Just saw on Facebook that Eli Boling has left the compiler team. I'm happy for Eli, but not pleased by the timing. Doesn't bode well.

Comments

  1. just don't know its a bad news or good ones for all of us - in meaning of current compilers quality...

    ReplyDelete
  2. Maybe that depends on how many other people know how the compiler works...

    ReplyDelete
  3. I wonder how hard it is to recruit people who enjoy working with compilers? I guess it is not the world's biggest niche...

    ReplyDelete
  4. Eli Boling -- he had been at EMBT for about five years.  He was with Borland back in the day as well.  He is a very, very, very, very good compiler developer, and he is a big loss to the team. 

    As I said, the timing is.........interesting.

    ReplyDelete
  5. It could easily just be coincidence - people do move companies all the time. As for working on compilers, personally I suspect it's a small niche, but one very much in demand. It's about as interesting work as it's possible to get, I'd think.

    ReplyDelete
  6. Let's slow down with the speculation. No-one mentioned Linux in any way.

    ReplyDelete
  7. Sometimes one wishes to bring the band back together.

    ReplyDelete
  8. the new owners do have some strange idealogical ideas about people from certain countries it has turned out..just ask Babok...

    ReplyDelete
  9. David Millington , If I'm not wrong Eli was the compiler guy behind Kylix compiler, also when he back to work at EMB he was responsible on the MacOS compiler, so it's logical he will be involved in the new Linux compiler.
    Nick Hodges Please correct me if said anything wrong, you were there and know more than us.

    ReplyDelete
  10. David Millington When I was there, he worked on the MacOS stuff, yes.  I believe you are correct about Kylix as well.

    Brian Hamilton Not sure about what your point is regarding Babok.

    ReplyDelete
  11. Javier Hernández You are a veritable ray of sunshine today... Who peed in your Wheaties?

    ReplyDelete
  12. Lars Fosdal People bang on GCC, FPC, etc. for FREE, so I don't think it's that hard.

    ReplyDelete
  13. Joseph Mitzen  but these people don't need/don't want work on closed source "walking-dead" compiler. IMO today's Delphi compiler quality is much lower than FPC => Embarcadero/Idera should invest time into FPC, they can develop only Delphi mode without permission whole community. That would be a win/win solution with big gain for all of us.

    ReplyDelete
  14. Maciej Izak Why do you want the language feature set to go backwards? Last time I looked FPC didn't have closures, in particular, and Delphi-style generics weren't implemented fully (e.g. generic methods).

    ReplyDelete
  15. Joseph Mitzen Contributing to an open source compiler doesn't necessarily make you a good compiler developer? (Note that I am  not evaluating the FPC contributors here - just making an observation)

    ReplyDelete
  16. Chris Rolliston Sooner or later FPC will cover mentioned features. With Embarcadero/Idera is possible to get amazing synergy. More platforms, better quality and single code base (pascal code base!) for compiler. More brains engaged for compiler development. IMO Delphi commercial compiler has no sense in current market. FPC becomes strong competitor, efforts should be merged.

    ReplyDelete
  17. Maciej Izak 'Sooner or later' - well it hasn't happen yet, has it? If you're using modern language features now, it's a bit of a deal breaker if you don't care for the additional platforms FPC targets.

    ReplyDelete
  18. Maciej Izak Rather later. I've asked for anonymous methods like 3 years ago or so and got told that "someone was working on it" - but that never saw the light of the day afaik. Generics have been worked on even longer and they still only support generic types but not parameterized methods. RTTI just saw a bit of improvement because mormot needed it afair but it is still lightyears away of what Delphi can do (not just reading typeinfo but dynamic invokation of methods and all that stuff) because the low level implementation needs to be there for all the different platforms and architectures which nobody is working on so far.
    How helpful would all these features be if they would not work on half the platforms.

    ReplyDelete
  19. Stefan Glienke I'd love to see all those implemented. Can you point us at QPs? The reason is so that we know they're in the system and can point people at specific items when there's an opportunity to point out missing or important features (eg, if someone emails Marco/Atanas/whoever, for example.)

    ReplyDelete
  20. David Millington I was talking about FPC, you know?

    ReplyDelete
  21. Stefan Glienke I missed that. But Delphi's RTTI (for example) doesn't have all of those things, does it?

    ReplyDelete
  22. David Millington Of course it does, how do you think are things like a DI container or mocks working? :)

    ReplyDelete
  23. Yeah, ok. I'll sidle off feeling a little bit stupid now :)

    ReplyDelete
  24. Lars Fosdal I guess we could ask the same question: being an off-shored Romanian developer for Embarcadero doesn't mean you're a good compiler developer either I guess.

     With open source people judge you not by your asking salary or your degree or background, but solely on the quality of your code. 

    I know that gcc, at least on Linux, is extremely optimized and one of the most-used compilers out there. People working on FPC seem to at least have a high interest in compilers, since the language targets so many different OSes and architectures.

