Bruce Forstall [Tue, 31 Jan 2017 23:55:42 +0000 (15:55 -0800)]
Merge pull request #9202 from BruceForstall/FixRetBufMarking
Mark RetBuf as SIMD variable
Jonghyun Park [Tue, 31 Jan 2017 23:16:57 +0000 (08:16 +0900)]
Use Explicit Field Reference in FillRegDisplay (for x86) (#9212)
Stephen Toub [Tue, 31 Jan 2017 22:54:43 +0000 (17:54 -0500)]
Merge pull request #9224 from stephentoub/spinlock_tryenter_perf
Make SpinLock.TryEnter(ref bool) fail faster
Bruce Forstall [Mon, 30 Jan 2017 23:34:28 +0000 (15:34 -0800)]
Mark RetBuf as SIMD variable
I see changes where storeIndir is used now instead of
a BLK copy, which is an IR shape improvement.
This improves codegen for SIMD8, degrades it for SIMD12 (due
to encoding size differences with SSE instructions, though
the IR looks better), and is neutral for SIMD16, in the
VectorReturn test.
William Godbe [Tue, 31 Jan 2017 21:33:34 +0000 (13:33 -0800)]
Merge pull request #9217 from wtgodbe/syncNativeBins
Download native binaries into a folder named 'RID'
Stephen Toub [Tue, 31 Jan 2017 19:50:44 +0000 (14:50 -0500)]
Make SpinLock.TryEnter(ref bool) fail faster
One uses SpinLock.TryEnter(ref bool) when they want to do a quick test to either acquire the lock or know it's held by someone else. We can make the latter path much faster so that it fails faster.
William Godbe [Tue, 31 Jan 2017 21:12:35 +0000 (13:12 -0800)]
Merge pull request #9222 from wtgodbe/fileNameTooLong
Remove all '.requires_signing' files before zipping tests
wtgodbe [Tue, 31 Jan 2017 20:09:36 +0000 (12:09 -0800)]
Remove all '.requires_signing' files before zipping tests
wtgodbe [Tue, 31 Jan 2017 18:56:32 +0000 (10:56 -0800)]
Download native binaries into a folder named 'RID'
Aditya Mandaleeka [Tue, 31 Jan 2017 18:26:49 +0000 (10:26 -0800)]
Merge pull request #9207 from adityamandaleeka/ci_dump_improvements
Switch to a newer version of the dump management service we use
Gaurav Khanna [Tue, 31 Jan 2017 16:55:02 +0000 (08:55 -0800)]
Merge pull request #9206 from dotnet-bot/master-UpdateDependencies
Update CoreFx to beta-24931-02 (master)
Bruce Forstall [Tue, 31 Jan 2017 16:51:09 +0000 (08:51 -0800)]
Merge pull request #9127 from BruceForstall/FixSimdFieldMorphing
Fix morphing of GT_FIELD of SIMD in IND context
dotnet-bot [Tue, 31 Jan 2017 14:58:57 +0000 (14:58 +0000)]
Update CoreFx to beta-24931-02
Jonghyun Park [Tue, 31 Jan 2017 09:43:27 +0000 (18:43 +0900)]
Enable EECodeManager::GetGSCookieAddr when CROSSGEN_COMPILE is not defined (#9209)
Bruce Forstall [Tue, 24 Jan 2017 01:40:57 +0000 (17:40 -0800)]
Fix morphing of GT_FIELD of SIMD in IND context
Bruce Forstall [Tue, 31 Jan 2017 07:19:28 +0000 (23:19 -0800)]
Merge pull request #9183 from BruceForstall/FixBlockOpMorphing
Fix morphing of block op assignments
Aditya Mandaleeka [Thu, 26 Jan 2017 23:25:10 +0000 (15:25 -0800)]
Switch to a newer version of the dump management service we use.
