3b4e7b4 | Keno Fischer | 10 October 2018, 00:26:12 UTC | HACK: Option to ignore all inlining heuristics | 10 October 2018, 00:26:12 UTC |
86db845 | Keno Fischer | 10 October 2018, 00:25:27 UTC | cfg_simplify fixup | 10 October 2018, 00:25:46 UTC |
cdd945b | Keno Fischer | 09 October 2018, 22:13:52 UTC | Fix world age maybe? | 09 October 2018, 22:13:52 UTC |
7f29308 | Keno Fischer | 09 October 2018, 04:28:36 UTC | Rebase fixes | 09 October 2018, 04:28:36 UTC |
e727eb5 | Keno Fischer | 09 October 2018, 03:13:55 UTC | Enable aggressive inference | 09 October 2018, 03:13:55 UTC |
cc30c74 | Keno Fischer | 09 October 2018, 01:38:50 UTC | Merge remote-tracking branch 'origin/kf/inferencecomplexity' into kf/tpu2 | 09 October 2018, 01:38:50 UTC |
aa676a8 | Keno Fischer | 09 October 2018, 01:38:26 UTC | Merge remote-tracking branch 'origin/kf/codetypedconst' into kf/tpu2 | 09 October 2018, 01:38:26 UTC |
5e831cc | Keno Fischer | 09 October 2018, 01:38:08 UTC | Merge branch 'kf/cfgsimp' into kf/tpu2 | 09 October 2018, 01:38:08 UTC |
0248bc3 | Keno Fischer | 21 September 2018, 17:39:17 UTC | Inlining thing | 09 October 2018, 01:37:42 UTC |
03fbdf4 | Keno Fischer | 21 September 2018, 03:45:06 UTC | RFC: Allow structural recursion without triggering edge cycle limiting This attempts to fix inference for the case in #29293 (the one returning `Any`). It does not fix the cache poisoning part of that issue, which is a separate concern. The idea here is that we avoid applying limiting if the argtypes of the frame become strictly simpler (thus guaranteeing eventual termination). It is important that the complexity relation be transitive and anti-reflexive. | 09 October 2018, 01:28:28 UTC |
6adb50c | Keno Fischer | 18 September 2018, 21:50:33 UTC | Extend code_typed to be able to debug constant prop I'm currently debugging code that uses a lot of arrays with dimensions in their type parameters. This type of code heavily relies on constant propagation to lift dimensions from the value domain into the type domain. Unfortunately, it's a bit hard to discover what exactly causes inference to drop information from constants, because there's no way to feed in constants for a particular invocation using code_typed. This is a quick hack to remidy that, by making `$`-interpolated expressions available as constants to type inference, e.g. ```julia julia> @code_typed 1+1 CodeInfo( 53 1 ─ %1 = (Base.add_int)(x, y)::Int64 └── return %1 ) => Int64 julia> @code_typed $(1)+$(1) CodeInfo( 53 1 ─ return 2 ) => Int64 ``` Additionally, this extends the same mechanism to apply to get/setproperty!, even in the absence of `$` for the symbol argument, reflecting their special casing inference. | 09 October 2018, 00:57:41 UTC |
38612e0 | Klaus Crusius | 08 October 2018, 19:30:36 UTC | flatten empty tuple - fix #29112 (#29548) | 08 October 2018, 19:30:36 UTC |
15843d5 | Benoît Legat | 08 October 2018, 14:28:36 UTC | Code-format some function names in doc headers (#29485) | 08 October 2018, 14:28:36 UTC |
361c5e7 | Klaus Crusius | 08 October 2018, 07:52:45 UTC | added sparse multiplication and division for triangular matrices. Fix… (#28507) * added sparse multiplication and division for triangular matrices. Fix #28451 * merge with master * merge with master 2 * improved find diagonal part * refactored to purge name space of SparseArrays * additional test cases and bug fix * reformulated tests | 08 October 2018, 07:52:45 UTC |
c03d1dc | Iblis Lin | 07 October 2018, 15:44:06 UTC | test FreeBSD CI status reporter (#29518) | 07 October 2018, 15:44:06 UTC |
cfc967b | Iblis Lin | 06 October 2018, 14:48:46 UTC | cleanup history file for generate_precompile (#29068) | 06 October 2018, 14:48:46 UTC |
62e9ba4 | Rami | 06 October 2018, 08:44:49 UTC | Update performance-tips.md (#29543) My apologies if this small edit is somewhere else waiting to be merged. I couldn't find it... | 06 October 2018, 08:44:49 UTC |
