[libc][math][c23] Add fmodf16 C23 math function (#94629)

Part of #93566.

6 files changed, 132 insertions, 43 deletions

diff --git a/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h b/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h
index 504d1be..3027932 100644
--- a/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h
+++ b/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h
@@ -9,6 +9,7 @@
 #include "src/__support/FPUtil/FPBits.h"
 #include "test/src/math/performance_testing/Timer.h"
 
+#include <cstddef>
 #include <fstream>
 
 namespace LIBC_NAMESPACE {
@@ -25,7 +26,10 @@ public:
 
   static void run_perf_in_range(Func myFunc, Func otherFunc,
                                 StorageType startingBit, StorageType endingBit,
-                                StorageType N, std::ofstream &log) {
+                                size_t N, size_t rounds, std::ofstream &log) {
+    if (endingBit - startingBit < N)
+      N = endingBit - startingBit;
+
     auto runner = [=](Func func) {
       volatile T result;
       if (endingBit < startingBit) {
@@ -33,13 +37,15 @@ public:
       }
 
       StorageType step = (endingBit - startingBit) / N;
-      for (StorageType bitsX = startingBit, bitsY = endingBit;;
-           bitsX += step, bitsY -= step) {
-        T x = FPBits(bitsX).get_val();
-        T y = FPBits(bitsY).get_val();
-        result = func(x, y);
-        if (endingBit - bitsX < step) {
-          break;
+      for (size_t i = 0; i < rounds; i++) {
+        for (StorageType bitsX = startingBit, bitsY = endingBit;;
+             bitsX += step, bitsY -= step) {
+          T x = FPBits(bitsX).get_val();
+          T y = FPBits(bitsY).get_val();
+          result = func(x, y);
+          if (endingBit - bitsX < step) {
+            break;
+          }
         }
       }
     };
@@ -49,7 +55,7 @@ public:
     runner(myFunc);
     timer.stop();
 
-    double my_average = static_cast<double>(timer.nanoseconds()) / N;
+    double my_average = static_cast<double>(timer.nanoseconds()) / N / rounds;
     log << "-- My function --\n";
     log << "     Total time      : " << timer.nanoseconds() << " ns \n";
     log << "     Average runtime : " << my_average << " ns/op \n";
@@ -60,7 +66,8 @@ public:
     runner(otherFunc);
     timer.stop();
 
-    double other_average = static_cast<double>(timer.nanoseconds()) / N;
+    double other_average =
+        static_cast<double>(timer.nanoseconds()) / N / rounds;
     log << "-- Other function --\n";
     log << "     Total time      : " << timer.nanoseconds() << " ns \n";
     log << "     Average runtime : " << other_average << " ns/op \n";
@@ -71,22 +78,24 @@ public:
     log << "     Mine / Other's  : " << my_average / other_average << " \n";
   }
 
-  static void run_perf(Func myFunc, Func otherFunc, const char *logFile) {
+  static void run_perf(Func myFunc, Func otherFunc, int rounds,
+                       const char *logFile) {
     std::ofstream log(logFile);
     log << " Performance tests with inputs in denormal range:\n";
     run_perf_in_range(myFunc, otherFunc, /* startingBit= */ StorageType(0),
                       /* endingBit= */ FPBits::max_subnormal().uintval(),
-                      10'000'001, log);
+                      1'000'001, rounds, log);
     log << "\n Performance tests with inputs in normal range:\n";
     run_perf_in_range(myFunc, otherFunc,
                       /* startingBit= */ FPBits::min_normal().uintval(),
                       /* endingBit= */ FPBits::max_normal().uintval(),
-                      10'000'001, log);
+                      1'000'001, rounds, log);
     log << "\n Performance tests with inputs in normal range with exponents "
            "close to each other:\n";
-    run_perf_in_range(
-        myFunc, otherFunc, /* startingBit= */ FPBits(T(0x1.0p-10)).uintval(),
-        /* endingBit= */ FPBits(T(0x1.0p+10)).uintval(), 1'001'001, log);
+    run_perf_in_range(myFunc, otherFunc,
+                      /* startingBit= */ FPBits(T(0x1.0p-10)).uintval(),
+                      /* endingBit= */ FPBits(T(0x1.0p+10)).uintval(),
+                      1'000'001, rounds, log);
   }
 