Bruce Forstall [Mon, 30 Jan 2017 23:46:39 +0000 (15:46 -0800)]
Merge pull request #9199 from BruceForstall/FixBadIfCondition
Fix misspelling of FEATURE_MULTIREG_ARGS
Frederik Carlier [Mon, 30 Jan 2017 23:21:49 +0000 (00:21 +0100)]
Support cross-building CoreCLR for arm64 Android (#9173)
* Support cross-building CoreCLR for arm64 Android
* Link gcc_s before libunwind
Koundinya Veluri [Mon, 30 Jan 2017 22:47:29 +0000 (14:47 -0800)]
Fix Thread constructor with stack size argument (#9186)
Fix Thread constructor with stack size argument
Functional fix for #9158 and #9167
William Godbe [Mon, 30 Jan 2017 22:30:43 +0000 (14:30 -0800)]
Merge pull request #9200 from wtgodbe/FixPublish
Change '.' to '-' in publishing test native binaries
wtgodbe [Mon, 30 Jan 2017 21:24:16 +0000 (13:24 -0800)]
Change '.' to '-' in publishing test native binaries
Bruce Forstall [Mon, 30 Jan 2017 20:26:25 +0000 (12:26 -0800)]
Fix misspelling of FEATURE_MULTIREG_ARGS
Dan Moseley [Mon, 30 Jan 2017 19:47:31 +0000 (12:47 -0700)]
Merge pull request #9188 from justinvp/fixed_addressof
Avoid unnecessary branches when fixing non-null/empty arrays
Dan Moseley [Mon, 30 Jan 2017 19:47:04 +0000 (12:47 -0700)]
Merge pull request #9197 from dotnet-bot/master-UpdateDependencies
Update CoreClr to beta-24930-03 (master)
William Godbe [Mon, 30 Jan 2017 18:46:55 +0000 (10:46 -0800)]
Merge pull request #9198 from wtgodbe/FixReleaseHelix
Set BuildType in HelixPublish.proj
wtgodbe [Mon, 30 Jan 2017 18:30:48 +0000 (10:30 -0800)]
Set BuildType in HelixPublish.proj
William Godbe [Mon, 30 Jan 2017 18:27:20 +0000 (10:27 -0800)]
Merge pull request #9156 from dotnet/revert-9155-revert-9152-SpeedUpSync
Speed up test native binary syncing in pipeline
Carol Eidt [Mon, 30 Jan 2017 18:26:15 +0000 (10:26 -0800)]
Merge pull request #9170 from CarolEidt/Fix9159
Fix GT_LCL_FLD case in gtGetStructHandleIfPresent
dotnet-bot [Mon, 30 Jan 2017 17:44:04 +0000 (17:44 +0000)]
Update CoreClr to beta-24930-03
Dan Moseley [Mon, 30 Jan 2017 17:41:30 +0000 (10:41 -0700)]
Merge pull request #9117 from jamesqo/copy.names
Fix parameter names for Array.Copy
Dan Moseley [Mon, 30 Jan 2017 17:38:28 +0000 (10:38 -0700)]
Merge pull request #8438 from JonHanna/fix_1764
Support filter and fault blocks in DynamicILGenerator
Dan Moseley [Mon, 30 Jan 2017 17:37:48 +0000 (10:37 -0700)]
Merge pull request #8564 from lucenticus/x86-gdbjit
[x86/Linux] Initial support of GDB JIT interface
Dan Moseley [Mon, 30 Jan 2017 17:37:11 +0000 (10:37 -0700)]
Merge pull request #9140 from dotnet-bot/master-UpdateDependencies
Update CoreClr, CoreFx to beta-24930-02, beta-24930-01, respectively (master)
Frederik Carlier [Mon, 30 Jan 2017 16:52:12 +0000 (17:52 +0100)]
Support building lldbplugin on ARM64 (#9195)
Stephen Toub [Mon, 30 Jan 2017 16:37:27 +0000 (11:37 -0500)]
Remove allocations from Environment.GetResourceString (#9194)
* Remove delegate allocations from ResourceHelper.GetResourceString
* Remove state object allocation from GetResourceString
Now that we're no longer using ExecuteCodeWithGuaranteedBackout, we don't need the try/catch/finally functionality separated out into separate methods, and thus we don't need to allocate an object to pass into them.
