- 01 6月, 2022 16 次提交
-
-
由 Aaron Robinson 提交于
* Remove wprintf from eventpipe. * Remove messagebox logic from utils and place its single usage under debug/. * Convert Profiler GUID logging to use UTF-8
-
由 Buyaa Namnan 提交于
Co-authored-by: NJan Kotas <jkotas@microsoft.com>
-
由 Tanner Gooding 提交于
* Adding support for vector constants via GenTreeVecCon * Responding to PR feedback * Support tracking the underlying simdBaseJitType for GenTreeVecCon * Applying formatting patch * Handle the scenario where TYP_SIMD8 is retyped to TYP_LONG or TYP_DOUBLE * Ensure GenTreeVecCon tracks the simdSize so that TYP_SIMD8 ABI retyping still works * Ensure the create call and not the cloned local is lowered * Resolving PR feedback * Ensure morph and rationalize check `OBJ(ADDR(CNS_VEC))` * Update src/coreclr/jit/gentree.cpp Co-authored-by: NSingleAccretion <62474226+SingleAccretion@users.noreply.github.com> Co-authored-by: NSingleAccretion <62474226+SingleAccretion@users.noreply.github.com>
-
由 Vladimir Sadov 提交于
-
由 Huo Yaoyuan 提交于
* Convert CompareUnmanagedFunctionPtrs to managed * Move simple type match to managed * Add more check in managed portion * Rename method for FCall * Change FCall to directly call IsEquivalentTo * Add helper method frame * Fix uninitialized result variable * Change to QCall for type equivalence
-
由 Sergio Pedri 提交于
* Add managed fast path for SZ Array.GetValue(int) This is only used when the element type is a reference type * Move faster GetValue(int) to CoreCLR-specific file * Add Array.GetFlattenedIndex(int) overload * Add managed stub for InternalGetValue in CoreCLR This is used whenever the element type is a class * Initial (partial) port of ArrayNative::GetValue to C# * Simplify TypeDesc checks in managed Array.GetValue * Remove Array.InternalGetValue CoreCLR FCall * Add RuntimeHelpers.Box FCall on CoreCLR * Implement value type support for managed Array.InternalGetValue * Remove leftover MethodTable.IsElementTypeClass property * Minor API tweaks, add XML docs * Fix some typos and oversights * Remove managed TypeDesc type (not needed anymore) * Revert name change for MethodTable.ElementType field * Simplify Array.GetFlattenedIndex (CoreCLR/NativeAOT) * Apply suggestions from code review * Move FCIMPL to comutilnative.cpp * Remove ExplicitLayout from TypeHandle * Tweak comment on unsupported path for Array.GetValue * Use MemoryMarshal.GetArrayDataReference instead of manual addressing * Add HELPER_METHOD_FRAME to FCall implementation * Move MethodTable FCall decl to comutilnative.h * Simplify legacy behavior check in ArrayNative::SetValue * Change RuntimeHelpers.Box to just call JIT_Box * Simplify GetArrayElementTypeHandle(), add an assert
-
由 Jakob Botsch Nielsen 提交于
* Split the IR validation after tailcalls into a separate function. Previously it was intertwined with updating of profile weights for follow-up blocks. * Generalize the validation to use a tree walk and handle more cases. This fixes an assertion failure seen in some PGO runs. * Generalize the updating of profile weights for follow-up blocks. Previously this was only updating profile weights for one follow up blocks, but there can be an arbitrary number of successor blocks due to inlining. Fix #69939
-
由 Bruce Forstall 提交于
Change it to take a dst, not a writeable dst, pointer.
-
由 Dan Moseley 提交于
* Improve tests for offset start * improve * comment
-
由 Carlos Sanchez 提交于
-
由 Tom Deseyn 提交于
Co-authored-by: NKevin Jones <vcsjones@github.com>
-
由 Matt Galbraith 提交于
-
由 Fan Yang 提交于
* Add a new LLVMArgStorage type * Fix llvm signature * Fix emit_call * Enable passing SIMD register on arm64 and clean up debugging stuff * Clean up unnecessary format change * Update code after rebase * Only use SIMD register when element type is primitive type * Disable SIMD intrinsics for generic sharing, since it doesn't work at all. * Check if element type is primitive on arm64 * Not using SIMD register when it is full aot * Address review feedbacks
-
由 Radek Zikmund 提交于
-
由 Vlad Brezae 提交于
* [mono][mini] Throw exception if we have conflicts with static virtual interface call Small refactoring to avoid duplicated code * [mono][interp] Throw exception if we have conflicts with static virtual interface call * [mono][mini] Fix method access check during ldftn From IL, it is possible to change the access modifier when overriding a method. In such case we need to check the accessibility of the IL method and not the one that we resolve from the constrained class.
-
由 Johan Lorensson 提交于
* Fixes all 4244 warnings on Mono runtime x86/x64 Windows build, aligning with SDL requirements (#66154). There will be additional warnings in the cross compilers that needs to be fixed separately, but this should fix the major bulk of 4244 warnings. All changes done in this PR are mitigating compiler issued data type truncation 4244 warnings: 'conversion' conversion from 'type1' to 'type2', possible loss of data and PR will align to the same type as the compiler issued in the warning. PR introduce a new set of macros to cast between types following patter of existing GPOINTER_TO_INT, GPOINTER_TO_UINT, GTYPE1_TO_TYPE2. The idea of use these macros is that it will be clear where casts are done, and we will have ability to intercept and add validation logic in specialized builds to catch truncation errors. The PR also introduce the needed inline functions under ENABLE_CHECKED_CASTS that could be used for extended validation when using the macros. Failure actions in these inline functions are not currently fully implemented, since that will be different from case to case, but can be added when needed since all infrastructure is prepared. If it is possible to change types used in code to mitigate the warnings, that will be the initial strategy used by this PR, if that however would cause ripple effects through the source code (unfortunately quite common), we will fallback using the GTYPE1_TO_TYPE2 macros to clearly show the intent, keep trackability of the change as well as adding ability to do additional validation of casts in the future.
