Skip to content

Commit 1fa2b37

Browse files
author
matu3ba
authored
compiler_rt: document math functions (#14552)
This documents the currently implemented math functions in alphabetical order and sorts the remaining items.
1 parent d4bd1b1 commit 1fa2b37

File tree

1 file changed

+146
-10
lines changed

1 file changed

+146
-10
lines changed

‎lib/compiler_rt/README.md‎

Lines changed: 146 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -552,20 +552,156 @@ TODO brief explanation + implementation
552552
| ------ | ------------- | --------- | --------- | --------- | -------------------------- |
553553
| | | | | | **Fixed-Point Fractional** |
554554

555-
Further content:
555+
Math functions according to C99 with gnu extension sincos. f16, f80 and f128 functions
556+
are additionally supported by Zig, but not part of C standard. Alphabetically sorted.
557+
558+
| Done | Name | a | b | Out | Comment |
559+
| ---- | ------- | --------- | --------- | --------- | -------------------------- |
560+
| ✓ | __ceilh | f16 | ∅ | f16 |smallest integer value not less than a|
561+
| ✓ | ceilf | f32 | ∅ | f32 |If a is integer, +-0, +-NaN, or +-infinite, a itself is returned.|
562+
| ✓ | ceil | f64 | ∅ | f64 | .. |
563+
| ✓ | __ceilx | f80 | ∅ | f80 | |
564+
| ✓ | ceilf128 | f128 | ∅ | f128 | .. PPC |
565+
| ✓ | ceilq | f128 | ∅ | f128 | .. |
566+
| ✓ | ceill |long double| ∅ |long double| .. |
567+
| ✓ | __cosh | f16 | ∅ | f16 | `cos(a)=(e^(ia)+e^(-ia))/2`|
568+
| ✓ | cosf | f32 | ∅ | f32 | .. |
569+
| ✓ | cos | f64 | ∅ | f64 | .. |
570+
| ✓ | __cosx | f80 | ∅ | f80 | .. |
571+
| ✓ | cosf128 | f128 | ∅ | f128 | .. |
572+
| ✓ | cosq | f128 | ∅ | f128 | .. PPC |
573+
| ✓ | cosl |long double| ∅ |long double| .. |
574+
| ✓ | __exph | f16 | ∅ | f16 | `e^a` with e base of natural logarithms|
575+
| ✓ | expf | f32 | ∅ | f32 | .. |
576+
| ✓ | exp | f64 | ∅ | f64 | .. |
577+
| ✓ | __expx | f80 | ∅ | f80 | .. |
578+
| ✓ | expf128 | f128 | ∅ | f128 | .. |
579+
| ✓ | expq | f128 | ∅ | f128 | .. PPC |
580+
| ✓ | expl |long double| ∅ |long double| .. |
581+
| ✓ | __exp2h | f16 | ∅ | f16 | `2^a` |
582+
| ✓ | exp2f | f32 | ∅ | f32 | .. |
583+
| ✓ | exp2 | f64 | ∅ | f64 | .. |
584+
| ✓ | __exp2x | f80 | ∅ | f80 | .. |
585+
| ✓ | exp2f128 | f128 | ∅ | f128 | .. |
586+
| ✓ | exp2q | f128 | ∅ | f128 | .. PPC |
587+
| ✓ | exp2l |long double| ∅ |long double| .. |
588+
| ✓ | __fabsh | f16 | ∅ | f16 | absolute value of a |
589+
| ✓ | fabsf | f32 | ∅ | f32 | .. |
590+
| ✓ | fabs | f64 | ∅ | f64 | .. |
591+
| ✓ | __fabsx | f80 | ∅ | f80 | .. |
592+
| ✓ | fabsf128 | f128 | ∅ | f128 | .. |
593+
| ✓ | fabsq | f128 | ∅ | f128 | .. PPC |
594+
| ✓ | fabsl |long double| ∅ |long double| .. |
595+
| ✓ | __floorh | f16 | ∅ | f16 |largest integer value not greater than a|
596+
| ✓ | floorf | f32 | ∅ | f32 |If a is integer, +-0, +-NaN, or +-infinite, a itself is returned.|
597+
