nikic wrote:

> @nikic Doubling the analysis depth for binary operations seems to solve the 
> compile-time regression: 
> https://llvm-compile-time-tracker.com/compare.php?from=d852c040c510fb452284a8d04fb8bc0d45013f61&to=c69dce81a3bb663078009ed9b102213c2c6f2625&stat=instructions:u
> 
> And even slightly better when the depth is set to `Max - 1`: 
> https://llvm-compile-time-tracker.com/compare.php?from=c69dce81a3bb663078009ed9b102213c2c6f2625&to=761207338c54d90e193873107da61b5f5efae7cb&stat=instructions:u
> 
> Do you think the current approach converges or should I refactor the users 
> nonetheless?

I was a bit confused by these results, but apparently those are just the 
differences to the baseline regression. All the commits combined together are: 
https://llvm-compile-time-tracker.com/compare.php?from=3bd7dc3d65d3a690721d9ec11d2af02f87323840&to=761207338c54d90e193873107da61b5f5efae7cb&stat=instructions%3Au

So this remains a significant regression.

https://github.com/llvm/llvm-project/pull/181110
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to