07b47a0 | Jeff Bezanson | 23 May 2020, 21:51:11 UTC | make `getindex(::AbstractRange, i)` specific to `StepRange` The (i > 0) check and (i - 1) code imply this assumes the range is indexed 1:len, which may not be true for all subtypes of AbstractRange. Also fix its boundscheck block. | 02 December 2022, 11:20:58 UTC |
89671ae | Philip Swannell | 02 December 2022, 07:39:10 UTC | doc: fix a typo in the FAQ (#47767) | 02 December 2022, 07:39:10 UTC |
b572839 | Shuhei Kadowaki | 02 December 2022, 07:07:46 UTC | minor perf optimization on `abstract_call` (#47765) | 02 December 2022, 07:07:46 UTC |
cc25a13 | Simeon Schaub | 01 December 2022, 20:42:20 UTC | fix unescaping in `global` expressions (#47719) This fixes some issues around macro hygiene in `global` expressions. Apparently we always treat l-values in global expressions as being escaped, but we still need to be careful to handle type annotations and destructuring correctly. | 01 December 2022, 20:42:20 UTC |
293ab47 | Dilum Aluthge | 01 December 2022, 16:48:27 UTC | Update the libuv checksums (#47763) | 01 December 2022, 16:48:27 UTC |
8a75bcb | Oscar Smith | 01 December 2022, 15:06:56 UTC | faster `BigFloat(::Float64)` (#47546) * faster `BigFloat(::Float64)` This is roughly 2x faster for precision<10000. It's a little slower for precisions >10^6 (about 20%) but IMO this doesn't really matter since for high precision things like multiplication will be way slower than construction. | 01 December 2022, 15:06:56 UTC |
a9e6b5a | Shuhei Kadowaki | 01 December 2022, 09:48:24 UTC | inlining: follow up #47371, remove unused `override_effects` keyword arg (#47762) This keyword argument was introduced by #47305, which was then reverted by #47371. Now it's dead, so let's remove it. | 01 December 2022, 09:48:24 UTC |
04a4edb | Fons van der Plas | 30 November 2022, 21:28:15 UTC | Use `checked_mul` in `length(::ProductIterator)` (#43429) Co-authored-by: Steven G. Johnson <stevenj@mit.edu> | 30 November 2022, 21:28:15 UTC |
c46834b | Shuhei Kadowaki | 30 November 2022, 20:32:46 UTC | follow up #47739 (#47746) | 30 November 2022, 20:32:46 UTC |
7842335 | Shuhei Kadowaki | 30 November 2022, 20:32:32 UTC | add `raise` optional keyword argument to `Base._which` (#47747) | 30 November 2022, 20:32:32 UTC |
f4534d1 | Jameson Nash | 30 November 2022, 19:19:17 UTC | strengthen setglobal to default to release-consume ordering (#47742) In looking at a TSAN report recently, I noticed that globals were getting stored as atomic-unordered (since c92ab5e79ea #44182), instead of atomic-release as intended (since 46135dfce9074e5bf94eb277de28a33cad9cc14f #45484). | 30 November 2022, 19:19:17 UTC |
060a492 | Max Horn | 30 November 2022, 17:16:18 UTC | Provider cycleclock() for 32bit ARM targets (#47358) Based on https://github.com/google/benchmark/blob/main/src/cycleclock.h | 30 November 2022, 17:16:18 UTC |
ee54dd5 | Daniel Karrasch | 30 November 2022, 13:13:41 UTC | reduce ambiguity in *diagonal multiplication code (#47683) | 30 November 2022, 13:13:41 UTC |
5da8d5f | Michael Abbott | 30 November 2022, 11:11:07 UTC | Add a warning about inadvertent Diagonal(::Matrix) (#46864) | 30 November 2022, 11:11:07 UTC |
41b73e2 | Tim Besard | 30 November 2022, 06:37:13 UTC | Bump libuv. (#47707) Adapts to the final version of the constrained/available memory APIs. | 30 November 2022, 06:37:13 UTC |
