James Newton-King [Tue, 23 May 2023 17:06:00 +0000 (01:06 +0800)]
Improve Microsoft.Extensions.Configuration debugging (#86624)
* Improve Microsoft.Extensions.Configuration debugging
* Clean up
* Clean up
* Clean up + tests
Jakob Botsch Nielsen [Tue, 23 May 2023 16:54:24 +0000 (18:54 +0200)]
JIT: Use unsigned native return types for small structs (#86631)
Prefer the return type to be unsigned when small structs are returned in
registers. This causes the backend to use zero extension instead of sign
extension in a few cases, which has a smaller encoding on xarch.
Peter Sollich [Tue, 23 May 2023 16:37:03 +0000 (18:37 +0200)]
Fix assert failure in grow_heap_segment in no GC region. (#86644)
This addresses issue #86612.
We won't actually decommit anything once we enter a no GC region, so the assert was overeager for this case, but it seems better to keep the assert and turn off the gradual_decommit_in_progress_p flag when we enter a no GC region.
Peter Sollich [Tue, 23 May 2023 15:43:41 +0000 (17:43 +0200)]
Dynamic heap count (#86245)
This is an initial implementation for changing the GC heap count dynamically in response to changing load conditions.
Using more heaps will increase memory footprint, but in most cases also improve throughput because more GC work is parallelized, and lock contention on the allocation code path is reduced by spreading the load.
We try to keep the measured overhead from lock contention and GC pauses at 5% or below, and only reduce the heap count if the measured overhead is below 1% and we estimate a significant reduction in memory footprint. There is some more fine tuning for the cases where smaller reductions in either footprint or overhead are possible without impacting the other side of the equation too much.
Because the input data for GC pause etc. are quite noisy, we use a median of 3 filter before the data is used to make decisions. Preliminary data suggests this is effective, but we may need to do more filtering if gains in either direction appear to be small.
Vlad Brezae [Tue, 23 May 2023 15:31:31 +0000 (18:31 +0300)]
[mono][interp] Attempt to devirtualize call if we have type information about `this` (#85528)
* [mono][interp] Preserve klass information when generating MINT_CKNULL
This information will be useful to later devirtualize the call
* [mono][interp] Add more specific type information for EqualityComparer.Default
* [mono][interp] Devirtualize calls if we can find a final implementation
Remove asserts from mono_class_get_virtual_method so we can check a virtual method on any klass, returning NULL if no implementation is found.
* [mono][interp] Optimize out null check on ldloca
* [mono][interp] Constant fold unop conditionals applied to non null values
* [mono][interp] Optimize out unused MINT_BOX
* [mono] Remove duplicated code between interp and jit
* [mono][metadata] Allow method to be called for abstract classes
It will just return null if a concrete implementation wasn't found.
* [mono][interp] Don't populate wrong type information into the stack state
* [mono][interp] Don't devirtualize final methods
They can still be overriden
* [mono][interp] Ensure class is initialized before getting aot method
Rupesh Ghosh [Tue, 23 May 2023 13:41:12 +0000 (19:11 +0530)]
Fix Incorrect RequiresDynamicCode attribute message (#86602)
* Fix Incorrect RequiresDynamicCode attribute message
Replace GetEnumValues<TEnum> in the RequiresDynamicCode attribute message on System.Type.GetEnumValues, System.Reflection.SignatureType.GetEnumValues and System.Runtime.GetEnumValues with Enum.GetValues<T> with Enum.GetValues<T>
Fix #86585
* Update attribute message as suggested in PR
* Add warning suppression
Vlad-Alexandru Ionescu [Tue, 23 May 2023 11:11:48 +0000 (13:11 +0200)]
[mono][AOT] Enabled tests which fail to JIT compile during AOT (#86581)
* Enabled 2 tests
Signed-off-by: Vlad - Alexandru Ionescu <vlad-alexandruionescu@Vlads-MacBook-Pro-5.local>
---------
Signed-off-by: Vlad - Alexandru Ionescu <vlad-alexandruionescu@Vlads-MacBook-Pro-5.local>
Co-authored-by: Vlad - Alexandru Ionescu <vlad-alexandruionescu@Vlads-MacBook-Pro-5.local>
Marek Fišera [Tue, 23 May 2023 08:54:46 +0000 (10:54 +0200)]
[browser] Turn webcil off for .NET < 8.0 (#86505)
Vitek Karas [Tue, 23 May 2023 07:53:01 +0000 (00:53 -0700)]
Unreachable filter end (#86492)
Fixes the problem described in https://github.com/dotnet/roslyn/issues/67494 for illink.
`endfilter` instruction must be kept if the filter block is kept, even if the instruction itself is unreachable - per the ECMA spec. Fix the branch removal code to preserve the instruction unconditionally.
Adds a test to validate the new behavior.
I also enabled the test to run on NativeAOT, since it has code to handle this as well. We only validate that the compiler will not crash/warn, the Kept validation is disabled, since we can't really inspect IL of the produced methods in AOT.
Also fixes a bug in the illink test infra where the `By` property on `SkipKeptItemsValidationAttribute` was ignored.
Vitek Karas [Tue, 23 May 2023 07:50:10 +0000 (00:50 -0700)]
Fix endless loop in illink analyzer (#86449)
The root cause is the fact that `ArrayValue` is mutable, so it needs to be `DeepCopy` everywhere. So the first set of fixes is to make sure that even hoisted locals are correctly copied.
The second problem is that we use `HashSet` to store multiple values in `ValueSet`/`MultiValue`. HashSet assumes that items don't change identity (equality group) while they're stored inside. But `ArrayValue` does exactly that. So we can't rely on pure `HashSet` functionality to implement `Equals`.
