Markus Westerlind
Markus Westerlind
This is what I have come up with so far which is to just convert the UTF-8 encoded args in rust into wide strings (UCS-2) and then back to the...
Is it really corrode's job to calculate in which scope to put each variable? As long as it does a decent enough job (probably don't want all variables placed at...
> Thanks for taking this on! It would be interesting to know why these changes reduce the llvm ir bloat. Every combination of types that these functions gets called with...
Did a couple of more tweaks to re-use code when the output type is the same.
With the new changes (and using a different/newer compiler) we go from 47011 to 33329 (-29%)
I tried factoring out the non-visitor parts of https://github.com/serde-rs/json/blob/92ddbdf493f338eead7d2632f72deeec2f94776d/src/de.rs#L1480-L1511 and https://github.com/serde-rs/json/blob/92ddbdf493f338eead7d2632f72deeec2f94776d/src/de.rs#L1430-L1459 but did not see any improvements.
@dtolnay Rebased and re-ran my criterion hack for "json-benchmark", there is some variance from run to run but it seems like this may actually improve performance (at least improvements seem...
@dtolnay Are you able to take a look at this? Is it something that might be merged at some point?
I did try this earlier with no improvements to compile time, but I think this is an improvement regardless https://github.com/serde-rs/json/issues/313#issuecomment-349049626
Rebased and add a few more changes which totals up the improvement in my testing to 5.5% (up from 1.7%)