0a685d3 | Keno Fischer | 04 October 2018, 22:56:35 UTC | Turn some asserts into runtime traps instead Unfortunately, we cannnot always rely on :invokes to have argument values that match the declared ssa types (e.g. if the :invoke is dynamically unreachable). Because of that, we cannot assert here, but must instead emit a runtime trap. | 06 October 2018, 03:10:31 UTC |
cdddf35 | Keno Fischer | 05 October 2018, 17:48:56 UTC | Don't eliminate PiNodes for mismatched types E.g. if we had `PiNode(1, CartesianIndex)`, we would eliminate that because `1` was a constant. However, leaving this in allows the compiler to realize that this code is unreachable, as well as guarding codegen against having to wrok through invalid IR. | 06 October 2018, 03:10:31 UTC |
58dc4d1 | Jarrett Revels | 05 October 2018, 20:49:32 UTC | add tests for new PartialTuple capabilities | 06 October 2018, 03:10:31 UTC |
55b500c | Jarrett Revels | 05 October 2018, 20:39:43 UTC | Replace `InferenceResult` varargs hack with new inference lattice element for "partially constant" tuples Previously, we hacked in an additional `InferenceResult` field to store varargs type information in order to facilitate better constant propagation through varargs methods. There were many other places, however, where constants moving in/out of tuples/varargs thwarted constant propagation. This commit removes the varargs hack, replacing it with a new inference lattice element (`PartialTuple`) that represents tuples where some (but not all) of the elements are constants. This allows us to follow through with constant propagation in more situations involving tuple construction/destructuring, and also enabled a clean-up of the `InferenceResult` caching code. | 06 October 2018, 03:10:31 UTC |
ec574fb | Douglas Bates | 06 October 2018, 03:01:19 UTC | mul!, rmul! and lmul! methods for UniformScaling (#29506) * mul!, rmul! and lmul! methods for UniformScaling As per https://discourse.julialang.org/t/in-place-multiplication-methods-for-uniformscaling-type/15837 * Add tests and another mul! method | 06 October 2018, 03:01:19 UTC |
841c2a8 | Keno Fischer | 05 October 2018, 18:57:39 UTC | Add CFG simplify pass tests | 05 October 2018, 23:06:37 UTC |
ea7b1a7 | Jameson Nash | 05 October 2018, 20:21:06 UTC | task: drop parent and module references (#29483) The parent link can unnecessarily hold memory references live, and the current-module function is gone. Their uses have been deprecated in favor of the macros @`__FILE__` and @`__MODULE__`. The `workspace()` function is also gone now, so we can remove the support code for it. To implement detection of broken incremental compilation (and help ensure the module is permanently rooted as is generally expected), keep track of all "open" modules in a hashtable (jl_current_modules). We can further refine this later, if desired (e.g. also allow any submodule and of any item in `jl_module_init_order`). Also removes the global `module_stack` state. This makes us a bit more robust to async module definition. We could perhaps do even more here, if necessary (e.g. by checking first that all parents are initialized, instead of just the direct ancestor). But this is hopefully a vanishingly rare situation anyways. | 05 October 2018, 20:21:06 UTC |
d461d56 | Keno Fischer | 05 October 2018, 19:55:53 UTC | Merge pull request #28368 from rbehrends/rb/gc-extensions GC Extensions | 05 October 2018, 19:55:53 UTC |
1b18f12 | Keno Fischer | 05 October 2018, 18:45:43 UTC | Backport CFG simplification pass from XLA backend Right now we don't do too much CFG simplification (esp since we turned off doing so for constant folded branches). This isn't too much of a problem, because LLVM is very good at cutting down any excess basic block we happen to emit. However, for non-SSA backends, excess CFG can be a significant problem that the backend may not be able to optimize away (even if it's trivial at the Julia IR level). Plus it's annoying for humans to read. The XLA backend had a simple CFG simplification pass. Backport this pass to Base, so it can live alongside the code it depends on (it has a fairly close dependency on the details of the CFG and IncrementalCompact). As it stands, I don't think it's useful to have this pass in the default compiler pipeline (both because LLVM can handle it easily and because our round-trip to statement based representations cleans up some of this), but I do think it's useful interactively and for non-standard compiler backends. We should re-evaluate whether to put this in the standard compiler pipeline once we re-enable the CFG transformations of constant folded conditions. If that ends up leaving a lot of basic block chains around, it might yet be worth putting this in. | 05 October 2018, 19:24:46 UTC |
