[pull] master from ruby:master - #1436
Merged
Merged
Conversation
When running benchmarks, sometimes there is a overhead in the statistics collection, so it can be useful to disable `zjit-stats`. In my case I was investigating Shopify#1051 and was confused why I was not seeing any new instruction being generated in the LIR but there was still performance difference when running zjit-diff. The cause for the performance difference in that PR was new instructions to increment stats counters. So this PR allow running the tool without stats enabled. I'm even wondering if the default should not be with it disabled.
We should not ever hit this from Ruby code; user code cannot modify ivars that do not start with @. This makes `num_members` leaf and therefore Struct class allocation also leaf.
Now fibers can be collected by multiple ractors at once when those Ractors are running local GCs. Fibers are created in the Ractor that owns them, so if multiple Ractors have unmarked fibers during local GC sweeping, we can now collect them concurrently instead of deferring their collection to a postponed job. Each fiber pool is protected by its own native mutex. In the future, maybe fiber pools will be ractor-local, but we have to experiment with it first and see if it's enough of a performance win to justify the change. We also want to minimize resource acquisition for each Ractor as they're supposed to be lighweight.
Preserve `ELOOP` instead of hiding symbolic-link resolution errors with the file-enumeration fallback, so existence checks return false.
Cover equal and different file contents in `test_cmp` and verify that `compare_file` and `identical?` remain aliases. Fix ruby/fileutils#180 Fix ruby/fileutils#184 ruby/fileutils@01ac124ce4
Cover directory changes, block behavior, and verbose output through `cd`, and verify that `chdir` remains an alias. Fix ruby/fileutils#182 ruby/fileutils@8476509760
Avoid duplicating working-directory coverage already in `test_pwd`. Fix ruby/fileutils#183 ruby/fileutils@d937a3d5ba
Verify that `ln` creates hard links and that `link` remains an alias. Fix ruby/fileutils#185 ruby/fileutils@46db9d557b
Check `mkdir_p` return values and verify the `makedirs` and `mkpath` aliases without duplicating directory creation tests. Fix ruby/fileutils#186 ruby/fileutils@72ff748e26
Check the result and effects of `mv` and verify the `move` alias. Fix ruby/fileutils#187 ruby/fileutils@3bed6e74a1
Cover removal of existing and missing trees through `rm_rf` and verify the `rmtree` alias. Fix ruby/fileutils#188 ruby/fileutils@07cc858f16
Cover removal of existing and missing files through `rm_f` and verify the `safe_unlink` alias. Fix ruby/fileutils#189 ruby/fileutils@902bcce36b
Cover single, multiple, and dangling links through `ln_s`, keeping valid-link tests available where dangling links are unsupported. Fix ruby/fileutils#190 ruby/fileutils@189aefb679
Windows reports self-referential links as existing, so use a missing target to test a dangling link rather than a symbolic-link loop. ruby/fileutils@803fe3be07
I saw that YJIT has a fast-path for this so I figured we should too.
`rb_obj_dup` skips immediates:
```c
VALUE
rb_obj_dup(VALUE obj)
{
VALUE dup;
if (special_object_p(obj)) {
return obj;
}
// ...
}
static inline int
special_object_p(VALUE obj)
{
if (SPECIAL_CONST_P(obj)) return TRUE;
switch (BUILTIN_TYPE(obj)) {
case T_BIGNUM:
case T_FLOAT:
case T_SYMBOL:
case T_RATIONAL:
case T_COMPLEX:
/* not a comprehensive list */
return TRUE;
default:
return FALSE;
}
}
```
(ruby/prism#4232) Followup to ruby/prism@ba16ae2 I can't run rake `cargo:test`: ``` $ cargo build Compiling ruby-prism-sys v1.9.0 (ruby-prism/rust/ruby-prism-sys) Compiling ruby-prism v1.9.0 (ruby-prism/rust/ruby-prism) warning: ruby-prism-sys@1.9.0: In file included from ruby-prism/rust/ruby-prism-sys/vendor/prism-1.9.0/include/prism/ast.h:22, warning: ruby-prism-sys@1.9.0: from ruby-prism/rust/ruby-prism-sys/vendor/prism-1.9.0/include/prism/internal/static_literals.h:6, warning: ruby-prism-sys@1.9.0: from ruby-prism/rust/ruby-prism-sys/vendor/prism-1.9.0/src/static_literals.c:1: warning: ruby-prism-sys@1.9.0: ruby-prism/rust/ruby-prism-sys/vendor/prism-1.9.0/include/prism/arena.h:28:1: warning: 'nodiscard' attribute ignored [-Wattributes] warning: ruby-prism-sys@1.9.0: 28 | PRISM_EXPORTED_FUNCTION PRISM_NODISCARD pm_arena_t * pm_arena_new(void); warning: ruby-prism-sys@1.9.0: | ^~~~~~~~~~~~~~~~~~~~~~~ warning: ruby-prism-sys@1.9.0: ruby-prism/rust/ruby-prism-sys/vendor/prism-1.9.0/include/prism/arena.h:28:41: error: expected identifier or '(' before 'pm_arena_t' warning: ruby-prism-sys@1.9.0: 28 | PRISM_EXPORTED_FUNCTION PRISM_NODISCARD pm_arena_t * pm_arena_new(void); warning: ruby-prism-sys@1.9.0: | ^~~~~~~~~~ error: failed to run custom build command for `ruby-prism-sys v1.9.0 (ruby-prism/rust/ruby-prism-sys)` ``` It expands to `__attribute__((__visibility__("default"))) extern [[nodiscard]] pm_arena_t * pm_arena_new(void);`, which is not valid in that order. ruby/prism@3942928d8f
We can use `ObjectAllocClass` for Structs because they are basically a `__members__` ivar lookup for sizing plus the default allocator. Requires #18888 to make sure the constructor does not raise.
