ragmani [Wed, 8 Mar 2017 09:22:19 +0000 (18:22 +0900)]
Fix wrong gcc inline assembly code. (#10024)
It was coded as "rbx" instead of "ebx" register in clobber list position.
Mike McLaughlin [Wed, 8 Mar 2017 08:29:50 +0000 (00:29 -0800)]
Fix assert in portable pdb reader running the clrstack command (#10011)
Issue #9969
Atsushi Kanamori [Wed, 8 Mar 2017 08:23:30 +0000 (00:23 -0800)]
Move two sharable classes into the shared partition. (#10009)
The corert side of this was commited by
https://github.com/dotnet/corert/pull/2915#event-
990346611
William Godbe [Wed, 8 Mar 2017 06:33:45 +0000 (22:33 -0800)]
Merge pull request #10006 from wtgodbe/parallelWrapperGen
Zip all test wrappers in parallel
Aditya Mandaleeka [Wed, 8 Mar 2017 05:30:02 +0000 (21:30 -0800)]
Remove TRACK_CXX_EXCEPTION_CODE_HACK and related code. (#10018)
Daniel Podder [Wed, 8 Mar 2017 03:14:57 +0000 (21:14 -0600)]
Port Windows PGO support to master (#9985)
The bulk of this PR is a cherry-pick of commit fa02660 that shipped in
release/1.1.0, updating the build system support for PGO to support
consuming PGDs properly during release builds on Windows.
Also included are the following new changes:
* Skip restore of opdata if the requisite project.json is missing
* If the optdata package restore fails, fail the build.
* Add new build option: 'skiprestoreoptdata'
Note: This change doesn't by itself enable PGO in master yet, because
training data (optdata packages) for master don't exist on myget yet.
However, with these changes, the only step remaining to enable PGO
optimizations is to add a project.json referencing the correct optdata
package.
Aditya Mandaleeka [Wed, 8 Mar 2017 01:54:45 +0000 (17:54 -0800)]
Merge pull request #10021 from adityamandaleeka/fix_broken_ci_badges
Fix broken CI badges
Aditya Mandaleeka [Wed, 8 Mar 2017 01:49:12 +0000 (17:49 -0800)]
Fix broken badges.
Alex Perovich [Wed, 8 Mar 2017 01:49:42 +0000 (19:49 -0600)]
Add shared changes from corert (#10010)
Stephen Toub [Wed, 8 Mar 2017 01:02:56 +0000 (20:02 -0500)]
Merge pull request #10005 from stephentoub/taskextensions
Add TaskExtensions to CoreLib
Frederik Carlier [Tue, 7 Mar 2017 22:46:27 +0000 (23:46 +0100)]
Add scripts for building for arm Android (#9961)
* Add scripts for building for arm Android
* Use arm soft float for bringup
* Use NDK r14
* Set HAVE_PT_REGS to false, as they are not available on Android
* _aligned_alloc is not available on clang
* Detect Visual C++ instead of clang
* Detect Android instead of clang
ragmani [Tue, 7 Mar 2017 22:28:58 +0000 (07:28 +0900)]
[x86/linux] Fix "Undefined Reference" error in CLR release build. (#10000)
Replace "IncCantStopCount()" and "DecCantStopCount()" functions with
nothing(#define IncCantStopCount()) to call in methods of UTSemReadWrite
on x86/Linux.
Jan Vorlicek [Tue, 7 Mar 2017 21:41:22 +0000 (22:41 +0100)]
Fix incremental build of mscordaccore (#9982)
* Fix incremental build of mscordaccore
This change sets up proper dependency for the mscordaccore build and stops
using PRE_LINK command which is executed always.
The way it solves the problem is that it creates an object library from the
mscordac.cpp and then lets the mscordaccore.exp be generated from that object
and finally use both that object library and the mscordaccore.exp while building
mscordaccore.
