float x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_ceilf(x);
+#else
+  return fputil::ceil(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/ceilf16.cpp b/libc/src/math/generic/ceilf16.cpp
index 1d17daf283d07..708bc4cfd4860 100644
--- a/libc/src/math/generic/ceilf16.cpp
+++ b/libc/src/math/generic/ceilf16.cpp
@@ -10,9 +10,17 @@
 #include "src/__support/FPUtil/NearestIntegerOperations.h"
 #include "src/__support/common.h"
 #include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/architectures.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float16, ceilf16, (float16 x)) { return fputil::ceil(x); }
+LLVM_LIBC_FUNCTION(float16, ceilf16, (float16 x)) {
+#if defined(__LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC) &&                       \
+    defined(LIBC_TARGET_ARCH_IS_AARCH64)
+  return static_cast<float16>(__builtin_ceilf(x));
+#else
+  return fputil::ceil(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/floor.cpp b/libc/src/math/generic/floor.cpp
index 417d6aca3ffe5..86aed6c61a7cc 100644
--- a/libc/src/math/generic/floor.cpp
+++ b/libc/src/math/generic/floor.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(double, floor, (double x)) { return fputil::floor(x); }
+LLVM_LIBC_FUNCTION(double, floor, (double x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_floor(x);
+#else
+  return fputil::floor(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/floorf.cpp b/libc/src/math/generic/floorf.cpp
index ca059e6d02d53..22739eff68ec2 100644
--- a/libc/src/math/generic/floorf.cpp
+++ b/libc/src/math/generic/floorf.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float, floorf, (float x)) { return fputil::floor(x); }
+LLVM_LIBC_FUNCTION(float, floorf, (float x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_floorf(x);
+#else
+  return fputil::floor(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/floorf16.cpp b/libc/src/math/generic/floorf16.cpp
index 46068c2a548a9..84e4b0730ac68 100644
--- a/libc/src/math/generic/floorf16.cpp
+++ b/libc/src/math/generic/floorf16.cpp
@@ -10,9 +10,17 @@
 #include "src/__support/FPUtil/NearestIntegerOperations.h"
 #include "src/__support/common.h"
 #include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/architectures.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float16, floorf16, (float16 x)) { return fputil::floor(x); }
+LLVM_LIBC_FUNCTION(float16, floorf16, (float16 x)) {
+#if defined(__LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC) &&                       \
+    defined(LIBC_TARGET_ARCH_IS_AARCH64)
+  return static_cast<float16>(__builtin_floorf(x));
+#else
+  return fputil::floor(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/rint.cpp b/libc/src/math/generic/rint.cpp
index f7837df133484..5defa60ddac1c 100644
--- a/libc/src/math/generic/rint.cpp
+++ b/libc/src/math/generic/rint.cpp
@@ -14,7 +14,11 @@
 namespace LIBC_NAMESPACE_DECL {
 
 LLVM_LIBC_FUNCTION(double, rint, (double x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_rint(x);
+#else
   return fputil::round_using_current_rounding_mode(x);
+#endif
 }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/rintf.cpp b/libc/src/math/generic/rintf.cpp
index 29a2845cc85d4..2fe7788241168 100644
--- a/libc/src/math/generic/rintf.cpp
+++ b/libc/src/math/generic/rintf.cpp
@@ -14,7 +14,11 @@
 namespace LIBC_NAMESPACE_DECL {
 
 LLVM_LIBC_FUNCTION(float, rintf, (float x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_rintf(x);
+#else
   return fputil::round_using_current_rounding_mode(x);
+#endif
 }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/rintf16.cpp b/libc/src/math/generic/rintf16.cpp
index 69b89de31bf40..0e8c091efcf9b 100644
--- a/libc/src/math/generic/rintf16.cpp
+++ b/libc/src/math/generic/rintf16.cpp
@@ -10,11 +10,17 @@
 #include "src/__support/FPUtil/NearestIntegerOperations.h"
 #include "src/__support/common.h"
 #include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/architectures.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
 LLVM_LIBC_FUNCTION(float16, rintf16, (float16 x)) {
+#if defined(__LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC) &&                       \
+    defined(LIBC_TARGET_ARCH_IS_AARCH64)
+  return static_cast<float16>(__builtin_rintf(x));
+#else
   return fputil::round_using_current_rounding_mode(x);
+#endif
 }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/round.cpp b/libc/src/math/generic/round.cpp
index d8b171a2ef092..6ed5be5065962 100644
--- a/libc/src/math/generic/round.cpp
+++ b/libc/src/math/generic/round.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(double, round, (double x)) { return fputil::round(x); }
+LLVM_LIBC_FUNCTION(double, round, (double x)) {
+#ifdef __LIBC_USE_BUILTIN_ROUND
+  return __builtin_round(x);
+#else
+  return fputil::round(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/roundeven.cpp b/libc/src/math/generic/roundeven.cpp
index e6f599991bc0f..42f19f38b5385 100644
--- a/libc/src/math/generic/roundeven.cpp
+++ b/libc/src/math/generic/roundeven.cpp
@@ -14,7 +14,11 @@
 namespace LIBC_NAMESPACE_DECL {
 