d3399ef | Keno Fischer | 24 September 2018, 22:34:56 UTC | IncrementalCompact additions for CFG simplify | 05 October 2018, 18:54:22 UTC |
50c27fd | Tim Holy | 05 October 2018, 11:14:26 UTC | one(::CartesianIndex) => oneunit(::CartesianIndex) (#29442) | 05 October 2018, 11:14:26 UTC |
1968b23 | Nathan Daly | 05 October 2018, 06:49:18 UTC | Add NEWS entry for #28156: add `splitpath(p)` (#29448) | 05 October 2018, 06:49:18 UTC |
3cc4b80 | Qi Zhang | 05 October 2018, 06:36:40 UTC | Fix a typo in parallel computing doc (#29521) | 05 October 2018, 06:36:40 UTC |
601ef78 | Qi Zhang | 05 October 2018, 06:35:17 UTC | Remove unecessary mention of Julia version in docs (#29519) fix #29510 | 05 October 2018, 06:35:17 UTC |
0210b1d | Simon Byrne | 05 October 2018, 04:32:58 UTC | at-kwdef support for parametric types and subtypes (#29316) * at-kwdef support for parametric types and subtypes Fixes #29307. | 05 October 2018, 04:32:58 UTC |
7942395 | Reimer Behrends | 05 October 2018, 00:48:19 UTC | Add a test for the task scanner. | 05 October 2018, 00:48:19 UTC |
8ff75ad | Keno Fischer | 11 September 2018, 23:21:11 UTC | Better dominators This commit replaces the naive algorithm for replacing dominator trees by a faster implementation based on the Semi-NCA algorithm (reference in the code comments). LLVM recently switched to this algorithm and found it to be faster in practice than SLT (which it used before). It is also slightly easier to implement. More importantly though, it should easily extend to dynamic dominators. This fixes the preformance problems in dominator construction noted in #25927 and should provide a basis for a dynamic dominator implementation to fix #29107. | 05 October 2018, 00:15:29 UTC |
83bc53b | Reimer Behrends | 04 October 2018, 23:40:11 UTC | Add functionality to find a task's stack buffer. | 04 October 2018, 23:40:11 UTC |
37ab710 | Rafael Fourquet | 04 October 2018, 16:32:20 UTC | REPL: show "failed" when i-search fails, like in readline (#29413) The i-search prompt becomes for example: `(failed forward-i-search)`. | 04 October 2018, 16:32:20 UTC |
0f920ab | Keno Fischer | 22 September 2018, 21:12:12 UTC | Fix phinode rename bug during domsort Since the edge list of a phi may be incomplete with respect to predecessors, it is not safe to replace a one-value phi by its value, unless we know that there is only one predecessor. Remove the incorrect logic from domsort. We do also do the correct transformation in compaction right afterwards these days. | 04 October 2018, 16:14:36 UTC |
f9b9d85 | Frank Brütting | 04 October 2018, 14:52:22 UTC | Typo (#29482) | 04 October 2018, 14:52:22 UTC |
9280bfc | Rafael Fourquet | 04 October 2018, 14:00:32 UTC | REPL: make paste work in search mode (#29373) | 04 October 2018, 14:00:32 UTC |
32e3851 | Reimer Behrends | 06 September 2018, 04:36:40 UTC | Use atomic operations to initialize conservative GC support. | 04 October 2018, 13:59:53 UTC |
62d9cde | Reimer Behrends | 05 September 2018, 14:43:43 UTC | Copy mark stack struct to thread-local storage. Copying the mark stack struct instead of aliasing it ensures that structural replacement of aggregates for its members can still take place in the mark loop. | 04 October 2018, 13:59:53 UTC |
64b019d | Max Horn | 10 September 2018, 16:23:30 UTC | Rename gc_mark_sp_t to jl_gc_mark_sp_t | 04 October 2018, 13:59:15 UTC |
3821314 | Reimer Behrends | 23 August 2018, 06:59:06 UTC | Improve comments for the GC extension functionality. | 04 October 2018, 13:58:37 UTC |
