3 ms·
>Yeah, but this project is actually delivering stuff that's been shipped. The initial selling point was ~ 5x speedup over 5 years. This is over a year now, wi
by coldtea 3y ago
>Yeah, but this project is actually delivering stuff that's been shipped.
The initial selling point was ~ 5x speedup over 5 years.
This is over a year now, with far more people and resources plus the backling of a major industry-player (MS), and it's been like 20%-30% improvements at best, and even those in danger to be wiped out by the no-gil change.
>As for your fantasy, we already have PyPy.
That's nowhere near Lars Bak level fantasy results...
- aeyes 3y agoIt was actually 5x in 4 years (1.5x per year). As you said we see 20-30% at best. And several important workloads have important slowdowns (for example coveragepy runs 30-50% slower). The GIL is an academic problem which has little real world relevance.