 LLVM_LIBC_FUNCTION(double, roundeven, (double x)) {
+#ifdef __LIBC_USE_BUILTIN_ROUNDEVEN
+  return __builtin_roundeven(x);
+#else
   return fputil::round_using_specific_rounding_mode(x, FP_INT_TONEAREST);
+#endif
 }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/roundevenf.cpp b/libc/src/math/generic/roundevenf.cpp
index 0b63a093edf6c..98bdc6545d94e 100644
--- a/libc/src/math/generic/roundevenf.cpp
+++ b/libc/src/math/generic/roundevenf.cpp
@@ -14,7 +14,11 @@
 namespace LIBC_NAMESPACE_DECL {
 
 LLVM_LIBC_FUNCTION(float, roundevenf, (float x)) {
+#ifdef __LIBC_USE_BUILTIN_ROUNDEVEN
+  return __builtin_roundevenf(x);
+#else
   return fputil::round_using_specific_rounding_mode(x, FP_INT_TONEAREST);
+#endif
 }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/roundevenf16.cpp b/libc/src/math/generic/roundevenf16.cpp
index 8a27d81b895ae..b45670bd24ff1 100644
--- a/libc/src/math/generic/roundevenf16.cpp
+++ b/libc/src/math/generic/roundevenf16.cpp
@@ -10,11 +10,17 @@
 #include "src/__support/FPUtil/NearestIntegerOperations.h"
 #include "src/__support/common.h"
 #include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/architectures.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
 LLVM_LIBC_FUNCTION(float16, roundevenf16, (float16 x)) {
+#if defined(__LIBC_USE_BUILTIN_ROUNDEVEN) &&                                   \
+    defined(LIBC_TARGET_ARCH_IS_AARCH64)
+  return static_cast<float16>(__builtin_roundevenf(x));
+#else
   return fputil::round_using_specific_rounding_mode(x, FP_INT_TONEAREST);
+#endif
 }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/roundf.cpp b/libc/src/math/generic/roundf.cpp
index 68be4b1a51987..d25f7128cb9ce 100644
--- a/libc/src/math/generic/roundf.cpp
+++ b/libc/src/math/generic/roundf.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float, roundf, (float x)) { return fputil::round(x); }
+LLVM_LIBC_FUNCTION(float, roundf, (float x)) {
+#ifdef __LIBC_USE_BUILTIN_ROUND
+  return __builtin_roundf(x);
+#else
+  return fputil::round(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/roundf16.cpp b/libc/src/math/generic/roundf16.cpp
index 06f9a79cc21f3..cb668c0e76388 100644
--- a/libc/src/math/generic/roundf16.cpp
+++ b/libc/src/math/generic/roundf16.cpp
@@ -10,9 +10,16 @@
 #include "src/__support/FPUtil/NearestIntegerOperations.h"
 #include "src/__support/common.h"
 #include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/architectures.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float16, roundf16, (float16 x)) { return fputil::round(x); }
+LLVM_LIBC_FUNCTION(float16, roundf16, (float16 x)) {
+#if defined(__LIBC_USE_BUILTIN_ROUND) && defined(LIBC_TARGET_ARCH_IS_AARCH64)
+  return static_cast<float16>(__builtin_roundf(x));
+#else
+  return fputil::round(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/trunc.cpp b/libc/src/math/generic/trunc.cpp
index 0bf2ac5962e7b..603750f11c9f6 100644
--- a/libc/src/math/generic/trunc.cpp
+++ b/libc/src/math/generic/trunc.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(double, trunc, (double x)) { return fputil::trunc(x); }
+LLVM_LIBC_FUNCTION(double, trunc, (double x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_trunc(x);
+#else
+  return fputil::trunc(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/truncf.cpp b/libc/src/math/generic/truncf.cpp
index 371cb9f8229a6..d7b0ffd96da9c 100644
--- a/libc/src/math/generic/truncf.cpp
+++ b/libc/src/math/generic/truncf.cpp
@@ -13,6 +13,12 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float, truncf, (float x)) { return fputil::trunc(x); }
+LLVM_LIBC_FUNCTION(float, truncf, (float x)) {
+#ifdef __LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC
+  return __builtin_truncf(x);
+#else
+  return fputil::trunc(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/truncf16.cpp b/libc/src/math/generic/truncf16.cpp
index ea7695dc95f0b..b931053e53438 100644
--- a/libc/src/math/generic/truncf16.cpp
+++ b/libc/src/math/generic/truncf16.cpp
@@ -10,9 +10,17 @@
 #include "src/__support/FPUtil/NearestIntegerOperations.h"
 #include "src/__support/common.h"
 #include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/architectures.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-LLVM_LIBC_FUNCTION(float16, truncf16, (float16 x)) { return fputil::trunc(x); }
+LLVM_LIBC_FUNCTION(float16, truncf16, (float16 x)) {
+#if defined(__LIBC_USE_BUILTIN_CEIL_FLOOR_RINT_TRUNC) &&                       \
+    defined(LIBC_TARGET_ARCH_IS_AARCH64)
+  return static_cast<float16>(__builtin_truncf(x));
+#else
+  return fputil::trunc(x);
+#endif
+}
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/test/src/math/performance_testing/CMakeLists.txt b/libc/test/src/math/performance_testing/CMakeLists.txt
index c7241317687e0..a75becba04d07 100644
--- a/libc/test/src/math/performance_testing/CMakeLists.txt
+++ b/libc/test/src/math/performance_testing/CMakeLists.txt
@@ -21,7 +21,7 @@ function(add_perf_binary target_name)
     "PERF"
     "" # No optional arguments
     "SUITE;CXX_STANDARD" # Single value arguments
-    "SRCS;HDRS;DEPENDS;COMPILE_OPTIONS" # Multi-value arguments
+    "SRCS;HDRS;DEPENDS;COMPILE_OPTIONS;LINK_LIBRARIES" # Multi-value arguments
     ${ARGN}
   )
   if(NOT PERF_SRCS)
@@ -64,9 +64,13 @@ function(add_perf_binary target_name)
     )
   endif()
 
