Page MenuHomePhabricator

ille (ille)
UserEmail Not Verified

Projects

User does not belong to any projects.

User Details

User Since
Oct 21 2020, 11:25 AM (6 w, 1 d)

Recent Activity

Tue, Nov 10

ille added inline comments to D90434: [CodeGen] Correct codegen for self-capturing __block var.
Tue, Nov 10, 9:49 AM · Restricted Project

Fri, Nov 6

ille added a comment to D90434: [CodeGen] Correct codegen for self-capturing __block var.

Ping :)

Fri, Nov 6, 8:44 AM · Restricted Project

Oct 30 2020

ille updated the diff for D90434: [CodeGen] Correct codegen for self-capturing __block var.

Satisfy clang-format bot, at the cost of formatting a few adjacent lines.

Oct 30 2020, 11:24 AM · Restricted Project

Oct 29 2020

ille requested review of D90434: [CodeGen] Correct codegen for self-capturing __block var.
Oct 29 2020, 3:13 PM · Restricted Project

Oct 22 2020

ille updated the diff for D89903: [CodeGen] Crash instead of generating broken code with self-capturing __block var.

Move the check to cover the atomic case as well.

Oct 22 2020, 8:50 AM · Restricted Project
ille added a comment to D89903: [CodeGen] Crash instead of generating broken code with self-capturing __block var.

We do not actually support allocation failure for a lot of things around blocks. I don't think the copy-helper functions even have a way to propagate out a failure in copying a field. I have never seen any code in the wild that would handle Block_copy returning a null pointer. Effectively it is assumed to not happen.

Oct 22 2020, 8:31 AM · Restricted Project

Oct 21 2020

ille added a comment to D89903: [CodeGen] Crash instead of generating broken code with self-capturing __block var.

One downside of that approach is performance. It's somewhat idiosyncratic, but I work on codebases that use blocks heavily in performance-sensitive paths. Typically, the blocks are never copied (they don't escape) and are expected to be inlined. Any implicit heap allocations added by the compiler would have an undesirable performance cost, and would likely also hinder inlining. For those codebases, I would much prefer to be prompted to refactor the code to avoid the self-capture. I suppose this could be implemented as a warning, perhaps off by default.

Oct 21 2020, 12:58 PM · Restricted Project
ille requested review of D89903: [CodeGen] Crash instead of generating broken code with self-capturing __block var.
Oct 21 2020, 11:32 AM · Restricted Project