Jarret Shook [Fri, 8 Dec 2017 23:21:20 +0000 (15:21 -0800)]
Merge pull request #15445 from jashook/arm_archive_obj
Archive the obj directory for arm builds
Andy Ayers [Fri, 8 Dec 2017 23:10:25 +0000 (15:10 -0800)]
Jit: fix long shift helper for overly long shift counts (#15443)
Reduce shift amount modulo 64 to match behavior on other platforms and the
jit optimizer.
Also, fix IL in related test case so it is valid for 32 bits too.
Also, make these two tests pri-0 so they get run with regular CI testing.
Fixes #15442.
jashook [Fri, 8 Dec 2017 22:29:16 +0000 (14:29 -0800)]
Archive the obj directory for arm builds
This will allow us to correctly populate the core_root for eventual
arm linux testing.
Konstantin Baladurin [Fri, 8 Dec 2017 20:26:00 +0000 (23:26 +0300)]
Fix uaf in DestroyThread function. (#15437)
Mark thread's buffers as no longer owned before calling
Thread::OnThreadTerminate because thread can delete itself in
this method.
Bruce Forstall [Fri, 8 Dec 2017 19:59:55 +0000 (11:59 -0800)]
Merge pull request #15439 from BruceForstall/FixTimeout
Fix timeouts. Up timeouts for non-default (Pri-1) builds.
Bruce Forstall [Fri, 8 Dec 2017 18:49:15 +0000 (10:49 -0800)]
Fix timeouts. Up timeouts for non-default (Pri-1) builds.
Ahson Khan [Fri, 8 Dec 2017 11:07:16 +0000 (03:07 -0800)]
Add optional integer offset to OwnedMemory Pin (#15410)
Ahson Khan [Fri, 8 Dec 2017 11:04:57 +0000 (03:04 -0800)]
Add GetReference and TryGetArray to MemoryMarshal (#15417)
* Add GetReference and TryGetArray to MemoryMarshal
* Marking GetReference with AggressiveInlining
* Do not use ByReference as a return type.
* Addressing PR feedback.
Bruce Forstall [Fri, 8 Dec 2017 07:39:04 +0000 (23:39 -0800)]
Merge pull request #15342 from BruceForstall/FixLongNeg
Fix NEG decomposition to mark instructions that set and use flags
Bruce Forstall [Fri, 8 Dec 2017 07:34:30 +0000 (23:34 -0800)]
Merge pull request #15407 from BruceForstall/ReduceArmGcStress3JobFrequency
Reduce arm/armlb/arm64 GCStress=3 and HeapVerify=1 jobs to weekly
Sean Gillespie [Fri, 8 Dec 2017 02:39:02 +0000 (18:39 -0800)]
Cast lp1 to the ScanContext; lp2 is a function pointer. (#15433)
Ben Adams [Fri, 8 Dec 2017 00:26:23 +0000 (01:26 +0100)]
Externalize exception boxing (#15418)
Michelle McDaniel [Fri, 8 Dec 2017 00:04:22 +0000 (16:04 -0800)]
Turn off outerloop perf and make job run periodically (#15416)
This is mostly for testing purposes.
