Re: [fpc-pascal] "Is nested" vs "of object" compatibility with global functions
> On May 29, 2022, at 2:26 PM, Martin Frb via fpc-pascal > wrote: > > Actually afaik you push an extra param in for "is nested" as well as for "of > object. > > But for "of object" it is the first param, for "is nested" it is the last. So you can cast nested or object to each other and the compiler doesn’t complain the parameters have bad values so this confirms what you’re saying. Anyways the lesson learned today playing with these is that we still shouldn’t just make all our function pointers “reference to” because each time you add one of these anonymous functions in a function it will add some setup code at runtime, even if it’s not called. If you’re doing something realtime the penalty will add up fast and be a real problem if you’re not careful. Luckily anonymous functions are available on the other types so we can use those without incurring runtime penalties but we still can’t write caller-agnostic functions which have no runtime penalty and accept global, nested or object function pointers. Regards, Ryan Joseph ___ fpc-pascal maillist - fpc-pascal@lists.freepascal.org https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-pascal
Re: [fpc-pascal] "Is nested" vs "of object" compatibility with global functions
> On May 29, 2022, at 2:26 PM, Martin Frb via fpc-pascal > wrote: > > Actually afaik you push an extra param in for "is nested" as well as for "of > object. > > But for "of object" it is the first param, for "is nested" it is the last. Then “references to” is the only universal function pointer type now (which is great) but since it’s a heap constructed class it’s about 4 times slower than the other ones. I still think FPC needs a universal stack based function pointer type which works with objects, nested and global functions. Michael hinted to how that could be done but it sounds like this all legacy baggage from Delphi. Ideally the programmer only has to consider “does it leave the scope or not” and not have to worry about these other 2 is nested/of object types. I can’t even think of any other language that has so many function pointer types which are visible to the programmer (even c++???). Regards, Ryan Joseph ___ fpc-pascal maillist - fpc-pascal@lists.freepascal.org https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-pascal
Re: [fpc-pascal] "Is nested" vs "of object" compatibility with global functions
On 29/05/2022 07:21, Michael Van Canneyt via fpc-pascal wrote: On Sun, 29 May 2022, Hairy Pixels via fpc-pascal wrote: I’ve been testing out all the different function pointer types in FPC to test their compatibility with each other and I noticed that “is nested” can accept a global function declaration but “of object” can not. What is the reason for this exactly? I wouldn’t expect nested function types to accept global functions but since they do I wonder why “of object” is different. Because you're pushing a parameter which does not exist in the case of a global function. Actually afaik you push an extra param in for "is nested" as well as for "of object. But for "of object" it is the first param, for "is nested" it is the last. procedure TFoo.Bar(self: TBar; a: integer); procedure NestedBar(a: integer; parentFp: pointer); So in the code at the end of the mail, with a global function assigned to "p", it calls p(1, SomeValueForParentFp). But the global "procedure f" just ignores the extra param. If it was "of object", the "self" value would be first, and it would be in place of the parameter "a". program Foo; type TProc = procedure(a:integer) is nested; var p: TProc; procedure f(a:integer); begin end; begin p:= @f; p(1); end. ___ fpc-pascal maillist - fpc-pascal@lists.freepascal.org https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-pascal
Re: [fpc-pascal] "Is nested" vs "of object" compatibility with global functions
On Sun, 29 May 2022, Hairy Pixels via fpc-pascal wrote: I’ve been testing out all the different function pointer types in FPC to test their compatibility with each other and I noticed that “is nested” can accept a global function declaration but “of object” can not. What is the reason for this exactly? I wouldn’t expect nested function types to accept global functions but since they do I wonder why “of object” is different. Because you're pushing a parameter which does not exist in the case of a global function. I think they both have a hidden first parameter (like self) and nested types simply ignore this for global functions so I would think objects types could do the same. I cannot comment on the 'is nested', I don't know what the extra field is for. I assume a parent stack frame pointer, which can perfectly be nil, presumably. A method pointer is actually 2 fields, both pointers: the data (self) and the address (location of method). Because you cannot distinguish between 'self = nil' and 'there is no self' when actually calling the method, it's not possible to assign a global method: you would be pushing a 'self' parameter to a procedure that is not expecting a 'self' parameter, wreaking havoc on the stack. Theoretically one could add a third field, or maybe encode a special value to distinguish between the two, but then you would need to add some logic at every call of an 'of object' method to determine what is needed. That would slow things down, and would be not be Delphi compatible. There is plenty of code out there that assumes the current mechanism to 'just work' Similarly, you could let all global methods accept a 'self' pointer which is simply Nil, but same argument applies: it slows things down (extra parameter to handle) and is not TP nor Delphi compatible. When starting from zero, one could probably make all 3 methods (plain procedure, 'of object' and 'is nested' compatible - maybe even 'reference to', but for historical reasons it is not possible. Michael.___ fpc-pascal maillist - fpc-pascal@lists.freepascal.org https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-pascal
