Define dot product between vectors of arrays/numbers #22220

Make CoreIO thread-safer

Add a missing return in doInitialization

Load module on remote workers from master (v2)

turn on deprecation warnings for `type` and `immutable`

give a syntax error for repeated keyword args. fixes #16937

fix #21848, bug in widening done by limit_type_depth

fix #20575, syntax error for juxtaposing a string literal

Build fixes for ASAN

Use where syntax in umfpack.jl

downloads of nightlies

fix #21906, `ccall` causes unnecessary variable `Box`

Create show_supertypes function

Fix pool lookup function

This is undefined behavior in LLVM and can actually segfault on real hardware including x86 in certain cases. Also add missing endianess test since it is changed in this commit.

base/digits: allow negative bases

Use cached urls for appveyor and contrib/prepare_release.sh

Currently `gvim` is not recognized by `edit(path, line)`. I recommend to add it with option to run in background turned on.

re-enable ndigits(::Bool, b) (fix #21919)

fix #21923, regressions in circular type definitions

`remote_do` is already an async call.

* Add GitDescribe functions and tests * Add docs for new functions * "git-describe" => "git describe" * Fix hypen

Fix and enforce reinterpret/unsafe_wrap alignment

group ccall'ed MPZ library functions in a module

…cache fix #21963

Make structure of Cmd more obvious in printing

access Cmd elements as an array of strings

)

…#21931)

Hopefully this will make it easier to compile C code as C++ with GCC/Clang

Fix #10442

…21995)

…#21994) by only looking at the "active" triangle of the memory

Fix fallthrough warning on gcc 7

Fix all pedantic warnings in C

* Remove `volatile`. The thread safety is provided by atomics and not volatile * Update comment. We are now relying on the GC bits to be `GC_OLD_MARKED` in order to not update it's metadata similar to other ptrfree permgen objects allocations.

Fix reuse of client port on Linux. Implement for OSX.

getsockname mirrors UNIX getsockname now. Added getpeername.

* Return socket correctly in socket_reuse_port * Run reuseport tests only if SO_REUSEPORT is supported. * Test fix - test for successful addition of workers in distributed test

…timeHeader Introduce USE_POLLY_ACC

would segfault when reaching the jl_datatype_nfields call

TODO: this is helping to avoid a type-system bug mis-computing sparams during intersection, but that can already cause significant problems elsewhere too

…to be non-terminating this same issue applied to several other similar functions similarly, need to avoid nesting the use of functional code (like map) too deeply inside the array code to avoid the appearance of indeterminate recursion in inference

Make julia_alignment actually return the jl_datatype_align if `alignment == 0` and check that the requested alignment is coherent with the heap alignment.

…where N>1 (#19737). (#21988)

Minor clean up in symbol allocation and lookup

Fixed broken link.

…ents are Real. (#22010) * Only try to call specialized findin method for sorted input when elements are Real. * Fix termination cirteria in _sortedfindin Use total order `isless` for tests Add tests * Define eachindex for numbers * Add a few comments to the algorithm

Add LLVM patches for new GC rooting

fix #22021

fix jl_gc_realloc_string size miscomputation

…tion <= (#22027)

…ing traits rename (#16378). (#21997)

More flexible perm-gen allocation alignment control

Under certain circumstances, we emit loads/stores of large LLVM structs. A lot of these can be trivially folded to memcpy and memcpyopt is capable of doing so, but we weren't running it until after SROA. SROA unfortunately, likes to take these apart, causing exponential compile-time blow up and reduced runtime peroformance. In one particular case (2000 element tuple), this change results in a 100x improvement in compile time.

Update README.md

improve inferability of base

@generated

…ll macros also emit an explicit push_loc in @generated functions rather than depending on the existence of a LineNumberNode and other lowering heuristics to produce it

limit maximum vector alignment to heap alignment

With the highlighting updates in Documenter v0.10.1 it is now possible to have specialized highlighting for REPL blocks via the julia-repl attribute on code blocks. Doctests already use this, so this makes the highlighting for non-doctest REPL blocks consistent.