Jan Kotas [Mon, 30 Jan 2017 16:36:29 +0000 (08:36 -0800)]
Delete unused encodings from CoreLib (#9196)
Andy Ayers [Mon, 30 Jan 2017 16:19:25 +0000 (08:19 -0800)]
JIT: fixup finally target flag after finally opts run (#9171)
The finally target flag can't be incrementally maintaned during finally
opts, so flag if any of these opts run and then run a post-opts fixup pass
to properly determine the flag value.
Only impacts arm targets.
Closes #9015.
dotnet-bot [Mon, 30 Jan 2017 15:18:54 +0000 (15:18 +0000)]
Update CoreClr, CoreFx to beta-24930-02, beta-24930-01, respectively
Frederik Carlier [Mon, 30 Jan 2017 10:53:45 +0000 (11:53 +0100)]
Add support for lldb-3.9 (#9190)
Evgeny Pavlov [Mon, 30 Jan 2017 10:12:17 +0000 (13:12 +0300)]
[GDB-JIT][Linux] Add support of base classes introspection (#9083)
* [GDB-JIT] Initial support of inherited fields displaying for classes
* [GDB-JIT] Fix link to already existed parent class
* [GDB-JIT] Fix incorrect processing of links to parent class with RefTypeInfo
* [GDB-JIT] Ignore inheritance from System.ValueType
* [GDB-JIT] Insert additional check whether we already dump parent class
Jan Kotas [Sun, 29 Jan 2017 15:22:30 +0000 (07:22 -0800)]
Merge pull request #9181 from dotnet/revertevent
Revert recent changes of #if blocks in Eventing code
Jan Kotas [Sun, 29 Jan 2017 03:54:57 +0000 (19:54 -0800)]
Remove invalid asserts on unsupported underlying enum type (#9192)
The underlying enum type can be floating point type or native integer for historic reasons.
However, not all framework methods support these special enum types. They should throw exception
for them instead of asserting.
Fix issue #7670
Sivarv [Sat, 28 Jan 2017 17:54:45 +0000 (09:54 -0800)]
Merge pull request #9184 from AndyAyersMS/RegexDna
Add BenchmarkGames regexdna test
Justin Van Patten [Sat, 28 Jan 2017 17:31:32 +0000 (09:31 -0800)]
Avoid unnecessary branches when fixing non-null/empty arrays
Karel Zikmund [Sat, 28 Jan 2017 17:06:43 +0000 (09:06 -0800)]
Update glossary.md
Andy Ayers [Sat, 28 Jan 2017 00:45:38 +0000 (16:45 -0800)]
Add BenchmarkGames regexdna test
Based on the "version 6" test base. As with revcomp, keeping the input
files as unix line ending, and relatively small.
Bruce Forstall [Sat, 28 Jan 2017 01:05:05 +0000 (17:05 -0800)]
Fix morphing of block op assignments
During morphing, block op assignments had their children set to
the "IND" morph context. This is incorrect. Previously, block
ops would "incorporate" the "IND-ness" of their children. However,
now, the operands will be a IND/BLK/OBJ node explicitly. Deleting
this incorrect code enables other changes, not yet merged, to work.
This causes no amd64 SPMI desktop asm diffs.