70bda2c | Valentin Churavy | 30 November 2022, 01:44:53 UTC | Merge pull request #44527 from JuliaLang/teh-vc/serialize_partial Support external linkage in "sysimages" | 30 November 2022, 01:44:53 UTC |
328dd57 | Keno Fischer | 29 November 2022, 23:48:15 UTC | Fix generator-invocation legality check for varargs generators (#47739) This code was introduced by me back in #31025 to speed up evaluation of generated functions that didn't make use of all of their arguments to make generation decisions. However, it neglected to take into account the possibility that the generator could be varargs. As a result, an unfortunate coincidence of an unused slot in the correct position could have allowed expansion of generators that were not supposed to be expandable. This can cause incorrect inference with all the usual consequences. However, fortunately this coincidence appears to be pretty rare. Fixes https://github.com/JuliaDebug/CassetteOverlay.jl/issues/12 | 29 November 2022, 23:48:15 UTC |
e06a591 | Valentin Churavy | 27 November 2022, 23:26:31 UTC | Allow re-initialization and caching of foreign types (#47407) Co-authored-by: Tim Holy <tim.holy@gmail.com> Co-authored-by: Max Horn <max@quendi.de> | 29 November 2022, 20:56:22 UTC |
cbfdb3f | Tim Holy | 07 February 2022, 11:19:24 UTC | Replace the `.ji` serialization with sysimage format This unifies two serializers, `dump.c` (used for packages) and `staticdata.c` (used for system images). It adopts the `staticdata` strategy, adding support for external linkage, uniquing of MethodInstances & types, method extensions, external specializations, and invalidation. This lays the groundwork for native code caching as done with system images. Co-authored-by: Valentin Churavy <v.churavy@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Tim Holy <tim.holy@gmail.com> | 29 November 2022, 20:56:22 UTC |
54a9c2f | Oscar Smith | 29 November 2022, 16:36:50 UTC | improve inlining cost analysis for invoke (#47671) | 29 November 2022, 16:36:50 UTC |
edc8241 | Dilum Aluthge | 29 November 2022, 07:38:07 UTC | Revert "improve @(s)printf docstrings (#47583)" (#47734) This reverts commit 4d5fd91463ee59e759e627b3d6fc2ef7c6cdb22c. | 29 November 2022, 07:38:07 UTC |
4d5fd91 | st-- | 29 November 2022, 07:13:43 UTC | improve @(s)printf docstrings (#47583) | 29 November 2022, 07:13:43 UTC |
4fd26ba | Shuhei Kadowaki | 29 November 2022, 04:19:25 UTC | reflection: support additional call syntaxes for `@invoke[latest]` (#47705) Like `@invoke (xs::Xs)[i::I] = v::V` and `@invokelatest x.f = v`. Co-Authored-By: Jameson Nash <vtjnash@gmail.com> | 29 November 2022, 04:19:25 UTC |
f6e911a | Shuhei Kadowaki | 29 November 2022, 04:18:07 UTC | make `ml_matches` return `nothing` when there are too many matches (#44478) Continuation from <https://github.com/JuliaLang/julia/pull/44448#discussion_r820025006>. Previously `ml_matches` (and its dependent utilities like `_methods_by_ftype`) returned `false` for cases when there are too many matching method, but its consumer like `findall(::Type, ::InternalMethodTable)` usually handles such case as `missing`. This commit does a refactor so that they all use the more consistent value `nothing` for representing that case. | 29 November 2022, 04:18:07 UTC |
865007d | Logan Kilpatrick | 29 November 2022, 00:05:51 UTC | Add section headings to the Doc's home page & links (#47646) * Add sections heading to the Doc's home & links Co-authored-by: Viral B. Shah <ViralBShah@users.noreply.github.com> Co-authored-by: Fredrik Ekre <ekrefredrik@gmail.com> | 29 November 2022, 00:05:51 UTC |
