Evening Reading

Thread Truncated. Click to see all 81 replies.

• toggle

  wow. that is, quite simply, fucking retarded if true. informative post monsieur Chryx. : PlanetPerfecto
  • toggle

    yep :( ffs, x87 AND single threaded? I'm surprised Nvidia haven't figured out that what they're act... : Toraz Chryx
    • it's a classic issue with corporate bottom-line mentality. i doubt it will ever go away. im sure plenty o... : PlanetPerfecto
• toggle

  Ouch. Way to hamstring adoption by trying to sell a few more GPUs. : DinK
  • And without adoption, PhysX just disappears. You have to get your product out there before you can use it... : Robotic Puppy Eater
• Lol so there's really no reason for anyone to bother using PhysX at this point. nVidia is intentionally f... : Robotic Puppy Eater
• It's long been speculated, nice to see someone call them on it. : Visceral Monkey
• jesus christ, there must be a lawsuit in there somewhere : filthysock
• toggle

  It always was a farce. I wonder how valcan_s is going with his dedicated 470 nvidia card just for physx? : abrasion
  • toggle

    Sounds like the software version was a farce to make the hardware solution more appealing or impressive. ... : DinK
    • toggle

      Yeah, it's killing widespread support: http://en.wikipedia.org/wiki/Physx#Title_support : daVinci1980
      • toggle

        What percentage of games is this over how long a time period? How many of these games do anything signif... : DinK
        • this. there's basically no game that needs it. Lack of PhysX is not going to stop me from getting a ch... : MamiyaOtaru
  • I hear Sniper: Ghost Warrior looks ****AWESOME**** on it. : chrisschutz
• toggle

  None of you are programmers, and it's ridiculously obvious. Writing code that uses SSE instructions re... : daVinci1980
  • toggle

    But don't all CPUs capable of running games utilizing PhysX have at least SSE1 support? Maybe SSE2? : Prozium
    • toggle

      Minimum specification is an application decision, not a middleware decision. : daVinci1980
      • toggle

        I don't think anything that can actually run nvidias drivers can be run on a CPU that doesn't have SSE su... : bostrov
        • toggle

          It doesn't matter, it's not a decision middleware providers can make for all application providers. To... : daVinci1980
          • toggle

            can't Nvidia just change their middleware spec? Just basically say, if you want to use our PhysX tech you... : Prozium
            • toggle

              Sure, they could. But most middleware vendors target the least common denominator. Increasing the min spe... : daVinci1980
              • toggle

                So they would essentially lose what, zero developers? Is there anyone out there still developing physics ... : Prozium
                • toggle

                  Blizzard has amazingly low system requirements on all of their titles. UnrealEngine3 has amazingly low sy... : daVinci1980
                  • toggle

                    Is there a reason, like I suggested below, that they couldn't just leave the current version in there as ... : DinK
                    • toggle

                      Ah, now we get into a tricky deployment question. How are these backends loaded? Calling across DLL bound... : daVinci1980
                      • toggle

                        Anything using modern OpenGL extensions is calling across a DLL boundary at an even higher frequency than... : ThereCanBeOnlyOneCat
                        • That's true for Direct3D as well, but still not a reason to add an extra few thousand or hundred thousand... : daVinci1980
  • toggle

    i thought the intel compiler gave you switches for using SSE and being compatible with other cpus : filthysock
    • It isn't quite that easy. Back when I was regularly using the intel compiler the only thing it would auto... : bostrov
    • toggle

      PhysX is distributed to developers as mostly closed source, black box libraries. They are compiled agains... : daVinci1980
      • toggle

        I think he means that nvidias themselves should be able to vectorize physx via compiler switched with lit... : bostrov
        • #2802

          Refresh Thread

          By: daVinci1980 x Show Full Post

          Reply

          It's not little difficulty. The fastpath on the CPU doesn't look like the fastpath on the gpu. They're completely different architectures.

