[Win32]gmp-5.1.0-RC1: no assembly for 32-bit build

Sisyphus sisyphus1 at optusnet.com.au
Mon Dec 10 03:29:20 CET 2012

----- Original Message ----- 
From: "Torbjorn Granlund" <tg at gmplib.org>
To: "Sisyphus" <sisyphus1 at optusnet.com.au>
Cc: <gmp-bugs at gmplib.org>
Sent: Monday, December 10, 2012 8:39 AM
Subject: Re: [Win32]gmp-5.1.0-RC1: no assembly for 32-bit build

> "Sisyphus" <sisyphus1 at optusnet.com.au> writes:
>  Firstly, a big *thank you* to the person(s) responsible for enabling
>  the assembly support for 64-bit M$Windows. That appears to be working
>  well -I see lots of asm files specified at the end of the configure
>  output, and all tests passed.
> It was not just to "enable" it, or we'd have done that long ago.

I certainly didn't mean to imply that I thought this was trivial to achieve. 
I know it would have been far from that - which is the reason that I wanted 
to explicitly thank the person(s) involved for the time and effort it would 
have required.

>  For some reason, however, my 32-bit build is now a generic C build.
> I am happy you spotted this problem!  I think I've fixed it now.  Here
> is a new test release:
>  ftp://gmplib.org/pub/snapshot/gmp-5.1.0-RC2.tar.xz
> Please try it and report back.

Yes - this seems right, now. Thanks, Torbjorn.

I was only alerted to the problem because the mpz/reuse.exe test took around 
4 to 5 times longer to run on the 32-bit build than on the 64-bit build. 
(The timing was around 70 seconds for 32-bit, and 15seconds for 64-bit.)

With RC2, the timing for 32-bit mpz/reuse.exe has dropped from 70 seconds to 
30 seconds - which is still much slower than for the 64-bit build.
I assume that comparative slowness is *not* indicative of a problem with the 
gmp library. (I mention it only in case that assumption ought to be 

Thanks, again, to the gmp team.


More information about the gmp-bugs mailing list