Michelle McDaniel [Tue, 7 Mar 2017 19:24:10 +0000 (11:24 -0800)]
Merge pull request #9992 from adiaaida/fixjit32testing
Fix jit32 perf and throughput testing
Stephen Toub [Tue, 7 Mar 2017 17:50:17 +0000 (12:50 -0500)]
Add TaskExtensions to CoreLib
Our TaskExtensions.Unwrap implementation in corefx is more allocation-heavy than the implementation we had in desktop and we could have in CoreLib, where CreateUnwrapPromise is available. By moving this implementation down to CoreLib, the code:
```C#
Task<Task> toUnwrap = ...;
Task t = toUnwrap.Unwrap();
```
incurs 1 allocation instead of 4.
Michelle McDaniel [Mon, 6 Mar 2017 22:21:15 +0000 (14:21 -0800)]
Fix jit32 perf and throughput testing
For jit32, we need to download the compatjit to make sure we are using a
jit32 build, not legacy backend (which is what is built when we run a
regular build out of coreclr). This change updates the groovy scripting to
pull down the jit nuget package that contains compatjit.
Until the packaging is fixed (#9991), we are defaulting to the last
package that packaged compatjit along with clrjit.
Bruce Forstall [Tue, 7 Mar 2017 19:14:00 +0000 (11:14 -0800)]
Merge pull request #9996 from BruceForstall/FixCallFunFormatting
Un-clang-format-horrible-ify emitIns_Call() and genEmitCall()
wtgodbe [Tue, 7 Mar 2017 19:07:58 +0000 (11:07 -0800)]
Zip all test wrappers in parallel
Jonghyun Park [Tue, 7 Mar 2017 11:27:35 +0000 (20:27 +0900)]
[x86/Linux] CDECL as default P/Invoke Calling Convetion (#9977)
* [x86/Linux] CDECL as default P/Invoke Calling Convetion
Jan Vorlicek [Tue, 7 Mar 2017 11:03:08 +0000 (12:03 +0100)]
Enhance ICU version scan (#9993)
The scanning that tries to find locally installed ICU version was not
considering the case when only ICU library files with major version
component were present (like libicuuc.so.52). This is a rare case that
I've seen so far only when people built and installed ICU manually.
But it is worth to add checks for such case as a fallback, which is
what this change does.
I have also added checks for minorVersion and subVersion being -1 when
trying to detect the symbol versioning used. While the pre-existing code
would still work fine, it would unnecessarily try nonsensical version.
Hanjoung Lee [Tue, 7 Mar 2017 08:57:29 +0000 (17:57 +0900)]
[x86/Linux] Fix incorrect FP values (#9975)
* [x86/Linux] Fix incorrect FP values
Restoring FP by CallerSP could be incorrect for nested EH.
Now that we have #9820, we can use current context for 1st pass of EH.
Fix #9848
Bruce Forstall [Tue, 7 Mar 2017 07:34:31 +0000 (23:34 -0800)]
Merge pull request #9995 from BruceForstall/FixNetciForceRelocs
Add missing "break" in a forcerelocs case
Andy Ayers [Tue, 7 Mar 2017 03:31:24 +0000 (19:31 -0800)]
Harden resolveVirtualMethod against invalid types (#9989)
Split resolveVirtualMethod into a wrapper/helper to allow early
returns as failure reasons are discovered.
Before devirtualizing, ensure that the derived class claimed by the jit
is really a subclass of the class that introduces the virtual method.
Fail early if this is not the case.
Also add early out for interface call devirtualization. The jit currently
does not ask for this but asking for it should not trigger an assert.
Closes #9945 (along with #9959).
Jonghyun Park [Tue, 7 Mar 2017 01:31:47 +0000 (10:31 +0900)]
[x86/Linux] Check Stack Alignment in StubRareEnable/StubRareDisableTHROW (#9925)
William Godbe [Tue, 7 Mar 2017 00:36:06 +0000 (16:36 -0800)]
Merge pull request #9986 from wtgodbe/XunitWrapperGranularity
Create Xunit wrappers for 2nd-level test directories instead of top-level
Jan Vorlicek [Tue, 7 Mar 2017 00:35:14 +0000 (01:35 +0100)]
Fix ICU detection in System.Globalization.Native build (#9990)
The config.cmake was not getting the ICU library paths it requires to
build a code that detects presence of the ucol_setMaxVariable.
