[07:40] *** [Coke] left
[07:41] *** [Coke] joined
[13:31] *** soverysour joined
[13:31] *** soverysour left
[13:31] *** soverysour joined
[13:36] *** ShimmerFairy left
[13:36] *** soverysour left
[13:36] *** ShimmerFairy joined
[15:44] *** soverysour joined
[15:48] *** soverysour left
[20:16] <Geth> ¦ MoarVM/perf_jitdump_support: 5 commits pushed by (Timo Paulssen)++

[20:16] <Geth> ¦ MoarVM/perf_jitdump_support: e3423faf7d | Output JITDUMP format with MVM_JIT_PERF_DUMP

[20:16] <Geth> ¦ MoarVM/perf_jitdump_support: bf7658c3b6 | Take a path in MVM_JIT_PERF_DUMP instead of outputting to CWD

[20:16] <Geth> ¦ MoarVM/perf_jitdump_support: c641d45e77 | when jitdump is on, put jumps to all BBs at the end

[20:16] <Geth> ¦ MoarVM/perf_jitdump_support: 8b500a99fe | jit iscoderef to code instead of a call

[20:16] <Geth> ¦ MoarVM/perf_jitdump_support: 359cb727de | jit: allow runccode to emit a direct call

[20:16] <Geth> ¦ MoarVM/perf_jitdump_support: review: https://github.com/MoarVM/MoarVM/compare/c2dc9f4c3fcc...359cb727de30

[21:36] <Geth> ¦ MoarVM/main: 092f983b85 | timo++ (committed using GitHub Web editor) | 9 files

[21:36] <Geth> ¦ MoarVM/main: Output JITDUMP format with MVM_JIT_PERF_DUMP (#1982)

[21:36] <Geth> ¦ MoarVM/main: 

[21:36] <Geth> ¦ MoarVM/main: * Output JITDUMP format with MVM_JIT_PERF_DUMP

[21:36] <Geth> ¦ MoarVM/main: 

[21:36] <Geth> ¦ MoarVM/main: It's a bit of a hassle, but together with `perf inject` you can now get samples

[21:36] <Geth> ¦ MoarVM/main: along with the correct assembly code shown in `perf report`.

[21:36] <Geth> ¦ MoarVM/main: 

[21:36] <Geth> ¦ MoarVM/main: <…commit message has 21 more lines…>

[21:36] <Geth> ¦ MoarVM/main: review: https://github.com/MoarVM/MoarVM/commit/092f983b85

[21:36] <timo> I decided there's no reason not to have this in main

[21:49] <timo> patrickb: I approved the debugserver changes pull request. would you like to merge it into main?

[22:13] <lizmat> let me first bump MoarVM

[22:27] *** MasterDuke joined
[22:27] <MasterDuke> timo: do you know anything about the "deopts table"?

[22:28] <MasterDuke> while compiling the settings i see a bunch of time spent in `MVM_spesh_deopt_find_inactive_frame_deopt_idx`, which is just a loop doing a linear search

[22:31] <MasterDuke> https://gist.github.com/MasterDuke17/4b2dfcbeee9c5663d91848030399802c has some metrics of how far into the table it looked before finding a match

[22:37] <MasterDuke> looks like it isn't sorted, so can't do a smarter search

[23:30] *** MasterDuke left