Carol Eidt [Fri, 8 Dec 2017 00:02:10 +0000 (16:02 -0800)]
Merge pull request #15400 from CarolEidt/Fix15325
Fix GT_FIELD_LIST double passed as int
Bruce Forstall [Fri, 8 Dec 2017 00:00:34 +0000 (16:00 -0800)]
Merge pull request #15425 from BruceForstall/ImproveProgressOutput
Add progress output to runtest.cmd so you don't think it's hung
William Godbe [Thu, 7 Dec 2017 22:46:48 +0000 (14:46 -0800)]
Merge pull request #15420 from swgillespie/test-build-perf
Don't set BuildVersionFile when doing a test build
Bruce Forstall [Thu, 7 Dec 2017 21:39:33 +0000 (13:39 -0800)]
Add progress output to runtest.cmd so you don't think it's hung
Sean Gillespie [Thu, 7 Dec 2017 21:09:02 +0000 (13:09 -0800)]
Predicate overwrite of BuildVersionFile on OverwriteCoreClrPackageVersion == true
Sean Gillespie [Thu, 7 Dec 2017 19:37:03 +0000 (11:37 -0800)]
Fix a variable re-definition that resulted in two git invocations per project build
Tom Deseyn [Thu, 7 Dec 2017 18:08:56 +0000 (19:08 +0100)]
Don't change SIGINT/SIGQUIT handling when they are set to SIG_IGN (#15393)
* Don't change SIGINT/SIGQUIT handling when they are set to SIG_IGN
Andy Ayers [Thu, 7 Dec 2017 07:21:07 +0000 (23:21 -0800)]
JIT: fix remorph assert in cast long shift optimization (#15386)
In some cases reachable from IL we may simplify the shift amount,
which sets the MORPHED flag, and then later remorph, leading to an assert
in DEBUG/CHECK builds.
Fix is to clear the MORPHED flag.
Added test case.
dotnet-maestro-bot [Thu, 7 Dec 2017 04:27:45 +0000 (22:27 -0600)]
Update CoreClr, CoreFx to preview1-26006-09, preview1-26006-07, respectively (#15401)
Fei Peng [Thu, 7 Dec 2017 04:01:18 +0000 (20:01 -0800)]
new intrinsic type support (#15340)
Anirudh Agnihotry [Thu, 7 Dec 2017 03:16:50 +0000 (19:16 -0800)]
Null Check added to GetFullPath Function in Unix (#15399)
Null Check added to GetFullPath Function in Unix
Bruce Forstall [Thu, 7 Dec 2017 00:10:58 +0000 (16:10 -0800)]
Reduce arm/armlb/arm64 GCStress=3 and HeapVerify=1 jobs from daily to weekly
This will reduce our ARM64 hardware load.
Bruce Forstall [Thu, 7 Dec 2017 00:00:33 +0000 (16:00 -0800)]
Merge pull request #15385 from BruceForstall/EnableGcstressPeriodicJobs
Enable arm/armlb/arm64 periodic GCStress=C jobs
Michelle McDaniel [Wed, 6 Dec 2017 19:35:55 +0000 (11:35 -0800)]
Move all non-full opt, non-pgo perf legs to periodic (#15398)
We don't need the perf results for min-opt/no-pgo runs as frequently as
we need full-opt pgo runs (which we want to get per commit, if
possible), so move them to only running once daily.
Carol Eidt [Wed, 6 Dec 2017 00:32:52 +0000 (16:32 -0800)]
Fix GT_FIELD_LIST double passed as int
Correctly handle a double field passed in int regs under a `GT_FIELD_LIST`, especially in the case of a `PutArgSplit`.
Also, reduce the spew of dump info when `COMPlus_JitStressModeNames` or `COMPlus_JitStressModeNamesNot` are set.
Enhance the dumping of `fgArgInfo`.
Fix #15325
Bruce Forstall [Wed, 6 Dec 2017 17:22:20 +0000 (09:22 -0800)]
Merge pull request #15388 from BruceForstall/NoLongRunningTests
Don't run ARM/ARM64 tests marked LONG_RUNNING
dotnet-maestro-bot [Wed, 6 Dec 2017 16:48:13 +0000 (10:48 -0600)]
Update PgoData to master-
20171206-0036 (#15390)
Bruce Forstall [Wed, 6 Dec 2017 16:45:04 +0000 (08:45 -0800)]
Merge pull request #15387 from BruceForstall/FixArmTesting
Adjust BenchI/BenchF paths to account for recent file moves
Anirudh Agnihotry [Wed, 6 Dec 2017 07:09:24 +0000 (23:09 -0800)]
Address PR feedback for change adding key to KeyNotFoundException (#15234)
* Throwing KeyNotFoundException using Throw + get for JIT to generate better code for ii
Bruce Forstall [Wed, 6 Dec 2017 01:48:54 +0000 (17:48 -0800)]
Don't run ARM/ARM64 tests marked LONG_RUNNING
These tests can time out when run with the default job timeout.