+  set(link_libraries ${link_object_files})
+  foreach(lib IN LISTS PERF_LINK_LIBRARIES)
+    list(APPEND link_libraries ${lib}.unit)
+  endforeach()
   target_link_libraries(
       ${fq_target_name}
-      PRIVATE ${link_object_files} libc_diff_test_utils)
+      PRIVATE ${link_libraries} libc_diff_test_utils)
 
   set_target_properties(${fq_target_name}
     PROPERTIES RUNTIME_OUTPUT_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
@@ -385,12 +389,16 @@ add_perf_binary(
     libc.src.math.ceilf16
     libc.src.math.floorf
     libc.src.math.floorf16
-    libc.src.math.roundevenf
-    libc.src.math.roundevenf16
+    libc.src.math.rintf
+    libc.src.math.rintf16
     libc.src.math.roundf
     libc.src.math.roundf16
+    libc.src.math.roundevenf
+    libc.src.math.roundevenf16
     libc.src.math.truncf
     libc.src.math.truncf16
   COMPILE_OPTIONS
     -fno-builtin
+  LINK_LIBRARIES
+    LibcFPTestHelpers
 )
diff --git a/libc/test/src/math/performance_testing/nearest_integer_funcs_perf.cpp b/libc/test/src/math/performance_testing/nearest_integer_funcs_perf.cpp
index 24176a377e9d4..b7bd6636a72e1 100644
--- a/libc/test/src/math/performance_testing/nearest_integer_funcs_perf.cpp
+++ b/libc/test/src/math/performance_testing/nearest_integer_funcs_perf.cpp
@@ -11,17 +11,23 @@
 #include "src/math/ceilf16.h"
 #include "src/math/floorf.h"
 #include "src/math/floorf16.h"
+#include "src/math/rintf.h"
+#include "src/math/rintf16.h"
 #include "src/math/roundevenf.h"
 #include "src/math/roundevenf16.h"
 #include "src/math/roundf.h"
 #include "src/math/roundf16.h"
 #include "src/math/truncf.h"
 #include "src/math/truncf16.h"
+#include "test/UnitTest/RoundingModeUtils.h"
 #include "test/src/math/performance_testing/Timer.h"
 