-
- 31 5月, 2022 18 次提交
-
-
由 Simon Rozsival 提交于
[Android] Add UnmanagedCallersOnly attribute to SafeDeleteSslContext.ReadFromConnection/WriteToConnection methods (#69507)
-
由 Fan Yang 提交于
-
由 Theodore Tsirpanis 提交于
-
-
由 Michal Strehovský 提交于
Fixes issue reported in https://github.com/dotnet/runtime/issues/66716#issuecomment-1069310326. We create a hashset for the purpose of avoiding this problem, check it, but never add to it.
-
由 Michal Strehovský 提交于
This file was added for universal shared code. We can restore it again if it's needed, but we don't need it now. On NativeAOT we could leverage the CCC for reflection invoke, but this code was never updated to work outside Windows. It's safer to delete it.
-
由 Lakshan Fernando 提交于
* CI fix for native aot tests * parse notrait commandline args * try a different platform * FB * Remove unnecessary whitespace change These make `git blame` harder to follow. Plus the consistent thing in this file is no extra newline here. Co-authored-by: NMichal Strehovský <MichalStrehovsky@users.noreply.github.com>
-
由 Steve Dunn 提交于
* Change wording of exception * Update src/libraries/System.Text.Json/src/Resources/Strings.resx Co-authored-by: NEirik Tsarpalis <eirik.tsarpalis@gmail.com> * Update wording. Revert whitespace changes from before. * Really fix whitespace diffs Co-authored-by: NEirik Tsarpalis <eirik.tsarpalis@gmail.com>
-
由 Jakob Botsch Nielsen 提交于
Previously, for struct args, we would use the class handle of the argument (from the IL stack) instead of the class handle from the signature. These may sometimes differ in __Canon when inlining shared code. This change switches the class handle stored in CallArg to always be the type reported back by the EE when asking about the signature.
-
由 Jackson Schuster 提交于
-
由 dotnet-maestro[bot] 提交于
* Update dependencies from https://github.com/dotnet/hotreload-utils build 20220516.1 Microsoft.DotNet.HotReload.Utils.Generator.BuildTool From Version 1.1.0-alpha.0.22259.2 -> To Version 1.1.0-alpha.0.22266.1 * Update dependencies from https://github.com/dotnet/hotreload-utils build 20220523.2 Microsoft.DotNet.HotReload.Utils.Generator.BuildTool From Version 1.1.0-alpha.0.22259.2 -> To Version 1.1.0-alpha.0.22273.2 Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com> Co-authored-by: NLarry Ewing <lewing@microsoft.com> Co-authored-by: NAleksey Kliger (λgeek) <alklig@microsoft.com>
-
由 Lakshan Fernando 提交于
When a null ref return method is invoked. Fixes #69755
-
由 Michal Strehovský 提交于
We should have Roslyn that is recent enough to compile this now.
-
由 Vlad Brezae 提交于
When unwinding the stack during EH, we first populate a StackFrameInfo with various data, like native_offset and jinfo. Let's assume this data is for an untiered method. If this frame has multiple finally clauses that need to be invoked, a finally invocation could tier up the InterpFrame that it is executing. The future finally invocations would use the IPs from the previously resolved untiered method. This change makes the interpreter receive only the clause index (in interp_run_finally) so we correctly resolve the native offsets of the handler.
-
由 Andy Ayers 提交于
The JIT currently will aggressively reorder the flow graph before running its loop recognition phases. When there is PGO data this sometimes perturbs the block order so that loops are no longer recognized, and we miss out on some loop optimizations. This change defers most block reordering until after the JIT has gone through the optimization phases. There is still a limited form of flow cleanup done early on. There is also a compensating change in loop recognition in one place where it was relying on adjacent blocks being combined. Fixes #67318.
-
由 Qiao Pengcheng 提交于
-
由 Qiao Pengcheng 提交于
* [LoongArch64] Fix the segment fault within JIT_UpdateWriteBarrierState. * [LoongArch64] update with the upstream.
-
由 Egor Bogatov 提交于
-
- 30 5月, 2022 6 次提交
-
-
由 dotnet bot 提交于
-
由 Eirik Tsarpalis 提交于
-
由 Marek Fišera 提交于
-
由 SingleAccretion 提交于
* Do not depend on class handle in "fgMorphArgs" Use the signature's handle when making an outgoing arg copy. * Add a test
-
由 SingleAccretion 提交于
Seems to have been a workaround for an issue long gone.
-
由 Peter Sollich 提交于
Fix setting of DacpHeapSegmentData::highAllocMark on DAC side. Keep some logic for backward compat in dacprivate.h, the copy here and the one in the Diagnostics repo, so that a new SOS can work correctly with an older DAC. The PR in the diagnostics repo is this: dotnet/diagnostics#3062.
-