Changes:
* Make Maybe<T> an IDeepCopyValue and make sure it deep copies its value - because it can store values which require a deep copy
* Make ValueSet<T> an IDeepCopyValue since it can store values which require a deep copy (this causes an effective rename of MultiValue.Clone to MultiValue.DeepCopy - lot of files touched)
* Modify ValueSet<T> to be careful when comparing for equality - we can't rely on immutability of values and since we store them in a hashset, the hashset itself effectively breaks. Maybe we should consider switching to a simple List? But then we want to to unify same values.
* Modify array equality comparison - previously it caused an allocation of a `HashSet` - we compare values quite a bit (in analyzer during multi-pass analysis). Currently we rely on values in arrays to be immutable.
Future looking:
The data flow analysis simply doesn't have a solution for mutable values (mostly mutable refence types, but I think similar set of problems might occur with structs as well). ILLink implements by-ref values, which are probably the closest approximation yet, but the full CFG data flow doesn't support those yet. And ultimately it's probably more complex then just by-ref values.
Proper fix is complex, and requires some design work first.
Note that we will want to backport this to .NET 7 very likely - I haven't looked yet how well it will port.
Egor Bogatov [Tue, 23 May 2023 07:21:59 +0000 (09:21 +0200)]
Allow overbudget for small methods (#86473)
Co-authored-by: Andy Ayers <andya@microsoft.com>
Jan Kotas [Tue, 23 May 2023 05:04:38 +0000 (22:04 -0700)]
Add note about transient pointers to ECMA augments (#86622)
Fixes #62784
Buyaa Namnan [Tue, 23 May 2023 02:36:40 +0000 (19:36 -0700)]
Abstract out System.Reflection.Emit.ILGenerator (#86594)
* Abstract out ILGenerator
* Finish up abstracting
* Remove protected abstract methods, use more public abstract instead
* Use RuntimeILGenerator for Linq test
Tanner Gooding [Tue, 23 May 2023 01:53:29 +0000 (18:53 -0700)]
Update GetElement and ToScalar to lower IND, LCL_FLD, and LCL_VAR specially (#86400)
* Update GetElement and ToScalar to lower IND, LCL_FLD, and LCL_VAR specially
* Ensure GetElement still correctly handles too large indices
* Only optimize to LCL_FLD for lvDoNotEnregister
* Account for an extreme edge case in offset overflow
* Use simdSize rather than maxSIMDStructBytes
Katelyn Gadd [Tue, 23 May 2023 01:52:18 +0000 (18:52 -0700)]
[wasm] No longer discard jiterp inference information when crossing a branch (#86611)
* Discarding inferred state when crossing a branch is not strictly necessary and reduces performance
* Don't allow treating ldloca'd addresses as constant
SingleAccretion [Mon, 22 May 2023 23:33:11 +0000 (02:33 +0300)]
Add a test for the first chance exception event (#86560)
Elinor Fung [Mon, 22 May 2023 22:29:48 +0000 (00:29 +0200)]
Rename setting for using RID graph to System.Runtime.Loader.UseRidGraph (#86598)
Tanner Gooding [Mon, 22 May 2023 22:25:03 +0000 (15:25 -0700)]
Ensure Avx512Vbmi has [Intrinsic] on the right members (#86604)
* Ensure Avx512Vbmi has [Intrinsic] on the right members
* Ensure vpermb is covered
Jakob Botsch Nielsen [Mon, 22 May 2023 22:00:37 +0000 (00:00 +0200)]
JIT: Switch basic block successors to a visitor pattern (#86543)
BasicBlock::GetAllSuccs seems to be very slow.
Jakob Botsch Nielsen [Mon, 22 May 2023 21:53:06 +0000 (23:53 +0200)]
Switch physical promotion stress modes to only be enabled in checked builds (#86590)
Otherwise we may be running old preview 4 bits with physical promotion
enabled and running into issues that were already fixed.
Fix #86588
Stephen Toub [Mon, 22 May 2023 21:44:38 +0000 (17:44 -0400)]
Avoid capturing diagnostic objects in Regex source generator incremental values (#86606)
It's mostly a non-issue given how we use it, but Diagnostic's message arguments are compared with reference equality rather than value equality. Store the data into our own object instead, and only create the actual Diagnostic objects when we're about to report them.