So, disable them in normal runs.
dotnet-maestro-bot [Wed, 6 Dec 2017 01:29:56 +0000 (19:29 -0600)]
Update CoreClr to preview1-26005-01 (#15371)
Bruce Forstall [Wed, 6 Dec 2017 00:53:59 +0000 (16:53 -0800)]
Adjust BenchI/BenchF paths to account for recent file moves
Bruce Forstall [Wed, 6 Dec 2017 00:26:22 +0000 (16:26 -0800)]
Enable arm/armlb/arm64 periodic GCStress=C jobs
Steve MacLean [Tue, 5 Dec 2017 23:15:23 +0000 (18:15 -0500)]
[Arm64] Enable SIMD (#14633)
* [Arm64] EXTERNAL_FeatureSIMD_Default = 1
* [Arm64] Enable SIMD compilation
* [Arm64] Enable SIMD in crossgen
* [Arm64] Allow SIMD in altjit
Add flag SIMD16ByteOnly
Set COMPlus_SIMD16ByteOnly=1 in x64_arm64_altjit.cmd
If SIMD16ByteOnly, limit clear FLAG_USE_AVX2 to disable SIMD32 vector size
Enable SIMD in protonjit
* Fix #if per feedback
Matt Warren [Tue, 5 Dec 2017 17:08:36 +0000 (17:08 +0000)]
Fixed some broken formatting (#15369)
dotnet-maestro-bot [Tue, 5 Dec 2017 12:54:34 +0000 (06:54 -0600)]
Update BuildTools, CoreClr, PgoData to prerelease-02304-01, preview1-26004-04, master-
20171205-0037, respectively (#15356)
Stephen Toub [Tue, 5 Dec 2017 03:53:04 +0000 (22:53 -0500)]
Remove old methods using string format (#15363)
Jon Hanna [Tue, 5 Dec 2017 03:52:04 +0000 (03:52 +0000)]
Have a Type.IsConvertible property. (#15365)
dotnet/corefx#25671 asks for this as a public API, but this just seeks
to make it available to corefx.
Motivation discussed at dotnet/corefx#25663 and dotnet/corefx#25670
Mirror PR to corert will need and overload making runtime types return
false there.
Michelle McDaniel [Mon, 4 Dec 2017 22:01:07 +0000 (14:01 -0800)]
Separate large perf benchmarks into their own legs (#15231)
Separate large perf benchmarks into their own legs
This change splits the windows perf test stages into 6 pipelined legs per flavor to reduce the amount of time we spend running the perf tests and reduce the total time of the job. This change also decreases the size of the stashed bin directory by deleting the obj directory. Finally, we move the benchstones suite into one directory (moving BenchF and BenchI into a shared dir called Benchstones)
dotnet-maestro-bot [Mon, 4 Dec 2017 12:27:25 +0000 (06:27 -0600)]
Update CoreClr, PgoData to preview1-26004-01, master-
20171204-0047, respectively (#15350)
Zhukov Maxim [Sun, 3 Dec 2017 05:42:48 +0000 (08:42 +0300)]
Rename the method FloorLog2 to be more clear & fix spelling (#25335) (#15349)
dotnet-maestro-bot [Sat, 2 Dec 2017 20:35:44 +0000 (14:35 -0600)]
Update CoreClr, PgoData to preview1-26002-02, master-
20171202-0035, respectively (#15346)
Andy Ayers [Sat, 2 Dec 2017 17:12:09 +0000 (09:12 -0800)]
JIT: defer removing statements during opt CSE (#15323)
The statements may contain CSE defs which if removed can confuse subsequent
CSE processing.
Fixes #15319.
