pyscripter

@Darian MillerI must add that I am testing with the beta in case it makes a difference.

@Darian Miller Please try with this version. It uses the OnmiThreadLibrary CAS function in both 32-bit and 64-bit. Works solidly here. iaStressTest.TThreadedQueue.PopItem.pas

@Anders MelanderCurrently the only way to pass the tests in 64bits is to use the modified version of the CAS function which now looks like this: function CAS(const oldData: pointer; oldReference: NativeInt; newData: pointer; newReference: NativeInt; var destination): boolean; asm {$IFNDEF CPUX64} ... {$ELSE CPUX64} .pushnv rbx mov rax, oldData mov rbx, newData mov rcx, newReference mov r8, [destination] lock cmpxchg16b [r8] {$ENDIF CPUX64} setz al end; { CAS } Your implementation of InterlockedCompareExchange128 does not pass the tests, same as Delphi's own implementation. It would be nice to come up with a working InterlockedCompareExchange128 and submit that as a separate bug report with a working solution.

SetMinimumBlockAlignment does nothing in 64bits (see the source code). Alignment is fixed at mba16Byte always.

If I replace the push and pop rbx with .pushnv rbx and remove the .noframe the code crashes. How is one supposed to use .pushnv? function CAS(const oldData: pointer; oldReference: NativeInt; newData: pointer; newReference: NativeInt; var destination): boolean; asm {$IFNDEF CPUX64} {$ELSE CPUX64} .pushnv rbx //rsp := rsp - 8 ! mov rax, oldData mov rbx, newData mov rcx, newReference mov r8, [destination + 8] //+8 with respect to .noframe lock cmpxchg16b [r8] {$ENDIF CPUX64} setz al end; { CAS }

@David HeffernanThanks. Could you please post here what the function should be?

Well there is the Delphi implementation which does not work: function InterlockedCompareExchange128(Destination: Pointer; ExchangeHigh, ExchangeLow: Int64; ComparandResult: Pointer): Bool; stdcall; asm MOV R10,RCX MOV RBX,R8 MOV RCX,RDX MOV RDX,[R9+8] MOV RAX,[R9] LOCK CMPXCHG16B [R10] SETZ AL end; and the CAS function from OmniThreadLibrary which works well. function CAS(const oldData: pointer; oldReference: NativeInt; newData: pointer; newReference: NativeInt; var destination): boolean; asm {$IFNDEF CPUX64} push edi push ebx mov ebx, newData mov ecx, newReference mov edi, destination lock cmpxchg8b qword ptr [edi] pop ebx pop edi {$ELSE CPUX64} .noframe push rbx //rsp := rsp - 8 ! mov rax, oldData mov rbx, newData mov rcx, newReference mov r8, [destination + 8] //+8 with respect to .noframe lock cmpxchg16b [r8] pop rbx {$ENDIF CPUX64} setz al end; { CAS } @David Heffernanhas said that both are wrong (see earlier comments). By the way Allen Bauer explains .PUSHNV here.

@David HeffernanI know very little about assembler and I will leave this to experts to resolve. However I do see push rbx and pop rbx in the code. Does this not save and restore rbx? This code has been in the master branch of OmniThreadLibrary for ages and should have been tested extensively. It does appear to work well we the TMonitor fix.

Success! If I use the CAS function from OmniThreadLibrary it works. So it appears that Delphi's InterlockedCompareExchange128 is broken. Passed a test with 1000 threads and 50 million calls to NewWaitObj. I am attaching the revised test code which should now work in both 32bits and 64bits. We should now try to convince Embarcadero to include the fixes in the Delphi RTL. iaStressTest.TThreadedQueue.PopItem.pas

@Kas Ob.Thanks for the info about alignment. Actually Delphi does offer support for alignment (undocumented feature see @Uwe Raabe response in this Stackoverflow question). If you change the definition of TEventStack to TEventStack = record Head: PEventItemHolder; Counter: NativeInt; procedure Push(EventItem: PEventItemHolder); function Pop: PEventItemHolder; end align {$IFDEF CPUX64}16{$ELSE CPUX64}8{$ENDIF CPUX64}; There are not longer crashes. The test still fails though and I am investigating why. And by the way. There is also the [Align(8)] attribute which works with fields, global variables and records it seems.

For reference this is from fpc. function InterlockedCompareExchange128(var Target: Int128Rec; NewValue: Int128Rec; Comperand: Int128Rec): Int128Rec; assembler; { win64: rcx ... pointer to result rdx ... target r8 ... NewValue r9 ... Comperand } {$ifdef win64} asm pushq %rbx { store result pointer for later use } pushq %rcx { load new value } movq (%r8),%rbx movq 8(%r8),%rcx { save target pointer for later use } movq %rdx,%r8 { load comperand } movq (%r9),%rax movq 8(%r9),%rdx {$ifdef oldbinutils} .byte 0xF0,0x49,0x0F,0xC7,0x08 {$else} lock cmpxchg16b (%r8) {$endif} { restore result pointer } popq %rcx { store result } movq %rax,(%rcx) movq %rdx,8(%rcx) popq %rbx end; Any chance of anyone coming up with a working InterlockedCompareExchange128 for Delphi?

