This is the mail archive of the libc-alpha@sources.redhat.com mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] PPC64 Use hardware sqrt.


Kumar Gala wrote:

Understand, but future PPC64 from Freescale (former Motorola SPS) are unlikely to implement it.

If they do then likely they will also leave the other optional floating point operations unimplemented and in that case the correct way to build glibc is with the --nf option which is already the case for PPC32 processors without floating point. If the processor has all the other optional floating point instructions except sqrt, then I'd say that design needs to be revisited. Historically the optional floating point instructions are either all there (as is the case with ALL ppc64 processors ever made) or not in which case as pointed out to you the --nf option is for you.


I believe sjmunroe is right and his implementation is sound.

Regards,

Tom

- kumar

On Apr 30, 2004, at 11:00 AM, Steve Munroe wrote:

Kumar Gala wrote on 04/29/2004 05:55:21 PM:

I do not think its fair to say that all future PPC64 processors will
implement the fsqrt instruction.

- kumar


All server (i/pSeries) and 970 models do and will. The silicon required for this is so small as to be lost in the noise of a modern chip design.

So I will take 3-5X performance gain.

Steven J. Munroe
Linux on Power Toolchain Architect
IBM Corporation, Linux Technology Center



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]