bot upstream main merge 2021 11 30 #9
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Catch patterns like the one in dotnet#37904 where a trinary compare feeds a binary compare.
* We were not recording precise info in inlinees except for at IL offset 0 because most of the logic that handles determining when to attach debug info did not run for inlinees. There are no changes in what the EE sees since we were normalizing debug info back to the root anyway. * Propagate debug info even further than just until rationalization, to make it simpler to dump the precise debug info. This means we create some more GT_IL_OFFSET nodes, in particular when the inlinee debug info is valid but the root info is invalid. This is currently happening for newobj IL instructions when the constructor is inlined. We generate two statements: GT_ASG(GT_LCL_VAR(X), ALLOCOBJ(CLS)); GT_CALL(CTOR, GT_LCL_VAR(X)) and the first statement ends up "consuming" the debug info, meaning we end up with no debug info for the GT_CALL, which eventually propagates into the inline tree. I have held off on fixing this for now since it causes debug info diffs in the data reported back to the EE. The additional nodes in LIR result in 0.15% more memory use and 0.015% more instructions retired for SPMI over libraries. There is also a small fix in gtlist.h for GT_BFIZ when MEASURE_NODE_SIZES is defined. No SPMI diffs.
There are connectivity issues on some physical Android devices. We should disable the affected tests until the issue isn't resolved. Ref dotnet#61343
…net#60966) We reference entry #0 in the pinned plug queue even if there are no pinned plugs at all and thus the pinned plug queue contains left-over data from the mark phase. The fix is to initialize saved_pinned_plug_index to a value that is invalid as a pinned plug queue index, and only use saved_pinned_plug_index as an index if is valid.
Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
Don't run if the JIT-EE GUID has changed, since there won't be any collections to download, so the downloads will fail. Also update the superpmi-asmdiffs trigger comment.
… annotate individual methods as requiring AVX (dotnet#61259) * Updating src/tests/Interop/PInvoke/Generics/GenericsNative.Vector* to annotate individual methods as requiring AVX * Always use __m256i on XARCH
* XARCH: Remove redudant tests for GT_LT/GT_GE relops. We can now optimize cases such as `(x + y < 0)` or `for (int x = v; x >= 0; x--)` using the flag tracking logic during the emit stage. Notably, cases that would generate... ``` add reg0, reg1 test reg0, reg0 jge LABEL ``` now transform to ``` add reg0, reg1 jns LABEL ``` This transform is valid for signed GE and signed LT only. * Add a few asserts related to flag reuse optimizations.
* Moving Narrow to implemented using SIMDAsHWIntrinsic * Moving Widen to implemented using SIMDAsHWIntrinsic * Fix some handling of Narrow/Widen hwintrinsics * Ensure that Vector.Widen is still treated as an intrinsic * Fixing NI_VectorT128_WidenUpper on ARM64 to actually call gtNewSimdWidenUpper
Measure browser app start times, 2 measurements implemented.
First to measure till the JS window.pageshow event, second to measure
time when we reach managed C# code.
Example ouput:
| measurement | time |
|-:|-:|
| AppStart, Page show | 108.1400ms |
| AppStart, Reach managed | 240.2174ms |
* Environment.SystemPageSize returns cached value * we are no longer shipping MS.IO.Redist, so we can use Array.MaxLength directly * we are no longer shipping MS.IO.Redist, there is no need for File to be partial * we are no longer shipping MS.IO.Redist, there is no need for FileInfo to be partial * there is no need for .Win32.cs and .Windows.cs file anymore
* Remove all DLLIMPORTGENERATOR_ENABLED usage. * Explicitly set EnableDllImportGenerator to true in test project.
… the receiver is a gshared type constrained to a primitive type/enum. (dotnet#61513)
…trained types. (dotnet#61514) These are used for example in Span<T>:.ctor ().
This was preventing block compaction with loop entry blocks in loops that had been previously optimized away (and thus removed from the loop table). There are a few cases where we now delete dead code that was previously left in the function. There are a number of spurious local weighting and IG textual asm diffs changes, possibly due to how PerfScore is implemented (there are some surprisingly large PerfScore changes in a few cases, despite no change in (most) generated code).