| ✓ | floor | f64 | ∅ | f64 | .. |
598+
| ✓ | __floorx | f80 | ∅ | f80 | .. |
599+
| ✓ | floorf128 | f128 | ∅ | f128 | .. |
600+
| ✓ | floorq | f128 | ∅ | f128 | .. PPC |
601+
| ✓ | floorl |long double| ∅ |long double| .. |
602+
| ✓ | __fmah | f16 | 2xf16 | f16 | args a,b,c result `(a*b)+c`|
603+
| ✓ | fmaf | f32 | 2xf32 | f32 |Fused multiply-add for hardware acceleration|
604+
| ✓ | fma | f64 | 2xf64 | f64 | .. |
605+
| ✓ | __fmax | f80 | 2xf80 | f80 | .. |
606+
| ✓ | fmaf128 | f128 | 2xf128 | f128 | .. |
607+
| ✓ | fmaq | f128 | 2xf128 | f128 | .. PPC |
608+
| ✓ | fmal |long double|2xlong double|long double| .. |
609+
| ✓ | __fmaxh | f16 | f16 | f16 | larger value of a,b |
610+
| ✓ | fmaxf | f32 | f32 | f32 | .. |
611+
| ✓ | fmax | f64 | f64 | f64 | .. |
612+
| ✓ | __fmaxx | f80 | f80 | f80 | .. |
613+
| ✓ | fmaxf128 | f128 | f128 | f128 | .. |
614+
| ✓ | fmaxq | f128 | f128 | f128 | .. PPC |
615+
| ✓ | fmaxl |long double|long double|long double| .. |
616+
| ✓ | __fminh | f16 | f16 | f16 | smaller value of a,b |
617+
| ✓ | fminf | f32 | f32 | f32 | .. |
618+
| ✓ | fmin | f64 | f64 | f64 | .. |
619+
| ✓ | __fminx | f80 | f80 | f80 | .. |
620+
| ✓ | fminf128 | f128 | f128 | f128 | .. |
621+
| ✓ | fminq | f128 | f128 | f128 | .. PPC |
622+
| ✓ | fminl |long double|long double|long double| .. |
623+
| ✓ | __fmodh | f16 | f16 | f16 |floating-point remainder of division a/b|
624+
| ✓ | fmodf | f32 | f32 | f32 | .. |
625+
| ✓ | fmod | f64 | f64 | f64 | .. |
626+
| ✓ | __fmodx | f80 | f80 | f80 | .. |
627+
| ✓ | fmodf128 | f128 | f128 | f128 | .. |
628+
| ✓ | fmodq | f128 | f128 | f128 | .. PPC |
629+
| ✓ | fmodl |long double|long double|long double| .. |
630+
| ✓ | __logh | f16 | ∅ | f16 |natural (base-e) logarithm of a|
631+
| ✓ | logf | f32 | ∅ | f32 | .. |
632+
| ✓ | log | f64 | ∅ | f64 | .. |
633+
| ✓ | __logx | f80 | ∅ | f80 | .. |
634+
| ✓ | logf128 | f128 | ∅ | f128 | .. |
635+
| ✓ | logq | f128 | ∅ | f128 | .. PPC |
636+
| ✓ | logl |long double| ∅ |long double| .. |
637+
| ✓ | __log10h | f16 | ∅ | f16 |common (base-10) logarithm of a|
638+
| ✓ | log10f | f32 | ∅ | f32 | .. |
639+
| ✓ | log10 | f64 | ∅ | f64 | .. |
640+
| ✓ | __log10x | f80 | ∅ | f80 | .. |
641+
| ✓ | log10f128 | f128 | ∅ | f128 | .. |
642+
| ✓ | log10q | f128 | ∅ | f128 | .. PPC |
643+
| ✓ | log10l |long double| ∅ |long double| .. |
644+
| ✓ | __log2h | f16 | ∅ | f16 | base-2 logarithm of a |
645+
| ✓ | log2f | f32 | ∅ | f32 | .. |
646+
| ✓ | log2 | f64 | ∅ | f64 | .. |
647+
| ✓ | __log2x | f80 | ∅ | f80 | .. |
648+
| ✓ | log2f128 | f128 | ∅ | f128 | .. |
649+
| ✓ | log2q | f128 | ∅ | f128 | .. PPC |
650+
| ✓ | log2l |long double| ∅ |long double| .. |
651+
| ✓ | __roundh | f16 | ∅ | f16 | a rounded to next int away from zero|
652+
| ✓ | roundf | f32 | ∅ | f32 | .. |
653+
| ✓ | round | f64 | ∅ | f64 | .. |
654+
| ✓ | __roundx | f80 | ∅ | f80 | .. |
655+
