Rama krishnan Raghupathy [Sat, 14 May 2016 05:32:52 +0000 (22:32 -0700)]
Merge pull request #4927 from ramarag/ModifyBuildOnly
Removing Redundant Projects that are BuildOnly:
Fadi Hanna [Sat, 14 May 2016 03:35:54 +0000 (20:35 -0700)]
Initial implementation of generic dictionary access for R2R. (#4519)
For now, only TypeHandle dictionary entry slots are supported and encoded in a R2R version resilient format (the rest to come soon).
Support is only limited for x64 Windows platforms (rest is still TODO)
The basic idea: each dictionary access is initally a call to a DynamicHelper R2R cell that computes the dictionary signature, and patches the R2R cell address with an assembly stub that performs the dictionary lookup.
Rama krishnan Raghupathy [Thu, 12 May 2016 06:56:29 +0000 (23:56 -0700)]
Removing Redundant Projects that are BuildOnly :
RunOnly Projects should refer to only BuildAndRun projects
BuildOnly projects are to be used only for Non-Execution cases
and custom execution drivers
Bruce Bowyer-Smyth [Sat, 14 May 2016 02:25:22 +0000 (12:25 +1000)]
Port string.PadLeft and string.PadRight from CoreRT (#4968)
Jan Kotas [Sat, 14 May 2016 02:24:38 +0000 (19:24 -0700)]
Merge pull request #4970 from dotnet-bot/from-tfs
Merge changes from TFS
Pat Gavlin [Fri, 13 May 2016 23:39:52 +0000 (16:39 -0700)]
Fix a PCH-related issue in the jitstdout change.
[tfs-changeset: 1604771]
Jarret Shook [Fri, 13 May 2016 23:37:32 +0000 (16:37 -0700)]
Merge pull request #4901 from LLITCHEV/Issue4801
Add a issues target file with tests to be excluded on jit32 x86.
Lubomir Litchev [Tue, 10 May 2016 23:44:40 +0000 (16:44 -0700)]
Add a issues target file with thests to be excluded on jit32 x86.
Aditya Mandaleeka [Fri, 13 May 2016 21:45:27 +0000 (14:45 -0700)]
Merge pull request #4922 from adityamandaleeka/remove_getfilenamefromsymlink
Change how PAL file I/O functions treat symbolic links
Sean Gillespie [Fri, 13 May 2016 21:13:21 +0000 (14:13 -0700)]
Merge pull request #4958 from swgillespie/disable-gcstress
Exclude from GCStress a few tests that take an unreasonably long time to complete
Pat Gavlin [Fri, 13 May 2016 21:02:56 +0000 (14:02 -0700)]
Merge pull request #4942 from pgavlin/gh4832
Use a JIT-specific stdout `FILE*`.
Russ Keldorph [Fri, 13 May 2016 20:55:41 +0000 (13:55 -0700)]
Merge pull request #4957 from RussKeldorph/fix2235
Reenable passing tests
Andy Ayers [Fri, 13 May 2016 20:23:49 +0000 (13:23 -0700)]
Merge pull request #4961 from AndyAyersMS/RandomInlineBudgetLimit
Add budget check to RandomPolicy
Brian Sullivan [Fri, 13 May 2016 19:51:37 +0000 (12:51 -0700)]
Merge pull request #4955 from briansull/fix-asmdiff
Fix issue with Arm64 AsmDiffs
Matt Mitchell [Fri, 13 May 2016 19:02:43 +0000 (12:02 -0700)]
Merge pull request #4956 from mmitche/gcstress-no-push
Run gcstress pri1r2r only a few times a week
tijoytom [Fri, 13 May 2016 18:44:26 +0000 (11:44 -0700)]
Merge pull request #4959 from tijoytom/master
Renaming to StringToCoTaskMemUTF8
Andy Ayers [Fri, 13 May 2016 18:32:49 +0000 (11:32 -0700)]
Add budget check to RandomPolicy
Prevents RandomPolicy from being overly aggressive with small methods
that can lead to excessive inlining.
Closes #4711. Undoes #4704.
tijoytk [Fri, 13 May 2016 18:07:28 +0000 (11:07 -0700)]
Renaming the API StringToCoTaskMemUTF8 since we
need to be consistent with existing one.Also
fixing a merge conflict.
Sean Gillespie [Fri, 13 May 2016 17:38:19 +0000 (10:38 -0700)]
Exclude from GCStress a few tests that take an unreasonably long time to complete
Matt Mitchell [Fri, 13 May 2016 17:19:57 +0000 (10:19 -0700)]
Run gcstress pri1r2r only a few times a week
Currently this is set on push, but since it can take days to run this is a bad idea
Russ Keldorph [Fri, 13 May 2016 17:12:52 +0000 (10:12 -0700)]
Reenable passing tests
Many of the tests listed in #2235 are now passing since their dependencies
have been fixed. b57367 is a notable exception; it's static dependency is
now satisfied, but at runtime it loads a type that is not present in .NET
Core. I'm deleting it because I verified there are other tests that test
the same load/invoke pattern (on different types).
The rest of the tests listed in #2235 have already been deleted for other
reasons.
Fixes #2235
Brian Sullivan [Fri, 13 May 2016 17:11:58 +0000 (10:11 -0700)]
Fix issue with Arm64 AsmDiffs
tijoytom [Fri, 13 May 2016 17:08:23 +0000 (10:08 -0700)]
Merge pull request #4911 from tijoytom/master
Marshaling helpers for MarshalAs LPUTF8Str.
Pat Gavlin [Fri, 13 May 2016 13:30:46 +0000 (06:30 -0700)]
Merge pull request #4813 from pgavlin/gh3058
Disable FEATURE_MERGE_JIT_AND_ENGINE on *nix.
Andy Ayers [Fri, 13 May 2016 07:34:40 +0000 (00:34 -0700)]
Merge pull request #4921 from AndyAyersMS/InlineReplay
Inliner: introduce ReplayPolicy
Pat Gavlin [Fri, 13 May 2016 03:48:39 +0000 (20:48 -0700)]
Address code review feedback.
Flip the order of the PID and activation function checks in the activation
signal handler.
Pat Gavlin [Wed, 11 May 2016 20:41:07 +0000 (13:41 -0700)]
Chain activation signal handlers in the PAL.
The handler for the PAL's activation signal (`SIGRTMIN` on Linux) was
not properly chaining to the previous handler. This caused problems
with thread suspension on Linux in the case that multiple PALs were
loaded in a process: whichever PAL was loaded last would handle the
signal, and if that PAL was not configured to perform the appropriate
thread suspension, the runtime would hang.
Gaurav Khanna [Fri, 13 May 2016 03:42:13 +0000 (20:42 -0700)]
Merge pull request #4941 from gkhanna79/Fix4938
Enable SHA256 to be used as hashing algorithm for compiler/assembler
Matt Ellis [Fri, 13 May 2016 02:48:23 +0000 (19:48 -0700)]
Merge pull request #4945 from dotnet-bot/from-tfs
Merge changes from TFS
Aditya Mandaleeka [Fri, 13 May 2016 01:04:47 +0000 (18:04 -0700)]
Stop using SYS_delete syscall for deleting files.
Aditya Mandaleeka [Wed, 11 May 2016 23:58:54 +0000 (16:58 -0700)]
Add tests for file IO functions operating on symlinks.
John Chen [Fri, 13 May 2016 00:39:03 +0000 (17:39 -0700)]
Merge pull request #4932 from JohnChen0/r2r
Fix contract violations in Ready to Run code
Kyungwoo Lee [Fri, 13 May 2016 00:00:18 +0000 (17:00 -0700)]
Merge pull request #4933 from dotnet-bot/from-tfs
Merge changes from TFS
Andy Ayers [Tue, 10 May 2016 21:19:41 +0000 (14:19 -0700)]
Inliner: introduce ReplayPolicy
The ReplayPolicy reads an external script to determine which inlines
to perform. The script is the same Xml syntax that's produced by
the inliner when JitInlineDumpXml is enabled. This format can be edited
by hand or tool to force particular inlining patterns to occur. Methods
or calls sites not mentioned in the script are considered as noinline.
There's a bunch of work still left to make this fully robust, but in
testing it works well enough for my immediate use case that I'll hold
off on further polish until it's needed. But, for future reference,
here's a laundry list:
* Need better ways to identify methods. Token and hash are not enough.
* Need better ways to identify call sites. Callee token is not enough.
* Consider preparsing or mapping the script into memory.
* Consider caching node positions in the InlineContexts.
* Make it robust for multithreading.
* Handle the prejit root case somehow.
* Possibly allow overriding of inline attributes.
Carol Eidt [Thu, 12 May 2016 23:34:00 +0000 (16:34 -0700)]
Update first-class-structs.md
Update some progress and fix some formatting
Pat Gavlin [Thu, 12 May 2016 21:32:13 +0000 (14:32 -0700)]
Use a JIT-specific stdout `FILE*`.
Historically, the JIT has had issues with logging inside of processes
that change the output mode of stdout (e.g. crossgen). This change
replaces the previous solution (which relied on #ifdefs) by dup'ing
the stdout file and wrapping it in a new `FILE*` set to a known
output mode.
Gaurav Khanna [Thu, 12 May 2016 23:08:10 +0000 (16:08 -0700)]
Enable SHA256 to be used as hashing algorithm for compiler/assembler
Russ Keldorph [Thu, 12 May 2016 23:07:27 +0000 (16:07 -0700)]
Small formatting fix
Sean Gillespie [Thu, 12 May 2016 21:55:12 +0000 (14:55 -0700)]
Merge pull request #4899 from swgillespie/gcsimulator-fixup
Enable the GCSimulator CI run to run on a regular basis
tijoytk [Wed, 11 May 2016 17:00:52 +0000 (10:00 -0700)]
Marshaling helpers for MarshalAs LPUTF8Str.
Tarek Mahmoud Sayed [Thu, 12 May 2016 20:17:40 +0000 (13:17 -0700)]
Merge pull request #4915 from tarekgh/FixLowerAsciiCharChecks
Fix the optimization check to call ordinal
John Chen (CLR) [Thu, 12 May 2016 17:02:33 +0000 (10:02 -0700)]
Fix contract violations in Ready to Run code
Pat Gavlin [Thu, 12 May 2016 19:57:13 +0000 (12:57 -0700)]
Apply the JIT interface changes needed for Fadi's upcoming shared generics change.
[tfs-changeset: 1604316]
Aditya Mandaleeka [Thu, 12 May 2016 19:51:46 +0000 (12:51 -0700)]
Update unix-test-instructions.md
Add quotes around coreFxBinDir in unix-test-instructions document.
Brian Sullivan [Thu, 12 May 2016 18:21:17 +0000 (11:21 -0700)]
Fix trival build break, with warning treated as error for jmpDist usage
[tfs-changeset: 1604285]
Bruce Forstall [Thu, 12 May 2016 17:59:50 +0000 (10:59 -0700)]
Merge pull request #4931 from dotnet-bot/from-tfs
Merge changes from TFS
Matt Ellis [Thu, 12 May 2016 17:42:33 +0000 (10:42 -0700)]
Merge pull request #4419 from ellismg/change-default-locale
Change how we detect the default locale
Kyungwoo Lee [Thu, 12 May 2016 17:24:56 +0000 (10:24 -0700)]
Merge pull request #4919 from kyulee1/switch
ARM64: Switch Expansion Using Jump Table
Kyungwoo Lee [Thu, 12 May 2016 04:48:40 +0000 (21:48 -0700)]
ARM64: Switch Expansion Using Jump Table
Fixes #3332
To validate various addressing in #4896, I just enable this.
Previously, we only allow a load operation to JIT data (`ldr` or
`IF_LARGELDC`).
For switch expansion, jump table is also recorded into JIT data.
In this case, we only get the address of jump table head, and
load the right entry after computing offset. So, basically `adr` or
`IF_LARGEADR` is used to not only load label within code but also refer to
the location of JIT data.
The typical code sequence for switch expansion is like this:
```
adr x8, [@RWD00] // load address of jump table head
ldr w8, [x8, x0, LSL #2] // load jump entry from table addr + x0 * 4
adr x9, [G_M56320_IG02] // load address of current baisc block
add x8, x8, x9 // Add them to compute the final target
br x8 // Indirectly jump to the target
```
Jan Kotas [Thu, 12 May 2016 15:54:17 +0000 (08:54 -0700)]
Fix desktop build
[tfs-changeset: 1604234]
Kyungwoo Lee [Thu, 12 May 2016 15:46:53 +0000 (08:46 -0700)]
Merge pull request #4926 from mylibero/issue4925
Fixed a build error on ARM
Dongyun Jin [Thu, 12 May 2016 06:24:07 +0000 (15:24 +0900)]
Fixed a build error on ARM
The recent commit (61fe464) breaks ARM build.
It defines emitInsToJumpKind twice in src/jit/emit.h on ARM.
Signed-off-by: Dongyun Jin <dongyun.jin@samsung.com>
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
David Mason [Thu, 12 May 2016 04:45:31 +0000 (21:45 -0700)]
Merge pull request #4872 from vancem/FixETWArgs
Fix problem getting arguments from ETW to an EventSource.
Kyungwoo Lee [Thu, 12 May 2016 04:37:47 +0000 (21:37 -0700)]
Merge pull request #4896 from kyulee1/longjmp
ARM64: Enable Long Address
Pat Gavlin [Thu, 12 May 2016 04:21:47 +0000 (21:21 -0700)]
Merge pull request #4920 from pgavlin/JitStandaloneDefs
Remove the use of `CROSSGEN_COMPILE` in jitconfigvalues.h.
Kyungwoo Lee [Fri, 6 May 2016 15:11:08 +0000 (08:11 -0700)]
ARM64: Enable Long Address
Fixes https://github.com/dotnet/coreclr/issues/3668
Currently ARM64 codegen can have reference within +/-1 MB due to encoding
restriction in `b<cond>/adr/ldr` instructions. This is normally okay
assuming each function is reasonably small, but certainly not working for large method which also
can be formed with an aggressive inlining probably like crossgen/corert scenarios.
In addition, for hot/cold code separation long address is a prerequisite
since reference can be across different regions which are arbitrary.
In fact, we need additional relocations which are not in this change yet.
In details, this supports long address for conditional jump/address loading/constant
loading operations by default while they can be shortened later by
`emitJumpDistBind()` if they can fit into the smaller encoding. Logically
those operations now can reach within +/-4GB address range.
Note I haven't extended unconditional jump in this change for simplicity
so it can reach within +/-128MB same as before.
`emitOutputLJ` is extended to finally encode these operations.
There are 3 pseudo instructions introduced. These can be expanded either
short/long form.
1. Conditional jump. See `emitIns_J()`
a. Short form(`IF_BI_0B`): `b<cond> rel_addr`
b. Long form(`IF_LARGEJMP`):
```
b<rev cond> $LABEL
b rel_addr (unconditional jump)
$LABEL:
```
2. Load label(address computation). See `emitIns_R_L()`
a. Short form(`IF_DI_1E`): `adr x, [rel_addr]`
b. Long form(`IF_LARGEADR`):
```
adrp x, [rel_page_addr]
add x, x, page_offs
```
3. Load constant (from JIT data). See `emitIns_R_C()`
a. Short form(`IF_LS_1A`): `ldr x, [rel_addr]`
b. Long form(`IF_LARGLDC`):
```
adrp x, [rel_page_addr]
ldr x, [x, page_offs]
(fmov v, x in case loading vector constant)
```
In addition, JIT data is aligned on 8 byte to be accessible from large
load. Replaced JitLargeBranches by JitLongAddress to test stress on these
operations.
Tarekm Mahmoud Sayed [Wed, 11 May 2016 17:38:21 +0000 (10:38 -0700)]
Fix the optimization check to call ordinal
we have some optimization to check if we can perform the string search as ordinal
operation instead of lingustic operation. we used to check if the string is Ascii
which check if all chars in the string are in range 0~7F.
This check was not enough because in Linux, ICU do some special handling for some
characters in the Ascii range. so we need to avoid calling the ordinal operation
when encountering such chracters.
The fix is to take the advantage of String.IsFastSort which do exactly what we need
Lubomir Litchev [Thu, 12 May 2016 02:10:31 +0000 (19:10 -0700)]
Merge pull request #4917 from LLITCHEV/Issue2329
Increase timeout for GCStress jobs. Set the gcstress15_pri1r2r to be a
Pat Gavlin [Wed, 11 May 2016 22:18:54 +0000 (15:18 -0700)]
Remove the use of `CROSSGEN_COMPILE` in jitconfigvalues.h.
This symbol was previously used to ensure that noway failures did not trigger
debugger breaks when running non-release builds of the JIT in crossgen on the
desktop CLR. We believe that this behavior is no longer necessary nor desirable.
Bruce Forstall [Wed, 11 May 2016 22:23:37 +0000 (15:23 -0700)]
Merge pull request #4916 from BruceForstall/TweakLongs
Small tweaks to RyuJIT/x86 LONG handling
Lubomir Litchev [Wed, 11 May 2016 18:46:31 +0000 (11:46 -0700)]
Increase timeout for GCStress jobs. Set the gcstress15_pri1r2r to be a
GCStress job.
Aditya Mandaleeka [Wed, 11 May 2016 22:08:36 +0000 (15:08 -0700)]
Avoid following symlinks in PAL move/delete functions.
Rama krishnan Raghupathy [Wed, 11 May 2016 21:15:11 +0000 (14:15 -0700)]
Merge pull request #4809 from ramarag/ZapUnwindInfo
[Arm64]: Implement GetUnwindDataBlob
Bruce Forstall [Wed, 11 May 2016 18:39:04 +0000 (11:39 -0700)]
Merge pull request #1241 from mikedn/modopt
Extend the DIV/MOD dividend into RDX:RAX only if needed
Geoff Kizer [Wed, 11 May 2016 18:29:50 +0000 (11:29 -0700)]
Merge pull request #4762 from geoffkizer/ecsopt
Optimize async invocation
Carol Eidt [Wed, 11 May 2016 18:21:07 +0000 (11:21 -0700)]
Merge pull request #4848 from mikedn/lsra-counts
Remove unnecessary LSRA members
Bruce Forstall [Wed, 11 May 2016 17:35:47 +0000 (10:35 -0700)]
Small tweaks to RyuJIT/x86 LONG handling
1. Display GT_LONG trees in JitDump as gt_long, not just "long". This
helps distinguish them from the pervasive "long" type, making them easier
to see and search for.
2. Change the "hi" operator GT_* node definition to not include GTK_EXOP.
3. Dump the lva table after lvaPromoteLongVars().
Wes Haggard [Wed, 11 May 2016 17:22:29 +0000 (10:22 -0700)]
Merge pull request #4890 from AlexGhiondea/BuildRefAssemblyPerTargetOS
Fix the reference assembly for mscorlib to match the surface area of the OS
Bruce Forstall [Wed, 11 May 2016 17:13:06 +0000 (10:13 -0700)]
Merge pull request #4903 from BruceForstall/MoveToLegacy
Move some LEGACY_BACKEND code to codegenlegacy.cpp
Bruce Forstall [Wed, 11 May 2016 17:12:39 +0000 (10:12 -0700)]
Merge pull request #4900 from BruceForstall/PartialFix4817c
Fix kill set of rep stos to include ECX
tijoytom [Wed, 11 May 2016 16:57:46 +0000 (09:57 -0700)]
Merge pull request #4895 from tijoytom/master
Cleaning up a bunch of warnings.
Jan Kotas [Wed, 11 May 2016 11:17:04 +0000 (04:17 -0700)]
Merge pull request #4898 from dotnet-bot/from-tfs
Merge changes from TFS
Geunsik Lim [Wed, 11 May 2016 08:44:36 +0000 (17:44 +0900)]
Linux/ARM: Use -O1 level to avoid segfault in release-build for Linux/ARM (#4904)
We cannot still run 'hello world' console application even though
the cross compilation of CoreCLR is successfully completed.
In release-build, the segmentation fault of 'hello world'
is made by the misalignment of thread local storage (TLS) section
because of the aggressive optimization level of clang compiler for
code optimization(e.g., file size and execution speed).
It means that Clang/LLVM has a bug in case of the usage of O2/O3 flag.
Below is the major difference among the optimization levels of Clang.
* -O2 is based on -O1:
.adding: -gvn -constmerge -globaldce -slp-vectorizer -mldst-motion -inline
.removing: -always-inline
* -O3 is based on -O2:
.adding: -argpromotion
ver2:
- Corerun is loading 'libcoreclr.so' using dlopen() library call. So, we can not
use initial-exec TLS model (ver1) because of the thread-safe issue of
the multi-threaded. (Reported by @janvorli)
- In the release-build, Let's replace -O3 with -O1 to avoid the segmentation
fault due to the aggressive optimization level of the Clang until fixing
the bug of the Clang/LLVM.
ver1:
- The default value of clang is "global-dynamic". However, the thread-local
storage (TLS) model of Clang/LLVM does not guarantee the normal execution
of TLS's symbol relocation due to the misaligned __tls_get_addr symbol in
case of the aggressive optimizations of Clang on Linux/ARM.
- Let's enable initial-exec TLS model instead of the dynamic TLS model.
Signed-off-by: Geunsik Lim <geunsik.lim@samsung.com>
CC: Jan Kotas <jkotas@microsoft.com>
CC: Jan Vorlicek <janvorli@microsoft.com>
geoffkizer [Thu, 11 Feb 2016 19:44:19 +0000 (11:44 -0800)]
Improve performance of ExecutionContextSwitcher to help async invocation perf
Rama Krishnan Raghupathy [Thu, 5 May 2016 22:45:22 +0000 (15:45 -0700)]
[Arm64]: Implement GetUnwindDataBlob
Bruce Forstall [Wed, 11 May 2016 00:33:09 +0000 (17:33 -0700)]
Move some LEGACY_BACKEND code from codegencommon.cpp to codegenlegacy.cpp
And, move some header file changes from codegen.h to codegenclassic.h.
The code moved were whole functions with no apparent tight coupling with
other code already in codegencommon.cpp.
The code moved had some #ifndef LEGACY_BACKEND portions, no longer
needed, removed.
Jan Vorlicek [Wed, 11 May 2016 00:15:45 +0000 (02:15 +0200)]
Change hosting API string encoding to UTF-8 on Windows (#4894)
This change fixes a problem with the string encoding in the hosting API
using CP_ACP, which causes problem when the input is a string in other
language than the current Windows one. In such case, it loses some
characters.
The fix is to use UTF-8 encoding (which is what CP_ACP means on Unix
PAL too).
Bruce Forstall [Tue, 10 May 2016 23:43:55 +0000 (16:43 -0700)]
Fix kill set of rep stos to include ECX
There was a very strange condition in the existing code to only
conditionally include ECX in the kill set. I convinced myself that
for AMD64, this condition always holds in this code path (which
makes a lot more sense anyway), so I removed the condition, and
also made all this code execute for x86 as well.
Fixes #4817
Sean Gillespie [Tue, 10 May 2016 19:03:16 +0000 (12:03 -0700)]
Enable the GCSimulator CI run to run on a regular basis
Jan Kotas [Tue, 10 May 2016 23:26:42 +0000 (16:26 -0700)]
Fix razzle build breaks
[tfs-changeset: 1603547]
Alex Ghiondea [Tue, 10 May 2016 21:53:35 +0000 (14:53 -0700)]
Build the reference assembly and the facade for mscorlib.
Make sure that the facade projects generates pdbs when possible.
Alex Ghiondea [Tue, 10 May 2016 19:46:55 +0000 (12:46 -0700)]
Create a Nuget package that will contain the reference assembly for mscorlib.
tijoytk [Tue, 10 May 2016 22:19:24 +0000 (15:19 -0700)]
Cleaning up a bunch of warnings.
Alex Ghiondea [Tue, 10 May 2016 18:21:17 +0000 (11:21 -0700)]
Fix the reference assembly for mscorlib to match the surface area of the OS on which it builds.
Bruce Forstall [Tue, 10 May 2016 21:51:44 +0000 (14:51 -0700)]
Merge pull request #4891 from BruceForstall/PartialFix4817
Block ops using rep on x86 need to kill edi/esi/ecx, same as x64
Andy Ayers [Tue, 10 May 2016 20:54:13 +0000 (13:54 -0700)]
Merge pull request #4878 from AndyAyersMS/AddMissingBenchmarkAttrs
Add missing xunit performance attributes
Pat Gavlin [Fri, 6 May 2016 00:12:20 +0000 (17:12 -0700)]
Disable FEATURE_MERGE_JIT_AND_ENGINE on *nix.
Now that #3058 has been fixed, the JIT can be consumed as a dynamic
library on all platforms.
