Revision history - refs/heads/pc/codeinfo-args - origin: https://github.com/JuliaLang/julia

visit type:

Newer
Older

Revision	Author	Date	Message	Commit Date
5954572	Prem Chintalapudi	30 September 2023, 15:02:44 UTC	Handle 32-bit Int	30 September 2023, 15:02:59 UTC
36e7848	Prem Chintalapudi	25 September 2023, 01:59:41 UTC	nothing is not a boolean	30 September 2023, 15:02:59 UTC
7b98af3	Prem Chintalapudi	24 September 2023, 20:30:48 UTC	Add missing %	30 September 2023, 15:02:58 UTC
daaa20d	Prem Chintalapudi	24 September 2023, 20:19:07 UTC	Escape ssa names	30 September 2023, 15:02:58 UTC
d247782	Prem Chintalapudi	24 September 2023, 17:54:41 UTC	Fixup doctests and tests	30 September 2023, 15:02:58 UTC
e462fad	Prem Chintalapudi	24 September 2023, 14:05:37 UTC	Add arguments to codeinfo printing, prefix with %	30 September 2023, 15:02:58 UTC
d988f8f	Shuhei Kadowaki	30 September 2023, 06:44:19 UTC	[REPLCompletions] suppress false positive field completions (#51502) Field completions can be wrong when `getproperty` and `propertynames` are overloaded for a target object type, since a custom `getproperty` does not necessarily accept field names. This commit simply suppresses field completions in such cases. Fixes the second part of #51499.	30 September 2023, 06:44:19 UTC
0a4bea6	Shuhei Kadowaki	30 September 2023, 04:36:37 UTC	compiler: remove unused `:noinbounds` argument (#51506)	30 September 2023, 04:36:37 UTC
ab992b9	Rafael Fourquet	29 September 2023, 20:06:36 UTC	MersenneTwister: hash seeds like for `Xoshiro` (#51436) This addresses a part of #37165: > It's common that sequential seeds for RNGs are not as independent as one might like. This clears out this problem for `MersenneTwister`, and makes it easy to add the same feature to other RNGs via a new `hash_seed` function, which replaces `make_seed`. This is an alternative to #37766.	29 September 2023, 20:06:36 UTC
3a85776	Frames White	29 September 2023, 19:28:59 UTC	fix out of place entry in NEWS.md (#51459) `pmap` is part of the Distributed std lib	29 September 2023, 19:28:59 UTC
8436f68	adienes	29 September 2023, 15:24:10 UTC	add rand dispatch for tuple types (#50251) as suggested by https://github.com/JuliaLang/julia/issues/50236	29 September 2023, 15:24:10 UTC
8ab635d	Rafael Fourquet	29 September 2023, 10:52:59 UTC	Random: allow negative seeds (#51416) Alternative to #46190, see that PR for background. There isn't a strong use-case for accepting negative seeds, but probably many people tried something like `seed = rand(Int); seed!(rng, seed)` and saw it failing. As it's easy to support, let's do it. This might "break" some random streams, those for which the upper bit of `make_seed(seed)[end]` was set, so it's rare.	29 September 2023, 10:52:59 UTC
b74daf5	Rafael Fourquet	29 September 2023, 08:30:58 UTC	define `Random.GLOBAL_RNG = TaskLocalRNG()` (#51388) `GLOBAL_RNG` is a relic from pre-v1.3 when our global RNG was not thread-safe: ``` const GLOBAL_RNG = MersenneTwister() ``` `GLOBAL_RNG` was never really specified, but was referred to in docstrings (of `rand` etc.), and people had to use it in packages in order to provide a default RNG to their functions. So we have to keep defining `Random.GLOBAL_RNG` for the time being. In v1.3, we didn't have anymore a unique global RNG, but instead one per thread; in order to keep code using `GLOBAL_RNG` working, we got this new definition as a clever hack: ``` struct _GLOBAL_RNG <: AbstractRNG end const GLOBAL_RNG = _GLOBAL_RNG() ``` I.e. `GLOBAL_RNG` is just a "handle", which refers to the actual threaded-RNG when passed to `rand` functions. This entails defining most functions taking a `default_rng()` to also accept `_GLOBAL_RNG`. See #41123, and #41235 which is not even merged. But since v1.7, we have `Random.default_rng()` (our now official way to refer to the default RNG) returning `TaskLocalRNG()`, which is itself a "handle" (singleton). So we can as well redefine `GLOBAL_RNG` to be `TaskLocalRNG()` instead of `_GLOBAL_RNG()`, with the advantage that all the relevant methods are already defined for `TaskLocalRNG`. The only expected difference in behavior is `seed!(GLOBAL_RNG, seed)`, which previously would update `Random.GLOBAL_SEED` (a hack used by `@testset`), but now is equivalent to `seed!(TaskLocalRNG(), seed)`, which doesn't update this global seed. This precise behavior was never documented (`GLOBAL_SEED` is purely internal), so this should be fine.	29 September 2023, 08:30:58 UTC