Jan Kotas [Sat, 2 Dec 2017 09:21:49 +0000 (01:21 -0800)]
Delete sortversioning.cpp/h (#15345)
Fixes #8403
dotnet-maestro-bot [Sat, 2 Dec 2017 04:13:49 +0000 (22:13 -0600)]
Update CoreClr, CoreFx to preview1-26002-01, preview1-26002-01, respectively (#15344)
Bruce Forstall [Sat, 2 Dec 2017 00:27:24 +0000 (16:27 -0800)]
Fix NEG decomposition to mark instructions that set and use flags
Fixes bad codegen in System.Math.Sign(long):
```
return unchecked((int)(value >> 63 | (long)((ulong)-value >> 63)));
```
where the flag-setting low part of the decomposed NEG was removed
as dead by the constant right shift with constant >= 32, even though
the flag setting is needed by the upper part.
Fixes the MathSign5 failure of #14860.
Carol Eidt [Fri, 1 Dec 2017 23:56:43 +0000 (15:56 -0800)]
Merge pull request #15315 from CarolEidt/Fix14548
Unmark unused fields as not lvOnFrame
Stephen Toub [Fri, 1 Dec 2017 23:27:45 +0000 (18:27 -0500)]
Improve throughput of String.Split(char / char[], ...) (#15322)
* Improve throughput of String.Split(char / char[], ...)
Also reduces the amount of unsafe code used.
* Address PR feedback
Victor "Nate" Graf [Fri, 1 Dec 2017 23:18:02 +0000 (15:18 -0800)]
Change name to perf jobs to reflect the actual OS they run on (#15314)
* Change name fo perf jobs to reflect the actual OS they run on
* readd Ubuntu 14.04
Carol Eidt [Fri, 1 Dec 2017 23:03:18 +0000 (15:03 -0800)]
Merge pull request #15190 from sdmaclea/PR-ARM64-IsContainableMemoryOp
[Arm64] Allow IsContainableMemoryOp() use
Bruce Forstall [Fri, 1 Dec 2017 22:58:34 +0000 (14:58 -0800)]
Merge pull request #15281 from sdmaclea/PR-ARM64-Fix-UNROLL_LIMIT
[Arm64] Fix UNROLL_LIMIT logic
Bruce Forstall [Fri, 1 Dec 2017 22:57:24 +0000 (14:57 -0800)]
Merge pull request #15270 from sdmaclea/PR-ARM64-JITSTRESS-lvaStressLclFldCB
[ARM64] Align Compiler::lvaStressLclFldCB padding
Andy Ayers [Fri, 1 Dec 2017 22:08:31 +0000 (14:08 -0800)]
JIT: handle boundary cases for casts of long shifts (#15294)
* JIT: handle boundary cases for casts of long shifts
Remove the assert that the shift count is non-negative.
Don't try and optimize if the shift count is >= 64 or < 0.
Update test case to cover these values.
Updates the fix from #15236.
Closes #15291.
dotnet-maestro-bot [Fri, 1 Dec 2017 22:04:54 +0000 (16:04 -0600)]
Update BuildTools, CoreClr, CoreFx to prerelease-02301-01, preview1-26001-02, preview1-26001-02, respectively (#15321)
Bruce Forstall [Fri, 1 Dec 2017 20:30:33 +0000 (12:30 -0800)]
Merge pull request #15262 from sdmaclea/PR-ARM64-JITSTRESS-genEpilogRestoreReg
[Arm64] Use REG_IP1 for genEpilogRestoreReg*
Bruce Forstall [Fri, 1 Dec 2017 20:06:57 +0000 (12:06 -0800)]
Merge pull request #15275 from sdmaclea/PR-ARM64-SIMD-genMultiRegCallStoreToLocal-produceReg
[Arm64] SIMD genMultiRegCallStoreToLocal add missing produceReg
Bruce Forstall [Fri, 1 Dec 2017 17:48:12 +0000 (09:48 -0800)]
Merge pull request #15311 from BruceForstall/FixArmFlowTriggering
Don't add triggers for arm/armlb/arm64 non-flow jobs
Stephen Toub [Fri, 1 Dec 2017 11:38:30 +0000 (06:38 -0500)]
Merge pull request #15317 from dotnet-bot/from-tfs
Merge changes from TFS
Tom Deseyn [Fri, 1 Dec 2017 11:38:13 +0000 (12:38 +0100)]
Unix: Remove MaxPath, MaxName (#15229)
* Unix: Remove MaxPath, MaxName
PATH_MAX is not an upper bound on path lengths on Unix. system function return
values should be used to determine behavior, instead of limiting the path length.