Over here win 32 300 threads 10 minutes: 2020-05-17 01.01.08: StressTestPopItem Start: Waiting up to [600] seconds for PopItem to prematurely timeout. 2020-05-17 01.01.08: Note: Using [10] as PopTimeout on TThreadedQueue creation 2020-05-17 01.01.09: TThreadCreator Start: Creating up to [300] threads 2020-05-17 01.01.09: Note: Creating threads with a StackSize of [65536] 2020-05-17 01.01.09: TThreadCreator End: [300] worker threads created ... 2020-05-17 01.11.05: New Wait objects created: 14800000 2020-05-17 01.11.08: StressTestPopItem End: Overall maximum time limit reached for this test without an error detected...we will call this a success! 2020-05-17 01.11.08: Hit <enter> to exit NewWaitObj was called 15 million times. and Win32 2000 threads 30 seconds. 2020-05-17 01.30.28: StressTestPopItem Start: Waiting up to [30] seconds for PopItem to prematurely timeout. 2020-05-17 01.30.28: Note: Using [10] as PopTimeout on TThreadedQueue creation 2020-05-17 01.30.29: TThreadCreator Start: Creating up to [2000] threads 2020-05-17 01.30.29: Note: Creating threads with a StackSize of [65536] 2020-05-17 01.30.29: TThreadCreator End: [2000] worker threads created ... 2020-05-17 01.30.58: New Wait objects created: 5000000 2020-05-17 01.30.58: StressTestPopItem End: Overall maximum time limit reached for this test without an error detected...we will call this a success! 2020-05-17 01.30.58: Hit <enter> to exit NewWaitObj was called 5 million times in 30 seconds. Sometimes with many threads say more than 300 I get an early failure before all threads are created and I am not sure why. Once all the threads are created I get no errors. If you add Sleep(1000) at the top of TTestThread.Execute() to give it time for all the threads to be created these startup errors are avoided.

I am attaching a new version of the test code incorporating the suggestions of @Anders Melander in case anyone wants to try see attached iaStressTest.TThreadedQueue.PopItem that can be used with the original stress test of @Darian Miller). iaStressTest.TThreadedQueue.PopItem.pas

function InterlockedCompareExchage(var Dest: TEventStack; NewValue, CurrentValue: TEventStack): Boolean; begin {$IFDEF CPUX64} Result := InterlockedCompareExchange128(@Dest, Int64(NewValue.Head), NewValue.Counter, @CurrentValue); {$ELSE CPUX64} Result := InterlockedCompareExchange64(Int64(Dest), Int64(NewValue), Int64(CurrentValue)) = Int64(CurrentValue); {$ENDIF CPUX64} end; I have tried the above. Works well in 32 bits, but crashes in 64bits. The implementation of InterlockedCompareExchange128 in WinApi.Windows.pas is: function InterlockedCompareExchange128(Destination: Pointer; ExchangeHigh, ExchangeLow: Int64; ComparandResult: Pointer): Bool; stdcall; asm MOV R10,RCX MOV RBX,R8 MOV RCX,RDX MOV RDX,[R9+8] MOV RAX,[R9] LOCK CMPXCHG16B [R10] SETZ AL end;

Then I hope the weather will be bad 😀

If you have a look at the posted file, this is what I tried to do. Please have a go if you have the time.

Absolutely right. This is why I am passing the challenge to people like you, with much deeper knowledge than mine. I would very much hope that @Primož Gabrijelčičfor instance, has a go at providing a solution, since he has faced the very same issue in OmniThreadLibrary.

The following code program ThreadOSError; {$APPTYPE CONSOLE} uses System.SysUtils, System.Classes; begin Assert(GetLastError=0); TThread.CreateAnonymousThread(procedure begin try CheckOSError(GetLastError); except On E: Exception do WriteLn(E.Message); end; end).Start; ReadLn; end. produces the following output in 10.3,3 in Windows. System Error. Code: 87. The parameter is incorrect Same is true whichever way you run Thread code. Is this a known issue? Any idea why the OS error is raised?

Let me thank again everyone that responded. You said that the error-code is thread specific, this is not the way to check whether a specific API call failed etc., things I fully agree with, but which I knew already. My post just made an observation. Whenever you run code in a thread, GetLastError always returns 87 which corresponds to a call with invalid parameters, The original post asked two questions: Was this a known fact? More importantly why? In other words, what is the API call that sets this error? Even if it is inconsequential, and I believe it is, I had the curiosity to find out. I don't think I got an answer to these questions.

@Anders Melander @Lars FosdalSo you do not think this OS error is the result of Delphi making a call to an OS (Windows) function with invalid parameters.

@Der schöne GüntherI am not sure about the relevance of your quote. No matter what your thread code is or how you run it (by inheriting from TThread, creating an anonymous thread or whatever), something results in an OS error 87 (it is always the same code) that corresponds to "parameter is incorrect". This OS error has been raised before your thread code has started running. And this has nothing to do with the return code of a thread. This is a problem because if you do any OS stuff in your thread code, and then you want to check whether an error was raised using CheckOSError, an exception will be raised. A workaround would be to always start your thread code with the statement: SetLastError(0);

His legacy lives on through people like me using his code...

If you want type inference then add a new method: function SmartPtr<T>.Value: T; begin if FGuard <> nil then Result := FGuard.Value else Result := nil; end; Then you can write: var Bob := SmartPtr<TTalking>(TTalking.Create('Bob')).Value; slightly more elegant than having three times TTalking on the same line.

Funnily enough some of the most popular languages today, Python, JavaScript R and Ruby are single-threaded and you have to go out-of-your-way to use more than one cores.