Eirik Tsarpalis [Mon, 22 May 2023 21:39:42 +0000 (22:39 +0100)]
Avoid capturing diagnostic objects in SG incremental values. (#86605)
dotnet-maestro[bot] [Mon, 22 May 2023 21:02:11 +0000 (14:02 -0700)]
[main] Update dependencies from dnceng/internal/dotnet-optimization (#86436)
* Update dependencies from https://dev.azure.com/dnceng/internal/_git/dotnet-optimization build
20230517.3
optimization.linux-arm64.MIBC.Runtime , optimization.linux-x64.MIBC.Runtime , optimization.windows_nt-arm64.MIBC.Runtime , optimization.windows_nt-x64.MIBC.Runtime , optimization.windows_nt-x86.MIBC.Runtime , optimization.PGO.CoreCLR
From Version 1.0.0-prerelease.23260.3 -> To Version 1.0.0-prerelease.23267.3
Dependency coherency updates
Microsoft.SourceLink.GitHub,Microsoft.DotNet.XliffTasks
From Version 8.0.0-beta.23252.2 -> To Version 8.0.0-beta.23218.3 (parent: Microsoft.DotNet.Arcade.Sdk
* Update dependencies from https://dev.azure.com/dnceng/internal/_git/dotnet-optimization build
20230518.7
optimization.linux-arm64.MIBC.Runtime , optimization.linux-x64.MIBC.Runtime , optimization.windows_nt-arm64.MIBC.Runtime , optimization.windows_nt-x64.MIBC.Runtime , optimization.windows_nt-x86.MIBC.Runtime , optimization.PGO.CoreCLR
From Version 1.0.0-prerelease.23260.3 -> To Version 1.0.0-prerelease.23268.7
* Update dependencies from https://dev.azure.com/dnceng/internal/_git/dotnet-optimization build
20230519.15
optimization.linux-arm64.MIBC.Runtime , optimization.linux-x64.MIBC.Runtime , optimization.windows_nt-arm64.MIBC.Runtime , optimization.windows_nt-x64.MIBC.Runtime , optimization.windows_nt-x86.MIBC.Runtime , optimization.PGO.CoreCLR
From Version 1.0.0-prerelease.23260.3 -> To Version 1.0.0-prerelease.23269.15
* Update dependencies from https://dev.azure.com/dnceng/internal/_git/dotnet-optimization build
20230520.4
optimization.linux-arm64.MIBC.Runtime , optimization.linux-x64.MIBC.Runtime , optimization.windows_nt-arm64.MIBC.Runtime , optimization.windows_nt-x64.MIBC.Runtime , optimization.windows_nt-x86.MIBC.Runtime , optimization.PGO.CoreCLR
From Version 1.0.0-prerelease.23260.3 -> To Version 1.0.0-prerelease.23270.4
* Require clang-16 for PGO optimization
* Update dependencies from https://dev.azure.com/dnceng/internal/_git/dotnet-optimization build
20230521.4
optimization.linux-arm64.MIBC.Runtime , optimization.linux-x64.MIBC.Runtime , optimization.windows_nt-arm64.MIBC.Runtime , optimization.windows_nt-x64.MIBC.Runtime , optimization.windows_nt-x86.MIBC.Runtime , optimization.PGO.CoreCLR
From Version 1.0.0-prerelease.23260.3 -> To Version 1.0.0-prerelease.23271.4
* Revert sourcelink and xliff-tasks downgrades
---------
Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
Co-authored-by: Larry Ewing <lewing@microsoft.com>
Co-authored-by: Sven Boemer <sbomer@gmail.com>
Stephen Toub [Mon, 22 May 2023 19:51:08 +0000 (15:51 -0400)]
Avoid ToString allocations for each ISpanFormattable value in string.Concat/Join(..., IEnumerable<T>) (#86521)
We can use a T's ISpanFormattable implementation to avoid the individual ToStrings. We also don't need to maintain a separate implementation for Concat; at the cost of one branch per value, we can just reuse the Join implementation and pick up all of its optimizations for Concat.
Brant Burnett [Mon, 22 May 2023 19:37:52 +0000 (15:37 -0400)]
Vectorize IEnumerable<T>.Sum where possible (#84519)
* Vectorize IEnumerable<T>.Sum where possible
* Remove unnecessary CreateChecked and fix long test
* Add more assertions
* Improve comments, don't use Unsafe.Add, use T.MinValue
Eirik Tsarpalis [Mon, 22 May 2023 17:30:34 +0000 (18:30 +0100)]
Apply the SourceWriter pattern to the STJ source generator. (#86526)
* Use the SourceWriter pattern in the STJ SG emitter.
* Make helper method static.
* Minimize the size of the generic helper method.
Xu Liangyu [Mon, 22 May 2023 17:17:31 +0000 (01:17 +0800)]
[LoongArch64] Fix some bugs in the test sets 'System.Runtime.Tests' and 'System.Linq.Expressions.Tests' for loongarch64. (#86435)
* Fix the error of the Truncate/Saturate convert from Float/Double NAN to UINT32/UINT64.
* fix some random failed testcases in System.Runtime.Tests.
* Update codegenloongarch64.cpp
Fix the RangeCheck bug of the TYP_INT case in CodeGen::genRangeCheck().
Layomi Akinrinade [Mon, 22 May 2023 16:38:43 +0000 (09:38 -0700)]
Rename a couple model specs for config binding generator (#86459)
Andrew Moskevitz [Mon, 22 May 2023 16:09:07 +0000 (12:09 -0400)]
Update createdumpwindows to not appear to succeed after getting ERROR_PARTIAL_COPY 5 times (#85462)
* Update createdumpwindows.cpp
create dump has same bug as documented and fixed in dotnet-dump (https://github.com/dotnet/diagnostics/issues/3829)
* Update createdumpwindows.cpp
changed retry count to 10 to match dotnet-dump
Aleksey Kliger (λgeek) [Mon, 22 May 2023 15:21:21 +0000 (11:21 -0400)]
[mono] When creating an array type, don't do full initialization of the element type (#85828)
Consider code like this:
```csharp
public class Node<T>
{
public Node<T>[][] Children;
}
Console.WriteLine (typeof(Node<int>));
```
In this case, when we're JITing ``ldtoken class Node`1<int32>`` we first have to parse the type ``Node`1<int32>`` and initialize it.
When we're initializing it, we need to resolve the types of all the fields in the class in order to to figure out if its instance size.