Dan Moseley [Sat, 28 Jan 2017 00:46:56 +0000 (17:46 -0700)]
Merge pull request #9164 from justinvp/string_split
Remove unnecessary private String.SplitInternal method
Karel Zikmund [Fri, 27 Jan 2017 23:58:18 +0000 (15:58 -0800)]
Update glossary.md
danmosemsft [Fri, 27 Jan 2017 23:20:20 +0000 (15:20 -0800)]
Revert `Remove #if false blocks` just for Eventing
danmosemsft [Fri, 27 Jan 2017 23:19:04 +0000 (15:19 -0800)]
Revert `Remove FEATURE_MANAGED_ETW_CHANNELS`
danmosemsft [Fri, 27 Jan 2017 23:17:25 +0000 (15:17 -0800)]
Revert `Remove disabled FEATURE_ADVANCED_MANAGED_ETW_CHANNELS`
Bruce Forstall [Fri, 27 Jan 2017 22:42:26 +0000 (14:42 -0800)]
Merge pull request #9163 from BruceForstall/FixAddrModeBuilding
Stop address mode building in TreeNodeInfoInit pass
Carol Eidt [Fri, 27 Jan 2017 18:09:58 +0000 (10:09 -0800)]
Fix GT_LCL_FLD case in gtGetStructHandleIfPresent
This method was incorrectly returning the parent struct handle instead of the handle for the field. This is showing up in remorphing of a SIMD field that was assigned in a field-by-field assignment.
Fix #9159
Andy Ayers [Fri, 27 Jan 2017 17:36:09 +0000 (09:36 -0800)]
JIT: Remove empty try regions (#8949)
In runtimes that do not support thread abort, a try-finally with
an empty try can be optimized to simply the content of the finally.
See documentation update for more details.
Closes #8924.
Jonghyun Park [Fri, 27 Jan 2017 11:58:40 +0000 (20:58 +0900)]
Enable IsInPrologOrEpilog only if GC Info Decoder is not used (#9136)
* Enable IsInPrologOrEpilog only if GC Info Decoder is not used
Russ Keldorph [Fri, 27 Jan 2017 10:59:28 +0000 (02:59 -0800)]
Merge pull request #9130 from RussKeldorph/corefx
Support new corefx workflow
Frederik Carlier [Fri, 27 Jan 2017 09:38:32 +0000 (10:38 +0100)]
Use a constant to define the location of '/tmp' (#9150)
* The temp folder is in /data/local/tmp on Android, not /tmp
Frederik Carlier [Fri, 27 Jan 2017 09:37:49 +0000 (10:37 +0100)]
Support platforms where pthread is part of libc (#9154)
* Make threading code compatible with Android
Ahson Ahmed Khan [Fri, 27 Jan 2017 04:33:48 +0000 (20:33 -0800)]
Adding trivial implementations of Span Fill and Clear (#9162)
Justin Van Patten [Fri, 27 Jan 2017 03:42:29 +0000 (19:42 -0800)]
Remove unnecessary private String.SplitInternal method
Avoid the unnecessary intermediate method call.
Sivarv [Fri, 27 Jan 2017 03:23:26 +0000 (19:23 -0800)]
Merge pull request #9054 from kouvel/SpanGcReporting
Fix GC info for `ByReference<T>`
Bruce Forstall [Fri, 27 Jan 2017 02:44:31 +0000 (18:44 -0800)]
Stop address mode building in TreeNodeInfoInit pass
During Lower, TryCreateAddrMode() will try to find address
modes and convert them to GT_LEA by calling genCreateAddrMode().
Later, the TreeNodeInfoInit() pass walks the IR annotating
nodes with register requirements. It was also calling
genCreateAddrMode() and then clearing various previously
created register requirements, but not actually changing the
node to a GT_LEA. In some cases (in new code I have not checked
in yet), this was leading to code that codegen couldn't handle.
It turns out this second call to genCreateAddrMode() is
unnecessary (and undesireable), so I've removed it. This causes
no asm diffs on amd64 or x86 (run on desktop: superpmi
and ngen, respectively).
Carol Eidt [Fri, 27 Jan 2017 01:54:31 +0000 (17:54 -0800)]
Merge pull request #9123 from CarolEidt/FixCoreFx15309
Fix SIMD field issues
Frederik Carlier [Fri, 27 Jan 2017 01:11:40 +0000 (02:11 +0100)]
PAL Tests: Centralize common code (#9151)
Create a single COMMON_TEST_LIBRARIES variable which contains all libraries
with which the tests link.
