• 0 Posts
  • 13 Comments
Joined 2 年前
cake
Cake day: 2023年7月3日

help-circle

  • For me what generally happens if I stop at 9PM, I will work through the problem in my sleep (and it will prevent me from getting a good night sleep), but I will often find a breakthrough the next morning during shower time.

    I’m talking about those hard, multi-days debugging problems that nobody can figure out, but as someone else raised, that’s why I get paid good money for it.

    It still sucks though. That first response in the thread rings so true, ok now I get it, no you don’t…


  • it’s not an optimization if you don’t measure an improvement.

    This, so much. I often see this at work, theories about this and that being slow, or how something new « should » be better. As the resident code elder I often get to reply « why don’t you measure and find out »…

    For some reason it seems nobody uses sampling profilers anymore, those tell you exactly where the time is spent, you don’t have to guess from a trace. Optimizing code is kind of a lost art.

    Sometimes I wish compilers were better at data cache optimization, I work with C++ all the time but all of it needs to be done by hand (for example SoA vs AoS…). On the upside, I kind of have a good job security, at least until I retire.


  • I am not convinced bots would fill the list with hypothetical purchases, I don’t think scalpers are interested in waiting or having money tied up in backorders.

    The point is to eliminate the scalper advantage by ensuring one can buy the product « at some point ». If you need it by Christmas or whatever then you are kind of screwed.

    I remember for the SteamDeck OLED, stock was enabled in waves over at least a month, so even though the first batch was sold out in minutes, there was no rush to refresh the store page to try and finish the transaction before it ran out. This is in direct contrast to (say) the PS5 which sold out in minutes then still wasn’t available anywhere over a year after it launched.

    I don’t really understand how Valve solved the problem, it should have followed the same pattern of being sold out in minutes then scalpers would be the only option for months, but interestingly that’s not what happened.


  • It seems everyone forgot that not too long ago, you could back order something and the store would call/ship whenever they got it.

    Just let me give you my credit card number in exchange for a spot in the waiting list, then the scalpers lose and I get my new launch thing whenever they get around to it. But no, that would be too simple, gotta get the crowd riled up and race for the available units!

    I suppose this could be abused like everything else but it wouldn’t be worse than what we have now with fucking scalpers buying up the little stock that trickles in via automated bots.

    It’s not about getting your fix sooner for the new shiny, sometimes you really need a new GPU to replace the one you’ve had for 5 years! Why should you settle for the previous generation if the new one just came out and you are willing to pay launch MSRP for that privilege (not 2-3x MSRP for scalpers!!).





  • That is my recollection as well. Also I remember Psycho Mantis would make some comments about the content of your memory card before the fight.

    Another favorite of mine is The End in MGS3: Snake Eater, he is a sniper hiding in the woods and it takes like an hour of game time to beat him down. I seem to recall if you save during the fight and change the date forward a few years on your PS3, when you reload he has died of old age waiting for you to come back.



  • Vulkan and DirectX could already share shaders, because the input for both was already HLSL. The difference is the intermediate representation of the compiled shaders that will now be the same in the future (SPIR-V for both).

    The real winners here are driver programmers at NVIDIA/AMD/Intel, since they will no longer have to develop support for both DXIL and SPIR-V (which are similar in concept but different in implementation). How much of that will be true in practice remains to be seen, but I am hopeful.

    There are tools to analyze, process and transform SPIR-V bytecode already, presumably those will work for DX12 shader model 7 too. It might make performance analysis easier, same with debugging via a tool like RenderDoc that supports SPIR-V but not DXIL.

    As for the overhead of DirectX, with DX12 this is largely not true anymore, both are high performance APIs with comparable overhead (i.e. as little as possible).


  • I remember your previous post, congrats on not giving up.

    Whipping up a script to solve a very specific problem is super satisfying, but I found that anything you write quickly becomes a liability. Debugging Perl can be super difficult, especially when returning to something you wrote a while back.

    Personally I grew tired of the punishment and left it all behind! If I need a quick script I’ll use Python instead, and if it doesn’t work I can use a real debugger to fix it.

    In any case it’s always fun learning new things, I hope this experience ends up being useful to you in the future and you get to easily solve a problem that stumps everyone else involved.

    Cheers!