I have excluded the libraries detection from the CMakeLists.txt
some time ago since they are not used in the build itself anymore,
but I have not realized that they are needed for this detection
script that's included from the CMakeLists.txt.
Bruce Forstall [Tue, 7 Mar 2017 00:29:58 +0000 (16:29 -0800)]
Un-clang-format-horrible-ify emitIns_Call() and genEmitCall()
Bruce Forstall [Tue, 7 Mar 2017 00:22:07 +0000 (16:22 -0800)]
Add missing "break" in a forcerelocs case
Michelle McDaniel [Mon, 6 Mar 2017 23:25:57 +0000 (15:25 -0800)]
Merge pull request #9991 from adiaaida/fixx86NugetPackageCreation
Change PackagePlatform to Platform for x86 jit pkg
Michelle McDaniel [Mon, 6 Mar 2017 22:36:19 +0000 (14:36 -0800)]
Change PackagePlatform to Platform for x86 jit pkg
PackagePlatform was removed in
8dcf9665feabee339029c79b9d86dd24c71905d1,
but the x86 jit package props file was never updated to not use that
property. This change updates the props file to use Platform instead of
PackagePlatform to make sure we package compatjit.dll for x86.
wtgodbe [Mon, 6 Mar 2017 21:44:21 +0000 (13:44 -0800)]
Create Xunit wrappers for 2nd-level test directories instead of top-level
James Ko [Mon, 6 Mar 2017 18:44:34 +0000 (13:44 -0500)]
Implementation for Dictionary.TryAdd (#9923)
* Add Dictionary.TryAdd
* Fix the conflicts for real
* Remove branch from Add(TKey, TValue)
* Tweak naming
* Remove a slow assert
* PR feedback
Sivarv [Mon, 6 Mar 2017 18:25:03 +0000 (10:25 -0800)]
Merge pull request #9398 from mikedn/switch-cast
Change switch value widening
Davis Goodin [Mon, 6 Mar 2017 17:40:56 +0000 (11:40 -0600)]
Add stripSymbols argument; refactor symbol package build (#9664)
* Add argument to explicitly enable symbol stripping
* Refactor symbol package build
* Remove unnecessary native file distinctions
* Remove unnecessary WindowsSymbolFile step
Jonghyun Park [Mon, 6 Mar 2017 11:31:37 +0000 (20:31 +0900)]
[x86/Linux] Stack align aware unwinder (#9928)
Jan Kotas [Mon, 6 Mar 2017 05:58:13 +0000 (21:58 -0800)]
Delete a few more unused hosting methods (#9973)
Rich Lander [Mon, 6 Mar 2017 02:09:31 +0000 (18:09 -0800)]
Add 3PN for GCBench (#9971)
Rich Lander [Mon, 6 Mar 2017 02:09:17 +0000 (18:09 -0800)]
Add 3PN entry for Livermore Loops coded in C (#9972)
Austin Wise [Mon, 6 Mar 2017 01:51:40 +0000 (17:51 -0800)]
Remove hosting methods that always return false (#9930)
* Remove CLRIoCompletionHosted as it always returns false.
* Remove CLRSyncHosted as it always returns false.
* Remove CLRMemoryHosted as it always returns false.
* Remove CLRTaskHosted as it always returns false.
* Remove CLRAssemblyHosted, CLRGCHosted,and CLRSecurityHosted.
They are not called.
* Remove IsThreadPoolHosted as it always returns false.
* Remove EnterRuntime and LeaveRuntime, as they do nothing.
* Add back calls to RevertIfImpersonated and GCX_PREEMP.
I accidentally deleted the call to RevertIfImpersonated instead of just
removing an extra parameter.
When I removed the HR_LEAVE_RUNTIME_HOLDER macro from windowsruntime.h,
I not only removed a LeaveRuntimeHolder but also a GCX_PREEMP. So I added
it back. The holder and GCX_PREEMP where only inserted when the
FEATURE_LEAVE_RUNTIME_HOLDER macro was defined. Since it is always defined,
I removed it. Also as I understand it, you would always want to have a
GCX_PREEMP before calling into the Windows API as not to block the GC,
so I'm not sure why you would want to disable it.