…ardless of what they point at (dotnet#61538)
Implement method name wildcard matching for method descriptions Globbing doesn't work because we don't have g_pattern_match_simple in eglib. But a plain '*' wildcard does work.
When both OSR and PGO are enabled: * Enable instrumenting OSR methods, so that the combined profile data from Tier0 plus any OSR variants provide a full picture for subsequent Tier1 optimization. * Use block profiles for both Tier0 methods that are likely to have patchpoints and OSR methods. * Fix phase ordering so partially jitted methods don't lose probes. * A few more fixes for partial compilation, because the number of things we think we might instrument and the number of things we end up instrumenting can differ. * Also improve the DumpJittedMethod output for OSR, and allow selective dumping of a particular OSR variant by specifying its IL offset. The updates on the runtime side are to pass BBINSTR to OSR methods, and to handle the (typical) case where the OSR method instrumentation schema is a subset of the Tier0 method schema. We are still allowing OSR methods to read the profile data. So they are both profile instrumented and profile optimized. Not clear if this is going to work well as the Tier0 data will be incomplete and optimization quality may be poor. Something to revisit down the road.
Emit it in the interpreter when a method is inlined or replaced with an intrinsic. This is needed so the AOT profiler can track these methods.
…traints. (dotnet#59437) If a generic argument is a primitive type, and it has an interface constraint that enums don't implement, then partial sharing for that instance is not useful, since only the specific primitive type instance will be able to use the shared version. Fixes dotnet#54910.
* Avoid switching GC mode in COM's HasValidTarget Co-authored-by: Jan Kotas <[email protected]>
Co-authored-by: JamesNK <[email protected]>
in high memory load situations, when we had to wait for a BGC to finish, we switched to preemptive mode which means GCs could have occurred and replenished the budget. but when we come back from the wait, we use the previous decision on whether the budget was exceeded or not. this causes us to trigger GC incorrect and you would see GCs triggered when the allocated bytes are tiny.
Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com> Co-authored-by: Premek Vysoky <[email protected]> Co-authored-by: Alexander Köplinger <[email protected]>
* Fix miscellaneous typos * Cleanup trailing whitespaces ```sh # git remote add dotnet https://github.com/dotnet/runtime && git pull --rebase dotnet main if uname 2>/devnull | grep -q Darwin; then space=" " fi git show --name-only --pretty="" HEAD...dotnet/main |\ xargs -I{} sh -c "test -f {} && sed -i$space'' 's/[[:space:]]*$//' {}" ```
…62003) * mono_runtime_delegate_invoke wrapper with MONO_ENTER_GC_UNSAFE/MONO_EXIT_GC_UNSAFE; mono_runtime_delegate_invoke wrapper with MONO_ENTER_GC_UNSAFE/MONO_EXIT_GC_UNSAFE; * Update object.c
The CSE logic maintains a custom hashtable implementation. It triggers a resize at the same time as adding a new CSE, but forgets the hash of the new CSE needs to be updated from its pre-resize value. Failing to do so can lead to losing some CSEs, though not in a correctness-impacting way.
Otherwise we can end up not seeing the loop has memory havoc. Also added an assert that will prevent this issue from arising in the future.
…et#60054) * CI matrix change: add Windows.Server.Core.20H2 (dotnet#59572) * Add Windows.Server.Core.20H2 to CI matrix * add windows server 20H2 * Remove 20H2 from PR * Remove 20H2 from PR run * change helix queue for Windows.Server.Core.20H2
* Remove active issue. * Use --web-server-use-cop.
Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
* Allow zero byte reads on raw HTTP/1.1 response streams * Allow zero byte reads on the rest of HTTP/1.1 response streams * Allow zero byte reads on HTTP/2 response streams * Allow zero byte reads on HTTP/3 response streams * Enable sync zero-byte reads * Add zero-byte read tests * Fully enable zero-byte reads on HTTP/2 and 3 * Add zero-byte read tests for HTTP/2 and HTTP/3 * Remove unsafe-ish code from PeekChunkFromConnectionBuffer * Add comments when we do extra zero-byte reads * Update MockQuicStreamConformanceTests to allow zero-byte reads * Update ConnectedStream tests to allow zero-byte reads * Skip zero-byte read tests on Browser * Update comment on explicit zero-byte reads in ChunkedEncodingReadStream
…t#61707)" (dotnet#62157) This reverts commit 1997bee.