2642c4e | Shuhei Kadowaki | 28 November 2022, 23:12:37 UTC | allow `Base._which` to take overlay method table (#47722) | 28 November 2022, 23:12:37 UTC |
1badb9d | Allen Hill | 28 November 2022, 22:36:07 UTC | doc: add example of unwrapping `Some(nothing)` in `something` docstring (#47682) | 28 November 2022, 22:36:07 UTC |
3f9409c | apaz | 28 November 2022, 20:33:45 UTC | Fix nth_methtable tparam of -1 when n==0 (#47666) Fixes #47625 | 28 November 2022, 20:33:45 UTC |
c8ea33d | Jishnu Bhattacharya | 28 November 2022, 17:50:17 UTC | Add link to alloc profiler issue in Profile docs (#47728) | 28 November 2022, 17:50:17 UTC |
902e8a7 | Hendrik Ranocha | 28 November 2022, 11:17:27 UTC | fix 5-arg `mul!` for vectors of vectors (#47665) Co-authored-by: N5N3 <2642243996@qq.com> | 28 November 2022, 11:17:27 UTC |
ea23403 | Daniel Karrasch | 28 November 2022, 06:48:17 UTC | Return early in `axp[b]y` if alpha and beta is zero, fixes #47531. (#47557) | 28 November 2022, 06:48:17 UTC |
20d13da | C. Brenhin Keller | 28 November 2022, 06:45:13 UTC | Reuse `du2` of `Tridiagonal` matrix for pivoting in `lu!` if extant (#47564) | 28 November 2022, 06:45:13 UTC |
7514bcf | Rashid Rafeek | 28 November 2022, 06:44:05 UTC | Fix REPL keybinding CTRL-Q for stdlib methods (#47637) | 28 November 2022, 06:44:05 UTC |
99be295 | Adrian Hill | 28 November 2022, 06:28:45 UTC | Fix length of Markdown header underlines (#47708) | 28 November 2022, 06:28:45 UTC |
b34fe1d | Jakob Nybo Nissen | 28 November 2022, 06:26:07 UTC | Allow crc32c hashing of SubString{String} (#47693) | 28 November 2022, 06:26:07 UTC |
f9662b8 | Eugene Toder | 28 November 2022, 06:24:11 UTC | Remove dead code in Base.retry (#47718) y === nothing is always false inside of while y !== nothing loop. | 28 November 2022, 06:24:11 UTC |
42ac4b1 | t-bltg | 28 November 2022, 06:18:13 UTC | `Test`: fix `allowed_undefineds` typo (#47572) | 28 November 2022, 06:18:13 UTC |
5495b8d | Valentin Churavy | 27 November 2022, 23:21:46 UTC | Fix GCExt test (#47699) * Add test/gcext to out-of-tree * Disable gcext test that uses jl_gc_internal_obj_base_ptr | 27 November 2022, 23:21:46 UTC |
60668c5 | pchintalapudi | 27 November 2022, 01:35:55 UTC | Add loose compilation time test (#47716) | 27 November 2022, 01:35:55 UTC |
88a0627 | pchintalapudi | 26 November 2022, 09:58:10 UTC | Fix and simplify inference timing logic (#47711) * Fix and simplify inference timing logic * Reduce task struct size | 26 November 2022, 09:58:10 UTC |
02aa0b0 | Lilith Orion Hafner | 26 November 2022, 00:47:30 UTC | Fix overflow in pow5 (#47511) Fixup for #46764 | 26 November 2022, 00:47:30 UTC |
04214ec | Kristoffer Carlsson | 25 November 2022, 10:49:03 UTC | add string literal continuation syntax to docs for `"` (#47690) | 25 November 2022, 10:49:03 UTC |
9d25932 | Shuhei Kadowaki | 25 November 2022, 09:14:36 UTC | sort out argument order of `maybe_erase_unused!` (#47701) | 25 November 2022, 09:14:36 UTC |
039d8fd | Shuhei Kadowaki | 25 November 2022, 09:11:50 UTC | effects: add `:removable` convenient setting for `Base.@assume_effects` (#47700) This setting is similar to `:foldable` but for dead call elimination. | 25 November 2022, 09:11:50 UTC |