    ReplyDelete
  25. Maciej Izak I don't think they should normalize around FPC. The people contributing to FPC love compilers. As a consequence they have no interest in using LLVM as a backend. That leads to lots of time spent on targeting different architectures and less time spent on optimizations. I don't think Delphi needs to target Haiku or OS/2. 

    EMBT made the right decision using LLVM as the backend for their new compilers - what they need to do is introduce one compiler across all their architectures now, getting rid of the legacy desktop compiler and unifying the language differences between desktop and mobile.

    Now if FPC and EMBT adopted a single codebase that would be a good idea, but only if it were based on LLVM. To do otherwise is simply not efficient.

    ReplyDelete
  26. Chris Rolliston " If you're using modern language features now" 

    That works both ways. Delphi is in the dark ages in terms of modern language features compared to Oxygene for instance.

    The criticism's valid but Delphi is just as vulnerable to it. Neither FPC nor Delphi have focused on modernization. At one point FPC was ahead of Delphi, now Delphi is ahead of FPC. Both communities have loud contingents who argue against many forms of language modernization and seem like they'd be happy if the language didn't evolve much past Delphi 7. These are probably the (large) subset of people maintaining mature codebases vs. developing new code.

    ReplyDelete
  27. Joseph Mitzen LLVM target for FPC is developed.
    Chris Rolliston Stefan Glienke Allen Bauer Embarcadero/Idera can help with implementing missing pieces. Their current closed source model for compiler is outdated and as we can see in last days - loss of control on quality. Open Source + commercial support is very successful. They still can earning money by providing IDE, FMX and other libraries like FireDAC. They can't exterminate open source and they can't beat passionate people. They only can join and get profits (IIRC something like this was mentioned by Nick Hodges ).

    Anyway as was said - sooner or later FPC will beat Delphi compiler in mentioned fields :) Maybe later but IMO it will happen.

    ReplyDelete
  28. For now LLVM is way to slow for big projects.

    ReplyDelete
  29. Maciej Izak   No, Nick Hodges never mentioned open sourcing the compiler.

    ReplyDelete
  30. but something in that style about open source was mentioned by Nick Hodges am I right? :) where is said something about open sourcing the Delphi compiler? I mean only support for FPC project for example as Linux target.

    ReplyDelete
  31. Maciej Izak  I happen to be the world's foremost authority on what Nick Hodges has said, and I can assure you that he never has said that the Delphi compiler should be open sourced.

    ReplyDelete
  32. Nick Hodges Maybe my message was unclear.
    1. I don't want to open sourcing Delphi compiler. Only some help for FPC project.
    2. I mean your words about current open source world trend and cooperation with that trend - that can be profitable (or maybe something I misunderstood).

    Sorry :\

    ReplyDelete
  33. Maciej Izak Oh that will never happen, I assure you.  Why would they take resources and apply them to another compiler?

    ReplyDelete
  34. Nick Hodges philanthropy? :) and they can gain good quality Linux compiler with free bug fixes.

    ReplyDelete
  35. Maciej Izak Why should EMBT or Idera help or use FPC, their attitude is and has always been shit towards Delphi and EMBT. Also the passion I have seen over the past few years was mostly in arguing why not to implement something.

    ReplyDelete
  36. Stefan Glienke On the other hand I tested FPC generics this weekend and was able to do math with generic values, which I couldn't do with Delphi 10 Seattle (unless I missed something with D10). 

    But you're right that the sentiment in FPC right now seems to be on more platforms rather than more features - because the people contributing to the code love compilers and it's fun for them to target more platforms. Unless someone steps up willing to do the work to add new features, it's not likely to happen in the forseeable future. FPC used to be about blazing ahead of Delphi (64bit, generics, cross-platform), then it became about Delphi 7 compatibility, now it seems to be about targeting as many obscure architectures as possible. There really isn't a unifying vision/direction for it right now.

    ReplyDelete
  37. Jeroen Wiert Pluimers How do you mean "too slow"? It seems development of languages is too slow if one doesn't leverage all of the work (especially optimizations!) already being put into LLVM. That would seem to be why Embarcadero itself used it for its new mobile compilers. 

    Or do you mean the speed of the compiler? I thought LLVM was intended to be a fast compiler and already took much less time to compile than GCC.

    http://www.phoronix.com/scan.php?page=news_item&px=mtgznde

    ReplyDelete
  38. Nick Hodges No, as already established in the forums, I am the world's foremost authority on what Nick Hodges has said, and I believe the confusion here is that Nick Hodges once said that Delphi's unit tests should be open sourced.

    ReplyDelete
  39. Stefan Glienke I know you're sour on FPC, and I don't entirely disagree, but let's also remember EMBT used FPC for a commercial product and never contributed a line of code or a penny of profit back to the project. When EMBT bought AnyDAC they discontinued FPC support and Marco Cantu called FPC a competitor. I think it's fair to say that the attitude from Borland et al towards FPC has never been friendly either (they came up with their own generic syntax, etc.).

    ReplyDelete

Post a Comment