NAME_MAX depends on the file system used. Its value depends on the path, it is
not constant for the entire system.
* PR feedback
* Sync with corefx Interop.ReadLink
* Deal with /etc/localtime link containing relative path
Jan Kotas [Fri, 1 Dec 2017 07:21:40 +0000 (23:21 -0800)]
Jumpstub fixes (#15296)
- Reserve space for jump stubs for precodes and other code fragments at the end of each code heap segment. This is trying
to ensure that eventual allocation of jump stubs for precodes and other code fragments succeeds. Accounting is done
conservatively - reserves more than strictly required. It wastes a bit of address space, but no actual memory. Also,
this reserve is not used to allocate jump stubs for JITed code since the JITing can recover from failure to allocate
the jump stub now. Fixes #14996.
- Improve algorithm to reuse HostCodeHeap segments: Maintain estimated size of the largest free block in HostCodeHeap.
This estimate is updated when allocation request fails, and also when memory is returned to the HostCodeHeap. Fixes #14995.
- Retry JITing on failure to allocate jump stub. Failure to allocate jump during JITing is not fatal anymore. There is
extra memory reserved for jump stubs on retry to ensure that the retry succeeds allocating the jump stubs that it needs
with high probability.
- Respect CodeHeapRequestInfo::getRequestSize for HostCodeHeap. CodeHeapRequestInfo::getRequestSize is used to
throttle code heap segment size for large workloads. Not respecting it in HostCodeHeap lead to too many
too small code heap segments in large workloads.
- Switch HostCodeHeap nibble map to be allocated on regular heap as part. It simplied the math required to estimate
the nibble map size, and allocating on regular heap is overall goodness since it does not need to be executable.
Bruce Forstall [Fri, 1 Dec 2017 06:00:58 +0000 (22:00 -0800)]
Merge pull request #15312 from BruceForstall/FixSmartyArchive
Fix smarty output archiving
dotnet-maestro-bot [Fri, 1 Dec 2017 04:04:04 +0000 (22:04 -0600)]
Update CoreClr, CoreFx, PgoData to preview1-26001-01, preview1-26001-01, master-
20171130-0053, respectively (#15298)
Jan Kotas [Wed, 29 Nov 2017 06:31:23 +0000 (22:31 -0800)]
Merge pull request dotnet/corert#5038 from dotnet/nmirror
Merge nmirror to master
Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
Carol Eidt [Fri, 1 Dec 2017 01:51:36 +0000 (17:51 -0800)]
Unmark unused fields as lvOnFrame
If a dependently promoted (i.e. non-register candidate) field of a struct is unused, and its parent is not allocated onto the stack, we cannot allocate the dependent fields to the stack either (they were previously being assigned offsets from 0, which may be in the GC ptr area, which causes an assert).
