[libc] Generate math.h instead of the static file it is currently.

[sivachandra]

Summary

Just enough to help a follow up patch adding cosf and sinf has been
added.

[abrachet]

libc/config/linux/api.td

113

__builtin_nanf() takes a const char * argument. I have no idea why. https://github.com/llvm/llvm-project/blob/master/clang/lib/AST/ExprConstant.cpp#L12816

libc/spec/stdc.td

181–185

Should we include these even though we have no implementation of them?

[sivachandra]

• Add "" arg to __builtin_nanf.
• Remove cosf and sinf from linux API; will add them in the patch adding their implementation.

[sivachandra]

libc/spec/stdc.td

181–185

Good point. Will add them in the patch adding their implementation.

[PaulkaToast]

libc/config/linux/api.td

79

I think some of this might be gcc/glibc specific stuff and should be thought over. See:

• https://gcc.gnu.org/legacy-ml/gcc-patches/2014-11/msg01079.html
• https://github.com/bminor/glibc/commit/3c7d03129498e7426855b5d4cdd5b7109ecc2172

0 is actually not a value allowed by the C2011 standard for math_errhandling. See this excerpt:

The macros
MATH_ERRNO
MATH_ERREXCEPT
expand to the integer constants 1 and 2, respectively; the macro
math_errhandling expands to an expression that has type int and the value
MATH_ERRNO, MATH_ERREXCEPT, or the bitwise OR of both.

For all functions, a domain error occurs if an input argument is outside the domain over
which the mathematical function is defined. The description of each function lists any
required domain errors; an implementation may define additional domain errors, provided
that such errors are consistent with the mathematical definition of the function.228) On a
domain error, the function returns an implementation-defined value; if the integer
expression math_errhandling & MATH_ERRNO is nonzero, the integer expression
errno acquires the value EDOM; if the integer expression math_errhandling & MATH_ERREXCEPT
is nonzero, the ‘‘invalid’’ floating-point exception is raised.

__NO_MATH_ERRNO__ in particular is very gcc specific and clang doesn't set it regardless of -ffast-math or -fno-math-errno. This leads to this case where sqrt(-1.0) will not set errno despite us setting math_errhandling to MATH_ERRNO | MATH_ERREXCEPT: https://godbolt.org/z/ZhMT_z

For now it might be better just to define this conservatively as

#define math_errhandling (MATH_ERREXCEPT)

since it seems a little tough to determine whether math functions will set errno across both gcc and clang, but it is known that they will set the floating point exceptions.

113

It corresponds to the tagp argument for nan:

These functions shall return a quiet NaN, if available, with content indicated through tagp.

This allows the creation of differently "tagged" NaN values (the tags are in the trailing bits). Both musl and glibc use "" as the tag so this looks good.

[sivachandra]

libc/config/linux/api.td

79

There is a lot of information here so let me try to answer in parts.

Before I start, I would like to point out that a libc can itself be compiled with different values of math_errhandling. So, the options -ffast-math and/or -fno-math-errno are only applicable to your translation unit and not to your entire program in case of your godbolt examples.

About -ffast-math and -fno-math-errno: These are compiler options wherein the compilers are saying that irrespective of what the libc does, we will try to make your code run faster. The speed might come at the cost of non-compliance with standards. For example, -ffast-math implies -fno-math-errno and -funsafe-math-optimizations. The documentation for -funsafe-math-optimizations says, "Allow optimizations for floating-point arithmetic that (a) assume that arguments and results are valid and (b) may violate IEEE or ANSI standards." So, under -ffast-math, compilers *might* generate code which does not give any error information. Some instructions generated by the compilers can still set the floating point exceptions. But, in general, user code cannot except that any error information will be available. The way to surface this to the user code is to set math_errhandling to 0 under __FAST_MATH__.

Coming to the differences between clang and gcc you are seeing in your sqrt examples, it is because gcc and clang use different interpretations of the options and has nothing to do with the under lying libc they use if at all. After all, -ffast-math and -fno-math-errno are compiler options, and also best effort options. If you notice, both clang and gcc do not call sqrt at all when -ffast-math is used. Likewise, gcc chooses to call sqrt with or without -fno-math-errno. On the other hand, clang does not call sqrt even when -fno-math-errno is used. The important point is that these differences do not reflect or dictate what the underlying libc does or should do.

libc/config/linux/api.td

 
 def AssertAPI : PublicAPI<"assert.h"> {
   let Macros = [
     AssertMacro,
     StaticAssertMacro,
   ];
 }
 