| ✓ | roundf128 | f128 | ∅ | f128 | .. |
656+
| ✓ | roundq | f128 | ∅ | f128 | .. PPC |
657+
| ✓ | roundl |long double| ∅ |long double| .. |
658+
| ✓ | __sinh | f16 | ∅ | f16 | `sin(a)=(e^(ia)-e^(-ia))/2`|
659+
| ✓ | sinf | f32 | ∅ | f32 | .. |
660+
| ✓ | sin | f64 | ∅ | f64 | .. |
661+
| ✓ | __sinx | f80 | ∅ | f80 | .. |
662+
| ✓ | sinf128 | f128 | ∅ | f128 | .. |
663+
| ✓ | sinq | f128 | ∅ | f128 | .. PPC |
664+
| ✓ | sinl |long double| ∅ |long double| .. |
665+
| ✓ | __sincosh | f16 | 2x *f16 | ∅ |sin and cos of the same angle a|
666+
| ✓ | sincosf | f32 | 2x *f32 | ∅ |args a,*b,*c, `b.*=sin(x),c.*=cos(x)`|
667+
| ✓ | sincos | f64 | 2x *f64 | ∅ | .. |
668+
| ✓ | __sincosx | f80 | 2x *f80 | ∅ | .. |
669+
| ✓ | sincosf128 | f128 | 2x *f128 | ∅ | .. |
670+
| ✓ | sincosq | f128 | 2x *f128 | ∅ | .. PPC |
671+
| ✓ | sincosl |long double| 2x *long double|∅ | .. |
672+
| ✓ | __sqrth | f16 | ∅ | f16 | square root of a (find `r st. a=r^2`)|
673+
| ✓ | sqrtf | f32 | ∅ | f32 | .. |
674+
| ✓ | sqrt | f64 | ∅ | f64 | .. |
675+
| ✓ | __sqrtx | f80 | ∅ | f80 | .. |
676+
| ✓ | sqrtf128 | f128 | ∅ | f128 | .. |
677+
| ✓ | sqrtq | f128 | ∅ | f128 | .. PPC |
678+
| ✓ | sqrtl |long double| ∅ |long double| .. |
679+
| ✓ | __tanh | f16 | ∅ | f16 | `tan(x)=sin(x)/cos(x) |
680+
| ✓ | tanf | f32 | ∅ | f32 | .. |
681+
| ✓ | tan | f64 | ∅ | f64 | .. |
682+
| ✓ | __tanx | f80 | ∅ | f80 | .. |
683+
| ✓ | tanf128 | f128 | ∅ | f128 | .. |
684+
| ✓ | tanq | f128 | ∅ | f128 | .. PPC |
685+
| ✓ | tanl |long double| ∅ |long double| .. |
686+
| ✓ | __trunch | f16 | ∅ | f16 | a rounded to next int towards zero|
687+
| ✓ | truncf | f32 | ∅ | f32 | .. |
688+
| ✓ | trunc | f64 | ∅ | f64 | .. |
689+
| ✓ | __truncx | f80 | ∅ | f80 | .. |
690+
| ✓ | truncf128 | f128 | ∅ | f128 | .. |
691+
| ✓ | truncq | f128 | ∅ | f128 | .. PPC |
692+
| ✓ | truncl |long double| ∅ |long double| .. |
693+
694+
Further content (conditionally) exported with C abi:
556695
- aarch64 outline atomics
696+
- arm routines (memory routines + memclr [setting to 0], divmod routines and stubs for unwind_cpp)
557697
- atomics
558-
- msvc things like _alldiv, _aulldiv, _allrem
559-
- clear cache
560-
- tls emulation
561-
- math routines (cos, sin, tan, ceil, floor, exp, exp2, fabs, log, log10, log2, sincos, sqrt)
562698
- bcmp
563-
- ieee float routines (fma, fmax, fmin, fmod, fabs, float rounding, )
564-
- arm routines (memory routines + memclr [setting to 0], divmod routines and stubs for unwind_cpp)
699+
- clear cache
565700
- memory routines (memcmp, memcpy, memset, memmove)
701+
- msvc things like _alldiv, _aulldiv, _allrem
566702
- objective-c __isPlatformVersionAtLeast check
567703
- stack probe routines
704+
- tls emulation
568705

569-
Future work
570-
571-
Arbitrary length integer library routines
706+
Future work:
707+
- Arbitrary length integer library routines

0 commit comments

Comments
 (0)