To resolve the field types we have to parse the types of all the fields, and we eventually end up parsing the type ``Node`1<!0>[][]`` which ends up here:
https://github.com/dotnet/runtime/blob/
558345d16cf76525d0c7fdbafbfd3a2457142b39/src/mono/mono/metadata/metadata.c#L4023-L4027
When we get to line 4027 the second time (recursively), `etype` is ``Node`1<!0>`` as a `MonoType*`. And we want to set `type->data.klass` to its corresponding `MonoClass*`. That ends up calling `mono_class_from_mono_type_internal`, which does this:
https://github.com/dotnet/runtime/blob/
558345d16cf76525d0c7fdbafbfd3a2457142b39/src/mono/mono/metadata/class.c#L2214-L2215
When we call `mono_class_create_array` we end up here:
https://github.com/dotnet/runtime/blob/
558345d16cf76525d0c7fdbafbfd3a2457142b39/src/mono/mono/metadata/class-init.c#L1186-L1187
with `eclass` equal to ``Node`1<!0>` which we try to initialize and we get a TLE because we're going to try to initialize the same generic type definition ``Node`1`` twice.
Compare with this other class:
```csharp
public unsafe class Node2<T>
{
public Node2<T>[] Children;
}
```
In this case we only end up calling `mono_class_from_mono_type_internal` on ``Node2`1<int32>`` (not an array). And that branch does not do any initialization - it just returns `type->data.klass` (for ``Node2`1<!0>`` - which is seen as a gtd at this point) or `mono_class_create_generic_inst (type->data.generic_class)` (for ``Node2`1<int32>`` which is seen later when `ldtoken` inflates the gtd)) - without any extra initialization.
---
It seems the reason we get into trouble is because `mono_class_create_array` does more work than other `MonoClass` creation functions - it tries to initialize the `MonoClass` a little bit (for example by setting `MonoClass:has_references`) which needs at least a little bit of the element class to be initialized.
But note that the code has this:
https://github.com/dotnet/runtime/blob/
558345d16cf76525d0c7fdbafbfd3a2457142b39/src/mono/mono/metadata/class-init.c#L1186-L1189
I feel fairly certain we don't need the full class initialization if we're going to immediately initialize the field size information.
Fixes #85821
Brant Burnett [Mon, 22 May 2023 14:58:56 +0000 (10:58 -0400)]
Vectorize shorter buffers for CRC-32 on Intel (#86539)
Lakshan Fernando [Mon, 22 May 2023 14:45:13 +0000 (07:45 -0700)]
disable JitOptimizationSensitive check on NativeAOT testing (#86523)
Stephen Toub [Mon, 22 May 2023 14:34:54 +0000 (10:34 -0400)]
Lots of tweaks to FrozenDictionary/Set analysis and hashing (#86293)
- Some minor internal refactorings, e.g. changed the internal KeyAnalyzer.Analyze from being void returning with an out results to instead just return those results, and did some internal renaming.
- We were previously computing the min/max string length in KeyAnalyzer.Analyze twice, once in UseSubstring and then again after that in Analyze. We were also computing it in the length buckets analyzer. Changed the code to do the checks just once at the beginning and then pass that information around to everywhere that needs it.
- Once that information is passed in, AnalysisResults can then be made readonly, rather than being mutated to store the min/max after it was constructed.
- If the min length is 0, there's no point calling TryUseSubstring as it'll never find one, so skip the call to it entirely if there's an empty string in the mix.
- In TryUseSubstring, for a given substring length first check all the left justifications and then check all the right justifications, rather than intermingling them. Left justifications are a bit cheaper to look up, plus we can avoid creating any objects related to checking right justification if we end up finding a left justification for a given substring length first.
- We can also avoid doing any right justification checks if all of the input strings are of the same length, as at that point there's nothing right justification could find that left justification couldn't.
- When constructing the HashSets for evaluating uniqueness, when targeting more recent .NET versions we can presize the HashSets to avoid the expansion as we add all the items.
- Importantly, set a limit on the maximum length of substring we'll consider. This significantly curtails the worst-case analysis performance for very large inputs that don't yield any unique substrings. While it's certainly possible to construct cases where this will then fail to find a substring when it otherwise could, it's much more rare, and the longer we're dealing with the less we're saving on the hashing costs, which is the only thing this is avoiding.
- For hashing, create dedicated branchless hash checks for each of lengths 0 to 4.
- When we're doing the ASCII check to see if we can use an ASCII-optimized comparer, if OrdinalIgnoreCase was used, we can also check to see if the substring contains any ASCII letters; if the only the thing the substrings contain are ASCII non-letters, then we can switch to being case-sensitive, as no casing will impact the comparisons.
- The ASCII check downlevel was erroneously including 0x7f as being non-ASCII. Fixed it to be 0x80 instead of 0x7f.
- Changed GetHashCodeOrdinalIgnoreCase to delegate to GetHashCodeOrdinal after doing its non-ASCII casing work.
- On .NET 6+, we can use GetValueRefOrAddDefault to avoid some dictionary lookups.
- For the length-bucketing implementation, we can do a quick up-front check to rule out applicability of many inputs where we know just based on the number of input strings and the min/max lengths whether some bucket will be forced to be too big.