 #include <fstream>
 #include <math.h>
 
+using LIBC_NAMESPACE::fputil::testing::ForceRoundingMode;
+using LIBC_NAMESPACE::fputil::testing::RoundingMode;
+
 namespace LIBC_NAMESPACE::testing {
 
 template <typename T> class NearestIntegerPerf {
@@ -36,7 +42,7 @@ template <typename T> class NearestIntegerPerf {
                                 StorageType ending_bit, StorageType step,
                                 size_t rounds, std::ofstream &log) {
     auto runner = [=](Func func) {
-      volatile T result;
+      [[maybe_unused]] volatile T result;
       for (size_t i = 0; i < rounds; i++) {
         for (StorageType bits = starting_bit; bits <= ending_bit;
              bits += step) {
@@ -146,10 +152,10 @@ int main() {
                        FLOAT16_ROUNDS, "ceilf16_perf.log")
   NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::floorf16, ::placeholderf16,
                        FLOAT16_ROUNDS, "floorf16_perf.log")
-  NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::roundevenf16, ::placeholderf16,
-                       FLOAT16_ROUNDS, "roundevenf16_perf.log")
   NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::roundf16, ::placeholderf16,
                        FLOAT16_ROUNDS, "roundf16_perf.log")
+  NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::roundevenf16, ::placeholderf16,
+                       FLOAT16_ROUNDS, "roundevenf16_perf.log")
   NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::truncf16, ::placeholderf16,
                        FLOAT16_ROUNDS, "truncf16_perf.log")
 
@@ -157,12 +163,37 @@ int main() {
                        "ceilf_perf.log")
   NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::floorf, ::floorf, FLOAT_ROUNDS,
                        "floorf_perf.log")
-  NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::roundevenf, ::placeholderf,
-                       FLOAT_ROUNDS, "roundevenf_perf.log")
   NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::roundf, ::roundf, FLOAT_ROUNDS,
                        "roundf_perf.log")
+  NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::roundevenf, ::placeholderf,
+                       FLOAT_ROUNDS, "roundevenf_perf.log")
   NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::truncf, ::truncf, FLOAT_ROUNDS,
                        "truncf_perf.log")
 
+  if (ForceRoundingMode r(RoundingMode::Upward); r.success) {
+    NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::rintf16, ::placeholderf16,
+                         FLOAT16_ROUNDS, "rintf16_upward_perf.log")
+    NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::rintf, ::rintf, FLOAT_ROUNDS,
+                         "rintf_upward_perf.log")
+  }
+  if (ForceRoundingMode r(RoundingMode::Downward); r.success) {
+    NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::rintf16, ::placeholderf16,
+                         FLOAT16_ROUNDS, "rintf16_downward_perf.log")
+    NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::rintf, ::rintf, FLOAT_ROUNDS,
+                         "rintf_downward_perf.log")
+  }
+  if (ForceRoundingMode r(RoundingMode::TowardZero); r.success) {
+    NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::rintf16, ::placeholderf16,
+                         FLOAT16_ROUNDS, "rintf16_towardzero_perf.log")
+    NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::rintf, ::rintf, FLOAT_ROUNDS,
+                         "rintf_towardzero_perf.log")
+  }
+  if (ForceRoundingMode r(RoundingMode::Nearest); r.success) {
+    NEAREST_INTEGER_PERF(float16, LIBC_NAMESPACE::rintf16, ::placeholderf16,
+                         FLOAT16_ROUNDS, "rintf16_nearest_perf.log")
+    NEAREST_INTEGER_PERF(float, LIBC_NAMESPACE::rintf, ::rintf, FLOAT_ROUNDS,
+                         "rintf_nearest_perf.log")
+  }
+
   return 0;
 }