riscv/musl - musl - Gitea: Git with a cup of tea

Commit Graph

Author	SHA1	Message	Date
nsz	75483499da	new modff.c code, fix nan handling in modfl	14 years ago
nsz	2786c7d216	use scalbn or 2.0 instead of ldexp, fix fmal Some code assumed ldexp(x, 1) is faster than 2.0x, but ldexp is a wrapper around scalbn which uses multiplications inside, so this optimization is wrong. This commit also fixes fmal which accidentally used ldexp instead of ldexpl loosing precision. There are various additional changes from the work-in-progress const cleanups.	14 years ago
nsz	01fdfd491b	fix long double const workaround in cbrtl	14 years ago
nsz	2e8c8fbe7d	don't inline __rem_pio2l so the code size is smaller	14 years ago
nsz	c3587effe2	minor fix in __tanl (get sign properly)	14 years ago
nsz	eca1c35e5b	remove long double const workarounds Some long double consts were stored in two doubles as a workaround for x86_64 and i386 with the following comment: /* Long double constants are slow on these arches, and broken on i386. */ This is most likely old gcc bug related to the default x87 fpu precision setting (it's double instead of double extended on BSD).	14 years ago
nsz	9a810cb685	fix erfl wrapper for long double==double case	14 years ago
Rich Felker	b04b588791	asm for log1p	14 years ago
Rich Felker	9d82a15e15	asm for log2	14 years ago
Rich Felker	27deb53889	asm for remquo this could perhaps use some additional testing for corner cases, but it seems to be correct.	14 years ago
Rich Felker	02db27d9de	optimize exponential asm for i386 up to 30% faster exp2 by avoiding slow frndint and fscale functions. expm1 also takes a much more direct path for small arguments (the expected usage case).	14 years ago
Rich Felker	be5b01f855	fix broken modf family functions	14 years ago
Rich Felker	1bf4dad327	asm for modf functions	14 years ago
nsz	8051e08e10	simplify scalbn*.c implementations The old scalbn.c was wrong and slow, the new one is just slow. (scalbn(0x1p+1023,-2097) should give 0x1p-1074, but the old code gave 0)	14 years ago
Rich Felker	0b70a1e9a9	asm for floor/ceil/trunc	14 years ago
Rich Felker	58ff9e8eaf	asm for scalbn family unlike some implementations, these functions perform the equivalent of gcc's -ffloat-store on the result before returning. this is necessary to raise underflow/overflow/inexact exceptions, perform the correct rounding with denormals, etc.	14 years ago
Rich Felker	bc33e61704	asm for inverse trig functions unlike trig functions, these are easy to do in asm because they do not involve (arbitrary-precision) argument reduction. fpatan automatically takes care of domain issues, and in asin and acos, fsqrt takes care of them for us.	14 years ago
Rich Felker	495a52ae7b	asm for log functions	14 years ago
Rich Felker	aa1b4dff45	fix broken exponential asm infinities were getting converted into nans. the new code simply tests for infinity and replaces it with a large magnitude value of the same sign. also, the fcomi instruction is apparently not part of the i387 instruction set, so avoid using it.	14 years ago
Rich Felker	37eb14dd2b	asm for lrint family on i386	14 years ago
Rich Felker	6f26cf3dac	asm exponential functions for i386	14 years ago
nsz	682e471400	remove unnecessary TODO comments from fma.c	14 years ago
nsz	d09a83f613	fmal bug fix: nan input should not raise exception	14 years ago
nsz	b1cbd70743	add fma implementation for x86 correctly rounded double precision fma using extended precision arithmetics for ld80 systems (x87)	14 years ago
Rich Felker	b935147761	assembly optimizations for fmod/remainder functions	14 years ago
Rich Felker	8d9e948652	asm versions of some simple math functions for i386 and x86_64 these are functions that have direct fpu approaches to implementation without problematic exception or rounding issues. x86_64 lacks float/double versions because i'm unfamiliar with the necessary sse code for performing these operations.	14 years ago
nsz	afad262440	simplify lround and llround functions Simple wrappers around round is enough because spurious inexact exception is allowed.	14 years ago
nsz	65db00983f	make lrint and llrint functions work without fenv support	14 years ago
nsz	9b6899f2c5	faster lrint and llrint functions A faster workaround for spurious inexact exceptions when the result cannot be represented. The old code actually could be wrong, because gcc reordered the integer conversion and the exception check.	14 years ago
