- 10 3月, 2023 26 次提交
-
-
由 Vitek Karas 提交于
The `RequiresAssemblyFiles` attribute can be present on property or event (in which case it applies to all accessors). The analyzer already handles this case, but the NativeAOT compiler didn't. This change implements the correct behavior in NativeAOT as well. The implementation actually handles all Requires attributes this way now, but the other two are not allowed on properties or event (yet). Had to add an optimization to the `GetPropertyForAccessor`/`GetEventForAccessor` as these are now called very frequently and they can be really slow. Implements a new behavior with regard to IL3003 as described in the comment in the code as well as: #83235 Fixes https://github.com/dotnet/runtime/issues/71985.
-
由 Johan Lorensson 提交于
* Implement native container types used by shared EventPipe code. Use eglib's list, slist, array, ptr_array, queue, hash table as a starting point creating a new set of native container types shared between native runtime components like EventPipe. EventPipe shim container API used a STL like API, new container types follows STL style container API and match STL type names: vector forward list list queue unordered map Native container types are still included in runtime build artifacts (through EventPipe cmake files), this can be changed/done differently depending on how each runtime would like to build/use the source files. In order to make that intent more clear, container and EventPipe cmake files have been renamed to not use CMakeList.txt, but .cmake. A number of new native tests for added container types have been added into EventPipe native test runner.
-
由 Michal Strehovský 提交于
-
由 Miha Zupan 提交于
* Vectorize ProbabilisticMap.IndexOfAny on AVX2 * Use ResetLowestSetBit from BitOperations * Speed up Avx2 and add Vector128 support * Add Vector{128/256}.LoadUnsafe(ref char) and Vector128.ShuffleUnsafe * Use Vector128.ShuffleUnsafe in more places * PR feedback * Replace another ShiftRightLogical with '>>>' * Add WASM path to Vector128.ShuffleUnsafe * PR feedback
-
由 Steve Pfister 提交于
Co-authored-by: NSteve Pfister <steve.pfister@microsoft.com>
-
由 Larry Ewing 提交于
* Update dependencies from https://github.com/dotnet/llvm-project build 20230308.1 runtime.linux-arm64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.linux-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.linux-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-musl-arm64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-musl-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-musl-x64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-musl-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-x64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.linux-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.linux-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.osx-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.osx-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.osx-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.osx-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.win-arm64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.win-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.win-x64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.win-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.win-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.win-x64.Microsoft.NETCore.Runtime.ObjWriter From Version 14.0.0-alpha.1.23156.3 -> To Version 14.0.0-alpha.1.23158.1 * Update dependencies from https://github.com/dotnet/emsdk build 20230308.1 Microsoft.NET.Workload.Emscripten.Current.Manifest-8.0.100-preview.3 , Microsoft.NET.Workload.Emscripten.net6.Manifest-8.0.100-preview.3 , Microsoft.NET.Workload.Emscripten.net7.Manifest-8.0.100-preview.3 From Version 8.0.0-preview.3.23156.1 -> To Version 8.0.0-preview.3.23158.1 * try flowing a stable name * fix typo * Update since _ doesn't get removed in the package name * Update to new name --------- Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
-
https://github.com/dotnet/roslyn-analyzers由 dotnet-maestro[bot] 提交于
Microsoft.CodeAnalysis.Analyzers , Microsoft.CodeAnalysis.NetAnalyzers From Version 3.3.5-beta1.23124.1 -> To Version 3.3.5-beta1.23158.3 Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
-
https://github.com/dotnet/roslyn由 dotnet-maestro[bot] 提交于
Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset From Version 4.6.0-3.23157.14 -> To Version 4.6.0-3.23159.1 Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
-
由 Will Smith 提交于
* Do not emit overflow check for div if we know first operand is a constant that is a not a min value * Remove redundant branch * Update codegenarm64.cpp * Try to use subs, zr, dividendReg, 1 to determine overflow in division * Cleanup * Use IsNeverNegative to help not emit overflow checks * Fix check * Added helper functions * Use main node instead of op1 for type check * Add 64-bit check * Use cmp instead of subs since it is an alias * Handle 32-bit arch division for LONG. * Additional checks * More cleanup * More cleanup * Do not find exception blocks * Try using flags * Try using flags * More checks, more comments * Update src/coreclr/jit/gentree.cpp Co-authored-by: NSingleAccretion <62474226+SingleAccretion@users.noreply.github.com> * Flip meaning of check flags. Rename OperExceptions to GetExceptionSetFlags. Use GetExceptionSetFlags to determine to add div-by-zero or overflow blocks for arm64. * Use GetExceptionSetFlags in arm64 codegen * Revert name * Use GTF_DIV_MOD_NO_OVERFLOW_CHK in CanDivOrModPossiblyOverflow * Fix build * Rename flags by removing '_CHK' * Remove break --------- Co-authored-by: NSingleAccretion <62474226+SingleAccretion@users.noreply.github.com>
-
由 Carlos Sánchez López 提交于
* Add tar tests to confirm entries owned by deleted users can still be read and written. * Improve Process execution code to avoid potential deadlock.
