[prev in list] [next in list] [prev in thread] [next in thread] 

List:       glibc-alpha
Subject:    Re: libmvec requires x86?
From:       Steven Munroe <munroesj () linux ! vnet ! ibm ! com>
Date:       2015-12-21 22:16:27
Message-ID: 1450736187.10617.1.camel () oc7878010663
[Download RAW message or body]

On Mon, 2015-12-21 at 15:19 -0600, Steven Munroe wrote:
> On Mon, 2015-12-21 at 12:58 -0600, Timothy Pearson wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA224
> > 
> > All,
> > 
> > Are there currently any plans to port libmvec to architectures other than
> > x86?  I spent a rather fruitless day attempting to vectorize a floating
> > point calculation loop on POWER8 before realizing that libmvec is x86
> > only.  We cannot use modern x86 hardware for internal security reasons,
> > and it came as something of a surprise that relatively advanced features
> > of glibc are only supported on x86, even though functionally equivalent
> > SIMD capabilities exist on other architectures (e,g, POWER8).
> > 
> We (PowerPC) have been waiting for libmvec to stabilize and as
> Adhemerval has pointed out, powerpc already has existing solution using
> the IBM Mathematical Acceleration Subsystem (MASS) libraries.
> 
> The MASS library is available without cost from
> http://www-01.ibm.com/software/awdtools/mass/linux/mass-linux.html

I found a more direct link to the POWER8 MASS library here:
https://www.ibm.com/developerworks/community/blogs/fe313521-2e95-46f2-817d-44a4f27eba32/entry/c_c_cafe_performance_and_mass_libraries_updated?lang=en


> 
> We have direct GCC support for -mveclibabi=mass starting with GCC-4.6
> 
> This should vectorize your loop.


[prev in list] [next in list] [prev in thread] [next in thread]