dotnet-maestro[bot] [Mon, 22 May 2023 14:32:36 +0000 (09:32 -0500)]
[main] Update dependencies from dotnet/roslyn-analyzers (#86384)
* Update dependencies from https://github.com/dotnet/roslyn-analyzers build
20230516.2
Microsoft.CodeAnalysis.Analyzers , Microsoft.CodeAnalysis.NetAnalyzers
From Version 3.3.5-beta1.23262.2 -> To Version 3.3.5-beta1.23266.2
* Update dependencies from https://github.com/dotnet/roslyn-analyzers build
20230518.1
Microsoft.CodeAnalysis.Analyzers , Microsoft.CodeAnalysis.NetAnalyzers
From Version 3.3.5-beta1.23262.2 -> To Version 3.3.5-beta1.23268.1
* Update dependencies from https://github.com/dotnet/roslyn-analyzers build
20230519.1
Microsoft.CodeAnalysis.Analyzers , Microsoft.CodeAnalysis.NetAnalyzers
From Version 3.3.5-beta1.23262.2 -> To Version 3.3.5-beta1.23269.1
* Update dependencies from https://github.com/dotnet/roslyn-analyzers build
20230520.2
Microsoft.CodeAnalysis.Analyzers , Microsoft.CodeAnalysis.NetAnalyzers
From Version 3.3.5-beta1.23262.2 -> To Version 3.3.5-beta1.23270.2
---------
Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
dotnet-maestro[bot] [Mon, 22 May 2023 14:31:46 +0000 (09:31 -0500)]
[main] Update dependencies from dotnet/roslyn (#86497)
* Update dependencies from https://github.com/dotnet/roslyn build
20230518.11
Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
From Version 4.7.0-1.23259.11 -> To Version 4.7.0-1.23268.11
* Update dependencies from https://github.com/dotnet/roslyn build
20230519.2
Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
From Version 4.7.0-1.23259.11 -> To Version 4.7.0-1.23269.2
---------
Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
Peter Sollich [Mon, 22 May 2023 13:59:36 +0000 (15:59 +0200)]
This fixes an issue with gradual decommit in scenarios where we have … (#84975)
This fixes an issue with gradual decommit in scenarios where we have almost exclusively background GCs.
The problem is that we turn off the gradual_decommit_in_progress_p flag and rely on the distribute_free_regions flag to turn it back on. If we only trigger a BGC however, distribute_free_regions won't get called and the flag will stay off, despite the global_regions_to_decommit list being non-empty.
This will cause regions to accumulate in the global_regions_to_decommit list, and eventual will cause the process to run out of memory.
The fix is to just test for the no gc region case in decommit_step. This will also turn off the gradual_decommit_in_progress_p flag for the no gc region case, but not for background GCs.
Alexander Köplinger [Mon, 22 May 2023 13:56:02 +0000 (15:56 +0200)]
Use new queue for iOS arm64 device tests (#86575)
Vlad-Alexandru Ionescu [Mon, 22 May 2023 10:35:01 +0000 (12:35 +0200)]
[mono][aot] Enabling AOT wrappers for virtual delegates (#85643)
* Adding AOT virtual delegates wrappers
Signed-off-by: Vlad - Alexandru Ionescu <vlad-alexandruionescu@Vlads-MacBook-Pro-5.local>
---------
Signed-off-by: Vlad - Alexandru Ionescu <vlad-alexandruionescu@Vlads-MacBook-Pro-5.local>
Co-authored-by: Vlad - Alexandru Ionescu <vlad-alexandruionescu@Vlads-MacBook-Pro-5.local>
Stephen Toub [Mon, 22 May 2023 10:19:30 +0000 (06:19 -0400)]
Fix code coverage of corelib by updating coverlet (#86566)
Ilona Tomkowicz [Mon, 22 May 2023 10:17:59 +0000 (12:17 +0200)]
[wasm] Refactor name convention (#86446)
* Change case type restricted to threading files.
* Feedback - preserve consistent naming.
* Fix
* @pavelsavara's fix for samples on Windows (not connected with refactoring).
dotnet-maestro[bot] [Mon, 22 May 2023 07:40:07 +0000 (09:40 +0200)]
[main] Update dependencies from 7 repositories (#85913)
Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
Co-authored-by: Elinor Fung <elfung@microsoft.com>
Jiri Cincura ↹ [Mon, 22 May 2023 06:29:00 +0000 (08:29 +0200)]
Increase timeouts on perf-slow pipeline (because of current load on queues). (#86547)
Michal Strehovský [Mon, 22 May 2023 02:06:57 +0000 (11:06 +0900)]
Baseline a couple more JIT tests (#86429)
Stephen Toub [Mon, 22 May 2023 01:55:04 +0000 (21:55 -0400)]
Add DebuggerDisplay for SearchValues (#86559)
Michal Strehovský [Sun, 21 May 2023 23:53:52 +0000 (08:53 +0900)]
Make merged test wrappers single file friendly (#86428)
Assembly.Location returns an empty string in single file.
Egor Bogatov [Sun, 21 May 2023 20:41:51 +0000 (22:41 +0200)]
Use VEX xorps to zero a zmm reg (#86555)
Co-authored-by: Tanner Gooding <tagoo@outlook.com>
Egor Bogatov [Sun, 21 May 2023 16:09:07 +0000 (18:09 +0200)]
Remove ZMM_RECOMMENDED_THRESHOLD workaround (#86552)
Michal Strehovský [Sun, 21 May 2023 03:58:09 +0000 (12:58 +0900)]
Stop using LocalizationResources within framework (#86430)
Fixes #81459.
I moved the messages to `EventAttribute.Message` which I believe should be equivalent, but I would appreciate if someone could double-check.
I also removed all the now-unused resource strings. I believe they all need to be prefixed by one of well-known prefixes. I would appreciate if someone could double-check.