   static void run_diff(Func myFunc, Func otherFunc, const char *logFile) {
@@ -117,6 +126,15 @@ public:
 #define BINARY_OP_SINGLE_OUTPUT_PERF(T, myFunc, otherFunc, filename)           \
   int main() {                                                                 \
     LIBC_NAMESPACE::testing::BinaryOpSingleOutputPerf<T>::run_perf(            \
-        &myFunc, &otherFunc, filename);                                        \
+        &myFunc, &otherFunc, 1, filename);                                     \
     return 0;                                                                  \
   }
+
+#define BINARY_OP_SINGLE_OUTPUT_PERF_EX(T, myFunc, otherFunc, rounds,          \
+                                        filename)                              \
+  {                                                                            \
+    LIBC_NAMESPACE::testing::BinaryOpSingleOutputPerf<T>::run_perf(            \
+        &myFunc, &otherFunc, rounds, filename);                                \
+    LIBC_NAMESPACE::testing::BinaryOpSingleOutputPerf<T>::run_perf(            \
+        &myFunc, &otherFunc, rounds, filename);                                \
+  }
diff --git a/libc/test/src/math/performance_testing/CMakeLists.txt b/libc/test/src/math/performance_testing/CMakeLists.txt
index d1fb24e..4ea78f9 100644
--- a/libc/test/src/math/performance_testing/CMakeLists.txt
+++ b/libc/test/src/math/performance_testing/CMakeLists.txt
@@ -88,6 +88,8 @@ add_header_library(
   binary_op_single_output_diff
   HDRS
     BinaryOpSingleOutputPerf.h
+  DEPENDS
+    libc.src.__support.FPUtil.fp_bits
 )
 
 add_perf_binary(
@@ -344,6 +346,17 @@ add_perf_binary(
 )
 
