This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [RFC][PATCH 0/2] aarch64: Add optimized ASIMD versions of sinf/cosf
- From: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>
- To: Wilco Dijkstra <Wilco dot Dijkstra at arm dot com>, "Ashwin dot Sekhar at cavium dot com" <Ashwin dot Sekhar at cavium dot com>, Siddhesh Poyarekar <siddhesh at gotplt dot org>
- Cc: nd <nd at arm dot com>, "libc-alpha at sourceware dot org" <libc-alpha at sourceware dot org>, Szabolcs Nagy <Szabolcs dot Nagy at arm dot com>
- Date: Tue, 13 Jun 2017 15:49:08 -0300
- Subject: Re: [RFC][PATCH 0/2] aarch64: Add optimized ASIMD versions of sinf/cosf
- Authentication-results: sourceware.org; auth=none
- References: <VI1PR0802MB2621465F621A03FFA7C4607683C20@VI1PR0802MB2621.eurprd08.prod.outlook.com>
On 13/06/2017 14:49, Wilco Dijkstra wrote:
> Adhemerval Zanella wrote:
>
>> I think a good starting point I would be if Ashwin in could provide us with a C skeleton with same implementation done in assembly.
>
> I don't see the point of asking him to do that. It would be a significant amount of
> work which would be wasted once Szabolcs posts his implementation.
It was not a demand, but rather a suggestion if it were the case he has a
starting implementation based on C (which I know might not be the case).
Also, it would be helpful which kind of algorithm strategies Szabolcs is
planing to do different than Ashwin and current implementation that is
intended to supersede both implementation (like short algorithm description
sent by Ashwin).
>
> What I'd like to ask Ashwin is which benchmarks show a speedup due to
> his patch and whether it is essential they get a speedup this release
> (rather than a larger speedup later). I don't recall seeing sinf/cosf in profiles
> in popular benchmarks, so I don't understand the urgency of doing this now.
>
> What I think would be useful is to start collecting real traces of actual applications
> or large benchmarks like SPEC and creating representative microbenchmarks
> using that data.
>
> Wilco
>