Brian Sullivan [Fri, 27 Jan 2017 01:08:58 +0000 (17:08 -0800)]
Merge pull request #9157 from dotnet-bot/from-tfs
Merge changes from TFS
William Godbe [Fri, 27 Jan 2017 00:55:06 +0000 (16:55 -0800)]
Revert "Revert "Speed up test native binary syncing in pipeline""
William Godbe [Fri, 27 Jan 2017 00:55:03 +0000 (16:55 -0800)]
Merge pull request #9155 from dotnet/revert-9152-SpeedUpSync
Revert "Speed up test native binary syncing in pipeline"
William Godbe [Fri, 27 Jan 2017 00:54:39 +0000 (16:54 -0800)]
Revert "Speed up test native binary syncing in pipeline"
William Godbe [Fri, 27 Jan 2017 00:54:10 +0000 (16:54 -0800)]
Merge pull request #9152 from wtgodbe/SpeedUpSync
Speed up test native binary syncing in pipeline
Aditya Mandaleeka [Fri, 27 Jan 2017 00:53:54 +0000 (16:53 -0800)]
Merge pull request #9153 from adityamandaleeka/jitdisasm_fix
Fix another JIT-disasm check
Pat Gavlin [Thu, 26 Jan 2017 23:22:02 +0000 (15:22 -0800)]
Merge pull request #9141 from pgavlin/RemoveDeadCode
Remove dead code in the JIT.
Koundinya Veluri [Thu, 26 Jan 2017 22:21:03 +0000 (14:21 -0800)]
Don't require write barrier for assignment to ByRef
The legacy code gen was hitting an assert when trying to use a write barrier for ByRef to ByRef assignment. RyuJIT is also generating an unnecessary write barrier but the assert is not there. As ByRefs are never in GC locations, don't require a write barrier.
Aditya Mandaleeka [Thu, 26 Jan 2017 22:53:54 +0000 (14:53 -0800)]
Fix another JIT-disasm check.
Evgeny Pavlov [Thu, 26 Jan 2017 22:52:31 +0000 (01:52 +0300)]
[GDB-JIT][Linux] Fix incorrect DWARF offset for primitive types with typedef node (char, byte and sbyte) (#9106)
Frederik Carlier [Thu, 26 Jan 2017 22:35:37 +0000 (23:35 +0100)]
Clean up PAL functions (#9149)
Clean up PAL functions to enable cross-compiling for Android
* Rename wcsnlen to PAL_wcsnlen, avoids a name collision on Android
* Remove unuwed _getw and _putw functions
wtgodbe [Thu, 26 Jan 2017 22:01:18 +0000 (14:01 -0800)]
Speed up test native binary syncing in pipeline
Andy Ayers [Thu, 26 Jan 2017 21:18:55 +0000 (13:18 -0800)]
Merge pull request #9142 from AndyAyersMS/FixAsserts
JIT: fix some bad asserts
Russ Keldorph [Thu, 26 Jan 2017 21:01:52 +0000 (13:01 -0800)]
Merge pull request #9129 from RussKeldorph/defineenum
Disable assertions in ModuleBuilder.DefineEnum
Jan Kotas [Thu, 26 Jan 2017 20:41:45 +0000 (12:41 -0800)]
Fix encoding of native signatures in the presence of Enum.GetHashCode optimization (#9138)
Fixes #9118
dotnet-bot [Thu, 26 Jan 2017 18:51:17 +0000 (10:51 -0800)]
Fix CodeGenMirror build break due to
347243f80d62d5aaa5359ef8c8d. This includes the new file unwindx86.cpp to be compiled.
[tfs-changeset: 1645771]
Andy Ayers [Thu, 26 Jan 2017 18:48:35 +0000 (10:48 -0800)]
JIT: fix some bad asserts
Fix a few asserts that use = instead of ==.
Pat Gavlin [Thu, 26 Jan 2017 18:35:03 +0000 (10:35 -0800)]
Remove dead code in the JIT.
All code under `LONG_ASG_OPS` and `FANCY_ARRAY_OPT` is statically dead
(these symbols are never defined as anything besides 0). This change
removes this cruft.