29f2b2f	Rafael Fourquet	29 September 2023, 08:08:16 UTC	Random: reorganize some tests (#51480) This groups some toplevel tests together, and removes hardcoded random values generated by `MersenneTwister` for a given seed.	29 September 2023, 08:08:16 UTC
c099639	Katharine Hyatt	29 September 2023, 07:31:11 UTC	More tests for replace with AbstractDict (#50902) Co-authored-by: Rafael Fourquet <fourquet.rafael@gmail.com>	29 September 2023, 07:31:11 UTC
d3fb3b6	Lilith Orion Hafner	29 September 2023, 03:37:08 UTC	Update public keyword compat note (#51496) Co-authored-by: @IanButterworth Co-authored-by: Ian Butterworth <i.r.butterworth@gmail.com>	29 September 2023, 03:37:08 UTC
66fe51f	Ian Butterworth	28 September 2023, 19:57:19 UTC	Throw clearer ArgumentError for strip with two string args (#51491)	28 September 2023, 19:57:19 UTC
2d93567	Andrew Radcliffe	28 September 2023, 13:49:16 UTC	Add jump functions (`jump`, `long_jump`) for `Xoshiro` (#47743) Straightforward implementations given https://xoshiro.di.unimi.it/xoshiro256plusplus.c This is useful functionality which does not interfere with any existing code. Utility aside, it is arguable that the jump functions are necessary to complete the implementation of `xoshiro256++` tooling. In essence, given that the `xoshiro` family supports jump functions, we would be remiss to neglect this capability. For most users, I expect that the existing `TaskLocalRNG` is more than sufficient (hence, why I propose that this not be exported, just like `seed!`). However, if/when one does want to total control, jump functions are a requirement. Use cases arise when one wishes to utilize a single seed (with state advanced sufficiently far as to give non-overlapping subsequences) as the basis for a parallel computation. The alternative is manual seeding, which lacks the flexibility required for testing (imagine a program which requires a variable number of sub-sequences, one for each parallel portion). If further justification is needed, [good precedent](https://docs.rs/rand_xoshiro/latest/rand_xoshiro/struct.Xoshiro256PlusPlus.html) exists.	28 September 2023, 13:49:16 UTC
ed891d6	Nicholas Bauer	28 September 2023, 13:47:25 UTC	Remove fallback that assigns a module to inlined frames. (#51405) The work I did in #41099 introduced code which, if method information could not be found for an inlined frame, would fall back to use the module of the next-higher stack frame. This often worked there because the only frames that would not be assigned a module at this point would be e.g., `macro expansion` frames. However, due to the performance impact of the way method roots are currently encoded, the extra method roots were removed in #50546. The result is that inlined frames were being assigned a potentially incorrect module, rather than being left blank. Example: ``` julia> @btime plot([1 2 3], seriestype = :blah) ... [13] #invokelatest#2 @ BenchmarkTools ./essentials.jl:901 [inlined] [14] invokelatest @ BenchmarkTools ./essentials.jl:896 [inlined] ... ```	28 September 2023, 13:47:25 UTC
bf3eb09	Shuhei Kadowaki	28 September 2023, 13:04:21 UTC	REPLCompletions: remove unnecessary copy of top-level `CodeInfo` (#51490)	28 September 2023, 13:04:21 UTC
daeb1b4	Rafael Fourquet	28 September 2023, 12:49:20 UTC	Random.default_rng: clarify and expand documentation (#51385) Co-authored-by: Nathan Zimmerberg <39104088+nhz2@users.noreply.github.com>	28 September 2023, 12:49:20 UTC
50146b2	Daniel Karrasch	27 September 2023, 19:57:54 UTC	Handle `AbstractQ` in concatenation (#51132) Handling `AbstractQ`s in concatenation got lost in the overhaul. This brings it back (though it seemed to not be used anywhere), and even better: pre-1.9 `Q`s where handled via `AbstractMatrix` fallbacks, so elementwise. Now, it first materializes the matrix, and then copies to the right place in the destination array.	27 September 2023, 19:57:54 UTC