Andrew Au [Sun, 21 May 2023 01:46:51 +0000 (18:46 -0700)]
NoGCRegion Callback (#82045)
Co-authored-by: Jan Kotas <jkotas@microsoft.com>
Tomas Weinfurt [Sat, 20 May 2023 21:58:12 +0000 (14:58 -0700)]
avoid GCHandle in ConnectEx and SendTo on Windows (#86524)
* avoid GCHandle in ConnectEx and SendTo on Windows
* s/Span/ReadOnlySpan/ndd
Marc Gravell [Sat, 20 May 2023 21:36:24 +0000 (22:36 +0100)]
PipeReader.CopyToAsync(Stream) should not perform zero-length writes (#85768)
* add failing test for zero-length write
* simplify test
* apply fix
hrrrrustic [Sat, 20 May 2023 21:35:42 +0000 (00:35 +0300)]
Add new status code (#85841)
Egor Bogatov [Sat, 20 May 2023 19:49:01 +0000 (21:49 +0200)]
Clean up GenTreeCall's inline info (#86540)
Maksim Golev [Sat, 20 May 2023 17:51:37 +0000 (21:51 +0400)]
fix(66229): Fix invalid processing of empty Dictionary node. (#86485)
Co-authored-by: Maksim Golev <mgolev@htc-cs.ru>
Qiao Pengcheng [Sat, 20 May 2023 15:57:43 +0000 (23:57 +0800)]
[LoongArch64] delete the rsGetRsvdReg() for LA64. (#86319)
* [LoongArch64] delete the rsGetRsvdReg() for LA64.
* delete dead code for CR.
Katelyn Gadd [Sat, 20 May 2023 15:53:58 +0000 (08:53 -0700)]
[interp] Remove char element type support from transform-simd (#86534)
Remove char element type support from transform-simd
Also disable unusual types for packedsimd
Justin Anderson [Sat, 20 May 2023 07:45:57 +0000 (00:45 -0700)]
Allow using active host context in hostfxr_get_runtime_delegate (#86460)
These changes allow for passing `nullptr` for the handle parameter of the `hostfxr_get_runtime_delegate` function. This allows in-process callers who do not have the real host context handle to invoke this function and get a runtime delegate for the primary runtime instance. Calling hostfxr_get_runtime_delegate without the handle will only work if there is an active host context, otherwise it will return StatusCode::HostInvalidState.
Jeremy Koritzinsky [Sat, 20 May 2023 06:43:26 +0000 (23:43 -0700)]
Remove references to P/Invokes and DllImport from resource strings. (#85178)
Andrew Au [Sat, 20 May 2023 03:31:02 +0000 (20:31 -0700)]
Make the RefreshMemoryLimit API public (#85549)
Tanner Gooding [Sat, 20 May 2023 01:59:50 +0000 (18:59 -0700)]
Ensure Vector2/3/4, Quaternion, and Plane don't have a false dependency on Vector<T> (#86481)
* Ensure Vector2/3/4, Quaternion, and Plane don't have a false dependency on Vector<T>
* Apply JIT formatting patch
* Fixing a build issue
* Handle an SPMI assert
Jeremy Koritzinsky [Sat, 20 May 2023 01:59:29 +0000 (18:59 -0700)]
Add marshallers for GeneratedComInterface-based types (#86177)
Jeremy Koritzinsky [Fri, 19 May 2023 22:55:27 +0000 (15:55 -0700)]
Refactor our linear collection marshalling implementation (#86519)
Co-authored-by: Jackson Schuster <36744439+jtschuster@users.noreply.github.com>
Tanner Gooding [Fri, 19 May 2023 21:46:56 +0000 (14:46 -0700)]
Ensure we use bytes, not bits (#86510)
Sven Boemer [Fri, 19 May 2023 20:41:26 +0000 (13:41 -0700)]
Fix marking of saved assemblies (#86474)
This ensures the fix from
https://github.com/dotnet/runtime/pull/82197 works also for
assemblies which get the `save` action (which isn't supported
from the command-line, but may happen in custom steps).
It also ensures that `save` assemblies get fully marked
regardless of `DisableMarkingOfCopyAssemblies`.
---------
Co-authored-by: Jackson Schuster <36744439+jtschuster@users.noreply.github.com>
Egor Bogatov [Fri, 19 May 2023 20:38:33 +0000 (22:38 +0200)]
Remove fgMorphCastedBitwiseOp (#86491)
Ankit Jain [Fri, 19 May 2023 20:36:54 +0000 (16:36 -0400)]
[wasm] CI: Fix firefox debugger tests failure (#86476)
Tests are failing with:
```
Failed DebuggerTests.ArrayTests.InspectGenericValueTypeArrayLocals(line: 89, col: 8, method_name: "DebuggerTests.ArrayTestsClass.GenericValueTypeLoca"..., test_prev_frame: False, frame_idx: 0, use_cfo: True) [>
Error Message:
System.InvalidCastException : Cannot cast Newtonsoft.Json.Linq.JObject to Newtonsoft.Json.Linq.JToken.
Stack Trace:
at Newtonsoft.Json.Linq.Extensions.Convert[T,U](T token)
at DebuggerTests.FirefoxInspectorClient.HandleMessage(String msg, CancellationToken token) in /_/src/mono/wasm/debugger/DebuggerTestSuite/FirefoxInspectorClient.cs:line 184
at Microsoft.WebAssembly.Diagnostics.RunLoop.RunActualAsync(CancellationTokenSource x) in /_/src/mono/wasm/debugger/BrowserDebugProxy/Common/RunLoop.cs:line 142
at Microsoft.WebAssembly.Diagnostics.RunLoop.<>c__DisplayClass20_0.<<RunAsync>b__0>d.MoveNext() in /_/src/mono/wasm/debugger/BrowserDebugProxy/Common/RunLoop.cs:line 54
--- End of stack trace from previous location ---
at System.Threading.Channels.AsyncOperation`1.GetResult(Int16 token)
at System.Threading.Tasks.ValueTask`1.ValueTaskSourceAsTask.<>c.<.cctor>b__4_0(Object state)
```
One of the arguments of the message is a `JObject`, and thus fails with
`.Value<string>()`.