Jan Kotas [Mon, 6 Mar 2017 01:27:36 +0000 (17:27 -0800)]
Delete IsNonW8PFrameworkAPI checks (#9964)
Dead code in CoreCLR
Rich Lander [Sun, 5 Mar 2017 22:55:57 +0000 (14:55 -0800)]
Add 3PN for Burgers equation (#9966)
- Thanks to @taumuon for licensing the test, per our request!
- License is MIT.
Andy Ayers [Sat, 4 Mar 2017 04:01:52 +0000 (20:01 -0800)]
JIT: change codgen for dup to not lose type information
The jit has traditionally optimized most cases of a
zz; dup; stloc.x
sequence into
zz; stloc.x; ldloc.x
with a comment noting that this makes the resulting compiler IR
more amenable to CSE. Here zz is any IL operation that pushes
a value on the stack. However if zz's result is a ref type, the
dup'ed value produced by zz may be a subtype of the type of the
local x, so this transformation can lose type information.
In the past the loss of type information hasn't mattered, but now if
zz subsequently is used as the this object in a virtual call, the jit
may attempt devirtualization and ask an ill-posed question to the VM,
claiming the type of the this object is an impossible supertype.
This changes modifies the transformation by introducing a new local temp y
that has the type of zz. The result is now
zz; stloc.y; ldloc.y; ldloc.y; stloc.x
with the net result that a ldloc.y is left on the stack with the proper
type. As an optimization, if zz is ldloc.z then the expansion is simply
zz (ldloc.z); ldloc.z; stloc.y
and if zz is a simple "cheap" constant (say ldnull) then the constant
expression is duplicated:
zz (ldnull); ldnull; stloc.y
This resolves the jit side of #9945. Further work is needed on the VM
side since in unverifiable IL there may be other cases where the jit
can present an impossible type.
Diffs from this were minor and in many cases small improvements.
Paul Westcott [Sun, 5 Mar 2017 05:32:12 +0000 (16:32 +1100)]
Remove ComVisible as per #9371 (#9962)
When lazy was merged, this change was lost.
mikedn [Sat, 4 Mar 2017 17:58:56 +0000 (19:58 +0200)]
Remove unused bbAssertionKill (#9957)
Adeel Mujahid [Sat, 4 Mar 2017 17:50:24 +0000 (19:50 +0200)]
Update glossary.md (#9951)
* Added the definitions for AOT, UTC and UWP.
* Made R2R as the initialism with 'Read-to-Run' its abbreviation.
* Fixed a typo: contributers -> contributors.
* Format as Term-Description markdown table.
Varun Venkatesan [Sat, 4 Mar 2017 17:24:30 +0000 (09:24 -0800)]
Extending optimized JIT helpers to Buffer.MemoryCopy (#9786)
Jan Kotas [Sat, 4 Mar 2017 08:40:56 +0000 (00:40 -0800)]
Move System.Lazy to shared CoreLib partition (#9955)
Tanner Gooding [Sat, 4 Mar 2017 06:26:26 +0000 (22:26 -0800)]
Merge pull request #9954 from tannergooding/fcall
Updating `COMSingle` in `classlibnative` to use the appropriate FCALL declarations.
Alex Perovich [Sat, 4 Mar 2017 05:44:12 +0000 (23:44 -0600)]
Remove unreferenced file (#9952)
Paul Westcott [Sat, 4 Mar 2017 03:14:02 +0000 (14:14 +1100)]
Alternative System.Lazy implementation (#8963)
* Remove use of lock for PublicationOnly
This involves a little bit of trickery. In avoiding the lock we need to
go from m_implementation as PublicationOnly to null in a single step. We
can't do this though without another object, so we reuse out Lazy
object, which basically just spins its wheels until the value is ready.
This shouldn't be a very common scenerio (i.e. hitting that code path)
but it is possible.
* Minimize additional object creation
Storing factory in Lazy object, and passing Lazy as a parameter to
ILazyItem functions. This means that None & PublicationOnly now need no
creation of secondary object.