 add_perf_binary(
+  fmodf16_perf
+  SRCS
+    fmodf16_perf.cpp
+  DEPENDS
+    .binary_op_single_output_diff
+    libc.src.math.fmodf16
+    libc.src.__support.FPUtil.generic.fmod
+    libc.src.__support.macros.properties.types
+)
+
+add_perf_binary(
   fmodf128_perf
   SRCS
     fmodf128_perf.cpp
diff --git a/libc/test/src/math/performance_testing/fmodf16_perf.cpp b/libc/test/src/math/performance_testing/fmodf16_perf.cpp
new file mode 100644
index 0000000..ff01fa6
--- /dev/null
+++ b/libc/test/src/math/performance_testing/fmodf16_perf.cpp
@@ -0,0 +1,27 @@
+//===-- Performance test for fmodf16 --------------------------------------===//
+//
+// 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 "BinaryOpSingleOutputPerf.h"
+
+#include "src/__support/FPUtil/generic/FMod.h"
+#include "src/__support/macros/properties/types.h"
+
+#include <stdint.h>
+
+#define FMOD_FUNC(U) (LIBC_NAMESPACE::fputil::generic::FMod<float16, U>::eval)
+
+int main() {
+  BINARY_OP_SINGLE_OUTPUT_PERF_EX(float16, FMOD_FUNC(uint16_t),
+                                  FMOD_FUNC(uint32_t), 5000,
+                                  "fmodf16_u16_vs_u32_perf.log")
+
+  BINARY_OP_SINGLE_OUTPUT_PERF_EX(float16, FMOD_FUNC(uint16_t),
+                                  FMOD_FUNC(uint64_t), 5000,
+                                  "fmodf16_u16_vs_u64_perf.log")
+  return 0;
+}
diff --git a/libc/test/src/math/smoke/CMakeLists.txt b/libc/test/src/math/smoke/CMakeLists.txt
index 110fa1d..07e8b5d 100644
--- a/libc/test/src/math/smoke/CMakeLists.txt
+++ b/libc/test/src/math/smoke/CMakeLists.txt
@@ -3111,10 +3111,10 @@ add_fp_unittest(
   HDRS
     FModTest.h
   DEPENDS
+    libc.hdr.fenv_macros
     libc.src.errno.errno
     libc.src.math.fmodf
-    libc.src.__support.FPUtil.basic_operations
-    libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.FPUtil.fenv_impl
   # FIXME: Currently fails on the GPU build.
   UNIT_TEST_ONLY
 )
@@ -3128,10 +3128,10 @@ add_fp_unittest(
   HDRS
     FModTest.h
   DEPENDS
+    libc.hdr.fenv_macros
     libc.src.errno.errno
     libc.src.math.fmod
-    libc.src.__support.FPUtil.basic_operations
-    libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.FPUtil.fenv_impl
   # FIXME: Currently fails on the GPU build.
   UNIT_TEST_ONLY
 )
@@ -3145,10 +3145,27 @@ add_fp_unittest(
   HDRS
     FModTest.h
   DEPENDS
+    libc.hdr.fenv_macros
     libc.src.errno.errno
     libc.src.math.fmodl
-    libc.src.__support.FPUtil.basic_operations
-    libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.FPUtil.fenv_impl
+  # FIXME: Currently fails on the GPU build.
+  UNIT_TEST_ONLY
+)
+
+add_fp_unittest(
+  fmodf16_test
+  SUITE
+    libc-math-smoke-tests
+  SRCS
+    fmodf16_test.cpp
+  HDRS
+    FModTest.h
+  DEPENDS
+    libc.hdr.fenv_macros
+    libc.src.errno.errno
+    libc.src.math.fmodf16
+    libc.src.__support.FPUtil.fenv_impl
   # FIXME: Currently fails on the GPU build.
   UNIT_TEST_ONLY
 )
@@ -3162,10 +3179,10 @@ add_fp_unittest(
   HDRS
     FModTest.h
   DEPENDS
+    libc.hdr.fenv_macros
     libc.src.errno.errno
     libc.src.math.fmodf128
-    libc.src.__support.FPUtil.basic_operations
-    libc.src.__support.FPUtil.nearest_integer_operations
+    libc.src.__support.FPUtil.fenv_impl
   # FIXME: Currently fails on the GPU build.
   UNIT_TEST_ONLY
 )
diff --git a/libc/test/src/math/smoke/FModTest.h b/libc/test/src/math/smoke/FModTest.h
index f1015d6..405e310 100644
--- a/libc/test/src/math/smoke/FModTest.h
+++ b/libc/test/src/math/smoke/FModTest.h
@@ -9,13 +9,13 @@
 #ifndef LLVM_LIBC_TEST_SRC_MATH_FMODTEST_H
 #define LLVM_LIBC_TEST_SRC_MATH_FMODTEST_H
 
-#include "src/__support/FPUtil/BasicOperations.h"
-#include "src/__support/FPUtil/NearestIntegerOperations.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/errno/libc_errno.h"
 #include "test/UnitTest/FEnvSafeTest.h"
 #include "test/UnitTest/FPMatcher.h"
 #include "test/UnitTest/Test.h"
 
-#include "hdr/math_macros.h"
+#include "hdr/fenv_macros.h"
 