Bruce Forstall [Thu, 26 Jan 2017 18:32:20 +0000 (10:32 -0800)]
Merge pull request #9132 from BruceForstall/FixBuildTest
Fix build-test.cmd
Bruce Forstall [Thu, 26 Jan 2017 18:16:32 +0000 (10:16 -0800)]
Merge pull request #9125 from BruceForstall/ImproveAddrModeJitDump
Improve address mode detection JitDumps
Bruce Forstall [Thu, 26 Jan 2017 18:16:08 +0000 (10:16 -0800)]
Merge pull request #9126 from BruceForstall/CleanUnusedTailcallNodes
Clean up unused tailcall via helper nodes
Bruce Forstall [Thu, 26 Jan 2017 18:14:55 +0000 (10:14 -0800)]
Merge pull request #9133 from BruceForstall/ImproveSimdTests
Improve SIMD tests
Gaurav Khanna [Thu, 26 Jan 2017 16:32:50 +0000 (08:32 -0800)]
Merge pull request #9120 from dotnet-bot/master-UpdateDependencies
Update CoreClr, CoreFx to beta-24925-08, beta-24926-02, respectively (master)
dotnet-bot [Thu, 26 Jan 2017 15:15:11 +0000 (15:15 +0000)]
Update CoreClr, CoreFx to beta-24925-08, beta-24926-02, respectively
Jonghyun Park [Thu, 26 Jan 2017 14:13:23 +0000 (23:13 +0900)]
Extract GetStackParameterSize from ESPIncrOnReturn (#9135)
Jonghyun Park [Thu, 26 Jan 2017 14:10:59 +0000 (23:10 +0900)]
Show register name in UPDATEREG log (#9134)
Mike McLaughlin [Thu, 26 Jan 2017 14:07:00 +0000 (06:07 -0800)]
Implement PAL_ProbeMemory with write/read to pipe (issue #4094) (#9122)
The recommended way to validate/probe memory on Linux/xplat is to use the write/read file apis to a pipe instead
of the hacky fragile hardware exception catching based probing we have now.
This builds on previous cleanup work in the master branch that removed the PAL Read/WriteProcessMemory
functions and added/replaced them with PAL_ProbeMemory. The DbgTransportSession::CheckBufferAccess that
issue mentions was also replaced.
Eugene Rozenfeld [Thu, 26 Jan 2017 06:08:24 +0000 (22:08 -0800)]
Merge pull request #9114 from erozenfeld/SecurityPermissionAttribute
Remove System.Security.Permissions.SecurityPermissionAttribute from System.Private.CoreLib.
Bruce Forstall [Thu, 26 Jan 2017 05:48:36 +0000 (21:48 -0800)]
Merge pull request #9124 from BruceForstall/FixSimdDetection
Fix non-deterministic AVX2 detection
Justin Van Patten [Thu, 26 Jan 2017 05:30:02 +0000 (21:30 -0800)]
Reduce number of branches and IL size of string.Trim methods (#9115)
Koundinya Veluri [Mon, 23 Jan 2017 07:14:09 +0000 (23:14 -0800)]
Fix GC info for `ByReference<T>`
When `ByRefernece<T>` is directly on the stack, not inside a `Span<T>`, ComputeGCLayout was missing to include in the GC info. This was causing many span tests to fail in GCStress=c mode.
James Ko [Thu, 26 Jan 2017 03:28:40 +0000 (22:28 -0500)]
Merge branch 'copy.names' of github.com:jamesqo/coreclr into copy.names
James Ko [Thu, 26 Jan 2017 03:28:32 +0000 (22:28 -0500)]
Fix corelib.txt message
Bruce Forstall [Thu, 26 Jan 2017 02:53:44 +0000 (18:53 -0800)]
Improve SIMD tests
1. Since the F1/F2 tests of VectorReturn found an issue Vector2
on x86, I added variants of the same test for Vector3 and Vector4
(by basically copying F1/F2 and altering appropriately).
2. I noticed VectorMin for Vector4 wasn't checking the 'W' member,
so I fixed that.