* Remove Func<T> for default constructor invoking
Fixing startup performance concerns
* Moved non-generic functionality out of Lazy
...and into a helper class
* Expression-bodied functions
Tanner Gooding [Sat, 4 Mar 2017 03:08:33 +0000 (19:08 -0800)]
Fixing the FCALL 'common gotchas' comment to clarify that it applies to all types that are not passable via register in __fastcall.
Tanner Gooding [Sat, 4 Mar 2017 02:56:56 +0000 (18:56 -0800)]
Updating `COMSingle` in `classlibnative` to use the appropriate FCALL Declarations.
Jonghyun Park [Fri, 3 Mar 2017 23:57:35 +0000 (08:57 +0900)]
[x86/Linux] Clean up pContext access (#9933)
* [x86/Linux] Clean up pContext access
* Fix ARM/Windows build error
Jonghyun Park [Fri, 3 Mar 2017 23:56:06 +0000 (08:56 +0900)]
[x86/Linux] Revise VirtualMethodFixupStub (#9929)
* [x86/Linux] Revise VirtualMethodFixupStub
Tarek Mahmoud Sayed [Fri, 3 Mar 2017 22:50:38 +0000 (14:50 -0800)]
Clean up Normalization code (#9941)
Gaurav Khanna [Fri, 3 Mar 2017 21:32:05 +0000 (13:32 -0800)]
Merge pull request #9942 from dotnet-bot/master-UpdateDependencies
Update CoreClr to beta-25103-03 (master)
Frederik Carlier [Fri, 3 Mar 2017 21:16:55 +0000 (22:16 +0100)]
Support local stack unwinding on Android (#9940)
Joseph Tremoulet [Fri, 3 Mar 2017 20:28:57 +0000 (15:28 -0500)]
Merge pull request #9773 from mikedn/user-range-check
Generate OAK_NO_THROW assertions from (uint)i < (uint)a.len checks
Rich Lander [Fri, 3 Mar 2017 19:04:42 +0000 (11:04 -0800)]
Add 3PN entries (#9943)
* Fix spelling
* Add 3PN for SciMark
* Add ByteMark 3PN entry
* Remove Microsoft copyright
* Add 3PN for Richards benchmark
Bruce Forstall [Fri, 3 Mar 2017 18:55:06 +0000 (10:55 -0800)]
Merge pull request #9873 from mskvortsov/ryujit-arm32-long
[RyuJIT/ARM32][WIP] Enable CG for long nodes
Mike Danes [Fri, 3 Mar 2017 17:23:33 +0000 (19:23 +0200)]
Add/improve assertion propagation comments
dotnet-bot [Fri, 3 Mar 2017 17:40:48 +0000 (17:40 +0000)]
Update CoreClr to beta-25103-03
Mike Danes [Thu, 23 Feb 2017 06:45:02 +0000 (08:45 +0200)]
Generate OAK_NO_THROW assertions from (uint)i < (uint)a.len checks
If "(uint)i < (uint)a.len" is true then we know that "i" is always within bounds and so "a[i]" does not need a range check.
Mike Danes [Thu, 2 Mar 2017 21:45:45 +0000 (23:45 +0200)]
Refactor JTRUE handling during assertion gen/prop
- Pull JTRUE related code out of the inner loop of assertion gen to keep the loop small and simple
- Ignore JTRUE in the inner loop of assertion prop, there's nothing to propagate to or from it
- Avoid some unnecessary bitset clears and copies
Mike Danes [Thu, 2 Mar 2017 17:00:36 +0000 (19:00 +0200)]
Add tests for (uint)i < (uint)a.len generated assertions
Gaurav Khanna [Fri, 3 Mar 2017 17:12:27 +0000 (09:12 -0800)]
Merge pull request #9897 from hseok-oh/patch_rootfs
[ARM32/Linux] Prepare rootfs to build libcoreclrtraceptprovider.so
Gaurav Khanna [Fri, 3 Mar 2017 17:07:12 +0000 (09:07 -0800)]
Merge pull request #9934 from dotnet-bot/master-UpdateDependencies
Update CoreClr to beta-25103-02 (master)
Russ Keldorph [Thu, 2 Mar 2017 22:51:34 +0000 (14:51 -0800)]
Use CoreCLROverridePath property for CoreFx tests
Remove file copy hack and use the new msbuild property to allow
overriding the runtime when running CoreFx tests.