We previously only had this in `gen_insn` which left terminators un-commented.
It's an abrupt end to a block, but it definitely ends the block. No need to put a return instruction after it.
We use `HasType`+`CondBranch` a lot—it is how we do polymorphic dispatch in HIR—so it's pretty hot. Without smarter instruction fusion/instruction selection, this generates a lot of code. Kokubun identified this in Shopify#1039 Right now, `HasType` materializes a boolean just for `CondBranch` to go check it again. Since we only use `HasType` for `CondBranch` today, make a new fused instruction. This reduces interstitial values and also intermediate double branching. See for example this polymorphic method dispatch and its generated HIR and asm. Source: ```ruby class C def foo = 3 end class D def foo = 4 end def test(o) = o.foo ``` Before HIR: ``` Optimized HIR: fn test@../tmp/poly.rb:9: bb1(): EntryPoint interpreter v1:BasicObject = LoadSelf v2:CPtr = LoadSP v3:BasicObject = LoadField v2, :o@-0x20 Jump bb3(v1, v3) bb2(): EntryPoint JIT(0) v6:BasicObject = LoadArg :self@0 v7:BasicObject = LoadArg :o@1 Jump bb3(v6, v7) bb3(v9:BasicObject, v10:BasicObject): v16:CBool = HasType v10, ObjectSubclass[class_exact:C] CondBranch v16, bb5(), bb6() bb5(): PatchPoint NoSingletonClass(C@0x124574200) PatchPoint MethodRedefined(C@0x124574200, foo@0xec31, cme:0x12455e298) v37:Fixnum[3] = Const Value(3) Jump bb4(v37) bb6(): v22:CBool = HasType v10, ObjectSubclass[class_exact:D] CondBranch v22, bb7(), bb8() bb7(): PatchPoint NoSingletonClass(D@0x124574340) PatchPoint MethodRedefined(D@0x124574340, foo@0xec31, cme:0x12455e2e8) v40:Fixnum[4] = Const Value(4) Jump bb4(v40) bb8(): v28:BasicObject = Send v10, :foo # SendFallbackReason: Send: polymorphic call site Jump bb4(v28) bb4(v15:BasicObject): CheckInterrupts Return v15 ``` Before asm: ``` # Insn: v16 HasType v10, ObjectSubclass[class_exact:C] 0x124f4012c: tst x0, #7 0x124f40130: b.ne #0x124f4016c 0x124f40134: cmp x0, #0 0x124f40138: b.eq #0x124f40164 0x124f4013c: ldur x1, [x0, #8] 0x124f40140: ldr x2, #0x124f40148 0x124f40144: b #0x124f40150 0x124f40148: sel z0.d, p0, z26.d, z16.d 0x124f4014c: udf #1 0x124f40150: cmp x1, x2 0x124f40154: mov x1, #1 0x124f40158: mov x2, #0 0x124f4015c: csel x1, x1, x2, eq 0x124f40160: b #0x124f40170 0x124f40164: mov x1, #0 0x124f40168: b #0x124f40170 0x124f4016c: mov x1, #0 # Insn: v17 CondBranch v16, bb5(), bb6() 0x124f40170: tst x1, x1 0x124f40174: b.ne #0x124f40228 ``` After HIR: ``` Optimized HIR: fn test@../tmp/poly.rb:9: bb1(): EntryPoint interpreter v1:BasicObject = LoadSelf v2:CPtr = LoadSP v3:BasicObject = LoadField v2, :o@-0x20 Jump bb3(v1, v3) bb2(): EntryPoint JIT(0) v6:BasicObject = LoadArg :self@0 v7:BasicObject = LoadArg :o@1 Jump bb3(v6, v7) bb3(v9:BasicObject, v10:BasicObject): CondBranchHasType v10, ObjectSubclass[class_exact:C], bb5(), bb6() bb5(): PatchPoint NoSingletonClass(C@0x11fd14140) PatchPoint MethodRedefined(C@0x11fd14140, foo@0xec31, cme:0x11fcfe2f0) v35:Fixnum[3] = Const Value(3) Jump bb4(v35) bb6(): CondBranchHasType v10, ObjectSubclass[class_exact:D], bb7(), bb8() bb7(): PatchPoint NoSingletonClass(D@0x11fd14280) PatchPoint MethodRedefined(D@0x11fd14280, foo@0xec31, cme:0x11fcfe340) v38:Fixnum[4] = Const Value(4) Jump bb4(v38) bb8(): v26:BasicObject = Send v10, :foo # SendFallbackReason: Send: polymorphic call site Jump bb4(v26) bb4(v15:BasicObject): CheckInterrupts Return v15 ``` After asm: ``` # Insn: v16 CondBranchHasType v10, ObjectSubclass[class_exact:C], bb5(), bb6() 0x127e0412c: tst x0, #7 0x127e04130: b.ne #0x127e04168 0x127e04134: cmp x0, #0 0x127e04138: b.eq #0x127e04168 0x127e0413c: ldur x1, [x0, #8] 0x127e04140: ldr x2, #0x127e04148 0x127e04144: b #0x127e04150 0x127e04148: sel z0.d, p0, z25.d, z28.d 0x127e0414c: udf #1 0x127e04150: cmp x1, x2 0x127e04154: b.eq #0x127e0415c 0x127e04158: b #0x127e04168 ```
The precision in Time.new calls to_int which can modify the string passed
into Time.new. This can cause an use-after-free. For example, the following
script crashes:
str = "2000-01-01 00:00:00" + "0" * 1_000_000
obj = Object.new
obj.define_singleton_method(:to_int) { str.clear; 9 }
Time.new(str, precision: obj)
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.
7 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 : )