aadb507 | Reimer Behrends | 21 August 2018, 11:48:59 UTC | Update GC extension tests. * Fix formatting. * Make tests work with 1.0.0. * Enable conservative support. | 04 October 2018, 13:58:37 UTC |
680cb7a | Reimer Behrends | 21 August 2018, 11:47:12 UTC | Revise GC extensions implementation in accordance with code review. * Adjustments to conservative marking. * Remove symbol concatenation from gc_invoke_callbacks(). * Clean up handling of jl_buff_tag objects. * Add comments. * Fix formatting. | 04 October 2018, 13:58:37 UTC |
55c3c29 | Rafael Fourquet | 04 October 2018, 13:48:30 UTC | REPL: really fix #29347, searching a string not in history (#29412) | 04 October 2018, 13:48:30 UTC |
fa95de1 | Reimer Behrends | 09 August 2018, 10:58:35 UTC | Use age bits to determine liveness for conservative marking. | 04 October 2018, 13:46:16 UTC |
68836e5 | Reimer Behrends | 07 August 2018, 00:59:46 UTC | Revised support for conservative marking. | 04 October 2018, 13:46:16 UTC |
66c36d4 | Reimer Behrends | 31 July 2018, 03:15:07 UTC | Add tests for the new GC extensions. | 04 October 2018, 13:46:16 UTC |
3adc378 | Reimer Behrends | 31 July 2018, 03:15:02 UTC | Support optional conservative tracing of objects in the GC. | 04 October 2018, 13:46:16 UTC |
05cb43c | Reimer Behrends | 31 July 2018, 03:14:59 UTC | Support the tracing of additional roots in the GC. | 04 October 2018, 13:46:16 UTC |
5b38be1 | Reimer Behrends | 31 July 2018, 03:14:54 UTC | Support foreign types with custom mark and sweep functions. | 04 October 2018, 13:40:09 UTC |
2c751b8 | Reimer Behrends | 31 July 2018, 03:14:47 UTC | Initial framework for Julia GC extensions. | 04 October 2018, 13:38:19 UTC |
60cff80 | Tim Holy | 04 October 2018, 13:16:57 UTC | Merge pull request #29440 from JuliaLang/teh/colon_cartindex Add colon constructor for CartesianIndices | 04 October 2018, 13:16:57 UTC |
be0fceb | asbisen | 04 October 2018, 12:38:07 UTC | Update README.md (#29514) Fixed typo | 04 October 2018, 12:38:07 UTC |
4fa7161 | Tim Holy | 04 October 2018, 10:38:39 UTC | Fix NEWS-update.jl for cases of empty issues list | 04 October 2018, 10:38:39 UTC |
ba0add8 | Tim Holy | 30 September 2018, 12:34:33 UTC | Add colon constructor for CartesianIndices See https://discourse.julialang.org/t/psa-replacement-of-ind2sub-sub2ind-in-julia-0-7/14666/6 and posts below it. | 04 October 2018, 10:37:58 UTC |
b8ca1a4 | Jeff Bezanson | 03 October 2018, 21:09:38 UTC | remove vestigial code related to (removed) custom inlining of `^` | 04 October 2018, 03:29:41 UTC |
5bc3de4 | Mark | 04 October 2018, 01:34:52 UTC | Minor English fixes (#29337) | 04 October 2018, 01:34:52 UTC |
7dd1e4c | Mark | 04 October 2018, 01:32:56 UTC | English fixes (#29336) | 04 October 2018, 01:32:56 UTC |
d02f4f0 | Jeff Bezanson | 04 October 2018, 01:28:08 UTC | fix a transitivity error in method specificity (#29405) | 04 October 2018, 01:28:08 UTC |
7cc9209 | Simon Byrne | 04 October 2018, 01:20:31 UTC | Add --project option to man page (#29497) | 04 October 2018, 01:20:31 UTC |
8b5f5da | Keno Fischer | 02 October 2018, 22:16:07 UTC | Allow :new constant inference in more cases Previously we allowed only bitstypes. However, it is legal to do so for and the current code works for all concrete immutables, so weaken the guarding condition appropriately here. | 03 October 2018, 22:17:10 UTC |
a984ad0 | Sam O'Connor | 03 October 2018, 21:45:04 UTC | Fix #29451: parse(Int, s::AbstractString) when ncodeunits(s) > lastindex(s) (#29465) | 03 October 2018, 21:45:04 UTC |
6eb4921 | Keno Fischer | 22 September 2018, 21:52:24 UTC | Use recursion rather than iteration for error check in uncolon Inference has an easier time analyzing recursion (esp if structural over tuples) than iteration, which is why the rest of this function is using recursion. With sufficient inlining, LLVM does generally get rid of the extra error check and branch, but by switching to recursion, we make things easier on non-LLVM backends. | 03 October 2018, 20:50:23 UTC |
