Twitter | Search | |
Paul Dragoonis
's experimental JIT branch is currently 54% faster than PHP7.1. We are in for another huge performance jump in our next major release :)
Reply Retweet Like More
“feature” 8 Dec 16
Replying to @dr4goonis
. *synthetic tests*. Can you link the branch? Would love to run real apps against it
Reply Retweet Like
Benjamin Eberlei 8 Dec 16
Replying to @dr4goonis @Stof70
bench.php or Real world? Source?
Reply Retweet Like
Paul Dragoonis 8 Dec 16
Replying to @Ocramius
.. branch: "jit-dynasm"
Reply Retweet Like
Paul Dragoonis 8 Dec 16
Replying to @beberlei @Stof70
bench.php
Reply Retweet Like
“feature” 8 Dec 16
Replying to @dr4goonis
awesome, thanks!
Reply Retweet Like
Benjamin Eberlei 8 Dec 16
Replying to @dr4goonis @Stof70
that probably translates to 0 real world change :) jitting math intensive code is not often the use case
Reply Retweet Like
Daniel Noel-Davies 8 Dec 16
Reply Retweet Like
Paul Dragoonis 8 Dec 16
Replying to @beberlei @Stof70
we both know that. However the changes in the ZendEngine are not math focused, so you can still expect positive results ;)
Reply Retweet Like
Crashy McCiderface 8 Dec 16
Replying to @dr4goonis
is that the same benchmark that was accused of being cooked up last time or something else? 🤔
Reply Retweet Like
Paul Dragoonis 8 Dec 16
Replying to @philsturgeon
dont know what thing you're referring to ?
Reply Retweet Like
Jordan Pittman 8 Dec 16
this is amazing
Reply Retweet Like
jnr in life 8 Dec 16
Replying to @Ocramius @dr4goonis
would be great full to see the results from your stuff 🤔
Reply Retweet Like
“feature” 8 Dec 16
Replying to @dr4goonis
register '%r15' unsuitable for global register variables on this target register const zend_op* volatile opline __asm__("%r15");
Reply Retweet Like
“feature” 8 Dec 16
Replying to @dr4goonis
I guess it's probably compiling with local patches
Reply Retweet Like
Graham Campbell 8 Dec 16
Cool stuff, yes. Note that this is only an artificial benchmark. Also, who knows what the memory usage is like. :P
Reply Retweet Like
Jordan Pittman 8 Dec 16
True.
Reply Retweet Like
MP 8 Dec 16
Replying to @dr4goonis @CalEvans
Is that branch planned for a 7.2 release? If not, when?
Reply Retweet Like
Rob Colburn 8 Dec 16
Replying to @dr4goonis
how's the head2head w/ Hack JIT?
Reply Retweet Like
Paul Dragoonis 9 Dec 16
Replying to @mike_pearson @CalEvans
that fundamentally is decided upon if it breaks BC or not. If it doesn't break BC then 7.x, if it goes then PHP8
Reply Retweet Like