By: Jukka Larja (roskakori2006.delete@this.gmail.com), October 30, 2020 9:04 pm
Room: Moderated Discussions
Mark Roulo (nothanks.delete@this.xxx.com) on October 30, 2020 11:42 am wrote:
> Jukka Larja (roskakori2006.delete@this.gmail.com) on October 30, 2020 11:29 am wrote:
> > Mark Roulo (nothanks.delete@this.xxx.com) on October 30, 2020 10:21 am wrote:
> > > Jukka Larja (roskakori2006.delete@this.gmail.com) on October 30, 2020 9:52 am wrote:
> > > > me (me.delete@this.me.com) on October 30, 2020 9:47 am wrote:
> > > > > > And I presume you are talking about AVX2. Both PS4 and Xbox One support AVX,
> > > > > > and that alone obviously didn't gather any significant support (even though
> > > > > > consoles have much more benefit from CPU optimizations compared to PC).
> > > > > >
> > > > > > -JLarja
> > > > >
> > > > > IIRC Jaguar's AVX isn't any faster than using SSE so devs mostly ignored it I guess.
> > > >
> > > > Is any AVX faster than SSE?
> > > >
> > > > -JLarja
> > >
> > > For my employer’s loads, AVX2 runs about 1.7x faster than SSE for compute bound loads.
> >
> > Yes, but that's AVX2, which is not available on Jaguar.
> >
> > -JLarja
>
> It is AVX2 because many of our loads are short integer loads.
>
> AVX won’t be faster on Jaguar than SSE because the Jaguar vector units are 128 bits wide so the AVX
> instructions needs roughly 2x the clocks per instruction which cancels out doing 2x more work per op.
>
>
> I would expect AVX implementations that are 256 bits wide (eg. Sandy
> Bridge) to run ~1.7x faster than SSE on floating point code.
>
> The slowdown is AMD specific (and only on some cores), not inherent with AVX.
Thanks, I recalled falsely that AVX had only 128 bit registers (like SSE) and was thus pretty much useless (though I guess it's supposed to be more compiler friendly).
Anyway, the fact that even AVX (non-2) on many CPUs offers twice the bandwidth, yet gathers no interest from game developers, just points more directly that SIMD extensions aren't interesting in the field.
-JLarja
> Jukka Larja (roskakori2006.delete@this.gmail.com) on October 30, 2020 11:29 am wrote:
> > Mark Roulo (nothanks.delete@this.xxx.com) on October 30, 2020 10:21 am wrote:
> > > Jukka Larja (roskakori2006.delete@this.gmail.com) on October 30, 2020 9:52 am wrote:
> > > > me (me.delete@this.me.com) on October 30, 2020 9:47 am wrote:
> > > > > > And I presume you are talking about AVX2. Both PS4 and Xbox One support AVX,
> > > > > > and that alone obviously didn't gather any significant support (even though
> > > > > > consoles have much more benefit from CPU optimizations compared to PC).
> > > > > >
> > > > > > -JLarja
> > > > >
> > > > > IIRC Jaguar's AVX isn't any faster than using SSE so devs mostly ignored it I guess.
> > > >
> > > > Is any AVX faster than SSE?
> > > >
> > > > -JLarja
> > >
> > > For my employer’s loads, AVX2 runs about 1.7x faster than SSE for compute bound loads.
> >
> > Yes, but that's AVX2, which is not available on Jaguar.
> >
> > -JLarja
>
> It is AVX2 because many of our loads are short integer loads.
>
> AVX won’t be faster on Jaguar than SSE because the Jaguar vector units are 128 bits wide so the AVX
> instructions needs roughly 2x the clocks per instruction which cancels out doing 2x more work per op.
>
>
> I would expect AVX implementations that are 256 bits wide (eg. Sandy
> Bridge) to run ~1.7x faster than SSE on floating point code.
>
> The slowdown is AMD specific (and only on some cores), not inherent with AVX.
Thanks, I recalled falsely that AVX had only 128 bit registers (like SSE) and was thus pretty much useless (though I guess it's supposed to be more compiler friendly).
Anyway, the fact that even AVX (non-2) on many CPUs offers twice the bandwidth, yet gathers no interest from game developers, just points more directly that SIMD extensions aren't interesting in the field.