Andrey Akinshin [Fri, 3 Mar 2017 15:26:54 +0000 (20:26 +0500)]
Fix license header for test (GitHub_1323) (#9931)
Mikhail Skvortcov [Fri, 3 Mar 2017 12:20:04 +0000 (15:20 +0300)]
RyuJIT/ARM32: enable codegen for long nodes.
dotnet-bot [Fri, 3 Mar 2017 09:33:08 +0000 (09:33 +0000)]
Update CoreClr to beta-25103-02
dotnet bot [Fri, 3 Mar 2017 05:37:23 +0000 (21:37 -0800)]
Update CoreClr to beta-25103-01 (#9927)
Alex Perovich [Fri, 3 Mar 2017 02:11:47 +0000 (20:11 -0600)]
Merge shared changes from corert and use C# shared project to enable VS editing (#9922)
Bruce Forstall [Fri, 3 Mar 2017 02:10:39 +0000 (18:10 -0800)]
Merge pull request #9915 from BruceForstall/IncBuild1
Fix AsmConstants.inc to be incrementally rebuilt as necessary
Aditya Mandaleeka [Fri, 3 Mar 2017 00:56:08 +0000 (16:56 -0800)]
Merge pull request #9921 from adityamandaleeka/rt_dump_changes
Fix error in dumpling text file cleanup
Bruce Forstall [Thu, 2 Mar 2017 18:23:21 +0000 (10:23 -0800)]
Fix AsmConstants.inc to be incrementally rebuilt as necessary
Previously, it was rebuilt on every build.
Bruce Forstall [Fri, 3 Mar 2017 00:17:41 +0000 (16:17 -0800)]
Merge pull request #9917 from BruceForstall/BuildCrossAltJit
Build cross-compile altjits for arm32 and arm64
Aditya Mandaleeka [Thu, 2 Mar 2017 22:28:13 +0000 (14:28 -0800)]
Fix error in dumpling text file cleanup.
Bruce Forstall [Thu, 2 Mar 2017 17:14:52 +0000 (09:14 -0800)]
Build cross-compile altjits for arm32 and arm64
On x86, build an x86-host, arm32-target altjit.
On amd64, build an amd64-host, arm64-target altjit.
Each are named protononjit.dll, using our existing naming
convention (proto for "prototype", nonjit meaning the code
won't be executed).
Use these via:
```
set COMPlus_AltJit=*
set COMPlus_AltJitName=protononjit.dll
```
Of course, instead of `*`, you can use any function name.
They are very useful for debugging the JIT in a "nicer" environment
than the full native environment, especially if the full native
environment isn't complete, or if the tools (such as debuggers, editors)
are better in the cross environment than the native environment.
These are only built on Windows. It might be possible to build them
on Linux as well, but that will require some more build work.
