Sabitlenmiş Tweet
Pleased to share my first ever paper :) 'Improving Memory Dependence Prediction with Static Analysis': arxiv.org/abs/2403.08056
English
muke
1.3K posts
muke
@muke1010101
computer architecture phd student vroom vroom
Imperial Collage London Katılım Temmuz 2021
66 Takip Edilen141 Takipçiler
@loveryteks well how cross-arch is ffmeg code, technically? doesn't it just detect the target its running on and jump to appropriate code at run time already?
English
@muke1010101 then again, i don't keep up with the current state of the art with compilers or SIMD, the last time i was seriously into either wasin highschool.. maybe it's different now somehow, especially given that you can probably assume AVX2 is available everywhere now..
English
how empty are you gonna feel when compilers inevitably make your handrolled assembly redundant?
FFmpeg@FFmpeg
FFmpeg is moving to Rust 🦀 Our use of C and Assembly in FFmpeg has been an unacceptable violation of safety. FFmpeg will be running 10x slower - but we're doing it for your safety. All your videos will appear green - safety first, working software later.
English
@loveryteks yea I was definitely getting ahead of myself with superoptimisation there. there have been some interesting things in academia lately but nothing that translates to production yet (for instance arxiv.org/pdf/2306.00229)
I do wonder how good MLIR for vectorisation could be though.
English
@muke1010101 this isn't to imply endorsement of the perspective of the ffmpeg post, though.
English
@csjh__ You're probably right (though there might also be something in knowing int's can't overflow for ffmpeg), but that just goes to show how little the difference between C and Rust matters so you should go with the memory safe option anyway. The extra performance they boast comes -
English
muke retweetledi
Wow, RISC-V is really gaining traction again.
Alibaba just announced the Xuantie C950…which is basically claiming Apple M1 (ish) levels of performance.
I don’t see a lot of people talking about it!
(2.6/Ghz SPECint2017, Apple M1 P-core also around ~2.6)
English
@muke1010101 bruh check telegram do u even still love me
English
in case anyone's ever wanted it here are two mcpat forks with the phast and store set mem dep predictors implemented
gitlab.com/muke101/mcpat-…
gitlab.com/muke101/pnd-lo…
English
these also contain what is apparently the only gem5tomcpat.py parser script that works on simpoint checkpoint restores
English
Update: this was working really well really quickly until my supervisor reminded me about functions that don't return and exceptions, so now I gotta deal with all that.
muke@muke1010101
today's procrastination from my existing projects is implementing an open source version of something intel probably already does called push-pop acceleration
English
oh apple do this too and it probably is exactly what i just described but i cba to read patents patents.google.com/patent/WO20251…
English
