MahnoKropotkinvich/tinycc
1
0
Fork 0
mirror of git://repo.or.cz/tinycc.git synced 2026-11-04 21:26:36 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

win32: simplify tcc 0.9.27 self-host (part 2 of 2)
Some checks failed
build and test / test-x86_64-linux (push) Has been cancelled
Details
build and test / test-x86_64-osx (push) Has been cancelled
Details
build and test / test-aarch64-osx (push) Has been cancelled
Details
build and test / test-x86_64-win32 (push) Has been cancelled
Details
build and test / test-i386-win32 (push) Has been cancelled
Details
build and test / test-arm64-win32 (push) Has been cancelled
Details
build and test / test-armv7-linux (push) Has been cancelled
Details
build and test / test-aarch64-linux (push) Has been cancelled
Details
build and test / test-riscv64-linux (push) Has been cancelled
Details
build and test / test-i386-linux (push) Has been cancelled
Details

Browse Source 
tcc 0.9.27 does have a usable ldexp - but not ldexpl. however,
on windows these functions are identical, so map ldexpl to ldexp
and remove the MUSL-implementation which was copied from tcc_libm.h.
This commit is contained in:
Avi Halachmi (:avih) 2026-07-14 16:38:29 +03:00
parent add111e6fe
commit d9d02c5640
1 changed files with 8 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
tccpp.c
View File

@ -21,47 +21,19 @@
#define USING_GLOBALS
#include "tcc.h"
#if defined(_WIN32) && defined (__TINYC__)
#if defined(_WIN32) && defined(__TINYC__)
/* allow self-host build with tcc 0.9.27 - doesn't have ldexpl in tcc_libm.h .
*
* we can't feature-test whether we have ldexpl, and our ./configure also
* can't perform such tests (and it's also not used by win32/build-tcc.bat),
* so define it unconditionally if we're compiling using tcc on windows
* (otherwise the platform's math.h should already have ldexpl).
* in tcc 0.9.27 both ldexp and ldexpl are declared in win32/include/math.h,
* but only ldexp can be linked - via win32/lib/msvcrt.def. we can't test
* whether we have ldexpl or not, so map ldexpl to ldexp unconditionally.
*
* we could instead include win32/include/math.h (specifically tcc_libm.h),
* but this requires working around include guards and redefinitions,
* so KISS - copy the tcc_libm.h definitions here with a different name.
* note that ldexpl takes long double while ldexp takes double, however, on
* windows these types are identical, and current ldexpl in tcc_libm.h also
* uses the "normal" double scalbn - just like ldexp, so do the same here.
*/
#undef ldexpl
#define ldexpl ldexpl_local
/* MUSL scalbn - (renamed) copied from win32/include/tcc/tcc_libm.h */
static double scalbn_local(double x, int n) {
union {double f; uint64_t i;} u;
if (n > 1023) {
x *= 0x1p1023, n -= 1023;
if (n > 1023) {
x *= 0x1p1023, n -= 1023;
if (n > 1023) n = 1023;
}
} else if (n < -1022) {
x *= 0x1p-1022 * 0x1p53, n += 1022 - 53;
if (n < -1022) {
x *= 0x1p-1022 * 0x1p53, n += 1022 - 53;
if (n < -1022) n = -1022;
}
}
u.i = (0x3ffull + n) << 52;
return x * u.f;
}
/* (windows long double is the same type as double - 64bit) */
/* (renamed) copied from win32/include/tcc/tcc_libm.h */
static long double ldexpl_local(long double x, int exp) {
return scalbn_local(x, exp);
}
#define ldexpl ldexp
#endif
/* #define to 1 to enable (see parse_pp_string()) */