[9fans] IEEE floats

[9fans] IEEE floats

[Steve Simon]

Porting stuff from the net (under ape) which wants to use IEEE floats.
i.e. ieee754_float32_t, powf(), fabsf() and log10f().

I could map these to plan9 floats and call the double version
of the transient functions, but is there a better solution?

-Steve

Re: [9fans] IEEE floats

[erik quanstrom]

On Thu Feb 27 06:54:22 EST 2014, steve@quintile.net wrote:
> Porting stuff from the net (under ape) which wants to use IEEE floats.
> i.e. ieee754_float32_t, powf(), fabsf() and log10f().
> 
> I could map these to plan9 floats and call the double version
> of the transient functions, but is there a better solution?

plan 9 uses ieee floating point on machines that support it
(all current arches) so ieee754_float32_t ≡ float.
substituting the double version should work.

did they abolish the rule that float is promoted to double
on function call?  or is this the equivalent of char fu(char) vs
int fu(int).

- erik

Re: [9fans] IEEE floats

[Charles Forsyth]

[-- Attachment #1: Type: text/plain, Size: 624 bytes --]

On 27 February 2014 12:09, erik quanstrom <quanstro@quanstro.net> wrote:

> did they abolish the rule that float is promoted to double
> on function call?  or is this the equivalent of char fu(char) vs
> int fu(int).
>

if you use the original declaration syntax of 40 years ago,
float must be promoted to double because the caller won't know the type,
but the ANSI declarator syntax that everyone ought to be using allows
the distinction between float and double to be maintained, and it is.
compare

void
g(float a)
{
g(a);    /* a is float */
}

void
h(a)
float a;
{
h(a);    /* a is double */
}

[-- Attachment #2: Type: text/html, Size: 1690 bytes --]

Re: [9fans] IEEE floats

[erik quanstrom]

> if you use the original declaration syntax of 40 years ago,
> float must be promoted to double because the caller won't know the type,
> but the ANSI declarator syntax that everyone ought to be using allows
> the distinction between float and double to be maintained, and it is.
> compare
>
> void
> g(float a)
> {
> g(a);    /* a is float */
> }
>
> void
> h(a)
> float a;
> {
> h(a);    /* a is double */
> }

thanks!

- erik