Marek Materzok

Results 90 comments of Marek Materzok

Could you elaborate?

I had a hunch that the round-robin might interact badly with new features in the transaction subsystem. Yet another thing to be investigated (with low priority I think).

Integrating this functionality into the core might be nontrivial! And here is why. * So far, every operation has at most one destination register. Fused multiplication has two! Currently, we...

I've backported binutils 2.40 for Debian stable, and I can confirm that it understands Zmmul.

1. There is a Sphinx directive `inherited-members` which shows inherited members for a class. When applied indiscriminately through `sphinx-apidoc`, we get the opposite problem: docs show methods and attributes which...

Indeed, that's what it means. Nobody opposed this API when it was merged ;) The alternative API would be to add a keyword Boolean parameter to `schedule_before`, which says if...

Well... * PEP 563 proposes that annotations should be saved as parseable strings. * Obviously, environment at the annotation site is lost, and dynamic scoping is introduced. * Just before...

If you have a branch changing transaction schedulers to arbiters, could you make a pull request out of that? I think this is fairly uncontroversial.

Waiting until comments are addressed.

> Shift registers are linear. A shift register is linear in size, but constant in critical path. I fail to understand the point.