-
由 Jakob Botsch Nielsen 提交于
Avoid creating 0-sized block ops in the importer (but keep the side effects of the addresses). Add asserts that we don't create these nodes and get rid of some downstream checks for the case. Fix #12807 Also merge CEE_INITBLK and CEE_CPBLK importer cases. As part of this fix a bug where we weren't setting GTF_EXCEPT on the destination in the CEE_INITBLK case.
-
由 Katelyn Gadd 提交于
The jiterpreter was failing to add unconditional branch targets to the branch target table, which could cause branches targeting them to fail and make the trace bail out.
-
由 Larry Ewing 提交于
* Update the workload description * Update the path * Update the description for net6.0 * Add more descriptions * Add descriptions for the windows runtime packs * Update WorkloadManifest.en.json fix review comment
-
由 Pavel Savara 提交于
* cold start benchmark with download throttling and latency * make throttling shared on the same session * sequential * removed PageShowCold
-
由 Andy Ayers 提交于
Once synthesis arrives on the scene, we're not going to want phases in the JIT to arbitrarily modifying block weights. There is already a guard of this sort for regular profile data, so it makes sense to extend that to synthesized data as well. When synthesizing counts, propagate counts to finallies from the associated trys. This needs to be done carefully as we have make sure not to visit the finally until the count in the try is set. We rely on some of the properties of DFS pre and post number bracketing to do this efficiently, without needing to track extra state. Contributes to #82964.
-
由 Michal Strehovský 提交于
This will compile the NativeAOT compiler with the LKG build of the NativeAOT compiler that the repo is building with. At this point NativeAOT-compiled compiler is significantly better than the R2R+Trimmed+SingleFile-compiled combo we're shipping right now. ilc.exe size before: 31 MB. ilc.exe size after: 14 MB. Time to compile hello world before: 2.3 seconds. Time to compile hello world after: 1.3 seconds.
-
由 Andy Gocke 提交于
* Enable DwarfDump AOT test in Release There are some reports (#82802) that debugging might be particularly bad in Release. * Update warning counts for Release
-
由 dotnet-maestro[bot] 提交于
* Update dependencies from https://github.com/dotnet/llvm-project build 20230308.1 runtime.linux-arm64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.linux-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.linux-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-musl-arm64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-musl-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-musl-x64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-musl-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-x64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.linux-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.linux-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.linux-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.osx-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.osx-arm64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.osx-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.osx-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.win-arm64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.win-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.win-x64.Microsoft.NETCore.Runtime.JIT.Tools , runtime.win-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Sdk , runtime.win-x64.Microsoft.NETCore.Runtime.Mono.LLVM.Tools , runtime.win-x64.Microsoft.NETCore.Runtime.ObjWriter From Version 14.0.0-alpha.1.23156.3 -> To Version 14.0.0-alpha.1.23158.1 * Update dependencies from https://github.com/dotnet/emsdk build 20230308.1 Microsoft.NET.Workload.Emscripten.Current.Manifest-8.0.100-preview.3 , Microsoft.NET.Workload.Emscripten.net6.Manifest-8.0.100-preview.3 , Microsoft.NET.Workload.Emscripten.net7.Manifest-8.0.100-preview.3 From Version 8.0.0-preview.3.23156.1 -> To Version 8.0.0-preview.3.23158.1 --------- Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
-
由 Jakob Botsch Nielsen 提交于
Fix #83140
-
由 Katelyn Gadd 提交于
* Add getNN_unaligned memory.ts APIs for cases where we cannot be certain the source address is aligned * Remove CU64 typescript APIs since they were unused
-
由 Badre BSAILA 提交于
* provide custom marshaller for USER_INFO_1
-
由 Steve Pfister 提交于
* [workloads] Add win-arm64 in the net6 manifest This was added in release/7.0, but not in main. Fixes https://github.com/dotnet/sdk/issues/30019 * alias the aot compiler for win-arm64 --------- Co-authored-by: NSteve Pfister <steve.pfister@microsoft.com>
-
由 Egor Bogatov 提交于
-
由 Stephen Toub 提交于
If the directory watching encounters a failure error code other than abort, the Windows implementation raises the Error event and sets EnableRaisingEvents to false. However, because it does so in that order, the Error event handler may or may not see the FSW's EnableRaisingEvents as false, based on whether there's a SynchronizingObject that results in the callback being queued. By swapping the order to always disable first, we can ensure consistency regardless of SynchronizingObject, which also means the Error handler can choose based on the error to set EnableRaisingEvents to true again if it desires.