Fix #14548
Bruce Forstall [Fri, 1 Dec 2017 00:23:55 +0000 (16:23 -0800)]
Fix smarty output archiving
Need to unset AltJit environment variables (and possibly others)
before running powershell.
dotnet-bot [Fri, 1 Dec 2017 00:04:03 +0000 (16:04 -0800)]
delete the second definition of getMethodNameFromMetadata2
[tfs-changeset: 1682581]
dotnet-bot [Thu, 30 Nov 2017 23:37:53 +0000 (15:37 -0800)]
delete the second definition of getMethodNameFromMetadata
[tfs-changeset: 1682574]
Bruce Forstall [Thu, 30 Nov 2017 21:58:08 +0000 (13:58 -0800)]
Don't add triggers for arm/armlb/arm64 non-flow jobs
Roman Artemev [Thu, 30 Nov 2017 21:27:10 +0000 (13:27 -0800)]
Fixed DCE of call nodes (#15192)
Fixed DCE of call nodes which affect stack level
Bruce Forstall [Thu, 30 Nov 2017 20:40:28 +0000 (12:40 -0800)]
Merge pull request #15305 from BruceForstall/AddMoreSmartyZipFileLogging
Add temporary logging to track down failure to ZIP smarty output
Michal Strehovský [Thu, 30 Nov 2017 20:23:34 +0000 (21:23 +0100)]
Use ILAsm from buildtools, take 2 (#15300)
Since #15172 got rolled back in #15292, trying this again.
This time also deleting garbage `.permission` attributes from the failing tests. Permissions are not supported in CoreCLR.
Filed #15299 on some more ILAsm from CoreCLR frustrations.
Bruce Forstall [Thu, 30 Nov 2017 20:12:18 +0000 (12:12 -0800)]
Add temporary logging to track down failure to ZIP smarty output
Anirudh Agnihotry [Thu, 30 Nov 2017 19:54:26 +0000 (11:54 -0800)]
Removed Check Invalid Path Chars (#15191)
Removed Check Invalid Chars from Coreclr
chcosta [Thu, 30 Nov 2017 17:49:16 +0000 (09:49 -0800)]
Update build definition execute conditions for api v3.2 (#15285)
Tom Deseyn [Thu, 30 Nov 2017 15:42:16 +0000 (16:42 +0100)]
Expose Environment.ProcessorCount via EnvironmentAugments (#15297)
Sean Gillespie [Thu, 30 Nov 2017 05:25:20 +0000 (21:25 -0800)]
[Local GC] Fail fast on exceptions within a standalone GC (#15290)
dotnet-maestro-bot [Thu, 30 Nov 2017 04:38:08 +0000 (22:38 -0600)]
Update CoreFx to preview1-25930-02 (#15295)
Bruce Forstall [Thu, 30 Nov 2017 02:43:31 +0000 (18:43 -0800)]
Merge pull request #15292 from BruceForstall/Revert15172
Revert "Use ILAsm from buildtools (#15172)"
Sergey Andreenko [Thu, 30 Nov 2017 02:37:26 +0000 (18:37 -0800)]
[RyuJit] Delete contained workaround. (#15257)
* Move CodeGen::indirForm and IntForm to common.
* GenTreeFieldList: SetContained only for the head.
* delete the workaround.
* move check from isContained to CheckLIR
`isContained` can't determinate does node has user or not. Move this check to the appropriate place.
* check that SetContained and IsContained are consistent.
Bruce Forstall [Thu, 30 Nov 2017 01:12:01 +0000 (17:12 -0800)]
Merge pull request #15282 from BruceForstall/EnableMoreArmGcStressModes
Allow arm/armlb/arm64 to run additional jobs
Bruce Forstall [Thu, 30 Nov 2017 00:37:38 +0000 (16:37 -0800)]
Revert "Use ILAsm from buildtools (#15172)"
This reverts commit
e864120d14c829df6f6d4c6cff6f3d23db19606c.
William Godbe [Wed, 29 Nov 2017 23:43:34 +0000 (15:43 -0800)]
Merge pull request #15288 from wtgodbe/Crossgen
Ensure CoreCLR Package version overwrite works correctly in test build
Jan Vorlicek [Wed, 29 Nov 2017 23:22:05 +0000 (00:22 +0100)]
Fix GC stress C bug for arm (#15269)
* Fix GC stress C bug for arm
The OnGcCoverageInterrupt function was not removing the "thumb" bit from
the address before putting back the original instruction that was
previously replaced by an invalid instruction that invokes the
OnGcCoverageInterrupt. That resulted in the original instruction being
restored to incorrect location (to the original address + 1) and later
to the GC stress failure.