d0a211a | Elliot Saba | 24 November 2022, 15:42:49 UTC | Filesystem: `rm(; recursive=true)` should ignore `UV_EACCES` (#47668) The command-line program `rm` has no problem deleting an empty directory that we do not have listing permissions on, so we should follow suit. Example: ``` mktempdir() do dir mkpath("$(dir)/foo") chmod("$(dir)/foo", 0o200) rm(dir; recursive=true) end ``` | 24 November 2022, 15:42:49 UTC |
726bbd7 | Tim Besard | 24 November 2022, 11:30:00 UTC | Fix regression in generic_bitcast with Union{} arguments. (#47605) | 24 November 2022, 11:30:00 UTC |
25b2746 | Daniel Karrasch | 24 November 2022, 09:33:32 UTC | LinearAlgebra: Speed up the trace function (#47585) | 24 November 2022, 09:33:32 UTC |
113efb6 | pchintalapudi | 23 November 2022, 22:11:39 UTC | Remove typeinfer lock altogether (#46825) * Remove typeinfer lock altogether * Don't remove the typeinf lock functions * Track reentrancy in current task state * Fix up some git status * Initialize task variables * Promise that jl_typeinf_func is rooted somewhere | 23 November 2022, 22:11:39 UTC |
27ebaa7 | Gabriel Baraldi | 23 November 2022, 19:26:11 UTC | Print the detailed type on heap snapshot (#47503) Fixes https://github.com/JuliaLang/julia/issues/47502 | 23 November 2022, 19:26:11 UTC |
3200219 | Petr Vana | 23 November 2022, 04:02:38 UTC | build: add get-lld target (#47589) Fixes `make -C deps getall` | 23 November 2022, 04:02:38 UTC |
95165bb | Keno Fischer | 22 November 2022, 23:48:33 UTC | Add an inference option to assume no new bindings being added (#47674) This is currently intended for use with external AbstractInterpreters that want to be able to constant fold `:isdefined` away. We always support the `true` case, since we don't support removing bindings, but the `false` case is currently not foldable. In the future, we might partition bindings by world age, in which case we would likely get this for free, but for now, just add this as a hook for external AbstractInterpreters to use. | 22 November 2022, 23:48:33 UTC |
7262534 | st-- | 22 November 2022, 10:28:02 UTC | Fix mapreduce_first docstring error (#42832) The docstring of Base.mapreduce_first referred `reduce_first(op, f, x)`, but `reduce_first` is a 2-argument function, and it should refer to `mapreduce_first(f, op, x)` instead. | 22 November 2022, 10:28:02 UTC |
4fa07cd | Michael Abbott | 22 November 2022, 01:25:15 UTC | Add compat note for `sortperm(x; dims)` (#47657) | 22 November 2022, 01:25:15 UTC |
3966d5c | Shuhei Kadowaki | 21 November 2022, 22:44:17 UTC | minor NFC cleanups on slot2ssa (#47652) | 21 November 2022, 22:44:17 UTC |
5996520 | Simon Byrne | 21 November 2022, 21:31:33 UTC | update MPFR (#47659) checksums updated via approach described in #47174. | 21 November 2022, 21:31:33 UTC |
1e2ba33 | Keno Fischer | 21 November 2022, 20:28:29 UTC | Cleanup: Use `issingletontype` consistently (#47618) Rather than reaching for `isdefined(x, :instance)`. They are currently equivalent, but I was playing with an extension that would have made them not be. I don't currently have plans to finish that particular PR, but this cleanup seems good regardless to specify exactly what is meant rather than reaching for the implementation. | 21 November 2022, 20:28:29 UTC |
23160a1 | Shuhei Kadowaki | 21 November 2022, 19:43:36 UTC | correct throw `ArgumentError` from `lbt_find_backing_library` (#47651) | 21 November 2022, 19:43:36 UTC |