…tnet#61853) This PR does two things: 1. Reorganizes how we keep track of deltas, and how we do metadata lookups after updates 2. Adds an ifdefed-out initial draft of support for adding methods. (It worked for a simple test but it's not ready for general use yet). Details of the delta changes. * Previously Mono used a fully immutable model of metadata changes. Logically each delta brings in modifications to existing table rows and addition of new rows. Previously mono never formed the fully mutated tables - instead, each lookup was essentially responsible for playing all the changes forward every time it needed to lookup a row. This was fine when we primarily supported only row additions (because we could just skip past all the updates that hadn't added the row we wanted yet, and just look in a single delta). But as we started to support more and more modifications (of custom attributes, of property getters & setters, of parameters, etc) we now had to look through the whole list of deltas to make sure we found the latest modification of every row. * The new approach is for each `DeltaInfo` (representing a summary of a particular update to a single assembly) to contain a collection of `mutants` - the fully modified tables with all the newly added rows and all the new modifications. So lookups are fast now - we just go to the latest generation that is visible to a particular thread and look at its mutants tables. The downside is increased memory use. So we allocate the mutants from memory pools owned by each `DeltaInfo`. Right now we never dealloc the pools, but in the future we could. We already iterate over all the threads during our stop-the-world phase - we can record the earliest generation still needed by every thread and delete the pools for all earlier copies. In practice this means we only keep 3 sets of tables: the mmaped baseline tables, the newest mutants, and the mutants from the prior generation for any still-executing methods. * Additionally instead of storing a list of delta images in the `BaselineInfo`, we now store a list of `DeltaInfo` structs which each `DeltaInfo` pointing to a delta image. This means that we can avoid repeated hash table lookups to map from a delta image to its `DeltaInfo` every time the runtime queries us for a table row or for a heap element. --- * checkpoint: allow adding methods to existing classes. Doesn't do anything yet. And needs more sanity checking (roslyn won't give us virtual or abstract methods, but we should check). * fixme in loader * [metadata] Add mono_metadata_table_num_rows Returns the number of rows in a metadata table, taking into account metadata update deltas * add table to ptr table helper * Param attr lookups for deltas can have param_index == 0 The params are added with a subsequent enclog "add param" function. * WIP: start adding support for parameter additions. It "works" in that calling methods appears to work (ie direct token references seem to do the right thing). Primarily this is because the param table additions are not that interesting. All the good stuff is in the method signature (which is just in the blob heap). Presumably anything that actually needs parameter attributes, or anything that uses reflection to look at the parameters, will break. * WIP: add MethodDef -> TypeDef lookup Allows calling non-public methods, which are now assigned the correct parent * Add hot reload test for lambda capturing this Lambdas that only capture `this` (and not local variables or arguments) compile to a private instance method in the enclosing class. So it is enough to support EnC deltas that add methods. * clarify comments about MONO_METHOD_PARAMLIST * [hot_reload] Store debug info of updated methods * [hot_reload] Allocate modifiable tables in DeltaInfo This is the foundation for a new approach for metadata lookups. Instead of using an immutable model (each lookup traverses every delta to find the most up to date version of each table row), we are going to create a complete updated table for each generation and only do the lookup in the latest exposed generation directly. (This is essentially the CoreCLR model). This commit is just the first foundations: we allocate the tables and copy over the previous generations' rows and zero out any rows that will be inserted. Delta applications and lookups have not been updated yet. As a slight optimization, tables