          Jul 07, 2010 12:17am PDT
          toggle

          It's not little difficulty. The fastpath on the CPU doesn't look like the fastpath on the gpu. They're co... : daVinci1980
          • toggle

            I have written (while employed at intel) a number of physics libraries and various other code with x87, s... : bostrov
            • It's just another path that costs time, money and resources that could be better spent where customer dem... : daVinci1980
      • toggle

        http://msdn.microsoft.com/en-us/library/7t5yh4fd.aspx : filthysock
        • Automatic compiler switches do nothing to address the other issues I brought up. : daVinci1980
  • I suspect the code is already pretty close to the way it needs to be for SSE support and it does seem lik... : bostrov
  • toggle

    The guy who knows enough about this to have discovered it in the first place sure seems to think there ar... : DinK
    • toggle

      He's also enough of a fucktard to think that using vtune makes him Woodward and Bernstein. The fact i : uncleanmonkey
      • Yeah, I don't think it's strange or surprising at all. There's the desire to sell GPUs but consoles are ... : DinK
    • toggle

      I see, he's able to make a cost judgement based on his extensive work of looking through the proprietary ... : daVinci1980
      • When I said the guy quotes nvidia, I should have included "on his own site" which is linked in the articl... : DinK
  • toggle

    This might hold water if SSE was not supported by every chip manufactured since the Athlon XP. And there... : qsylvr
    • toggle

      http://www.shacknews.com/laryn.x?id=23409484 : daVinci1980
      • That seems like a reason to leave the current version available as a fall-back, not a reason to avoid doi... : DinK
    • toggle

      If it was just a compiler switch then developers, who have source licenses, would just enable it and the ... : uncleanmonkey
      • toggle

        according mr davinci they are distributed pre-compiled : filthysock
        • toggle

          Almost all middleware comes with a source license, although you can be limited in what you're allowed to ... : uncleanmonkey
          • I doubt the hardware abstraction/driver layer of physx is distributed to middleware licensees? : Toraz Chryx
        • Source licenses are granted to specific partners on a case-by-case basis. Some developers get source, mos... : daVinci1980
  • toggle

    This is just silly, compiler intrinsics are not hard. : Satyric0n
    • toggle

      And are not at all sufficient. If it was as easy as 'turn on this compiler switch', it'd be done. : daVinci1980
      • toggle

        Intrinsics aren't a compiler switch, they're simulated functions that serve as an alternative to handwrit... : Satyric0n
        • toggle

          They're not. They don't address the fallback path, they don't address providing multiple SSE backends, an... : daVinci1980
          • toggle

            They do address ease of development, which means there's basically no excuse for not providing an SSE pat... : Satyric0n
            • toggle

              Except for all of the reasons I've already outlined. But other than those, you're totally right. No reaso... : daVinci1980
              • toggle

                Your reasons are specious. You make it sound like one hobbiest developer that has to do the work. It's fu... : Satyric0n
                • toggle

                  Your argument is right up there with the "WoW makes $150M / mo, why isn't everything perfect?" Everything... : daVinci1980
                  • toggle

                    Except that I am a programmer (native HPC code, native, i.e. actually comparable to the code in question ... : Satyric0n
                    • native, : Satyric0n
                    • I've worked on the physx codebase for a couple of different projects when a source license was provided. ... : daVinci1980
      • If it was that easy AND they aren't intentionally leaving it in the current state to make the hardware so... : DinK
  • toggle

    You think that physx is handcoded in ASM? seriously? you think ANYTHING is handcoded in ASM? : Toraz Chryx
    • toggle

      of course important vector operations will be handcoded in ASM to support SSE and whatever instructions f... : moshman
      • toggle

        More realistically, compiler intrinsics will be used rather than handcoded asm, unless the compiler intri... : Satyric0n
        • toggle

          I'm sure that in general cases compiler intrinsics will be used, but for vector/matrix functions and the ... : moshman
          • Yes, absolutely. The single biggest time-sink related to using compiler intrinsics is verifying that outp... : Satyric0n
    • toggle

      01111001 01101111 01110101 00100000 01100001 01110010 01100101 00100000 01100001 00100000 01100110 011001... : [7]kami
      • toggle

        no u : Toraz Chryx
        • toggle

          haha someone went to the translator! : [7]kami
          • 0 : Toraz Chryx
    • not to shit on your parade but alot of shit is still being handcoded in asm. : italiano3d
    • No, that's not what I've said. When you write well optimized SSE and SSE2, you do write bits of it in ASM... : daVinci1980
• toggle

  oh man, i look forward to the pr rebuttal from this. ill expect something like 'but x87 lets us draw mor... : smacknca
  • I doubt they will say anything, it's not as if they have to justify anything. It's their product, they ca... : moshman
• nVidia being scummy? Shocked I say! : Sailor of Fortune
• toggle

  lol @ all the people saying nvidia is making bad business decisions, doesn't understand, is retarded etc.... : Pandilex
  • seems like a great strategy to avoid people using physx : moshman
  • No one in this thread is saying nvidia is making bad business decisions. It seems to be pretty standard f... : filthysock
  • Except that their product exists as part of a larger ecosystem.... by crippling the software implementati... : Toraz Chryx