bpo-45636: BINARY_OP (third time's the charm)#29482
Conversation
markshannon
left a comment
There was a problem hiding this comment.
I like this approach. Performance is acceptable and should improve with specialization.
Sorry, something went wrong.
|
🤖 New build scheduled with the buildbot fleet by @markshannon for commit 490b8da 🤖 If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again. |
Sorry, something went wrong.
|
👍 |
Sorry, something went wrong.
Fidget-Spinner
left a comment
There was a problem hiding this comment.
Awesome work!
Sorry, something went wrong.
|
🤖 New build scheduled with the buildbot fleet by @brandtbucher for commit bea2cff 🤖 If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again. |
Sorry, something went wrong.
|
The one buildbot failure looks unrelated, and seems to be affecting other PRs as well. |
Sorry, something went wrong.
This merges all numeric
BINARY_*andINPLACE_*instructions into oneBINARY_OPinstruction. As a consequence, theBINARY_ADDandBINARY_MULTIPLYspecialization families are merged into a newBINARY_OPfamily too.I prefer this new approach over the more complicated one taken in #29418. It simplifies the specialization logic for all operators (for example,
+=and*=now get specialized the same as+and*, basically for free).Performance is unchanged vs
main:Details
https://bugs.python.org/issue45636