diff options
| author | sebres <sebres@users.sourceforge.net> | 2019-06-25 10:56:14 (GMT) |
|---|---|---|
| committer | sebres <sebres@users.sourceforge.net> | 2019-06-25 10:56:14 (GMT) |
| commit | 5fe976a533abc428b1cf07bf2c5daa15b72416e5 (patch) | |
| tree | af98fb28621303338f02c11f7c5bec67bb282fbf | |
| parent | 9fb98b2d11c075c66b5cb297bcd700cf26c81eac (diff) | |
| parent | f310332a6ba0427e38582673f4d3fbe4b3d9eb5c (diff) | |
| download | tcl-5fe976a533abc428b1cf07bf2c5daa15b72416e5.zip tcl-5fe976a533abc428b1cf07bf2c5daa15b72416e5.tar.gz tcl-5fe976a533abc428b1cf07bf2c5daa15b72416e5.tar.bz2 | |
merge and rewrite fpclassify-mingw-x86: new TCL_FPCLASSIFY_MODE mode (3) for __builtin_fpclassify, actually not used directly
| -rw-r--r-- | generic/tclBasic.c | 26 |
1 files changed, 23 insertions, 3 deletions
diff --git a/generic/tclBasic.c b/generic/tclBasic.c index 41df33c..b67648c 100644 --- a/generic/tclBasic.c +++ b/generic/tclBasic.c @@ -24,7 +24,22 @@ #include <math.h> #include <assert.h> +/* + * TCL_FPCLASSIFY_MODE: + * 0 - fpclassify + * 1 - _fpclass + * 2 - simulate + * 3 - __builtin_fpclassify + */ + +#define TCL_FPCLASSIFY_MODE 3 +#warning mode: TCL_FPCLASSIFY_MODE + #ifndef TCL_FPCLASSIFY_MODE +/* + * MINGW x86 (tested up to gcc 8.1) seems to have a bug in fpclassify, + * [fpclassify 1e-314], x86 => normal, x64 => subnormal, so switch to _fpclass + */ # if ( defined(__MINGW32__) && defined(_X86_) ) /* mingw 32-bit */ # define TCL_FPCLASSIFY_MODE 1 # elif defined(fpclassify) /* fpclassify */ @@ -35,6 +50,7 @@ # else /* !fpclassify && !_fpclass (older MSVC), simulate */ # define TCL_FPCLASSIFY_MODE 2 # endif /* !fpclassify */ +/* actually there is no fallback to builtin fpclassify */ #endif /* !TCL_FPCLASSIFY_MODE */ @@ -8366,7 +8382,9 @@ ClassifyDouble( # define FP_SUBNORMAL 5 /* Value has lost accuracy */ #endif -# if TCL_FPCLASSIFY_MODE == 2 +# if TCL_FPCLASSIFY_MODE == 3 + return __builtin_fpclassify(FP_NAN, FP_INFINITE, FP_NORMAL, FP_SUBNORMAL, FP_ZERO, d); +# elif TCL_FPCLASSIFY_MODE == 2 /* * We assume this hack is only needed on little-endian systems. * Specifically, x86 running Windows. It's fairly easy to enable for @@ -8452,8 +8470,10 @@ ClassifyDouble( case _FPCLASS_SNAN: return FP_NAN; } -# endif /* REQUIRE_ANCIENT_WIN32_FPCLASSIFY_HACK */ -#endif /* fpclassify */ +# else /* unknown TCL_FPCLASSIFY_MODE */ +# error "unknown or unexpected TCL_FPCLASSIFY_MODE" +# endif /* TCL_FPCLASSIFY_MODE */ +#endif /* !fpclassify */ } static int |