Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 24 Jun 2015 00:24:12 -0400
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com
Subject: Re: [PATCH 2/5] dynlink.c: compute modulus via magic
 multiplication

On Wed, Jun 24, 2015 at 02:24:52AM +0300, Alexander Monakov wrote:
> +static uint32_t umod(uint32_t x, uint32_t div, struct udiv *p)
> +{
> +	if (!(div&(div-1)))
> +		return x&(div-1);
> +	uint32_t v = x;
> +	if (p->s1) v >>= p->s1;
> +	else if (v != ~0u) v += p->inc;
> +	int s32=32, s2=p->s2;
> +	if (sizeof(long) == 8)
> +		s32+=s2, s2=0;
> +	v = (1ull * v * p->mul) >> s32;
> +	v >>= s2;
> +	return x-v*div;
> +}

I think having the div argument here is a pessimization. Power-of-two
sizes are unlikely to be used, and it's possible for precompute_udiv
to setup the struct udiv to work perfectly fine for powers of two, so
that there is a single code path with no branches. The p->s1 branch
should not be needed either.

Rich

Powered by blists - more mailing lists

Your e-mail address:

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.