The full message:
```json
{
"from" : "server1.conn0.windowGlobal10737418242/windowGlobalTarget2",
"resources" : [
{
"message" : {
"arguments" : [
"mono_wasm_debug_event_raised:
aef14bca-5519-4dfe-b35a-
f867abc123ae",
{
"actor" : "server1.conn0.windowGlobal10737418242/longstractor35",
"initial" : "{\"eventName\":\"AssemblyLoaded\",\"assembly_name\":\"System.Private.CoreLib.dll\",\"assembly_b64\":\"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",
"length" : 7737462,
"type" : "longString"
},
"{}"
],
"columnNumber" : 12734,
"filename" : "http://localhost:9400/dotnet.runtime.js",
"innerWindowID" :
10737418242,
"level" : "debug",
"lineNumber" : 3,
"private" : true,
"sourceId" : null,
"timeStamp" :
1684452298362.82
},
"resourceType" : "console-message"
}
],
"type" : "resource-available-form"
}
```
Andrew Au [Fri, 19 May 2023 20:34:06 +0000 (13:34 -0700)]
We should not establish a heap hard limit for 32 bit platforms (#85844)
Adeel Mujahid [Fri, 19 May 2023 20:24:48 +0000 (23:24 +0300)]
Upgrade gcc CI to use gcc13 (#86051)
* Upgrade gcc CI to use gcc13
* Fix a Wstrict-prototypes
Jakob Botsch Nielsen [Fri, 19 May 2023 20:07:48 +0000 (22:07 +0200)]
JIT: Add a pass of liveness for the new locals inside physical promotion (#86043)
Used when decomposing stores to skip the fields that are dying and the remainder if it is dying. Also allows forward sub to kick in for the inserted replacement locals.
Morten Larsen [Fri, 19 May 2023 19:32:43 +0000 (21:32 +0200)]
Fix async XsdValidatingReader to not use synchronious IO (#84888)
* Fix async XsdValidatingReader to use GetValueAsync
* Remove async on ProcessReaderEventAsync
* Change to local static functions
Katelyn Gadd [Fri, 19 May 2023 18:41:46 +0000 (11:41 -0700)]
[wasm] Jiterpreter v128 misc optimizations (#86506)
* Use v128 opcodes when unrolling memsets/memmoves
* Introduce builder v128_const method
* v8 doesn't optimize splats so use the enormous encoding for v128 zero
* Fix fast memset for nonzero values
Jeremy Koritzinsky [Fri, 19 May 2023 17:48:28 +0000 (10:48 -0700)]
Upgrade build to use .NET 8 Preview 4 SDK (#86408)
Sven Boemer [Fri, 19 May 2023 17:25:45 +0000 (10:25 -0700)]
Fix mapping of redeclared type parameters (#86353)
We were skipping the type parameter mapping of compiler-generated
types whose only generic parameters were those implicitly created
for the declaring type's type parameters.
For the testcase in question, the nested state machine inherited
a generic parameter from the display class. This was causing
unnecessary warnings in a field assignment that assigned
`this` (an instance of the display class) to a field on the state
machine. In IL, that assignment references a field type like
`DisplayClass<T>` where `T` is the generic parameter on the state
machine. Here we were properly mapping type parameters of the
display class back to the annotated enclosing method's type
parameters, so we could tell that the "target" required
`PublicMethods`. But the substituted `T` from the state machine
was not mapped, causing the mismatch.
Kevin Jones [Fri, 19 May 2023 17:25:05 +0000 (13:25 -0400)]
Remove invalid assert in EvpPKeyDuplicate
EvpPKeyDuplicate has an invalid assert in it where it asserts that the handle it was passed is not IsInvalid. This may not be true in a multi-threaded scenario where an object is disposed by one thread while another thread is trying to use the resource.
Since we are marshaling the handle as a SafeHandle, just let it do the right thing of throwing an ObjectDisposedException when the marshaller attempts to add a reference to the handle.
Lakshan Fernando [Fri, 19 May 2023 16:33:25 +0000 (09:33 -0700)]
Enable event pipe tests in NativeAOT CI run (#86389)
* Test check on EventPipe NativeAOT
* exclude more tracing tests
* FB
Viktor Hofer [Fri, 19 May 2023 15:56:23 +0000 (17:56 +0200)]
Use net6.0 SqlClient TFM under source-build (#84782)
See discussion in https://github.com/dotnet/source-build-reference-packages/pull/633 for more context.
Ilona Tomkowicz [Fri, 19 May 2023 15:32:16 +0000 (17:32 +0200)]
[mono][debugger][docs] How we run debugger tests (#86495)
* Small fix.
Zoltan Varga [Fri, 19 May 2023 15:30:42 +0000 (08:30 -0700)]
[mono][jit] Enable Vector64 support on arm64 in jit mode. (#86188)
Vector64 is handled the same as the other SIMD types, only loads and stores
need to special case it.