that don't have modified or added rows are not copied from the base image. If a generation modifies or adds rows, from that point forward, each subsequent generation will copy the table. We could be a bit more thrifty with copying, but it will complicate lookups. Also eventually we will try to deallocate the pools for generations that are older than no thread needs anymore. Metadata heaps are still looked up in each delta directly - heap combining does not seem necessary yet. * [mini] Allow MONO_VERBOSE_METHOD='*:*' Implement method name wildcard matching for method descriptions Globbing doesn't work because we don't have g_pattern_match_simple in eglib. But a plain '*' wildcard does work. * populate mutated table rows leave suppressed columns unchanged * [hot_reload] Switch lookups to the mutant tables * cleanup: remove of effective_table calculation The row index doesn't change in the new scheme, pass it by value * cleanup: Remove relative_delta_index from component API It's only used internally by the update logic * cleanup: Pass DeltaInfo to relative_delta_index don't compute it from the dmeta image * cleanup: Store a list of DeltaInfo in the BaselineInfo Instead of storing a list of delta images and then relying on delta_info_lookup to find the DeltaInfo, just store them directly on the baseline info. This changes the cleanup responsibilities a bit. Now we destroy the DeltaInfo when we are iterate through the delta infos when we close the baseline image, instead of when we remove the delta_image_to_info hashtable entry. * Turn off method addition support, for now Just want to get the cleanups enabled for now * Fix null ptr when checking for updated ppdb info
…g corelib directly. (dotnet#62154)
The codespaces prebuild action can only run in the dotnet/runtime repo and not forks. The feature needs to be enabled for each repo and will fail if it isn't enabled. In order to minimize failure notifications, add a check to the action so it only runs against dotnet/runtime.
…tnet#62015) * A draft of doc for describing debugger behaviour. * Removing spaces.
The clang 13 optimizer started to assume that "this" pointer is always properly aligned. That lead to elimination of some code that was actually needed. It also takes pointer aliasing rules more strictly in one place in jit. That caused the optimizer to falsely assume that a callee with an argument passed by reference is not modifying that argument and used a stale copy of the original value at the caller site. This change fixes both of the issues. With this fix, runtime compiled using clang 13 seems to be fully functional.
In particular do not attach it to INS_align that do not end up producing any code.
* fix TimeSpan support in sourcegen * address feedback
yuc434
pushed a commit
that referenced
this pull request
Aug 8, 2022
* Initial implementation for contract customization fix build errors Move converter rooting to DefaultJsonTypeInfoResolver so that it can be used standalone Fix ConfigurationList.IsReadOnly Minor refactorings (#1) * Makes the following changes: * Move singleton initialization for DefaultTypeInfoResolver behind a static property. * Consolidate JsonSerializerContext & IJsonTypeInfoResolver values to a single field. * Move reflection fallback logic away from JsonSerializerContext and into JsonSerializerOptions * Update src/libraries/System.Text.Json/src/System/Text/Json/Serialization/JsonSerializerOptions.cs * remove testing of removed field Simplify the JsonTypeInfo.CreateObject implemenetation (#2) * Simplify the JsonTypeInfo.CreateObject implemenetation * Update src/libraries/System.Text.Json/src/System/Text/Json/Serialization/Metadata/JsonTypeInfoOfT.cs * Update src/libraries/System.Text.Json/src/System/Text/Json/Serialization/Metadata/JsonTypeInfoOfT.cs Co-authored-by: Krzysztof Wicher <[email protected]> Co-authored-by: Krzysztof Wicher <[email protected]> Tests and fixes for JsonTypeInfoKind.None TypeInfo type mismatch tests Allow setting NumberHandling on JsonTypeInfoKind.None test resolver returning wrong type of options JsonTypeInfo/JsonPropertyInfo mutability tests rename test file Move default converter rooting responsibility behind DefaultJsonTypeInfoResolver (#3) * Move default converter rooting responsibility behind DefaultJsonTypeInfoResolver * address feedback Add simple test for using JsonTypeInfo<T> with APIs directly taking it fix