ab262e7 | N5N3 | 21 November 2022, 18:27:33 UTC | Unroll `foldl/r` at julia level for small `NamedTuple` (#47109) And make `Tuple` dispatched similarly. This commit also renames `_reverse`, as it has a non-related 2-arg version in `Base`. | 21 November 2022, 18:27:33 UTC |
c5fe17b | Petr Vana | 21 November 2022, 14:45:21 UTC | Doc: The default sorting alg. is stable from 1.9 (#47579) * Update doc/src/base/sort.md * Update docs: The default sorting alg. is stable * Compat 1.9 for QuickSort to be stable * Specify the default algorithm * Use example from InlineStrings.jl * Change example to jldoctest * Remove "*appear* to be stable." as slightly misleading. Co-authored-by: Lilith Orion Hafner <lilithhafner@gmail.com> | 21 November 2022, 14:45:21 UTC |
d18fd47 | Gabriel Baraldi | 21 November 2022, 12:34:31 UTC | Make DemoteFloat16 a conditional pass (#43327) * add TargetMachine check * Add initial float16 multiversioning stuff * make check more robust and remove x86 check * move check to inside the pass * C++ is hard * Comment out the ckeck because it won't work inside the pass * whitespace in the comment * Change the logic not to depend on a TM * Add preliminary support for x86 test * Cosmetic changes | 21 November 2022, 12:34:31 UTC |
c9eccfc | Keno Fischer | 21 November 2022, 04:16:40 UTC | Algorithmic improvements for LateLowerGC pass (#47634) On a particularly pathological case I'm looking at (~100k calls, each a safepoint, similar number of allocations), I'm seeing LateLowerGC pass take the majority of the middle end time (around 80s out of 100s). This PR goes through and improves some of the algorithmics and data structure choices and gives a roughly 10x improvement on this particular pathological case (down to about 7.6s on my machine). That still represents about 1/3 of the total middle end time, which I'm not happy about, but perhaps that is reasonable given how pathological my particular test case is. For comparison, register allocation (which needs to solve a somewhat similar problem) on this IR takes about 20s. | 21 November 2022, 04:16:40 UTC |
67b8ac0 | Jameson Nash | 20 November 2022, 22:24:52 UTC | build: separate stdlib manifests by version (#47596) We install these by version, so it helps a lot if they reflect that in the manifest name. Also some other mild cleanup. | 20 November 2022, 22:24:52 UTC |
bba41d4 | Valentin Churavy | 20 November 2022, 19:45:20 UTC | Turn on Intel jitevents by default on Linux (#47586) | 20 November 2022, 19:45:20 UTC |
90934ff | Gabriel Baraldi | 19 November 2022, 21:56:42 UTC | Initialize padding in datatype layout (#47640) We use that memory for hashing it, so we need to ensure it is set to zero. | 19 November 2022, 21:56:42 UTC |
185b583 | Valentin Churavy | 19 November 2022, 11:14:12 UTC | Revert "Improve performance of toplevel code (#47578)" (#47635) This reverts commit 526cbf7260c4b31a7f83c27e56694096185519bb. | 19 November 2022, 11:14:12 UTC |
1d8f7e0 | Shuhei Kadowaki | 19 November 2022, 06:22:50 UTC | inference: implement type-based alias analysis to refine constrained field (#41199) This commit tries to propagate constraints imposed on object fields, e.g.: ```julia struct SomeX{T} x::Union{Nothing,T} end mutable struct MutableSomeX{T} const x::Union{Nothing,T} end let # o1::SomeX{T}, o2::MutableSomeX{T} if !isnothing(o1.x) # now inference knows `o1.x::T` here ... if !isnothing(o2.x) # now inference knows `o2.x::T` here ... end end end ``` The idea is that we can make `isa` and `===` propagate constraint imposed on an