* Reflect PR feedback
wtgodbe [Wed, 29 Nov 2017 22:56:19 +0000 (14:56 -0800)]
Ensure CoreCLR Package version overwrite works correctly in test build
dotnet-maestro-bot [Wed, 29 Nov 2017 22:43:36 +0000 (16:43 -0600)]
Update CoreClr, CoreFx, PgoData to preview1-25929-01, preview1-25929-02, master-
20171129-0045, respectively (#15266)
Bruce Forstall [Wed, 29 Nov 2017 22:03:16 +0000 (14:03 -0800)]
Allow arm/armlb/arm64 to run additional jobs
This improves stress test coverage and creates consistency with
other platforms.
The jobs:
```
tieredcompilation
forcerelocs
jitnosimd (will soon be interesting for ARM64)
heapverify1
gcstress0xc_zapdisable
gcstress0xc_zapdisable_jitstress2
gcstress0xc_zapdisable_heapverify1
gcstress0xc_minopts_heapverify1
```
Steve MacLean [Wed, 29 Nov 2017 21:14:04 +0000 (16:14 -0500)]
[Arm64] Fix UNROLL_LIMIT logic
Brian Sullivan [Wed, 29 Nov 2017 21:25:00 +0000 (13:25 -0800)]
[tfs-changeset: 1682429]
Steve MacLean [Wed, 29 Nov 2017 15:03:05 +0000 (10:03 -0500)]
[ARM64] Align Compiler::lvaStressLclFldCB padding
Michal Strehovský [Wed, 29 Nov 2017 18:16:18 +0000 (19:16 +0100)]
Use ILAsm from buildtools (#15172)
* Use ILAsm from buildtools
On Windows: this stops using the ILASM that ships with Windows in favor
of CoreCLR ILASM
On Unix: this stops using the live built ILASM.
I image this fixes some issues in crossbuild on both platforms but I
can't find where the workarounds for that are.
Necessary prerequisite for being able to build tests for the default
interfaces feature.
Steve MacLean [Wed, 29 Nov 2017 17:30:32 +0000 (12:30 -0500)]
[Arm64] SIMD genMultiRegCallStoreToLocal add missing produceReg
Stephen Toub [Wed, 29 Nov 2017 17:57:47 +0000 (12:57 -0500)]
Avoid provider-related costs for default integer formatting (#15272)
One of the biggest costs when invoking Int32.ToString() and similar methods is in fetching the current culture, accessing its number information, etc. And if a culture is provided explicitly, there are still costs related to casts. But when the default format is used for non-negative integer values, the number format information isn't used at all, and we can skip the associated costs. This improves formatting perf in such cases by upwards of ~25%.
Steve MacLean [Tue, 28 Nov 2017 20:10:38 +0000 (15:10 -0500)]
[Arm64] Use REG_IP1 for genEpilogRestoreReg*
Avoid conflict with FASTTAILCALL_TARGET reg
dotnet-maestro-bot [Wed, 29 Nov 2017 04:37:51 +0000 (22:37 -0600)]
Update BuildTools, CoreClr, CoreFx to prerelease-02228-01, preview1-25928-08, preview1-25929-01, respectively (#15252)
Stephen Toub [Wed, 29 Nov 2017 04:34:04 +0000 (23:34 -0500)]
Merge pull request #15246 from stephentoub/double_tryformat
Add double/float.TryFormat
Stephen Toub [Wed, 29 Nov 2017 03:22:50 +0000 (22:22 -0500)]
Address PR feedback
- Simplify FCALL signatures
- Simplify mantissa logic copied from corert
Stephen Toub [Tue, 28 Nov 2017 18:31:50 +0000 (13:31 -0500)]
Reduce duplicate code for span formatting in StringBuilder