and tests for untyped/typed CreateObject uncomment test cases, remove todo More tests and tiny fixes Add a JsonTypeInfoResolver.Combine test for JsonSerializerContext (#4) * Fix JsonTypeInfoResolver.Combine for JsonSerializerContext * Break up failing test Fix simple scenarios for combining contexts (#6) * Fix simple scenarios for combining contexts * feedback JsonSerializerContext combine test with different camel casing Remove unneeded virtual calls & branching when accessing Get & Set delegates (#7) JsonPropertyInfo tests everything minus ShouldSerialize & NumberHandling Update src/libraries/System.Text.Json/src/System/Text/Json/Serialization/JsonConverterOfT.cs Update src/libraries/System.Text.Json/src/System/Text/Json/Serialization/JsonConverterOfT.cs throw InvalidOperationException rather than ArgumentNullException for source gen when PropertyInfo.Name is assigned through JsonPropertyInfoValues tests for duplicated property names and JsonPropertyInfo.NumberHandling Add tests for NumberHandling and failing tests for ShouldSerialize disable the failing test and add extra checks disable remainder of the failing ShouldSerialize tests, fix working one Fix ShouldSerialize and IgnoreCondition interop Add failing tests for CreateObject + parametrized constructors Fix CreateObject support for JsonConstructor types (#10) * Fix CreateObject support for JsonConstructor types * address feedback Make contexts more combinator friendly (#9) * Make contexts more combinator friendly * remove converter cache * redesign test to account for JsonConstructorAttribute * Combine unit tests * address feedback * Add acceptance tests for DataContract attributes & Specified pattern (#11) * Add private field serialization acceptance test (#13) * tests, PR feedback (#14) * PR feedback and extra tests * Shorten class name, remove incorrect check (not true for polimorphic cases) * Make parameter matching for custom properties map property Name with parameter (#16) * Test static initialization with JsonTypeInfo (#17) * Fix test failures and proper fix this time (#18) * Fix test failures and proper fix this time * reinstate ActiveIssueAttribute * PR feedback and adjust couple of tests which don't set TypeInfoResolver * fix IAsyncEnumerable tests * Lock JsonSerializerOptions in JsonTypeInfo.EnsureConfigured() Co-authored-by: Eirik Tsarpalis <[email protected]> Co-authored-by: Eirik Tsarpalis <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
GT_AND_NOTfor ARM/ARM64 (ImplementGT_AND_NOTfor ARM/ARM64 dotnet/runtime#59881)GenTreeFlagsoftener for better debugging. dotnet/runtime#61272)GT_LCL_FLDdotnet/runtime#61209)m_hostAssemblyMapand ceremony around it (Removingm_hostAssemblyMapand ceremony around it dotnet/runtime#61292)Microsoft.Extensions.Logging.Generators.Roslyn3.11.Testsfailing due to OOM ([wasm][aot] Disable (again)Microsoft.Extensions.Logging.Generators.Roslyn3.11.Testsfailing due to OOM dotnet/runtime#61323)nuget.configas the base for installing manifests (InstallWorkload: Use the repo'snuget.configas the base for installing manifests dotnet/runtime#60699).versionfile in shared fx dotnet/runtime#60944)__DoCrossgen2(Delete__DoCrossgen2dotnet/runtime#61582)impCheckForNullPointerdotnet/runtime#61576)DisableImplicitNamespaceImports_DotNet(RemoveDisableImplicitNamespaceImports_DotNetdotnet/runtime#61656)Microsoft.Extensions.Configuration.Binder([Group 4] Enable nullable annotations forMicrosoft.Extensions.Configuration.Binderdotnet/runtime#57418)Microsoft.Extensions.Configuration.EnvironmentVariables([Group 4] Enable nullable annotations forMicrosoft.Extensions.Configuration.EnvironmentVariablesdotnet/runtime#57433)Microsoft.Extensions.Configuration.CommandLine([Group 4] Enable nullable annotations forMicrosoft.Extensions.Configuration.CommandLinedotnet/runtime#57432)GT_LIST) (Make intrinsic nodes multi op (aka deleteGT_LIST) dotnet/runtime#59912)GetPinnableReference()usage in multi-step custom type marshalling (UpdateGetPinnableReference()usage in multi-step custom type marshalling dotnet/runtime#61539)VNForMapStore(Do not explicitly pass type toVNForMapStoredotnet/runtime#61882)ThrowHelper([System.Text.Json] Move inline throw statements toThrowHelperdotnet/runtime#61746)InferStructOpSizeAligndotnet/runtime#61880)optComputeLoopSideEffectsto account for HWI stores (FixoptComputeLoopSideEffectsto account for HWI stores dotnet/runtime#61911)