object field if the _identity_ of that object. We can have such a lattice element that wraps return type of abstract `getfield` call together with the object _identity_, and then we can form a conditional constraint that propagates the refinement information imposed on the object field when we see `isa`/`===` applied the return value of the preceding `getfield` call. So this PR defines the new lattice element called `MustAlias` (and also `InterMustAlias`, which just works in a similar way to `InterConditional`), which may be formed upon `getfield` inference to hold the retrieved type of the field and track the _identity_ of the object (in inference, "object identity" can be represented as a `SlotNumber`). This PR also implements the new logic in `abstract_call_builtin` so that `isa` and `===` can form a conditional constraint (i.e. `Conditional`) from `MustAlias`-argument that may later refine the wrapped object to `PartialStruct` that holds the refined field type information. One important note here is, `MustAlias` expects the invariant that the field of wrapped slot object never changes. The biggest limitation with this invariant is that it can't propagate constraints imposed on mutable fields, because inference currently doesn't have a precise (per-object) knowledge of memory effect. | 19 November 2022, 06:22:50 UTC |
6707077 | Oscar Smith | 19 November 2022, 06:20:27 UTC | fix tab completion of invalid identifiers (#47594) fix #47593 | 19 November 2022, 06:20:27 UTC |
8201e8a | caipengxiang | 19 November 2022, 06:19:53 UTC | use jl_phicnode_type to create phicnode struct (#47600) Co-authored-by: pengxiangcai <pxcai@abeliancap.com> | 19 November 2022, 06:19:53 UTC |
e020ac8 | caipengxiang | 19 November 2022, 06:18:28 UTC | the variable assert should before it being used (#47624) Co-authored-by: pengxiangcai <pxcai@abeliancap.com> | 19 November 2022, 06:18:28 UTC |
ef6974d | Keno Fischer | 19 November 2022, 05:19:27 UTC | Add an option to not run compiler in IR2OC (#47633) I have a case where I'm generating an opaque closure that is only run once, but LLVM takes 200s to compile it. This adds an option to not run LLVM in the IR2OC path, which is useful for two things: 1. Getting a handle on the OC without waiting for LLVM to finish, so we can have it dump out the unoptimized LLVM IR for profiling purposes. 2. Just actually not running LLVM, since it's probably not worth it. | 19 November 2022, 05:19:27 UTC |
553047f | Shuhei Kadowaki | 19 November 2022, 00:56:02 UTC | InteractiveUtils: use quoted `typesof` in macroexpansion (#47626) fix #47606 | 19 November 2022, 00:56:02 UTC |
d12ac36 | Francesco Alemanno | 18 November 2022, 06:54:15 UTC | fix STABLE version number in README (#47582) | 18 November 2022, 06:54:15 UTC |
526cbf7 | Tim Besard | 17 November 2022, 22:07:28 UTC | Improve performance of toplevel code (#47578) * Demote the atomic world age load to monotonic. * Only load the world age before call instructions. | 17 November 2022, 22:07:28 UTC |
5f0da83 | Lilith Orion Hafner | 17 November 2022, 15:58:16 UTC | Fix error in docstring and improve comment in gcdx (#47573) The names `x` and `y` are internal. | 17 November 2022, 15:58:16 UTC |
e5ed5be | Oscar Smith | 17 November 2022, 13:57:01 UTC | Improve effect analysis of bitshifts by non `Int` and `UInt` `Integer`s (#47567) *Improve effect analysis of bitshifts by non Int and UInt Integers Co-authored-by: Thomas Christensen <tchr@mit.edu> | 17 November 2022, 13:57:01 UTC |
