search

uwplse / ruler

Rewrite Rule Inference Using Equality Saturation
https://dl.acm.org/doi/10.1145/3485496
MIT License
119 stars 9 forks source link

Fix halide baseline rules #178

Closed ajpal closed 1 year ago

ajpal commented 1 year ago

lmao we just weren't parsing these right

vee-effekt commented 1 year ago

What does this mean exactly? Are the current results wrong?

ajpal commented 1 year ago

What does this mean exactly? Are the current results wrong?

This won't change the results using enumo rules to derive the halide ones. But without the ? in the halide rules, the variables weren't being parsed as pattern variables (so for example the rule (! (! x)) ==> x would only match a variable enode literally named x). The halide->enumo derivability was saying 0 for everything (and I think we always knew this was suspect but just hadn't had a chance to dig into). This should fix that direction.

Screenshot 2023-04-08 at 10 13 24 PM
vee-effekt commented 1 year ago

ooooooooo okay great