+def MathErrHandlingMacro : MacroDef<"math_errhandling"> {
+  let Defn = [{
+    #ifndef math_errhandling
+    #ifdef __FAST_MATH__
+    #define math_errhandling 0
+    #elif defined __NO_MATH_ERRNO__
+    #define math_errhandling (MATH_ERREXCEPT)
+    #else
+    #define math_errhandling (MATH_ERRNO | MATH_ERREXCEPT)
+    #endif
+    #endif // math_errhandling not defined
+  }];
+}
+
+def IsFiniteMacro : MacroDef<"isfinite"> {
+  let Defn = [{
+    #define isfinite(x) __builtin_isfinite(x)
+  }];
+}
+
+def IsInfMacro : MacroDef<"isinf"> {
+  let Defn = [{
+    #define isinf(x) __builtin_isinf(x)
+  }];
+}
+
+def IsNanMacro : MacroDef<"isnan"> {
+  let Defn = [{
+    #define isnan(x) __builtin_isnan(x)
+  }];
+}
+
 def MathAPI : PublicAPI<"math.h"> {
+  let Macros = [
+    SimpleMacroDef<"MATH_ERRNO", "1">,
+    SimpleMacroDef<"MATH_ERREXCEPT", "2">,
+    MathErrHandlingMacro,
+
+    SimpleMacroDef<"INFINITY", "__builtin_inff()">,
+    SimpleMacroDef<"NAN", "__builtin_nanf(\"\")">,
+
+    IsFiniteMacro,
+    IsInfMacro,
+    IsNanMacro,
+  ];
   let Functions = [
-   "acos",
-   "acosl",
+   "round",
   ];
 }
 
 def StringAPI : PublicAPI<"string.h"> {
   let Functions = [
     "memcpy",
     "memmove",
     "memcmp",

libc/include/CMakeLists.txt

 add_header(
   ctype_h
   HDR
    ctype.h
   DEPENDS
     llvm_libc_common_h
 )
 
-add_header(
+add_gen_header(
   math_h
-  HDR
-    math.h
+  DEF_FILE math.h.def
+  GEN_HDR math.h
   DEPENDS
     llvm_libc_common_h
 )
 