3e7d796 | Shuhei Kadowaki | 16 November 2022, 23:01:26 UTC | optimizer: exclude `ConditionalsLattice` from the optimizer lattice (#47575) Since `Conditional`s (should) never appear within the optimization. Also added a missing widening of `Conditional`s in `slottypes` (a.k.a. `argtypes` in optimization) so that they never appear in the optimization. | 16 November 2022, 23:01:26 UTC |
b369511 | Jameson Nash | 16 November 2022, 21:17:00 UTC | ensure bindings handle write barriers for ty and globalref (#47580) This has probably been wrong for a long time (since being introduced in 79082468986). | 16 November 2022, 21:17:00 UTC |
5f256e7 | Jeff Bezanson | 16 November 2022, 16:17:34 UTC | fix #47410, syntax error with anonfn inside `elseif` and short-circuit op (#47499) | 16 November 2022, 16:17:34 UTC |
ee0f3fc | Shuhei Kadowaki | 16 November 2022, 00:27:26 UTC | lattice: fix minor lattice issues (#47570) I found some lattice issues when implementing `MustAlias` under the new extendable lattice system in another PR. | 16 November 2022, 00:27:26 UTC |
e805a18 | Shuhei Kadowaki | 16 November 2022, 00:27:05 UTC | minor compiler cleanups (#47571) | 16 November 2022, 00:27:05 UTC |
fe81138 | Jeff Bezanson | 16 November 2022, 00:01:25 UTC | fix #46778, precompile() for abstract but compileable signatures (#47259) | 16 November 2022, 00:01:25 UTC |
9b3f5c3 | Jeff Bezanson | 15 November 2022, 23:32:29 UTC | export `jl_gc_set_max_memory` (#47545) Also initialize it later, outside option parsing, so that modifying jl_options before calling jl_init works. | 15 November 2022, 23:32:29 UTC |
9413050 | Alex Arslan | 15 November 2022, 23:29:30 UTC | Define `parentmodule(::Method)` (#47548) Among the methods of `parentmodule` is one that looks for a matching method and retrieves the parent module of that method. However, there was no method of `parentmodule` for `Method`s themselves. The change made here introduces such a method and modifies code elsewhere to make use of it. | 15 November 2022, 23:29:30 UTC |
626f3b2 | apaz | 15 November 2022, 17:47:47 UTC | build: improve parsing of gfortran version (#47352) Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 15 November 2022, 17:47:47 UTC |
58b559f | Elliot Saba | 15 November 2022, 17:36:14 UTC | Limit initial OpenBLAS thread count (#46844) * Limit initial OpenBLAS thread count We set OpenBLAS's initial thread count to `1` to prevent runaway allocation within OpenBLAS's initial thread startup. LinearAlgebra will later call `BLAS.set_num_threads()` to the actual value we require. * Support older names | 15 November 2022, 17:36:14 UTC |
afe10f9 | Oscar Smith | 15 November 2022, 17:27:41 UTC | Faster and more accurate sinpi, cospi, sincospi for Float32, Float64 (#41744) faster and more accurate `sinpi` `cospi` and `sincospi` for `Float64`, `Float32`, and `Float16` | 15 November 2022, 17:27:41 UTC |
8d03330 | apaz | 15 November 2022, 14:53:14 UTC | UTF16 conversion in loader uses dynamic buffers (#47421) Follow up to #47406 | 15 November 2022, 14:53:14 UTC |
a5c5acb | Azamat Berdyshev | 15 November 2022, 06:59:43 UTC | Typofix in Asynchronous Programming docs (#47563) | 15 November 2022, 06:59:43 UTC |
18e7f40 | Shuhei Kadowaki | 15 November 2022, 03:40:41 UTC | lattice: not introduce `Conditional`s when external lattice doesn't handle it (#47555) This slightly increases the complexity but hopefully comes without any actual performance penalty. | 15 November 2022, 03:40:41 UTC |