Rich Felker	9e2a895aaa	fix loads of missing const in new libm, and some global vars (?!) in powl	14 years ago
Rich Felker	b60053e762	try fixing/optimizing x86_64 fenv exception code untested; may need followup-fixes.	14 years ago
Rich Felker	316e024f63	optimize x86 feclearexcept if all exception flags will be cleared, we can avoid the expensive store/reload of the environment and just use the fnclex instruction.	14 years ago
Rich Felker	9cb6878e74	fix x86_64 fe[gs]etround, analogous to nsz's x86 changes	14 years ago
Rich Felker	d5e576c752	minor 387 fenv optimizations	14 years ago
nsz	88cfaf8a14	fix i386 fegetround and make fesetround faster Note that the new fesetround has slightly different semantics: Storing the floating-point environment with fnstenv makes the next fldenv (or fldcw) "non-signaling", so unmasked and pending exceptions does not invoke the exception handler. (These are rare since exceptions are handled immediately and by default all exceptions are masked anyway. But if one manually unmasks an exception in the control word then either sets the corresponding exception flag in the status word or the execution of an exception raising floating-point operation gets interrupted then it may happen). So the old implementation did not trap in some rare cases where the new implementation traps. However POSIX does not specify anything like the x87 exception handling traps and the fnstenv/fldenv pair is significantly slower than the fnstcw/fldcw pair (new code is about 5x faster here and it's dominated by the function call overhead).	14 years ago
Rich Felker	d3fc724759	one more fenv availability issue: lround	14 years ago
Rich Felker	2e77dc13f8	make fma and lrint functions build without full fenv support this is necessary to support archs where fenv is incomplete or unavailable (presently arm). fma, fmal, and the lrint family should work perfectly fine with this change; fmaf is slightly broken with respect to rounding as it depends on non-default rounding modes to do its work.	14 years ago
Rich Felker	8c071f872b	other side of the signgam namespace fix: use the internal name	14 years ago
Rich Felker	1a3dce4184	make signgam a weak alias for an internal symbol otherwise, the standard C lgamma function will clobber a symbol in the namespace reserved for the application.	14 years ago
Rich Felker	de7db6e927	fix namespace issues for lgamma, etc. standard functions cannot depend on nonstandard symbols	14 years ago
Rich Felker	2cbb24bba3	remove junk sincos implementations in preparation to merge nsz's real ones	14 years ago
Rich Felker	9d507419db	remove special nan handling from x86 sqrt asm a double precision nan, when converted to extended (80-bit) precision, will never end in 0x400, since the corresponding bits do not exist in the original double precision value. thus there's no need to waste time and code size on this check.	14 years ago
Rich Felker	1295848efb	simplify nan check in sqrt (x86 asm); result of sqrt is never negative	14 years ago
nsz	40305f74bd	in math.h make lgamma_r and non-double bessel _GNU_SOURCE only long double and float bessel functions are no longer xsi extensions	14 years ago
nsz	0144b45b71	efficient sincos based on sin and cos	14 years ago
Rich Felker	5657cc58e5	implement sincosf and sincosl functions; add prototypes presumably broken gcc may generate calls to these, and it's said that ffmpeg makes use of sincosf.	14 years ago
Rich Felker	46702f68f9	avoid changing NaNs in sqrt (x86 asm) to satisfy c99 f.9 recommendation	14 years ago
Rich Felker	809556e60a	correctly rounded sqrt() asm for x86 (i387) the fsqrt opcode is correctly rounded, but only in the fpu's selected precision mode, which is 80-bit extended precision. to get a correctly rounded double precision output, we check for the only corner cases where two-step rounding could give different results than one-step (extended-precision mantissa ending in 0x400) and adjust the mantissa slightly in the opposite direction of the rounding which the fpu already did (reported in the c1 flag of the fpu status word). this should have near-zero cost in the non-corner cases and at worst very low cost. note that in order for sqrt() to get used when compiling with gcc, the broken, non-conformant builtin sqrt must be disabled.	14 years ago
Rich Felker	e0a54e6725	correct rounding for i387 sqrtf function	14 years ago
nsz	32ca5ef3ff	math cleanup: use 1.0f instead of 1.0F	14 years ago

1 2 3 4 5 ...

680 Commits (75483499dad38b97f5dabb710635e6a8bbbb1c84)