ea16391 | Jeff Bezanson | 03 October 2018, 20:12:45 UTC | fix #29468, bug in intersection of different-length vararg tuples (#29487) | 03 October 2018, 20:12:45 UTC |
37dc16e | Bogumił Kamiński | 03 October 2018, 18:41:36 UTC | correct language in documentation (#29502) | 03 October 2018, 18:41:36 UTC |
db100ad | Keno Fischer | 26 September 2018, 00:32:21 UTC | More robust show for invalid CFG Despite the fact that the compiler should never generate invalid data structures, it is a good practice for the show functions to be able to handle these and indicate what the problem is. In this case, if the IR has more instructions than its supposed to according to the CFG, put all trailing instructions into a "!" basic block to indicate an error. | 03 October 2018, 18:09:21 UTC |
590ee83 | Keno Fischer | 15 September 2018, 19:12:47 UTC | Align StmtRange field names with UnitRange | 03 October 2018, 18:09:21 UTC |
434c065 | Jeff Bezanson | 03 October 2018, 17:06:32 UTC | move 1.0 news to HISTORY file; prepare NEWS file for 1.1 (#29490) [ci skip] | 03 October 2018, 17:06:32 UTC |
a897365 | Fredrik Ekre | 03 October 2018, 14:45:26 UTC | Upgrade to new Documenter. (#29470) | 03 October 2018, 14:45:26 UTC |
a9550d5 | Marisuf | 03 October 2018, 13:47:19 UTC | Removed unnecessary extra line (#29476) Line 73 cleared. | 03 October 2018, 13:47:19 UTC |
4412dd4 | Keno Fischer | 02 October 2018, 19:03:41 UTC | Add an option to disable building libssh2 tests These fail to build if mbedtls is not available in the global linker search path. Add a simple flag to disable them, since we don't use them. | 03 October 2018, 01:06:14 UTC |
59b6645 | Keno Fischer | 02 October 2018, 19:00:41 UTC | Still set rpath-link in RPATH_ESCAPED_ORIGIN We use this setting for linking curl, which itself needs to find libssh2 and mbedtls. However, while curl will indeed eventually be in the same directory as these two libraries, it is not during linking, causing link failures (unless we happen to have global copies of the same libraries that the linker can pick up). Add an rpath-link command with the absolute path to our library directory to make sure the dependent libraries get picked up at link time. | 03 October 2018, 01:06:14 UTC |
b0a37b5 | Takafumi Arakaki | 02 October 2018, 22:08:25 UTC | Document scripting tips (#29423) | 02 October 2018, 22:08:25 UTC |
1f0f40f | Jeff Bezanson | 02 October 2018, 21:54:26 UTC | rm redundant `show` method for Enum types (#29460) | 02 October 2018, 21:54:26 UTC |
39117c9 | David Pérez-Suárez | 02 October 2018, 16:45:30 UTC | tiny formatting fix on docs (#29463) C and Julia code were inside the same code block | 02 October 2018, 16:45:30 UTC |
dee7d77 | DragonXDoom | 02 October 2018, 16:08:19 UTC | Closed issue #29454 (#29475) Changed `String` to `BitSet` as a `mutable struct` example. | 02 October 2018, 16:08:19 UTC |
50e9275 | Lyndon White | 02 October 2018, 16:07:29 UTC | fix unusual line spacing (#29474) | 02 October 2018, 16:07:29 UTC |
18652ea | Tanmay Gupta | 02 October 2018, 16:06:48 UTC | updated README.md (#29473) | 02 October 2018, 16:06:48 UTC |
9705c96 | Bogumił Kamiński | 02 October 2018, 14:48:03 UTC | add elseif to reserved-words (#29222) | 02 October 2018, 14:48:03 UTC |
d769ad2 | Valentin Churavy | 02 October 2018, 12:41:41 UTC | Merge pull request #13099 from JuliaLang/jn/no_copy_stacks no-copy stacks | 02 October 2018, 12:41:41 UTC |
a2df80e | Simon Byrne | 01 October 2018, 20:36:40 UTC | Don't error on help for keyword functions (#29308) Fixes #23991. Also allows for help where some arguments are undefined. | 01 October 2018, 20:36:40 UTC |
29b780e | Mark | 01 October 2018, 18:39:47 UTC | Improved English and egs (#29281) Improved English and egs The examples starting around line 80 showing how to create arrays with dimensions are unhelpful because in every case the dimensions are 2x2 so people don't get any sense of which dimension is which. So I've changed them all to 2x3 which doesn't take up any more vertical space but makes it much clearer which dimension is which. I don't like the Comprehensions example because the expression is much too complicated: the idea is to show comprehensions, so I think the expression should be a lot simpler. I changed the `searchsorted` eg to make the searched for number different from all the others to improve clarity. | 01 October 2018, 18:39:47 UTC |