Marek Fišera [Fri, 19 May 2023 14:49:23 +0000 (16:49 +0200)]
[browser] Update Wasm SDK pack version in workload manifest (#86382)
Blake Hensley [Fri, 19 May 2023 14:36:59 +0000 (07:36 -0700)]
Add ProcessStartInfo constructor that accepts IEnumerable<string> arguments (#86346)
Andy Ayers [Fri, 19 May 2023 13:23:34 +0000 (06:23 -0700)]
enable TieredPGO by default (#86225)
Katelyn Gadd [Fri, 19 May 2023 13:21:00 +0000 (06:21 -0700)]
[wasm] Generate jiterp feature detect modules dynamically (#86401)
Fixes dotnet/source-build#3442
Adam Sitnik [Fri, 19 May 2023 10:30:38 +0000 (12:30 +0200)]
remove [DebuggerDisplayAttribute] from ref assemblies (#86443)
Adam Sitnik [Fri, 19 May 2023 09:59:10 +0000 (11:59 +0200)]
FileConfigurationProvider.Dispose should dispose FileProvider when it ows it (#86455)
fixes #86146
Milos Kotlar [Fri, 19 May 2023 09:14:16 +0000 (11:14 +0200)]
Enable runtime tests on the Interpreter (#86375)
dotnet bot [Fri, 19 May 2023 09:03:59 +0000 (02:03 -0700)]
Localized file check-in by OneLocBuild Task: Build definition ID 679: Build ID 2183857 (#86468)
* Localized file check-in by OneLocBuild Task: Build definition ID 679: Build ID 2183518
* Localized file check-in by OneLocBuild Task: Build definition ID 679: Build ID 2183518
Tanner Gooding [Fri, 19 May 2023 04:22:12 +0000 (21:22 -0700)]
Remove getMaxIntrinsicSIMDVectorLength from the JIT/EE interface (#86479)
* Fixing a couple small typos
* Remove getMaxIntrinsicSIMDVectorLength from the JIT/EE interface
* Update src/coreclr/vm/methodtablebuilder.cpp
---------
Co-authored-by: Jan Kotas <jkotas@microsoft.com>
Katelyn Gadd [Fri, 19 May 2023 01:52:25 +0000 (18:52 -0700)]
[interp] Undo damage to the simd methods bsearch list; support vectors of char (#86475)
While inspecting the performance of some vectorized BCL code, I noticed that I accidentally damaged one of the binary search lists.
There also appear to be some cases where vectors of char are used, so I touched up the code to make sure that we will select the right intrinsic in that case if we already supported U2.
Jeremy Koritzinsky [Fri, 19 May 2023 01:48:30 +0000 (18:48 -0700)]
Move the GCC and CLR Tools Unit Tests legs to use the global build template (#86411)
* Move the GCC validation leg to use the global build job template instead of the CoreCLR build template.
* Move CLR Tools Unit Tests leg to use the global build job template instead of the CoreCLR build job template
* Add iltools for the CoreCLR tools unit tests job
* Pass -gcc flag to test build in GCC job.
* Update eng/pipelines/coreclr/templates/build-job.yml
James Newton-King [Fri, 19 May 2023 00:51:11 +0000 (08:51 +0800)]
Add debugging attributes to System.Security.Claims types (#86424)
* Add debugging attributes to System.Security.Claims types
* Update
* Update ref, add type proxies
* PR feedback
* PR feedback
Katelyn Gadd [Fri, 19 May 2023 00:30:22 +0000 (17:30 -0700)]
[wasm] Implement I2 and I4 shuffles in the jiterpreter (#86469)
Implement a wasm lowering for V128_I2_SHUFFLE and V128_I4_SHUFFLE in the jiterpreter so BCL char operations won't be terribly slow
Jan Kotas [Thu, 18 May 2023 22:59:52 +0000 (15:59 -0700)]
Exclude System.* reference assemblies in ILCompiler.Build.Tasks (#86423)
* Exclude System.* reference assemblies in ILCompiler.Build.Tasks
We expect the implementation of these assemblies to come as part of msbuild.
Fixes #83695
* Do not include any System.* assemblies as part of the task
* Delete custom resolver
Elinor Fung [Thu, 18 May 2023 22:35:07 +0000 (00:35 +0200)]
Switch RhConfig from TCHAR to char (#86393)
- Hide away the details of reading the environment variable as `wchar` vs `char`
- Keep embedded variables as as ASCII (they are embedded as such already)
- Stop using `TCHAR` for the `RhConfig` APIs, which seemed to just be more confusing to consume
Katelyn Gadd [Thu, 18 May 2023 22:28:09 +0000 (15:28 -0700)]
[wasm] Enable v128 and PackedSimd in the interpreter (#85920)
Fix the create scalar intrinsics in the jiterpreter
Fix trimming removing SIMD even if SIMD was enabled
Add wasm opcode mapping for extract msb
Fix some SN_ lists in transform-simd being ordered incorrectly
Due to mixed mode it is not possible to control interp SIMD with a runtime option, so remove the options
Disable linker substitutions for SIMD unless AOT is active, so the interp is responsible for the properties
Generate a MINT_NIY for unimplemented PackedSimd methods
Katelyn Gadd [Thu, 18 May 2023 22:20:58 +0000 (15:20 -0700)]
[wasm] Update jiterpreter documentation (#86461)
* Add missing line breaks
* Update jiterpreter documentation with a sample annotated trace
* I love github flavored markdown
Elinor Fung [Thu, 18 May 2023 22:13:59 +0000 (00:13 +0200)]
Make host look for RID-specific assets via a known list by default (#84100)
Miha Zupan [Thu, 18 May 2023 22:05:17 +0000 (00:05 +0200)]
Remove workaround for fixed Roslyn bug (#86464)
Tanner Gooding [Thu, 18 May 2023 21:25:00 +0000 (14:25 -0700)]
Add an undocumented switch to allow controlling the preferred vector width emitted implicitly by the JIT (#86457)
* Add an undocumented switch to allow controlling the preferred vector width emitted implicitly by the JIT
* Resolving issues and responding to PR feedback
* Simplifying the xarch cpu info check
Stephen Toub [Thu, 18 May 2023 19:10:13 +0000 (15:10 -0400)]
Remove noisy assert from Task (#86445)
Bruce Forstall [Thu, 18 May 2023 19:05:56 +0000 (12:05 -0700)]
Remove call to morph on hoisted trees (#86450)
It is unnecessary (this causes no diffs). And theoretically if it
did cause diffs, the morphed tree would then not match the tree
being hoisted, meaning CSE wouldn't "find" it and consider it a CSE.