 #define TEST_SPECIAL(x, y, expected, dom_err, expected_exception)              \
   EXPECT_FP_EQ(expected, f(x, y));                                             \
@@ -210,7 +210,8 @@ public:
   }
 
   void testRegularExtreme(FModFunc f) {
-
+    if constexpr (sizeof(T) < sizeof(float))
+      return;
     TEST_REGULAR(0x1p127L, 0x3p-149L, 0x1p-149L);
     TEST_REGULAR(0x1p127L, -0x3p-149L, 0x1p-149L);
     TEST_REGULAR(0x1p127L, 0x3p-148L, 0x1p-147L);
@@ -224,20 +225,20 @@ public:
     TEST_REGULAR(-0x1p127L, 0x3p-126L, -0x1p-125L);
     TEST_REGULAR(-0x1p127L, -0x3p-126L, -0x1p-125L);
 
-    if constexpr (sizeof(T) >= sizeof(double)) {
-      TEST_REGULAR(0x1p1023L, 0x3p-1074L, 0x1p-1073L);
-      TEST_REGULAR(0x1p1023L, -0x3p-1074L, 0x1p-1073L);
-      TEST_REGULAR(0x1p1023L, 0x3p-1073L, 0x1p-1073L);
-      TEST_REGULAR(0x1p1023L, -0x3p-1073L, 0x1p-1073L);
-      TEST_REGULAR(0x1p1023L, 0x3p-1022L, 0x1p-1021L);
-      TEST_REGULAR(0x1p1023L, -0x3p-1022L, 0x1p-1021L);
-      TEST_REGULAR(-0x1p1023L, 0x3p-1074L, -0x1p-1073L);
-      TEST_REGULAR(-0x1p1023L, -0x3p-1074L, -0x1p-1073L);
-      TEST_REGULAR(-0x1p1023L, 0x3p-1073L, -0x1p-1073L);
-      TEST_REGULAR(-0x1p1023L, -0x3p-1073L, -0x1p-1073L);
-      TEST_REGULAR(-0x1p1023L, 0x3p-1022L, -0x1p-1021L);
-      TEST_REGULAR(-0x1p1023L, -0x3p-1022L, -0x1p-1021L);
-    }
+    if constexpr (sizeof(T) < sizeof(double))
+      return;
+    TEST_REGULAR(0x1p1023L, 0x3p-1074L, 0x1p-1073L);
+    TEST_REGULAR(0x1p1023L, -0x3p-1074L, 0x1p-1073L);
+    TEST_REGULAR(0x1p1023L, 0x3p-1073L, 0x1p-1073L);
+    TEST_REGULAR(0x1p1023L, -0x3p-1073L, 0x1p-1073L);
+    TEST_REGULAR(0x1p1023L, 0x3p-1022L, 0x1p-1021L);
+    TEST_REGULAR(0x1p1023L, -0x3p-1022L, 0x1p-1021L);
+    TEST_REGULAR(-0x1p1023L, 0x3p-1074L, -0x1p-1073L);
+    TEST_REGULAR(-0x1p1023L, -0x3p-1074L, -0x1p-1073L);
+    TEST_REGULAR(-0x1p1023L, 0x3p-1073L, -0x1p-1073L);
+    TEST_REGULAR(-0x1p1023L, -0x3p-1073L, -0x1p-1073L);
+    TEST_REGULAR(-0x1p1023L, 0x3p-1022L, -0x1p-1021L);
+    TEST_REGULAR(-0x1p1023L, -0x3p-1022L, -0x1p-1021L);
   }
 };
 
diff --git a/libc/test/src/math/smoke/fmodf16_test.cpp b/libc/test/src/math/smoke/fmodf16_test.cpp
new file mode 100644
index 0000000..9a48c5a
--- /dev/null
+++ b/libc/test/src/math/smoke/fmodf16_test.cpp
@@ -0,0 +1,13 @@
+//===-- Unittests for fmodf16 ---------------------------------------------===//
+//
+// 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 "FModTest.h"
+
+#include "src/math/fmodf16.h"
+
+LIST_FMOD_TESTS(float16, LIBC_NAMESPACE::fmodf16)