[pull] master from ruby:master - #1400
Merged
Merged
Conversation
`RUBY_FIXNUM_MIN / -1` is the only Fixnum division that returns a Bignum. When either operand's known value rules out that pair, infer Fixnum instead of Integer. This allows downstream Fixnum operations to avoid guarding the result again. Division by zero side exits before producing a value, so typing its unreachable normal path as Fixnum is also safe. Co-authored-by: Hartley McGuire <skipkayhil@gmail.com>
`tr_trans_pairs_search` expect the pointer to be incremented by one only. Therefore, when matching a multibyte character, `search.s` must be incremented by 1, and not `clen`. Co-Authored-By: Federico Carrocera <fedecarrocera@gmail.com>
The disarm path MOD'd the registration to no events instead of deleting it, to save a syscall on the next wait. But epoll reports EPOLLHUP and EPOLLERR whatever the mask asks for, and a mask of 0 carries no EPOLLONESHOT, so those are reported level-triggered. A waiter that gave up without an event -- Thread#kill, an interrupt, a timeout -- therefore left a registration that fires forever once the fd hangs up, and epoll_wait returned it on every round. The timer thread then never took its timeout branch, which is the only place that mints a shared NT or signals the global ready queue. Under RUBY_MN_THREADS=2 the main thread goes dedicated while it waits in rb_ractor_terminate_all, so snt_cnt reaches 0, a ractor sits on the grq with nothing to run it, and the process never exits. Repro: kill a thread blocked on a pipe read, close the write end, then use a Ractor -- 4/6 to 8/20 runs hang, with the timer thread at 100% CPU. Under RUBY_MN_THREADS=1 the same spin burns a core silently. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
When the receiver is a `String` and both arguments are likely `Fixnum`, the annotation now emits a new `StringByteslice` HIR instruction that calls `rb_str_byte_substr` directly, instead of going through `CCallVariadic`. Any other shape, one argument, a Range, or non-Fixnum arguments, keeps using CCallVariadic.
) The ELFv2 ABI makes f14-f31 and v20-v31 callee-saved (OpenPOWER 64-bit ELF V2 ABI, 2.2.1 Register Roles), but coroutine_transfer saved only r14-r31, LR and CR. A coroutine resuming from a transfer therefore sees whatever the other side left in those registers, and GCC does keep GPR data there: it vectorizes "store the same pointer twice" as mtvsrdd vs63 / stxvx, and spills GPRs to VSX under register pressure. This never bit while nothing on either side of a transfer held a live value there. ce6f200 reshaped nt_start, and GCC 13 now keeps a pointer in v31 for the whole shared-nt loop, the transfer target of every M:N park. From then on rubyci ppc64le was red on every run: non-main Ractor threads resumed with a stale VALUE where the VM expected a block handler (rb_block_given_p() true without a block, "the block passed to ... may be ignored" warnings, SEGV in invoke_block_from_c_bh at 0x71/0x75 with the same address on every run). x86_64 has no callee-saved FP registers and the aarch64 Context.S already saves d8-d15, which is why only ppc64le failed. https://rubyci.s3.amazonaws.com/ppc64le/ruby-master/recent.html https://rubyci.s3.amazonaws.com/ppc64le/ruby-master/log/20260911T003005Z.fail.html.gz Save f14-f31 with stfd/lfd and v20-v31 with stvx/lvx (VMX only, no VSX requirement); the frame grows from 160 to 496 bytes and COROUTINE_REGISTERS from 24 to 66 words. The LR slot stays at index 18, so coroutine_initialize is otherwise unchanged. Fibers use the same transfer and are covered too. glibc's setjmp/longjmp on ppc64 already save and restore this set, so the EC tags and RB_VM_SAVE_MACHINE_CONTEXT were not affected. Verified on POWER9 (Ubuntu 24.04, gcc 13.3.0, 64K pages, the rubyci configure) at fef20ce: without this patch TestSetTraceFunc#test_tp_ractor_local_untargeted dies at 0x75 and TestEnv#test_delete_in_ractor at 0x71 as on rubyci, and TestTmpdir#test_ractor / TestEnv#test_fetch_in_ractor fail on the same warnings; with it all of them pass and bootstraptest/test_ractor.rb passes. Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
We can see some nice performance gains across the board with StickyImmix support: -------------- ------------- ------------ ---------------- ------------ ------------------- ----------------- --------------------- bench Immix (ms) RSS (MiB) StickyImmix (ms) RSS (MiB) StickyImmix 1st itr Immix/StickyImmix RSS Immix/StickyImmix activerecord 205.1 ± 1.4% 127.8 ± 0.0% 126.1 ± 3.8% 150.3 ± 0.6% 1.790 1.626 0.850 chunky-png 409.2 ± 1.6% 90.2 ± 0.1% 378.7 ± 1.0% 119.0 ± 0.4% 1.046 1.081 0.758 erubi-rails 1028.3 ± 1.0% 168.5 ± 0.0% 549.3 ± 0.8% 194.1 ± 0.0% 1.879 1.872 0.868 hexapdf 1014.9 ± 2.3% 534.7 ± 2.3% 885.5 ± 0.7% 743.4 ± 2.4% 1.075 1.146 0.719 liquid-c 35.1 ± 2.9% 93.5 ± 0.1% 25.1 ± 5.2% 107.6 ± 0.3% 1.502 1.399 0.869 liquid-compile 88.3 ± 3.4% 91.4 ± 0.1% 31.2 ± 3.9% 110.1 ± 0.3% 2.707 2.826 0.830 liquid-il 228.9 ± 1.7% 77.8 ± 0.1% 168.2 ± 0.7% 93.2 ± 0.0% 1.347 1.361 0.835 liquid-render 71.2 ± 2.0% 93.8 ± 0.1% 59.2 ± 2.4% 109.8 ± 0.2% 1.197 1.203 0.854 lobsters 425.8 ± 0.7% 387.6 ± 0.8% 368.6 ± 3.0% 514.6 ± 0.4% 1.238 1.155 0.753 mail 115.0 ± 1.7% 104.8 ± 0.0% 62.7 ± 2.2% 110.9 ± 0.1% 1.927 1.835 0.945 psych-load 1618.2 ± 1.1% 86.2 ± 0.0% 993.7 ± 0.6% 101.2 ± 0.0% 1.648 1.628 0.852 railsbench 1058.4 ± 0.4% 190.8 ± 0.0% 775.2 ± 0.5% 216.5 ± 0.0% 1.384 1.365 0.881 rubocop 84.3 ± 3.9% 169.5 ± 1.5% 74.6 ± 5.1% 202.4 ± 1.1% 1.381 1.130 0.837 ruby-lsp 106.8 ± 4.0% 126.8 ± 0.3% 75.2 ± 5.7% 151.1 ± 0.9% 1.377 1.420 0.839 sequel 54.2 ± 3.0% 81.1 ± 0.1% 27.7 ± 4.2% 92.5 ± 0.3% 2.065 1.953 0.876 shipit 667.4 ± 2.7% 240.0 ± 0.0% 630.6 ± 0.8% 301.3 ± 0.3% 1.097 1.058 0.796 -------------- ------------- ------------ ---------------- ------------ ------------------- ----------------- --------------------- ruby/mmtk@24f507513f
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
10 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )