Block ops using rep on x86 need to kill edi/esi/ecx, same as x64

Commit migrated from https://github.com/dotnet/coreclr/commit/3d3bebe396d1563d965093064829d63f66a885dd

Merge pull request dotnet/coreclr#4834 from gkhanna79/RemoveSxSJIT

Cleanup sxsJitStartup from codemanager

Commit migrated from https://github.com/dotnet/coreclr/commit/d3ff29a3e359b60c4c2fc0e21c1d6c99999157ea

Fix incremental build restore: move generated test_runtime project.json to bin (dotnet/coreclr#4764)

* Move generated test_runtime project.json to bin.

* Update helix tasks to point to moved test_runtime files.

* Remove unnecessary TestRuntimeDependenciesJson property.

Commit migrated from https://github.com/dotnet/coreclr/commit/388d58c3ff5e0792d6a4f7b7d00474f3f9b513df

Do NOT overwrite the stack frame pointer (dotnet/coreclr#4857)

CallEHFunclet currently overwrite the stack frame pointer (R7) while
recovering the registers.

This commit revises CallEHFunclet not to overwrite R7 when recoverint
the registers.

Commit migrated from https://github.com/dotnet/coreclr/commit/c55a8bd0b5949d5ed70c54a41aaab7d7b95fd8e5

Linux/ARM: fix stack alignment breaks (dotnet/coreclr#4858)

ehhelper.S: the function no more requires dummy stack add.
memcpy.S: C_FUNC(memcpy) has misaligned stack.
  (memcpy.S is not directly related with the issue dotnet/coreclr#4779.
  However, this requires the same fix with ehhelper.S)

Fix dotnet/coreclr#4779

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Commit migrated from https://github.com/dotnet/coreclr/commit/6acc1053342b308ab8f6e6f3c979d82a60641bdc

Merge pull request dotnet/coreclr#4845 from RussKeldorph/pending

Fix some tests where the only problem was incorrect retval

Commit migrated from https://github.com/dotnet/coreclr/commit/8ac5753f1bc07a0ee7daf6951d700261b6c23247

Update unix-test-instructions.md

Updated run coreclr tests instructions for Linux. Included the new semicolon separated paths for the coreFxBinDir.

Commit migrated from https://github.com/dotnet/coreclr/commit/6b92cff66f26c98bb3be88562626faaa54646118

Fix some tests where the only problem was incorrect retval

The following tests return values other than 100 on success.  Modify the
harness to expect the values that they return:
```
JIT\jit64\regress\vsw\102754\test1\test1.cmd
JIT\Regression\CLR-x86-JIT\V1-M09\b16102\b16102\b16102.cmd
```
The following test is identical to `test1` in the same directory.  Delete
it.
```
JIT\jit64\regress\vsw\102754\test2\test2.cmd
```
`rem_r4` returns a non-Unix-compatible 0xAAAA (>255).  Change it to the
standard 100.
```
JIT\IL_Conformance\Old\Conformance_Base\rem_r4\rem_r4.cmd
```
The following were using the non-functional Environment.ExitCode.
Switched to Environment.Exit(Environment.ExitCode):
```
JIT\Methodical\switch\switch1\switch1.cmd
JIT\Methodical\switch\switch10\switch10.cmd
JIT\Methodical\switch\switch11\switch11.cmd
JIT\Methodical\switch\switch2\switch2.cmd
JIT\Methodical\switch\switch3\switch3.cmd
JIT\Methodical\switch\switch4\switch4.cmd
JIT\Methodical\switch\switch5\switch5.cmd
JIT\Methodical\switch\switch6\switch6.cmd
JIT\Methodical\switch\switch7\switch7.cmd
JIT\Methodical\switch\switch8\switch8.cmd
JIT\Methodical\switch\switch9\switch9.cmd
JIT\Methodical\tailcall\_il_dbgrecurse_ep_void\_il_dbgrecurse_ep_void.cmd
JIT\Methodical\tailcall\_il_dbgtest_void\_il_dbgtest_void.cmd
JIT\Methodical\tailcall\_il_relrecurse_ep_void\_il_relrecurse_ep_void.cmd
JIT\Methodical\tailcall\_il_reltest_void\_il_reltest_void.cmd
JIT\Regression\CLR-x86-JIT\V1-M12-Beta2\b65423\b65423\b65423.cmd
JIT\Regression\CLR-x86-JIT\V1.2-M01\b08046\b08046\b08046.cmd
```
Note that `b08046.cmd` is left disabled because it is still failing.

Commit migrated from https://github.com/dotnet/coreclr/commit/fc4497b4679c4282b9e30bedf2c20060033cf437

Merge pull request dotnet/coreclr#4773 from kouvel/CardTableGrowFix3

Fix recently introduced timing issue when growing the card table

Commit migrated from https://github.com/dotnet/coreclr/commit/74798b5b95aca1b27050038202034448a523c9f9

ARM: Fix SOS stack dump command failure (dotnet/coreclr#4774)

Dumping address pointer for 32bit system needs adjusting for current implementation
while it assumes %p argument is 64bit. This patch changes (ULONG64) casting to SOS_PTR()
so that it will be safe for both 64 and 32bit.
This will fix DumpStack and EEStack segmentation fault problem.

Related issue: dotnet/coreclr#4672

Commit migrated from https://github.com/dotnet/coreclr/commit/70cfa355947ed82f7de8a56cac3bd38571cebfa5

Disable flaky test (dotnet/coreclr#4846)

Commit migrated from https://github.com/dotnet/coreclr/commit/e0df3ac5e4f6637f16160058eb92cdd935140fa0

RyuJIT - throughput improvements, PInvoke transitions tweaks: (dotnet/coreclr#4842)

* RyuJIT - throughput improvements, PInvoke transitions tweaks:

- Minor compilation throughput improvements:
     - Avoid redundant calls to getMethodAttribs in impMarkInlineCandidate by reusing the value from CALL_INFO
     - Avoid redundant calls to getFunctionEntryPoint in codegen by reusing the value computed during lower
     - Stop checking CORINFO_FLG_NATIVE flag for inliner observations. It is dead flag, not set anywhere
- Tweaks for helper-based PInvoke transitions (used CoreRT only today):
     - Stop emitting CORINFO_HELP_INIT_PINVOKE_FRAME. Just having the two simple begin/end helpers is better.
     - Stop passing size of arguments in secret register.
     - Enable direct calls for PInvoke targets

* Codereview feedback

Commit migrated from https://github.com/dotnet/coreclr/commit/0c4d3d92fe75ec011f2d2e95df2ceedd301de8ef

Merge pull request dotnet/coreclr#4747 from krytarowski/netbsd-support-75

NetBSD: Fix build with LLVM-3.9

Commit migrated from https://github.com/dotnet/coreclr/commit/c2c1abcb71065dd530e0b38045547603c38d1e86

Merge pull request dotnet/coreclr#4843 from dotnet-bot/from-tfs

Merge changes from TFS

Commit migrated from https://github.com/dotnet/coreclr/commit/1b9c70ca3d8fc86adc2bd9272a248db17394cebc

Delete left-over uses of FeatureHostedBinder

[tfs-changeset: 1602735]

Commit migrated from https://github.com/dotnet/coreclr/commit/4152ddcbcc8013858d3eca2b3f8cbbe3356cf224

Merge pull request dotnet/coreclr#4742 from lemmaa/fix-unit-test-script

Fix 'runtest.sh' script

Commit migrated from https://github.com/dotnet/coreclr/commit/59a9008ddbf7ec8b83d1950bf51ffa3b09c372b5

Merge pull request dotnet/coreclr#4836 from sivarv/simdPromotion

Mark simd type vars as lvRegStruct accurately.

Commit migrated from https://github.com/dotnet/coreclr/commit/ff7a57969edbb6da4983e823b690ca2c124e15d6

Fixes multidim array Get Accessor for enum datatype (dotnet/coreclr#4808)

Commit migrated from https://github.com/dotnet/coreclr/commit/417a98fb0f82b1c05010cdc2c41eb5056f14541c

Use Buffer.BlockCopy in System.Reflection.Emit (dotnet/coreclr#4812)

Commit migrated from https://github.com/dotnet/coreclr/commit/8eec3c5ac8975734836c333187a2724d50bd9af9

Fix a typo in NetBSD documentation. (dotnet/coreclr#4841)

Commit migrated from https://github.com/dotnet/coreclr/commit/8bee2889882af4826e1955e3a64a060ec9e41063

Buffer.BlockCopy: Avoid double-typechecks for arrays of same type (dotnet/coreclr#4807)

Commit migrated from https://github.com/dotnet/coreclr/commit/9a67ab7affc527168578a6749410c64ec3233f55

Remove FEATURE_HOSTED_BINDER definition (dotnet/coreclr#4838)

The FEATURE_HOSTED_BINDER is always on so remove it from all the sources.

Commit migrated from https://github.com/dotnet/coreclr/commit/29fcfe2938e3890cd6fb302d2f8c6dabc1a09fc4

Mark simd type vars as lvRegStruct accurately.

Commit migrated from https://github.com/dotnet/coreclr/commit/60d4c09a23dbc0e1ef5ccc348a05f6545bb6472d

Merge pull request dotnet/coreclr#4803 from pgavlin/RootBinDir

Add a `bindir` parameter to build.sh.

Commit migrated from https://github.com/dotnet/coreclr/commit/0ef1fc9aa3d65d41b5b9c8cd4911a1f8cdece049

Merge pull request dotnet/coreclr#4831 from pgavlin/gh2300

Move b12011 into testsUnsupportedOutsideWindows.

Commit migrated from https://github.com/dotnet/coreclr/commit/fbaa0827781be897a51b5d529cef305e569ca5cc

Merge pull request dotnet/coreclr#4826 from RussKeldorph/vararg

Move Windows-only vararg/arglist tests to the appropriate exclusion file

Commit migrated from https://github.com/dotnet/coreclr/commit/ef04d7d2083c8e6fb8f3cf6294f2ae06928dae10

Merge pull request dotnet/coreclr#4815 from AndyAyersMS/LegacyPolicyInlineLimit

Inliner: add JitInlineLimit check to LegacyPolicy under debug

Commit migrated from https://github.com/dotnet/coreclr/commit/5cdc97b5c80fe67f0ce06aced25b9d046e12a574

Merge pull request dotnet/coreclr#4770 from Maoni0/affi

enable affinity for GC threads and some bug fixes

Commit migrated from https://github.com/dotnet/coreclr/commit/726d1a3244b80bf963fd0d51e57d4bb90af1e426

Fix Windows x86 exception handling issue (dotnet/coreclr#4830)

This change fixes an exception handling issue that happens on x86 on Windows
when exception is raised in System.RuntimeType.MakeGenericType.
The problem was caused by GCPROTECT_HOLDER macro in RuntimeTypeHandle::GetTypeByName
that causes popping of GCFrame (and zeroing its m_next field) that's in the middle of the thread's frames list during
the stack unwinding.
That breaks the list and when UnwindFrames happen later and tries to walk the stack, the StackFrameIterator::NextRaw
asserts when checking the validity of the list.
The fix is to move the keepAlive to the managed caller of the RuntimeTypeHandle::GetTypeByName QCall, which removes
the need for the GCPROTECT_HOLDER.
Since it was the only usage of that holder and of the underlying FrameWithCookieHolder class, I've removed those.
In addition to that, I've modified COMModule::GetType and AssemblyNative::GetType to use the same pattern, since
they could also suffer from the problem the GCPROTECT_HOLDER was attempting to fix.

Commit migrated from https://github.com/dotnet/coreclr/commit/3c7e477ac50e50b616a64c72efb81388c045e63f

Cleanup sxsJitStartup from codemanager

Commit migrated from https://github.com/dotnet/coreclr/commit/c9cea125d28a9e465c37720a16427bcf88b03269

Merge pull request dotnet/coreclr#4690 from sejongoh/use_rax_for_ret_buf

Return RetBuf address via RAX for Windows 64

Commit migrated from https://github.com/dotnet/coreclr/commit/2308fbf2253abef97628bde2a04d8fb910ce1018

Merge pull request dotnet/coreclr#4795 from sivarv/simdret

[x64 Unix] Avoid SIMD types returned in two return registers from requiring to go through memory.

Commit migrated from https://github.com/dotnet/coreclr/commit/b3be35f379519120acd8427485524e92cf911918

Merge pull request dotnet/coreclr#4791 from kyulee1/fasttailcall

ARM64: Enable Fast Tail Call

Commit migrated from https://github.com/dotnet/coreclr/commit/06bc225c0117bbfd9ff31020e1f377ef33a561da

ARM64: Enable Fast Tail Call

Fixes https://github.com/dotnet/coreclr/issues/4420
Like x64, enable fast tail call but tail call opt is still off.
This means explicit tail call is converted to epilog + jmp IP0.
Tail call recursive/tail call via helper is still not enabled yet.

Commit migrated from https://github.com/dotnet/coreclr/commit/a9eee0c96fdb4eb8cc7c0312bb50bb923db177f4

Move b12011 into testsUnsupportedOutsideWindows.

This test requires COM interop.

Commit migrated from https://github.com/dotnet/coreclr/commit/4e0417947d47a5c632494353b943d44c33abdcc8

Add a `bindir` parameter to build.sh.

This parameter can be used to redirect the usual output path from
`./bin` to a different directory.

Commit migrated from https://github.com/dotnet/coreclr/commit/0e79232129f3a27305bc24141ff13dd54bc61e45

Merge pull request dotnet/coreclr#4814 from BruceForstall/FixGcRefAssert

Fix GC ref frame offset range assert

Commit migrated from https://github.com/dotnet/coreclr/commit/4c246c5b8182d284ded69c4a5bc48e09711f2438

Merge pull request dotnet/coreclr#4811 from BruceForstall/FixFunctionTrace

Fix COMPlus_JitFunctionTrace=1 to work better with NYI

Commit migrated from https://github.com/dotnet/coreclr/commit/f45cfa6dab8c5e40ebf613d98d1abf0a326727a5

Merge pull request dotnet/coreclr#4822 from dotnet-bot/from-tfs

Merge changes from TFS

Commit migrated from https://github.com/dotnet/coreclr/commit/17549875be841b032770e2e11ad027986e44a549

enable affinity and numa for GC and some bug fixes

double the threshold for balancing onto a remote NUMA node - this was proven to be best with the asp.net benchmark.

enable cpu group

Commit migrated from https://github.com/dotnet/coreclr/commit/73dfed727374982f0ce8804d82d9ca993a4b8bf1

Merge pull request dotnet/coreclr#4806 from AndyAyersMS/FixInlineDump

Fix inline dump format issue

Commit migrated from https://github.com/dotnet/coreclr/commit/1d561e7878288c109fab5183447d50b869fe8663

Move Windows-only vararg/arglist tests to the appropriate exclusion file

Our position on vararg/arglist is that it is supported on Windows only, so
I'm repurposing dotnet/coreclr#1440 to simply express that position in the exclusion
files rather than removing the tests.

Fixes dotnet/coreclr#1440

Commit migrated from https://github.com/dotnet/coreclr/commit/6a267e9e8ee18589fbcc73db3d19405126d9533b

Merge pull request dotnet/coreclr#4797 from JohnChen0/master

Remove an assert in AssemblySpec::EmitToken

Commit migrated from https://github.com/dotnet/coreclr/commit/9269fceccf0f134d312189f16d04037fede9d276

Merge pull request dotnet/coreclr#4810 from BruceForstall/Fix4666

Mark EDX as trashed by the no-gc helper calls

Commit migrated from https://github.com/dotnet/coreclr/commit/a6c18c6d3c499087ac43cdb3bced708164600425

Merge pull request dotnet/coreclr#4819 from BruceForstall/FixRyujitX86CheckedTestRun

Disable tests which fail in a RyuJIT/x86 Checked build test run

Commit migrated from https://github.com/dotnet/coreclr/commit/f4256993552b302c72df0afbc425ff5c05a9b208

Merge pull request dotnet/coreclr#4821 from ellismg/fix-unix-builds

Fix conditional in Unix build definition

Commit migrated from https://github.com/dotnet/coreclr/commit/bf61fa87d857381dfeeca258d1d366d62faf39be

Fix conditional in Unix build definition

Commit migrated from https://github.com/dotnet/coreclr/commit/cc284fb9c0ac4c2efbf1395efdbe8d8e0978dc92

Disable tests which fail in a RyuJIT/x86 Checked build test run

Commit migrated from https://github.com/dotnet/coreclr/commit/0c29217d7f0de620248c30d2db92ce29af6858ef

Fix COMPlus_JitFunctionTrace=1 to work better with NYI

Commit migrated from https://github.com/dotnet/coreclr/commit/4eb10dbf742604906019931ac4b1540b451f7322

Fix GC ref frame offset range assert

In emitSetFrameRangeGCRs(), on x86, there is the following assert:
  assert(offsHi <  0);
This is saying that the highest offset of a stack frame gc reference
is negative w.r.t. the frame pointer. This is not quite true: the
highest offset is exclusive of the range (the lowest offset is inclusive).
I found a test case where the 'this' pointer, a gc ref, was located
immediately below the frame pointer, in range [ebp-4,ebp), so offsHi==0.

The reason we don't hit this more is because this assert is only in
the JitDump code (under #ifdef DEBUG and 'if (verbose)'). (This
also seems wrong.)

I changed the assert to (offsHi <= 0).

Commit migrated from https://github.com/dotnet/coreclr/commit/16ca4a20ddd03479e101cea0d74705f1a2791ae5

This change ports the support files from NetFxDev1. It doesn't change behavior yet. Adding a feature for the app context switch for path compat as we don't intend to use it in core.

[tfs-changeset: 1602284]

Commit migrated from https://github.com/dotnet/coreclr/commit/2ae90a015933dc85192b73c2bf8361cf59cef035

Inliner: add JitInlineLimit check to LegacyPolicy under debug

Adds the ability to limit the number of inlines done by the LegacyPolicy.
Useful in trying to binary search for inlines that cause correctness
issues.

This limit impacts inlining in all methods, so effective isolation may
also require use of JitNoInlineRange to hone in on the root method that
is the source of problems.

Commit migrated from https://github.com/dotnet/coreclr/commit/0feeb4ff48a6f0beb614ee2e3d3604e85dfbca27

Mark EDX as trashed by the no-gc helper calls

Fixes dotnet/coreclr#4666

Commit migrated from https://github.com/dotnet/coreclr/commit/f6796220664d3bfa4ea42d2ab86cf5a6e2edeae9

Fix inline dump format issue

Lost some implict casts to float when I updated SIZE_SCALE to an enum
constant, so now make the casts explicit.

Just impacts dump output.

Also, add a newline to LegacyPolicy dump to keep messages distinct.

Commit migrated from https://github.com/dotnet/coreclr/commit/9614277d4fb85970871678601c9ffe0c4cbc6a97

Return RetBuf address via RAX for Windows 64

The AMD64 native calling convention requires return buffer address to be
returned by callee in RAX. This changeset makes JIT follow this rule.
Especially, in case that a managed method returns a struct value to its
native caller, the caller reads a pointer to the return value from RAX.

Asm diffs show expected differences such as 'mov rax, reg' at the end of
functions and some different register assignments.

Commit migrated from https://github.com/dotnet/coreclr/commit/9b54e5ebee7838b3eb9c7cacbd9f66ac78dfb968

Merge pull request dotnet/coreclr#4800 from jashook/x86_ci_fix

Fix x86 netci changes to build correctly.

Commit migrated from https://github.com/dotnet/coreclr/commit/3ddd9c43821bf617621f87d8d1519229e7d3b789

Avoid SIMD type return values from requiring to go through
memory for returning.

Commit migrated from https://github.com/dotnet/coreclr/commit/75a96376d9b6760b63eefbc245572e9274eeb347

Fixup netci changes to build correctly.

Also more verbosely show which jit and arch the job is.

Commit migrated from https://github.com/dotnet/coreclr/commit/5aca558718d3d57686bc301315034e76324d242c

Merge pull request dotnet/coreclr#4782 from BruceForstall/NyiLongExTree

RyuJIT/x86: convert SBCG to NYI for multi-level expression trees

Commit migrated from https://github.com/dotnet/coreclr/commit/3496df6d9dbfa5fded802ab90b99a9545a43bc74

Merge pull request dotnet/coreclr#4792 from dagood/remove-core_root-workaround

Remove CORE_ROOT workaround

Commit migrated from https://github.com/dotnet/coreclr/commit/b0bbac3d22969796e0c424c22242f255c1cbe044

Following the review comments

Commit migrated from https://github.com/dotnet/coreclr/commit/9290b53960f368ab9b6356ad029cb3cab69a12c6

Remove an assert in AssemblySpec::EmitToken

AssemblySpec::EmitToken has a PRECONDITION that prohibits emitting AssemblyRef
tokens to WinMD files, except inside CrossGen. However, there are scenarios,
such as XML serialization, that need to emit such tokens. Thus this PRECONDITION
is removed.

Commit migrated from https://github.com/dotnet/coreclr/commit/c377b53dca090869ea10d95fe7098e70ac3c0847

Merge pull request dotnet/coreclr#4790 from jashook/x86_netci_two

Add ryujit x86 pr trigger.

Commit migrated from https://github.com/dotnet/coreclr/commit/22738816cf92bd763c3e993c8acdef0640f9428c

Merge pull request dotnet/coreclr#4767 from ramarag/VSD_ResolveWorker_indirection

[Arm64] Initialize indirection cell for delegates

Commit migrated from https://github.com/dotnet/coreclr/commit/f5c775dd83954f2e815e8c1b3adc8fc2cb7bfefb

[Arm64] Initialize indirection cell for delegates

Commit migrated from https://github.com/dotnet/coreclr/commit/08b010a6cbd7803d178dd42da72d0aab072da9dc

Merge pull request dotnet/coreclr#4765 from adityamandaleeka/array_alloc_bounds_check

Add fwrapv compile option

Commit migrated from https://github.com/dotnet/coreclr/commit/dc71a169507810c1dafad6fa20dc3881c6e4d4ec

Merge pull request dotnet/coreclr#4780 from mkborg/fix_windows_linuxmscorlib_arm_build

Fix Windows 'linuxmscorlib arm' build

Commit migrated from https://github.com/dotnet/coreclr/commit/b889f007480b6818a5cb144e726d3e2940fd5cd3

Remove O_RDWR from sem_open and better error codes on failure (dotnet/coreclr#4772)

Remove O_RDWR from sem_open and better error codes on failure

Commit migrated from https://github.com/dotnet/coreclr/commit/98577b986fbb79b74088a5ab3804a3ff00efbe7f

Merge pull request dotnet/coreclr#4763 from AndyAyersMS/InlineIntoControlExamples

Add some tests for optimization of booleans

Commit migrated from https://github.com/dotnet/coreclr/commit/7995c7db6626d77a2deac664a1557157c89271fc

RyuJIT/x86: convert SBCG to NYI for multi-level expression trees

In general, we decompose longs by walking the tree bottom-up. So, e.g.,
we'll decompose 'add' as follows:

  + (gt_long(lo1, hi1), gt_long(lo2, hi2))
=>
  gt_long (+ (lo1, lo2), +Hi (hi1, hi2))

But what if lo1 is '+' and hi1 is '+Hi'? So,

  + (gt_long( + (lo3, lo4), +Hi (hi3, hi4) ), gt_long(lo2, hi2))
=>
  gt_long (+ ( + (lo3, lo4), lo2), +Hi ( +Hi (hi3, hi4), hi2))

But we can't put any instructions between '+ (lo3, lo4)' and '+Hi (hi3, hi4)'
or we'll lose the carry bit.

We need to model this differently, or introduce temps. Until this is
done, convert the silent bad codegen into NYI fallback.

Commit migrated from https://github.com/dotnet/coreclr/commit/d1d7a46db4d75f9ef323fc4ae5b4a1efc661476b

Merge pull request dotnet/coreclr#4504 from krytarowski/netbsd-support-72

Add initial NetBSD documentation

Commit migrated from https://github.com/dotnet/coreclr/commit/a12aed9422083394022752f282e7656d40206acd

Merge pull request dotnet/coreclr#4755 from sivarv/mcall

Multi-reg call node support for x64 Unix

Commit migrated from https://github.com/dotnet/coreclr/commit/7730864b3a7f0c0233c23c676e1a83e50cbd7116

Remove CORE_ROOT workaround. .NET CLI no longer uses the value.

Commit migrated from https://github.com/dotnet/coreclr/commit/eeece8acff8f0d724c82552ddc064d07400114c8

Add ryujit x86 pr trigger.

Adds ryujit x86 testing on pr, and collapses x86 lb to just run checked jobs.

Commit migrated from https://github.com/dotnet/coreclr/commit/53331c48fbff444568a67e73094d171f75e24b84

Merge pull request dotnet/coreclr#4723 from kyulee1/tailcall

ARM64: Enable Tail Call with Vararg

Commit migrated from https://github.com/dotnet/coreclr/commit/af1961073ba02f76c58a86f5528066e5813c2bb1

Merge pull request dotnet/coreclr#4727 from adiaaida/overflowLongArithmetic

Set overflow flag for add/sub hi on x86 TYP_LONG

Commit migrated from https://github.com/dotnet/coreclr/commit/777936dd947986ae54119ad6fccc1711f52f1daf

Multi-reg call node support.

Commit migrated from https://github.com/dotnet/coreclr/commit/a88ba565faa592e5aeb59df22640961e81d8d037

Merge pull request dotnet/coreclr#4757 from swgillespie/gcstress-disable-tests

Disable two tests under GCStress for dotnet/coreclr#4748 and dotnet/coreclr#4708

Commit migrated from https://github.com/dotnet/coreclr/commit/35c4b5bd36ac9e3e36dab3f8674a3e83b388aaf3

Merge pull request dotnet/coreclr#4760 from dotnet-bot/from-tfs

Merge changes from TFS

Commit migrated from https://github.com/dotnet/coreclr/commit/4047cc5464c287cacf981fdc4c2a2996ed4e6e86

ARM64: Enable Tail Call with Vararg

Fixes https://github.com/dotnet/coreclr/issues/4475
I've run into `IMPL_LIMITATION("varags + CEE_JMP doesn't work yet")` in
importer.cpp. This change enables ARM64 tail call path same as other
targets.
 1. Similar to amd64 `genFnEpilog`, I made the similar code under
 `!FEATURE_FASTTAILCALL`. Since `EC_FUNC_TOKEN_INDIR` is not defined for
 arm64, I've made NYI for such case.
 2. Added two pseudo branch instructions 'b_tail' and 'br_tail' which form
 jmp instruction encodings but follow call instruction semantics since
 they are used for tail-call.
 3. `GenJmpMethod` is enabled. Code is slightly changed to reflect correct
 float argument handlings and multi-reg support.

Commit migrated from https://github.com/dotnet/coreclr/commit/f63d726b01f42fbbceb497e5de6f36a8622f4000

Fix Windows 'linuxmscorlib arm' build

Commit migrated from https://github.com/dotnet/coreclr/commit/9b1f7b6e281c9164faa4280e7a8e5b3601b62efd

Merge pull request dotnet/coreclr#4759 from rahku/crossgen

Fix checking of correct errorlevel for crossgen

Commit migrated from https://github.com/dotnet/coreclr/commit/09d6a5d6536adc498917728d3224b1ee7f779d65

Merge pull request dotnet/coreclr#4771 from myungjoo/fix/4746

ildasm: incorrect signess usage with char

Commit migrated from https://github.com/dotnet/coreclr/commit/7eb8accc6846757925ab6f7896d76a204f9fa509

Unix/Build: use -fsigned-char

Some architectures including ARM assumes that plain char is
unsigned unlike x64.

Because CoreCLR has been assuming char is signed, we need to
explictly tell the toolchain the signedness of char type.

Fix dotnet/coreclr#4746

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Commit migrated from https://github.com/dotnet/coreclr/commit/7f203ae2ca9f7c14239b4e7228b8f9638b3388f3

Fix recently introduced timing issue when growing the card table

Observed issue:
- When software write watch is enabled, the runtime is suspended in grow_brick_card_tables() to copy the dirty state of pages into the new table
- The global card table pointer was updated before the attempt to suspend the runtime
- That thread gets blocked there, while the background GC thread successfully suspends the runtime, and calls copy_brick_card_table()
- Since the lowest and highest addresses of the GC heap were not updated, copy_brick_card_table() reads invalid memory

Fixes:
- Moved the change to g_card_table to after the point of suspend for software write watch. At the point of this suspend, the global state will be that prior to resize.
- Though much less frequent, this issue can also occur with the implicit suspend that occurs as part of StompWriteBarrierResize(), when the write barrier type needs to be changed. Moved the other StompWriteBarrierResize() down such that it is done after all global state is updated. At the point of this suspend, the global state will be that following resize.
- I don't think we can have a failure path (failing to commit the mark array) after updating global state such as g_card_table. Soon after that, there may be changes to the new table on paths that don't go through the write barrier, and upon failure to commit the mark array, those updates would be lost. Passed the new card table to functions that need them, and moved changes to global table pointers to after all of the failure points.

Fixes dotnet/coreclr#4750

Commit migrated from https://github.com/dotnet/coreclr/commit/52e149e3bac644087463f418a0516f02a6be0060

Merge pull request dotnet/coreclr#4478 from pgavlin/JitPackaging

Create JIT packages during the build.

Commit migrated from https://github.com/dotnet/coreclr/commit/59bc1154459251a5724f4263c582180894dcd2ab

Merge pull request dotnet/coreclr#4768 from adityamandaleeka/fix_ci_corefxbins

Add quotes around coreFxBinDir example in runtest.sh

Commit migrated from https://github.com/dotnet/coreclr/commit/e88dca4eeaf7c2eedb7027118ad44ba3cf2e23e5

Update per review feedback.

Commit migrated from https://github.com/dotnet/coreclr/commit/2121c05ffad2a686679a3d4c685bf5ce690f3c56

Merge pull request dotnet/coreclr#4740 from yizhang82/interop_x86_test

Fix interop x86 tests by fixing mismatching calling convention, and enable [NativeCallable] for every architecture

Commit migrated from https://github.com/dotnet/coreclr/commit/62845a015ec8e06e75229d9f5fbc786da0705f71

Add quotes around coreFxBinDir example.

Commit migrated from https://github.com/dotnet/coreclr/commit/fef43d61b50d6a7d2e365a3234eff1971f6224d2

Merge pull request dotnet/coreclr#4753 from AndyAyersMS/InlineForSize

Inliner: initial version of SizePolicy

Commit migrated from https://github.com/dotnet/coreclr/commit/d474c289096d7a2af96ab5690be37c9fa8aec10e

Add 'fwrapv' compile option.

Commit migrated from https://github.com/dotnet/coreclr/commit/e67df67213d1b4730d5d2c803c681bc55f009792

Add some tests for optimization of booleans

Test case to stress inlining, expression opts, and control flow
simplification for booleans.

Test case has 100 methods named Idxx. All 100 should generate identical
code. On x64 windows we expect to get the 4 byte sequence
```
       0FB6C1               movzx    rax, cl
       C3                   ret
```

Only 22 of the variants get this codegen; there are at least 12 other
sequences ranging in size from 9 to 32 bytes of code.

Likely touches on the same issues raised in dotnet/coreclr#914.

Commit migrated from https://github.com/dotnet/coreclr/commit/8354d2b9b4c2b5dd6f3b111a93f05322007bb239

Add initial NetBSD documentation

Thanks for @coypoop for improving the English wording.

Commit migrated from https://github.com/dotnet/coreclr/commit/229be2965ca36601980b5fcb398a76e5f6ca89ab

Merge pull request dotnet/coreclr#4705 from RussKeldorph/vector3interop

Disable Vector3Interop until failures are addressed

Commit migrated from https://github.com/dotnet/coreclr/commit/db6f6d7a04b15a376a40d434fea91e07b9b6f13f

Revert "Fix array allocation bounds checking."

This reverts commit dotnet/coreclr@16c9dfbae5e5ce18ad0519b31df515820fed2579.

Commit migrated from https://github.com/dotnet/coreclr/commit/835702ea450734c4ce1ac819ccafd99f5a790da8

ARM: fix stack frame management (dotnet/coreclr#4641)

ARM: fix stack frame management

Fix dotnet/coreclr#4638
This commit revises the implementation of CallEHFunclet to establish
stack frame appropriately.
This commit allows libunwind to work correctly for the frame created by
CallEHFunclet.

Commit migrated from https://github.com/dotnet/coreclr/commit/57185b4668e5b64855840655806c0d12524dbfb4

Merge pull request dotnet/coreclr#4693 from adityamandaleeka/delegateinvoke_platnotsup

Make Delegate BeginInvoke/EndInvoke throw PlatformNotSupportedException

Commit migrated from https://github.com/dotnet/coreclr/commit/2828f00b908866a08f0ce32a895bf93623b0e0e8

Add test for Begin/EndInvoke throwing PlatformNotSupportedException.

Commit migrated from https://github.com/dotnet/coreclr/commit/8590bdf2146b86c179fd4cbe78f721d4f0b581c7

Fix checking of correct errorlevel for crossgen

Commit migrated from https://github.com/dotnet/coreclr/commit/463322e4e0700d8d839acc4ceed36425bafa65c3