2460b57 | jaakkor2 | 01 October 2018, 17:42:35 UTC | win,init: if stdio is not available, do not exit (#29374) Fix #28820 | 01 October 2018, 17:42:35 UTC |
e721d35 | Jameson Nash | 01 October 2018, 15:53:31 UTC | slightly faster String allocation (#29254) Only a few % saved, but it essentially removes the cost of computing the sizeclass, so this should save on processor resources (frontend, BTP) also. And it makes the computation correct (previously the cutoff was slightly offset, so it would switch to the next larger size-class at 609 instead of at 624, for example). | 01 October 2018, 15:53:31 UTC |
61696f9 | Jameson Nash | 01 October 2018, 15:21:05 UTC | Merge pull request #29262 from JuliaLang/jn/interpret-phi interpreter: fix bugs with ssair phi-node handling | 01 October 2018, 15:21:05 UTC |
4851fab | Katharine Hyatt | 01 October 2018, 13:53:53 UTC | Merge pull request #29376 from JuliaLang/ksh/improvar Some improvements to variables and scoping manual | 01 October 2018, 13:53:53 UTC |
391f2dd | Valentin Churavy | 01 October 2018, 00:58:50 UTC | Merge pull request #29446 from eli-schwartz/system-nondefault-llvm build: work correctly when linking to a system llvm which is not the default | 01 October 2018, 00:58:50 UTC |
f07d461 | Keno Fischer | 21 September 2018, 19:39:56 UTC | Allow eliminating nothrow intrinsics during optimization | 30 September 2018, 23:59:04 UTC |
d1fab72 | Eli Schwartz | 30 September 2018, 19:30:53 UTC | build: work correctly when linking to a system llvm which is not the default It's erroneous to assume hardcoded -lLLVM works, when it will either not exist, or exist and be a symlink to -lLLVM-6.0, or exist and be a symlink to a completely invalid -lLLVM-7 Instead, use llvm-config for what it's meant to do, and acquire the linker --libs. In one case this means we can simplify on --libfiles rather than --libdir plus appending some hardcoded strings, and in another case we can simply drop a huge, confusingly repetitive invocation which was meant to work around very old llvm versions that are no longer supported. | 30 September 2018, 19:30:53 UTC |
ebfbe62 | Keno Fischer | 22 September 2018, 19:33:15 UTC | Add test for eliminating apply_type/TypeVar | 30 September 2018, 17:33:18 UTC |
8fa5296 | Keno Fischer | 22 September 2018, 18:44:20 UTC | Make TypeVar constructor a builtin Or more precisely make a builtin-equivalent of jl_new_typevar. This allows us to give it a proper nothrow tfunc, thus letting us eliminate it if it is not otherwise used. | 30 September 2018, 17:33:18 UTC |
f81749f | Keno Fischer | 22 September 2018, 18:05:42 UTC | Make apply_type nothrow tfunc more aggressive | 30 September 2018, 17:33:18 UTC |
2c86c1f | Adam B | 30 September 2018, 08:21:40 UTC | Remove link missed in #26432 (#29434) | 30 September 2018, 08:21:40 UTC |
4d06e00 | Valentin Churavy | 30 September 2018, 01:52:30 UTC | Merge pull request #29325 from JuliaLang/vc/restored check return from jl_restore_incremental correctly | 30 September 2018, 01:52:30 UTC |
f7f87c2 | Jameson Nash | 29 September 2018, 22:36:48 UTC | intrinsic casts: perform error checking (#29422) previously, we just let LLVM abort (even though this could potentially happen in dead-code). | 29 September 2018, 22:36:48 UTC |
8907aff | John Myles White | 29 September 2018, 19:21:09 UTC | Fix sentence fragment (#29431) Move whereas clause into subordinate position to avoid sentence fragment. | 29 September 2018, 19:21:09 UTC |
f42f7b7 | Jeff Bezanson | 29 September 2018, 19:20:25 UTC | fix #20872, handle UnionAll types in `which` (#29419) | 29 September 2018, 19:20:25 UTC |
9454c8d | Daniel Karrasch | 29 September 2018, 18:38:47 UTC | stdlib/SparseArrays: add rmul! and lmul! of sparse matrix with Diagonal (#29296) * Add rmul! and lmul! of sparse matrix with Diagonal | 29 September 2018, 18:38:47 UTC |