d0559c1 | Steve Kelly | 14 November 2022, 23:04:50 UTC | add `isunix` for BinaryPlatforms `AbstractPlatform` (#47543) This provides symmetry with the methods defined in `Sys`, and is convenient for use in e.g. BinaryBuilder platform filters. Note that there are other `is.*bsd` variants we omit since they are unsupported by BinaryPlatforms. The capitilization conventions are also different so we do not call the Sys methods directly. | 14 November 2022, 23:04:50 UTC |
7fe6b16 | Kristoffer Carlsson | 14 November 2022, 14:01:09 UTC | Set VERSION to 1.10.0-DEV (#47222) * Set VERSION to 1.10.0-DEV * move NEWS over to HISTORY | 14 November 2022, 14:01:09 UTC |
9b20aca | Navid C. Constantinou | 14 November 2022, 08:25:40 UTC | fix code alignment in example (#47514) | 14 November 2022, 08:25:40 UTC |
755775c | Thomas Christensen | 14 November 2022, 06:21:24 UTC | Improve performance of `isapprox(::AbstractArray, ::AbstractArray)` when `rtol = 0` (#47464) Co-authored-by: Lilith Orion Hafner <lilithhafner@gmail.com> | 14 November 2022, 06:21:24 UTC |
7116a8b | Carlos A. Michelén Ströfer | 14 November 2022, 03:30:51 UTC | Fix documentation mistake in Test.jl (#47552) Fix a mistake in the documentation: remove the interpolated loop indices from the `@testset` macro with a function call, since there are no loops in that example. | 14 November 2022, 03:30:51 UTC |
b44ce13 | Jeff Bezanson | 13 November 2022, 19:46:15 UTC | edit NEWS for v1.9 (#47535) | 13 November 2022, 19:46:15 UTC |
7650173 | Lilith Orion Hafner | 13 November 2022, 18:09:58 UTC | Update CONTRIBUTING.md (#47547) | 13 November 2022, 18:09:58 UTC |
f6f6e8f | Lilith Orion Hafner | 13 November 2022, 18:09:14 UTC | Switch Coveralls to Codecov in CONTRIBUTING.md (#47549) | 13 November 2022, 18:09:14 UTC |
3510eeb | Lilith Orion Hafner | 11 November 2022, 16:47:35 UTC | Don't show redundant typeinfo when printing LinRange (take 2) (#47521) | 11 November 2022, 16:47:35 UTC |
317211a | Andy Nowacki | 10 November 2022, 22:08:06 UTC | Dates: Error on construction/parsing with empty strings (#47117) When attempting to construct a `DateTime`, `Date` or `Time` from an `AbstractString`, throw an `ArgumentError` if the string is empty. Likewise, error when `parse`ing an empty string as one of these types, and return `nothing` from `tryparse`. This behavior differs from previously. Before, `Date` and `Time` would return default values of `Date(1)` and `Time(0)`, respectively, while `DateTime` would error without a `format` argument. With a `format` argument, it would return `DateTime(1)`. However, this appears not to have been explicitly intended, but rather a consequence of the way parsing was implemented; no tests for empty string parsing existed. This addresses #28090 and #43883; see discussion therein. Summary of changes: - Check for empty string in `Base.parse(::DateTime)` and throw if so. - Change documentation to mention this. - Add a compat notice to the docs contrasting the old behavior. | 10 November 2022, 22:08:06 UTC |
e0ba28a | Jakob Nybo Nissen | 10 November 2022, 18:37:38 UTC | Improve help for broadcasted && and ||, closes #47526. (#47527) Currently, the help text for `.&&` and `.||` says: `x .&& y` is akin to `broadcast(&&, x, y)`. However, that is invalid syntax. | 10 November 2022, 18:37:38 UTC |