-JLarja
Topic | Posted By | Date |
---|---|---|
Expiry of x86-64 patents | Beastian | 2019/04/19 09:05 AM |
Expiry of x86-64 patents | Gian-Carlo Pascutto | 2019/04/19 09:46 AM |
Expiry of x86-64 patents | Beastian | 2019/04/19 10:06 AM |
Expiry of x86-64 patents | Jukka Larja | 2019/04/19 10:44 AM |
Expiry of x86-64 patents | Gian-Carlo Pascutto | 2019/04/19 11:12 AM |
Expiry of x86-64 patents | Jukka Larja | 2019/04/19 12:41 PM |
Expiry of x86-64 patents | Robert Williams | 2019/04/19 01:18 PM |
Expiry of x86-64 patents | Gian-Carlo Pascutto | 2019/04/19 02:35 PM |
Expiry of x86-64 patents | IntelUser2000 | 2020/10/30 02:17 AM |
Expiry of x86-64 patents | Jukka Larja | 2020/10/30 07:49 AM |
Expiry of x86-64 patents | me | 2020/10/30 09:47 AM |
Expiry of x86-64 patents | Jukka Larja | 2020/10/30 09:52 AM |
Expiry of x86-64 patents | Mark Roulo | 2020/10/30 10:21 AM |
Expiry of x86-64 patents | Jukka Larja | 2020/10/30 11:29 AM |
Expiry of x86-64 patents | Mark Roulo | 2020/10/30 11:42 AM |
Expiry of x86-64 patents | Jukka Larja | 2020/10/30 09:04 PM |
SIMD syntax | hobold | 2020/10/31 06:54 AM |
SIMD syntax | Jukka Larja | 2020/10/31 09:14 AM |
SIMD syntax | hobold | 2020/11/01 08:22 AM |
SIMD syntax | Jukka Larja | 2020/11/01 11:11 AM |
SIMD syntax | hobold | 2020/11/02 05:33 AM |
Expiry of x86-64 patents | me | 2020/10/31 03:01 PM |
Expiry of x86-64 patents | Jukka Larja | 2020/10/31 09:23 PM |
Expiry of x86-64 patents | Foo_ | 2020/11/01 04:48 AM |
Expiry of x86-64 patents | Jukka Larja | 2020/11/01 07:01 AM |
Expiry of x86-64 patents | Adrian | 2020/10/30 12:02 PM |
Expiry of x86-64 patents | Bigos | 2020/10/30 01:20 PM |
Expiry of x86-64 patents | Geoff Langdale | 2019/04/19 02:52 PM |
Expiry of x86-64 patents | Jukka Larja | 2019/04/19 09:38 PM |
Expiry of x86-64 patents | Yuhong Bao | 2019/04/20 02:35 PM |
Expiry of x86-64 patents | Doug S | 2019/04/19 10:40 AM |
Expiry of x86-64 patents | Beastian | 2019/04/19 11:10 AM |
Expiry of x86-64 patents | Robert Williams | 2019/04/20 08:15 AM |
Expiry of x86-64 patents | Robert Williams | 2020/10/28 06:42 AM |
Expiry of x86-64 patents | anyone | 2019/04/20 07:11 AM |
Expiry of x86-64 patents | Groo | 2019/04/20 07:29 AM |
Expiry of x86-64 patents | wumpus | 2019/04/20 08:32 AM |
Expiry of x86-64 patents | blaine | 2020/10/30 12:03 PM |
Expiry of x86-64 patents | David Kanter | 2020/10/30 08:59 PM |
Intel vs AMD patents | Yuhong Bao | 2019/04/20 02:32 PM |
Intel vs AMD patents | Beastian | 2019/04/20 03:35 PM |
Expiry of x86-64 patents | Travis Downs | 2019/04/20 07:24 PM |
Expiry of x86-64 patents | none | 2019/04/21 07:36 AM |
Expiry of x86-64 patents | somebody | 2019/11/27 10:44 AM |
Expiry of x86-64 patents | Anon3 | 2019/11/27 05:16 PM |
Expiry of x86-64 patents | Travis Downs | 2019/11/27 06:17 PM |
Expiry of x86-64 patents | Montaray Jack | 2019/11/28 12:03 AM |
Expiry of x86-64 patents | none | 2019/11/28 01:57 AM |
Expiry of x86-64 patents | dmcq | 2019/11/28 11:20 AM |
Expiry of x86-64 patents | Montaray Jack | 2019/11/29 05:00 AM |