5e8f8a9	Gabriel Baraldi	27 September 2023, 15:17:00 UTC	codegen: add src alignment arg to emit_memcpy (#51152) This maybe allows for LLVM to optimize things a bit further since some passes check if both sides have correct/equal alignment Co-authored-by: Jameson Nash <vtjnash@gmail.com>	27 September 2023, 15:17:00 UTC
c105167	Tim Besard	27 September 2023, 15:00:57 UTC	Bump libunwind JLL to include revert. (#51477) Alternative to https://github.com/JuliaLang/julia/pull/51472	27 September 2023, 15:00:57 UTC
b6a748f	DilumAluthgeBot	27 September 2023, 13:23:50 UTC	🤖 [master] Bump the Pkg stdlib from cf0019fbb to 3960c692b (#51464)	27 September 2023, 13:23:50 UTC
7428d8a	tecosaur	27 September 2023, 11:03:49 UTC	Correct stat docstring on blocks' blocksize (#51460) The block size when reporting `blocks` is _always_ 512, regardless of the filesystem, on all systems we currently know about and support, and in particular is typically not the same as blksize (which is commonly reported as 4096). Fixes #51447	27 September 2023, 11:03:49 UTC
cde964f	Kiran	27 September 2023, 10:20:30 UTC	Fix segfault if root task is NULL (#51471) In `jl_print_task_backtraces()`. Follow-on to https://github.com/JuliaLang/julia/pull/51430.	27 September 2023, 10:20:30 UTC
0a82b71	Jameson Nash	27 September 2023, 09:31:25 UTC	inference: avoid inferring unreachable code methods (#51317)	27 September 2023, 09:31:25 UTC
8de80bd	Shuhei Kadowaki	26 September 2023, 13:55:08 UTC	inference: restrict type of type parameter from `Vararg` (#51449) Inference has been able to restrict type of `Vararg` type parameter `N` to `Int` for cases like `func(..., ::Vararg{T,N}) where {T,N}`, but this refinement was not available for signatures like `func(::Tuple{Vararg{T,N}}) where {T,N}`. This commit allows the later case to be inferred as well. Now the following kind of case will be inferred, e.g.: ```julia julia> function sub2ind_gen_fallback(dims::NTuple{N,Int}, I) where N # N is knonw to be ::Int ind = I[N] - 1 for i = (N - 1):-1:1 ind = I[i] - 1 + dims[i]*ind end return ind + 1 end; julia> only(Base.return_types(sub2ind_gen_fallback, (NTuple,Tuple{Vararg{Int}}))) Int64 ```	26 September 2023, 13:55:08 UTC
95c1dc5	Shuhei Kadowaki	26 September 2023, 13:52:34 UTC	[REPLCompletions] use inference for dict completions too (#51453) This allows us to get dict completions through nested `getindex`.	26 September 2023, 13:52:34 UTC
54bae4a	Shuhei Kadowaki	26 September 2023, 13:50:07 UTC	[REPLCompletions] get completions for `:toplevel`/`:tuple` expressions (#51454) This allows us to get completions for the following kind of cases: ```julia julia> some_issue36437 = Some(Issue36437(42)); julia> some_issue36437.value.a, some_issue36437.value. a b c julia> some_issue36437.value.a; some_issue36437.value. a b c ```	26 September 2023, 13:50:07 UTC
1db6bbd	Shuhei Kadowaki	26 September 2023, 13:46:38 UTC	add convenience entry for `typeinf_[code\|ircode\|frame]` (#51457) This change allows them to take `match::MethodMatch` and `mi::MethodInstance`, making debugging easier and reflection code a bit cleaner.	26 September 2023, 13:46:38 UTC
28d9f73	Shuhei Kadowaki	26 September 2023, 06:44:16 UTC	inference: make `throw` block deoptimization concrete-eval friendly (#49235) The deoptimization can sometimes destroy the effects analysis and disable [semi-]concrete evaluation that is otherwise possible. This is because the deoptimization was designed with the type domain profitability in mind (#35982), and hasn't been adequately considering the effects domain. This commit makes the deoptimization aware of the effects domain more and enables the `throw` block deoptimization only when the effects already known to be ineligible for concrete-evaluation. In our current effect system, `ALWAYS_FALSE`/`false` means that the effect can not be refined to `ALWAYS_TRUE`/`true` anymore (unless given user annotation later). Therefore we can enable the `throw` block deoptimization without hindering the chance of concrete-evaluation when any of the following conditions are met: - `effects.consistent === ALWAYS_FALSE` - `effects.effect_free === ALWAYS_FALSE` - `effects.terminates === false` - `effects.nonoverlayed === false` Here are some numbers: \| Metric \| master \| this commit \| #35982 reverted (set `unoptimize_throw_blocks=false`) \| \|-------------------------\|-----------\|-------------\|--------------------------------------------\| \| Base (seconds) \| 15.579300 \| 15.206645 \| 15.296319 \| \| Stdlibs (seconds) \| 17.919013 \| 17.667094 \| 17.738128 \| \| Total (seconds) \| 33.499279 \| 32.874737 \| 33.035448 \| \| Precompilation (seconds) \| 49.967516 \| 49.421121 \| 49.999998 \| \| First time `plot(rand(10,3))` [^1] \| `2.476678 seconds (11.74 M allocations)` \| `2.430355 seconds (11.77 M allocations)` \| `2.514874 seconds (11.64 M allocations)` \| \| First time `solve(prob, QNDF())(5.0)` [^2] \| `4.469492 seconds (15.32 M allocations)` \| `4.499217 seconds (15.41 M allocations)` \| `4.470772 seconds (15.38 M allocations)` \| [^1]: With disabling precompilation of Plots.jl. [^2]: With disabling precompilation of OrdinaryDiffEq. These numbers made me question if we are getting any actual benefit from the `throw` block deoptimization anymore. Since it is sometimes harmful for the effects analysis, we probably want to either merge this commit or remove the `throw` block deoptimization completely.	26 September 2023, 06:44:16 UTC
13d3efb	Troels Arnfred Bojesen	25 September 2023, 17:19:12 UTC	Fix typo in `Base.setindex` documentation (#51446) ... I also used `t` instead of `x`, since that seems to be more prevalent when talking about tuples. (I could be wrong about that, though...)	25 September 2023, 17:19:12 UTC
0287a00	Tim Besard	25 September 2023, 16:41:48 UTC	improve NTuple show and cleanup code for Vararg and NTuple type parameter construction (#51244) Various simplifications and improvements from investigating #51228. Improves the logic for showing of NTuple to handle constant lengths. Improves the logic for showing NTuple of bound length (e.g. NTuple itself). Also makes a choice to avoid showing non-types as NTuple, but instead try to write them out, to make it more visually obvious when the parameters have been swapped. --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com>	25 September 2023, 16:41:48 UTC
b44a95b	Shuhei Kadowaki	25 September 2023, 13:40:26 UTC	set `max_methods = 4` for `merge` (#51443) This enhances type/effects inference for calls of the `merge` function, especially when input types are known only as abstract `NamedTuple`. While we improved inference for the fallback `merge` path in JuliaLang/julia#48262, the enhancements have not been available to general cases as inference gives up attempts when there are ≥4 matching methods. This commit fixes it up.	25 September 2023, 13:40:26 UTC
3220565	Shuhei Kadowaki	25 September 2023, 13:38:05 UTC	reflection: remove the `func_for_method_checked` check (#51444) Introduced in 539224f, the `func_for_method_checked` check has been used for some of the reflection utilities like `code_typed`, but now it appears to be outdated. E.g. previously, when provided with abstract input types for optionally-generated functions, they would raise unnecessary errors, even though the inference would utilize the fallback code path. In such instances, the reflection utilities should present what the inference would perform, instead of raising errors. This commit simply removes the check. This works because all the reflection utilities now internally calls the `InferenceState()` constructor, where code generation occurs, and they gracefully handle cases when code generation fails. Now their behavior is more synced with the actual behavior of inference.	25 September 2023, 13:38:05 UTC
e5c6340	Jameson Nash	25 September 2023, 11:20:19 UTC	add support for async backtraces of Tasks on any thread (#51430)	25 September 2023, 11:20:19 UTC
66e83e7	Kristoffer Carlsson	25 September 2023, 10:34:25 UTC	remove "skipping mtime check..." debug output (#51441) This doesn't really serve any purpose to have and it gets very spammy.	25 September 2023, 10:34:25 UTC
f7d8bfb	Jameson Nash	22 September 2023, 21:18:24 UTC	deal with remaining race conditions in these APIs The jl_live_tasks API now reports all threads, instead of only Tasks first started by the current thread. There is a new abstraction called mtarraylist with adds functionality to small_arraylist (it is layout-compatible). In particular, it makes it safe for another thread to observe the content of the list concurrently with any mutations.	24 September 2023, 17:28:56 UTC