-
由 Jackson Schuster 提交于
-
由 Steve Pfister 提交于
https://github.com/dotnet/runtime/pull/81919 introduced a regression where the gss framework was omitted from the sample build. This change adds it back in. Fixes https://github.com/dotnet/performance/issues/2924
-
- 09 3月, 2023 14 次提交
-
-
由 Qiao Pengcheng 提交于
-
由 xtqqczze 提交于
* Replace `SafeLocalAllocHandle` in `Diagnostics.PerformanceCounter` * Update SharedPerformanceCounter.cs * Update System.Diagnostics.PerformanceCounter.csproj
-
由 Jakob Botsch Nielsen 提交于
This PR allows TYP_STRUCT locals to have block layouts and replaces uses of TYP_BLK and TYP_LCLBLK with such locals instead. There is still an invariant that any struct parameter local (even SIMD) has a non-block layout. Also fixes a bug related to GS cookie handle with jit32 GC encoder -- it is not allowed to be at fp+0, so we need to insert some padding in some cases because now that some GS cookie-requiring locals are TYP_STRUCT they can sometimes be removed from the frame. To fix a TP regression, also precache the 0-sized block layout in the layout table. This layout is used by all non-x86 compilations because the outgoing arg area starts out as a 0-sized block local that is always allocated.
-
由 Vitek Karas 提交于
This was caused by interaction between two PRs. AOT is now expected to warn on all single-file and dynamic code attributes, just like the analyzer.
-
由 Eirik Tsarpalis 提交于
* Fix handling of required & init-only properties with custom property names in source gen. * Address feedback.
-
由 dotnet bot 提交于
-
由 Aleksey Kliger (λgeek) 提交于
-
由 Michal Strehovský 提交于
-
由 Michal Strehovský 提交于
Should address https://github.com/dotnet/runtime/pull/82307#discussion_r1111436978.
-
由 Aaron Robinson 提交于
* Remove fragile NGen logic for precomputed hashcodes Hashing was confirmed to offer minor performance wins but not with lazy hash code generation. Pre hashing was needed to have a benefit on start-up but even then it was minor.
-
由 Carlos Sánchez López 提交于
Fix PAX extended attribute reading logic to treat '=' character as valid in the value strings. (#82810) * Move PaxExtendedAttribute_Roundtrips test to correct source code file. It is not handling any filesystem entries. * Bug fix: Do not fail when reading an extended attribute when the value contains an '=' character., * Add unit tests that verify extended attribute and global extended attribute roundtripping when the value contains an '=' character. Also add a null check for a subsequent GetNextEntry. * Convert duplicate InlineData to single shared MemberData method. * Apply suggestion --------- Co-authored-by: Ncarlossanlop <carlossanlop@users.noreply.github.com>
-
由 Tanner Gooding 提交于
* Ensure that the new Vector512 related ISAs are covered by ILLink Substitutions * Define the Avx512F and derived classes
-
由 Tanner Gooding 提交于
* Refactoring GenTreeVecCon to support TYP_SIMD64 * Apply formatting patch * Fixing a couple bugs in the refactoring * Respond to some PR feedback by deduplicating code
-
由 Katelyn Gadd 提交于
#82604 produced more regressions than it did improvements, so this PR relaxes one of its changes and disables the more suspect one temporarily. In the future the heuristic will be significantly revised, and the actual overhead of various parts of traces is going to change once the CFG is added, so it's not worthwhile to invest a bunch of time into tuning the parameters right now. Once things are more stable we will hopefully be able to identify the right parameter values by doing a bunch of benchmark runs.
-