A couple of code examples were not highlighted for some reason.

These blocks are not Julia code, but special REQUIRE file syntax instead.

Also add a newline to the end of the file.

UTF16.jl has been deprecated for a while. Use SHA.jl instead as the example.

The REPL example is basically exactly as above, and does not add anything to the latter bit about the actual convert methods.

Switching to julia-repl revealed several code blocks that could trivially be turned into doctests.

* fix at-deprecate with where syntax * tests for at-deprecate

Use `sincos` from libm in `cis`

Use julia-repl in manual & other edits

* Add complex randn

Add extra early memcpyopt pass

pass macro call location to macro function

fix #21900

check that reinterp on vector of tuples doesnt segfault

Fix #21200.

The values of `p` and `q` are not used in these two methods

Try to loosen cfunction optimization type check

Closes #13427.

Fix #17351.

* add AbstractArray to stdlib and moved constructors to the top * add some references

* document π, e, γ and φ * add type-documentation for number types - replace the uninformative list in doc/src/stdlib/numbers.md with these new docstrings - move constructor docstring for BigInt from helpdb/Base.jl to base/gmp.jl - add a doctest for BigInt constructor * add supertypes to number type docs * move irrational to concrete types

make sure aliasing is the same as for AbstracArrays for SparseArray{<:Real} add tests for this add shortcut for conj(SparseArray{<:Complex} which is faster than broadcasting

* references for Rational * references for Complex * references for Number * references to Real * references for AbstractFloat * references for Integer * references for Signed * references for Unsigned * references to Bool * references for Float16 * references for Float32 * references for Float64 * references for Int8 * references for UInt8 * references for Int16 * references for UInt16 * references for Int32 * references for UInt32 * references for Int64 * references for UInt64 * references for Int128 * references for UInt128 * references for BigInt * references for BigFloat

And change "dict" -> "collection" in the exception, as it is also used from rand(::Set).

Method overwriting by an ambiguity should also invalidate the method cache

removes performance bug added by 0292c42

this allows printing of the Exprs flowing through inference (which might instead have the field set to something like a Const object)

Removing actually may enable inference to get a sharper result, since it is no longer being directed to ignore backedges and correctness assumptions Replaces pure annotations in promotion with inline

fix #22105

replace pure annotations in promotion with inline

infers a missing method (MethodError) as returning `Union{}` and fix tests that were wrong, but not detected previously!

* Fix type sparse type parameter (fixes #22110) * No brackets for singular type param

Fix failing doctests

pass correct paths through to docstrings

* Use where syntax in sparsematrix.jl * Update sparsematrix.jl * Update sparsematrix.jl

* add info field to Cholesky type and delay throwing for non-positive definiteness * comment update [ci skip] * remove comment [ci skip]

add rand(::IntSet)

* Generate instead of hard coding factorial tables * Preallocate factorial tables * Add tests with old literal tables

Also open only one `/proc/self/mem`. There are apparently two instances of `RTDyldMemoryManagerJL` created. Fix #22082

inference: remove hacks around type-system incorrectness

ref #21653

Test if selfmem can write to RX page

…docs

Also, make `flipsign(::Bool, y)` typestable, and add `flipsign(x, Unsigned) = +x`.

prevent Docs.apropos from breaking on random interpolated stuff

… inference

inference: fix some cases of invalid age range updates

fix scope rules: implicitly using a global doesn't add it to the scope

If the requested size of array buffer is >=200% of its current size, grow exactly to the requested size, otherwise double the current size. This should make resize!() & sizehint!() allocate the exactly requested size in more situations (previously the exact resize was guaranteed for empty arrays only). This heuristic should reduce memory usage when arrays are growing non-incrementally (e.g. temporary buffers allocated by LAPACK wrappers).

windows readme: add automation tips

Fix bug with nul at end when printing BigFloat with >322 precision

Use uninitialized buffer for get_process_title()

Use VersionNumber for libllvm_version

reflection: move signature union-splitting logic under the control of inference

Check and return existing SharedArray objects when deserializing on the node which created the array.

Return original sharedarray object when deserializing on the creating node.

