[ANN] math32 - library for float32 and complex64

[David Turnbull]

math32 is a library for float32 and complex64 types. Everything from "math" and "math/cmplx" is available for 32-bit floats.

https://github.com/AE9RB/math32

Enjoy!

[simon place]

does this actually mean you can you just cut/paste float64 to float32 in existing code and have it work?

[simon place]

(reply to assembly code for float32 lib)

probably best thing to do, would be to simply copy as much as possible the float64 std lib, so as to have the same hardware dependencies.

mostly what seems to be needed is:

amd64:     ?????SD   ->   ?????SS

in x86;      F????   ->   F????S

and 8byes -> 4bytes in MOVes

for example: sqrt_amd64.s

     1	// Copyright 2009 The Go Authors.  All rights reserved.
     2	// Use of this source code is governed by a BSD-style
     3	// license that can be found in the LICENSE file.
     4	
     5	#include "../../cmd/ld/textflag.h"
     6	
     7	// func Sqrt(x float64) float64
     8	TEXT ·Sqrt(SB),NOSPLIT,$0
     9		SQRTSD x+0(FP), X0
    10		MOVSD X0, ret+8(FP)
    11		RET

needs this diff

     9		SQRTSS x+0(FP), X0
    10		MOVSS X0, ret+4(FP)

and 

sqrt_386.s

     1	// Copyright 2009 The Go Authors.  All rights reserved.
     2	// Use of this source code is governed by a BSD-style
     3	// license that can be found in the LICENSE file.
     4	
     5	#include "../../cmd/ld/textflag.h"
     6	
     7	// func Sqrt(x float64) float64	
     8	TEXT ·Sqrt(SB),NOSPLIT,$0
     9		FMOVD   x+0(FP),F0
    10		FSQRT
    11		FMOVDP  F0,ret+8(FP)
    12		RET

diff

     9		FMOVS   x+0(FP),F0
    10		FSQRTS
    11		FMOVSP  F0,ret+4(FP)

BTW bear in mind i've only been looking at assembler since last week, so i could be missing something important.

and

the 386 is currently just guesses and is untested, but i hope to test and will report back.(i guess the 386 will actually work on amd64) 

[David Turnbull]

On Saturday, March 15, 2014 9:18:25 AM UTC-7, simon place wrote:

does this actually mean you can you just cut/paste float64 to float32 in existing code and have it work?

For the most part, yes, that's the idea.

I'm surprised nobody commented about the release of a floating point math library on "Pi day".

[andrey mirtchovski]

> I'm surprised nobody commented about the release of a floating point math
> library on "Pi day".

I would have, but your library is insufficiently irrational.

[necl...@gmail.com]

Hi David,

I can no longer find the repository.

Can you tell me why it was taken? Using an old version I have, I am safe?

Thank you,

Ignazio