Evgeny Pavlov [Thu, 2 Mar 2017 21:27:36 +0000 (00:27 +0300)]
[x86/Linux] Add STDCALL for reverse P/Invoke delegates (#9909)
Igor Kulaychuk [Thu, 2 Mar 2017 20:41:09 +0000 (23:41 +0300)]
[Linux][GDB-JIT] Fix lldb type displaying issues (#9906)
* Make array data have fixed size of 1 element
* Add bounds info for multidimentional arrays
* Fix array element size for value types
* Fix ref types duplication in DWARF types cache map
* Add typedef for reference types
Example: make MyClass a typedef for __MyClass *
* Use type name with namespace in DWARF
* Fix duplication of typedef statements in DWARF
* Fix memory leak in ByteTypeInfo::DumpStrings
* Add typedefs for build-in types
Bruce Forstall [Thu, 2 Mar 2017 19:26:01 +0000 (11:26 -0800)]
Merge pull request #9916 from dotnet-bot/from-tfs
Merge changes from TFS
Bruce Forstall [Thu, 2 Mar 2017 19:23:20 +0000 (11:23 -0800)]
Add src\jit\protononjit directory to git mirror in preparation for future GitHub work
[tfs-changeset: 1649472]
Stephen Toub [Thu, 2 Mar 2017 19:18:19 +0000 (14:18 -0500)]
Merge pull request #9911 from dotnet-bot/master-UpdateDependencies
Update CoreClr to beta-25102-03 (master)
Michelle McDaniel [Thu, 2 Mar 2017 19:06:53 +0000 (11:06 -0800)]
Merge pull request #9775 from adiaaida/addThroughputPerfTesting
Add Throughput Perf testing
Fadi Hanna [Thu, 2 Mar 2017 19:04:22 +0000 (11:04 -0800)]
Fix LDVIRTFTN case for GVM target methods (#9890)
* Fix LDVIRTFTN case for GVM target methods
Brian Sullivan [Thu, 2 Mar 2017 18:43:01 +0000 (10:43 -0800)]
Update windows-instructions.md
Ahson Ahmed Khan [Thu, 2 Mar 2017 18:13:13 +0000 (10:13 -0800)]
Removing ref T Span.GetItem and updating Indexer to return ref (#9787)
dotnet-bot [Thu, 2 Mar 2017 17:28:13 +0000 (17:28 +0000)]
Update CoreClr to beta-25102-03
Gaurav Khanna [Thu, 2 Mar 2017 16:53:57 +0000 (08:53 -0800)]
Merge pull request #9894 from dotnet-bot/master-UpdateDependencies
Update CoreClr to beta-25102-02 (master)
Andy Ayers [Thu, 2 Mar 2017 03:47:18 +0000 (19:47 -0800)]
Devirtualize calls in some simple cases
Devirtualize calls where the this object type at a call site
is a subtype of the type described at the call site. Currently learns
types from local and arg references and a subset of other operators.
Will devirtualize if either the class or method is final (sealed in C#),
or if the type is known exactly (eg from a newobj).
Devirtualization is run twice, once during importation, and again in a
limited way after inlinining. Calls devirtualized during importation are
are subsequently eligible for inlining. Calls devirtualized during inlining
currently cannot be inlined.
Andy Ayers [Mon, 30 Jan 2017 23:45:11 +0000 (15:45 -0800)]
Jit interface support for devirtualization
Add new method to jit interface so the jit can determine what derived
method might be called for a given base method, derived class pair.
Implement support in the VM and in other places (zap, spmi).
Jonghyun Park [Thu, 2 Mar 2017 14:52:26 +0000 (23:52 +0900)]
[x86/Linux] Fix clang 3.9 build error (#9900)
dotnet-bot [Thu, 2 Mar 2017 10:12:44 +0000 (10:12 +0000)]
Update CoreClr to beta-25102-02
Jonghyun Park [Thu, 2 Mar 2017 09:58:55 +0000 (18:58 +0900)]
[x86/Linux] Initial filter support (#9820)
* [x86/Linux] Initial Filter Support
Austin Wise [Thu, 2 Mar 2017 09:28:56 +0000 (01:28 -0800)]
Remove CLRThreadpoolHosted as it always returns false. (#9899)
Jonghyun Park [Thu, 2 Mar 2017 07:50:11 +0000 (16:50 +0900)]
[x86/Linux] Clean up EXCEPTION_REGISTRATION_RECORD (#9895)
Ahson Ahmed Khan [Thu, 2 Mar 2017 07:49:08 +0000 (23:49 -0800)]
Temporarily disabling SpanBench tests that use indexer setter. (#9896)
Matt Warren [Thu, 2 Mar 2017 06:57:52 +0000 (06:57 +0000)]
Update links to point to valid BotR pages - One extra link that was missed out (#9878)
Michal Strehovský [Thu, 2 Mar 2017 06:15:34 +0000 (22:15 -0800)]
Use the correct class handle for fat call temporary (#9893)
`clsHnd` is the class handle of the type that owns the method for
`opcode != CEE_CALLI`. We need the class handle for the method's
return type though.
This is ending up with RyuJIT asking weird questions over JitInterface
(`getClassGClayout` of a reference type).
Also removing a line that was added in #9191 to support the old code.