 add_gen_header(
   assert_h
   DEF_FILE assert.h.def
   GEN_HDR assert.h

libc/include/math.h

-//===----------------- C standard library header math.h -----------------*-===//
-//
-// 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_MATH_H
-#define LLVM_LIBC_MATH_H
-
-#include <__llvm-libc-common.h>
-
-__BEGIN_C_DECLS
-
-double acos(double);
-
-float acosf(float);
-
-long double acosl(long double);
-
-double asin(double);
-
-float asinf(float);
-
-long double asinl(long double);
-
-double atan(double);
-
-float atanf(float);
-
-long double atanl(long double);
-
-double atan2(double, double);
-
-float atan2f(float, float);
-
-long double atan2l(long double, long double);
-
-double cos(double);
-
-float cosf(float);
-
-long double cosl(long double);
-
-double sin(double);
-
-float sinf(float);
-
-long double sinl(long double);
-
-double tan(double);
-
-float tanf(float);
-
-long double tanl(long double);
-
-double acosh(double);
-
-float acoshf(float);
-
-long double acoshl(long double);
-
-double asinh(double);
-
-float asinhf(float);
-
-long double asinhl(long double);
-
-double atanh(double);
-
-float atanhf(float);
-
-long double atanhl(long double);
-
-double cosh(double);
-
-float coshf(float);
-
-long double coshl(long double);
-
-double sinh(double);
-
-float sinhf(float);
-
-long double sinhl(long double);
-
-double tanh(double);
-
-float tanhf(float);
-
-long double tanhl(long double);
-
-double exp(double);
-
-float expf(float);
-
-long double expl(long double);
-
-double exp2(double);
-
-float exp2f(float);
-
-long double exp2l(long double);
-
-double expm1(double);
-
-float expm1f(float);
-
-long double expm1l(long double);
-
-double frexp(double, int);
-
-float frexpf(float, int);
-
-long double frexpl(long double, int);
-
-int ilogb(double);
-
-int ilogbf(float);
-
-int ilogbl(long double);
-
-double ldexp(double, int);
-
-float ldexpf(float, int);
-
-long double ldexpl(long double, int);
-
-double log(double);
-
-float logf(float);
-
-long double logl(long double);
-
-double log10(double);
-
-float log10f(float);
-
-long double log10l(long double);
-
-double log1p(double);
-
-float log1pf(float);
-
-long double log1pl(long double);
-
-double log2(double);
-
-float log2f(float);
-
-long double log2l(long double);
-
-double logb(double);
-
-float logbf(float);
-
-long double logbl(long double);
-
-double modf(double, double);
-
-float modff(float, float);
-
-long double modfl(long double, long double);
-
-double scalbn(double, int);
-
-float scalbnf(float, int);
-
-long double scalbnl(long double, int);
-
-double scalbln(double, long int);
-
-float scalblnf(float, long int);
-
-long double scalblnl(long double, long int);
-
-double cbrt(double);
-
-float cbrtf(float);
-
-long double cbrtl(long double);
-
-double fabs(double);
-
-float fabsf(float);
-
-long double fabsl(long double);
-
-double hypot(double, double);
-
-float hypotf(float, float);
-
-long double hypotl(long double, long double);
-
-double pow(double, double);
-
-float powf(float, float);
-
-long double powl(long double, long double);
-
-double sqrt(double);
-
-float sqrtf(float);
-
-long double sqrtl(long double);
-
-double erf(double);
-
-float erff(float);
-
-long double erfl(long double);
-
-double erfc(double);
-
-float erfcf(float);
-
-long double erfcl(long double);
-
-double lgamma(double);
-
-float lgammaf(float);
-
-long double lgammal(long double);
-
-double tgamma(double);
-
-float tgammaf(float);
-
-long double tgammal(long double);
-
-double ceil(double);
-
-float ceilf(float);
-
-long double ceill(long double);
-
-double floor(double);
-
-float floorf(float);
-
-long double floorl(long double);
-
-double nearbyint(double);
-
-float nearbyintf(float);
-
-long double nearbyintl(long double);
-
-double rint(double);
-
-float rintf(float);
-
-long double rintl(long double);
-
-long int lrint(double);
-
-long int lrintf(float);
-
-long int lrintl(long double);
-
-long long int llrint(double);
-
-long long int llrintf(float);
-
-long long int llrintl(long double);
-
-double round(double);
-
-float roundf(float);
-
-long double roundl(long double);
-
-long int lround(double);
-
-long int lroundf(float);
-
-long int lroundl(long double);
-
-long long int llround(double);
-
-long long int llroundf(float);
-
-long long int llroundl(long double);
-
-double trunc(double);
-
-float truncf(float);
-
-long double truncl(long double);
-
-double fmod(double, double);
-
-float fmodf(float, float);
-
-long double fmodl(long double, long double);
-
-double remainder(double, double);
-
-float remainderf(float, float);
-
-long double remainderl(long double, long double);
-
-double remquo(double, double, int);
-
-float remquof(float, float, int);
-
-long double remquol(long double, long double, int);
-
-double copysign(double, double);
-
-float copysignf(float, float);
-
-long double copysignl(long double, long double);
-
-double nan(const char);
-
-float nanf(const char);
-
-long double nanl(const char);
-
-double nextafter(double, double);
-
-float nextafterf(float, float);
-
-long double nextafterl(long double, long double);
-
-double nexttoward(double, long double);
-
-float nexttowardf(float, long double);
-
-long double nexttowardl(long double, long double);
-
-double fdim(double, double);
-
-float fdimf(float, float);
-
-long double fdiml(long double, long double);
-
-double fmax(double, double);
-
-double fmaxf(double, double);
-
-double fmaxl(double, double);
-
-double fmin(double, double);
-
-float fminf(float, float);
-
-long double fminl(long double, long double);
-
-double fma(double, double, double);
-
-float fmaf(float, float, float);
-
-long double fmal(long double, long double, long double);
-
-__END_C_DECLS
-
-#endif // LLVM_LIBC_MATH_H

libc/include/math.h.def

+//===----------------- C standard library header math.h -----------------*-===//
+//
+// 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_MATH_H
+#define LLVM_LIBC_MATH_H
+
+#include <__llvm-libc-common.h>
+
+%%public_api()
+
+#endif // LLVM_LIBC_MATH_H

libc/spec/stdc.td

HeaderSpec String = HeaderSpec<
               RetValSpec<SizeTType>,
               [ArgSpec<ConstCharPtr>]
           >,
       ]
   >;
 
   HeaderSpec Math = HeaderSpec<
       "math.h",
-      [], // Macros
+      [
+          Macro<"MATH_ERRNO">,
+          Macro<"MATH_ERREXCEPT">,
+          Macro<"math_errhandling">,
+
+          Macro<"INFINITY">,
+          Macro<"NAN">,
+
+          Macro<"isfinite">,
+          Macro<"isinf">,
+          Macro<"isnan">,
+      ],
       [
           NamedType<"float_t">,
           NamedType<"double_t">,
       ],
       [], // Enumerations
       [
-          FunctionSpec<"acos", RetValSpec<DoubleType>, [ArgSpec<DoubleType>]>,
-          FunctionSpec<"acosl", RetValSpec<LongDoubleType>, [ArgSpec<LongDoubleType>]>,
+          FunctionSpec<"cosf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
+          FunctionSpec<"sinf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
+
+          FunctionSpec<"round", RetValSpec<DoubleType>, [ArgSpec<DoubleType>]>,
       ]
   >;
 
   HeaderSpec StdIO = HeaderSpec<
       "stdio.h",
       [], // Macros
       [ // Types
           SizeTType,
       ],