Vance Morrison [Tue, 10 May 2016 19:32:13 +0000 (12:32 -0700)]
One for fix for non-windows platforms
Bruce Forstall [Tue, 10 May 2016 18:52:30 +0000 (11:52 -0700)]
Block ops using rep on x86 need to kill edi/esi/ecx, same as x64
AlexGhiondea [Tue, 10 May 2016 18:36:05 +0000 (11:36 -0700)]
Merge pull request #4833 from AlexGhiondea/StringBuilderFixes
Improve detection of error condition when using StringBuilder
Mike Danes [Mon, 9 May 2016 22:01:12 +0000 (01:01 +0300)]
Optimize integer div/mod by const power of 2 in lowering
Optimizing GT_DIV/GT_UDIV/GT_MOD/GT_UMOD by power of 2 in codegen is problematic because the xarch DIV instruction has special register requirements. By the time codegen decides to perform the optimization the rax and rdx registers have been already allocated by LSRA even though they're not always needed (as it happens in the case of unsigned division where CDQ isn't used).
Since the JIT can't represent a CDQ instruction in its IR an arithmetic shift (GT_RSH) has been instead to extract the dividend sign. xarch's SAR is larger than CDQ but it has the advantage that it doesn't require specific registers. Also, arithmetic shifts are available on architectures other than xarch.
Example: method "static int foo(int x) => x / 8;" is now compiled to
mov eax, ecx
mov edx, eax
sar edx, 31
and edx, 7
add edx, eax
mov eax, edx
sar eax, 3
instead of
mov eax, ecx
cdq
and edx, 7
add eax, edx
sar eax, 3
As a side-effect of this change the optimization now also works when the divisor is too large to be contained. Previously this wasn't possible because the divisor constant needed to be modified during codegen but the constant was already loaded into a register.
Example: method "static ulong foo(ulong x) => x /
4294967296;" is now compiled to
mov rax, rcx
shr rax, 32
whereas before a DIV instruction was used.
This change also fixes an issue in fgShouldUseMagicNumberDivide. The optimization that is done in lower can handle negative power of 2 divisors but fgShouldUseMagicNumberDivide handled those cases because it didn't check the absolute value of the divisor.
Example: method "static int foo(int x) => return x / -2;" is now compiled to
mov eax, ecx
mov edx, eax
shr edx, 31
add edx, eax
sar edx, 1
mov eax, edx
neg eax
instead of
mov eax, 0x7FFFFFFF
imul edx:eax, ecx
mov eax, edx
sub eax, ecx
mov edx, eax
shr edx, 31
add eax, edx
tijoytom [Tue, 10 May 2016 17:33:36 +0000 (10:33 -0700)]
Merge pull request #4793 from tijoytom/master
UTF8 Marshaling support(UnmanagedType.LPUTF8Str)
Vance Morrison [Tue, 10 May 2016 16:38:55 +0000 (09:38 -0700)]
Fix non-widows builds.
Hugh Bellamy [Tue, 10 May 2016 14:52:05 +0000 (15:52 +0100)]
Fix ArgumentNullException messages passed as parameter names (#2889)
Fixed several ArgumentNullExceptions that throw with the message as the parameter name, leading to potential confusion for developers.
Jonghyun Park [Tue, 10 May 2016 14:51:26 +0000 (23:51 +0900)]
Adjust relative offsets in UMThunkStub (#4888)
* Adjust relative offsets in UMThunkStub
This commit updates the relative offsets inside UMThunkStub according to the recent changes in
UMThunkStub (for stack unwinding).
* Uses 'UMThunkStub_StackArgsSize' instead of 'UMThunkStub_StackArgsOffset'
This commit revises 'UM2MThunk_WrapperHelper' to use 'UMThunkStub_StackArgsSize'
instead of 'UMThunkStub_StackArgsOffset'.
Jan Kotas [Tue, 10 May 2016 13:34:17 +0000 (06:34 -0700)]
Merge pull request #4713 from myungjoo/fix/4712
Match Inconsistent CMake/MSBuild Compiler Definitions 2
Jan Kotas [Tue, 10 May 2016 13:15:00 +0000 (06:15 -0700)]
Merge pull request #4870 from RussKeldorph/illicense
Update license header in most .il files.
Jan Kotas [Tue, 10 May 2016 13:10:55 +0000 (06:10 -0700)]
Merge pull request #4886 from dotnet-bot/from-tfs
Merge changes from TFS
Valery [Tue, 10 May 2016 12:57:47 +0000 (15:57 +0300)]
Fixed typo in jitinterface.cpp (#4889)
John Chen [Tue, 10 May 2016 06:20:49 +0000 (23:20 -0700)]
Merge pull request #4883 from JohnChen0/r2r
Fix CrossGen to skip some methods that require generic dictionary