Fixes #22173

Generalize UnitRange->AbstractUnitRange in sparse indexing

Fix TLS variant def on AArch64

== is supported between FixedPeriod and OtherPeriod objects, but equal periods were hashed to different values when their types were different. Overflows can happen when converting to a more precise period, and generate a hash collision, but only for really long periods. Collisions will only happen between periods which are congruent modulo a very large value (300,000 years or more), meaning this should not be a problem in practice.

fixes #22135, fixes #22122, fixes #22026, fixes #21581, fixes #16096, fixes #19351

Do exact array resize!() in more cases

fix some macro expander issues with new 0.6 syntax

fix #22032, `for f() in 1:10 ...`

* BLAS wrappers: resize!() instead of Vector{T}() - replace a = Vector{T}(n) with resize!(a, n) - add comments when BLAS API method is called twice * lapack.sygvd!(): fix rwork type

* Fix calculation of displayed time, show UTC offset Fixes #22103. * add space to argument

This fixes a regression introduced in 4f1b479. broadcast_indices() needs to be overloaded by packages for custom types, so it cannot be hidden under _broadcast_indices(). Also, ::Type is incorrect since the method only applies to scalars. Make the tests more complex to be closer to actual implementations in packages so that regressions like this will be noticed in the future.

* Add note on lightweight tags. Address #22102. * make note

…tion. (#21991)

inline some shift operators

Fix Period objects hashing to match ==

* workspace command: fix for missing incremental support * Add a test for #22101

…22174)

make flipsign(typemin(T), -1) == typemin(T) consistently

Show failing instruction on SIGILL

* Removing requirement for promotions from 0,1 * Using iszero() and added test * Tidy up spaces, tabs etc

* Move deps .PHONY together with others * Remove duplicate page from manual The "Interacting with Julia" page was listed twice in the sidebar. * Build docs with clean URLs Documenter has to be bumped to v0.11.1 for this. Also change two absolute URLs that will become outdated with this change to the more appropriate relative URLs. Also bump Compat to get rid of deprecation warnings. * Remove duplicate symm docstring reference There is only one 5-argument docstring: symm(side, ul, alpha, A, B)

As a consequence, using dynamically a version of GMP which does not match the compile time version is no longer an error (this patch makes this a warming). The code is shorter, but we break the MPZ API boundary; this also allows to put allocating pointer_to_array out of the loop, and get slightly better performances.

merge 2 versions (for GMP 5 & 6) of rand(::UnitRange{BigInt})

This fixes a bug in the patch that fixes #19976 causing encoding error on 32bit x86 and segfault when AVX/AVX2 is enabled. Ref LLVM bug report https://bugs.llvm.org//show_bug.cgi?id=29010 LLVM commit llvm-mirror/llvm@83260f2 Also ref where I saw this issue in #21849 (comment)

Make svd(x::Number) return numbers Implement hash, isequal, and == for Factorizations

Backport LLVM patches to fix AVX on i686

which is currently done for text/html this enables the jump-to-method thingy from #22007 to be used with methodswith(Foo) also This also prevents large spacing in the printing, since the general vector output prints with equal spacing for all elements, and some methods have very long signatures

Julia 0.3 is no longer maintained, so I can't imagine this note is useful anymore. [av skip]

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Define dot product between vectors of arrays/numbers #22220

Define dot product between vectors of arrays/numbers #22220

Commits on May 16, 2017

Commits on May 17, 2017

Commits on May 18, 2017

Commits on May 19, 2017

Commits on May 20, 2017

Commits on May 21, 2017

Commits on May 22, 2017

Commits on May 23, 2017

Commits on May 24, 2017

Commits on May 25, 2017

Commits on May 26, 2017

Commits on May 27, 2017

Commits on May 28, 2017

Commits on May 29, 2017

Commits on May 30, 2017

Commits on May 31, 2017

Commits on Jun 1, 2017

Commits on Jun 2, 2017

Commits on Jun 3, 2017

Commits on Jun 4, 2017

Commits on Jun 5, 2017

Commits on Jun 6, 2017