Announcement

Collapse
No announcement yet.

HyperZ: Errata & The Catalyst Command Stream

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #51
    Originally posted by bridgman View Post
    We haven't looked deeply into encrypted netlists but initial feedback wasn't particularly promising... at first glance they seem to really be "unencrypted netlists with encrypted debug info". We do need to spend more time looking into that area though...
    I'm not too optimistic. From what I understand, the ability to simulate encrypted netlists is something the simulator vendors cook into their products, precisely so that proprietary IP can be given to customers for simulation without revealing internal details. But I would think that depends on the simulator not spilling the beans, in addition to, as you suggest, that a lot of encryption is done - poorly. Add to that the fact that OSS tends to be repulsed by the idea of using encryption to hide proprietary details, and that you could hack the source code to reprint the newly decrypted source, etc...

    So let's look at a different tack... Is it possible to release "obsolete" netlists? They're smaller, so maybe it's possible so simulate in a week or two with a Beowulf cluster or GPU-at-home. Presumably there's also some "familial" relationship between your products, so that even in the new product, the old feature may work pretty much the same way. (Is Hyper-Z old enough to be on an "obsolete" chip yet?) Even obsolete netlists could also give OSS driver developers new insights into the hardware and how to code for it.

    Comment


    • #52
      would there be any use setting up a system like folding@home has been some time ago to provide simulation and processing power to developers via internet? like some huge cluster of linux users using cpu power, and when matured openCL to help the linux community?
      is it even possible? security issues? other "things"..?

      Comment


      • #53
        Originally posted by phred14 View Post
        ...
        So let's look at a different tack... Is it possible to release "obsolete" netlists? They're smaller, so maybe it's possible so simulate in a week or two with a Beowulf cluster or GPU-at-home. Presumably there's also some "familial" relationship between your products, so that even in the new product, the old feature may work pretty much the same way. (Is Hyper-Z old enough to be on an "obsolete" chip yet?) Even obsolete netlists could also give OSS driver developers new insights into the hardware and how to code for it.
        If the industry hadn't taken a bit of a step back (where interest in ultra-low power dx9-level GPUs is high again) that would be a lot easier. Unfortunately what might be considered obsolete netlists are suddenly real valuable again, in the same way that old CPU cores are coming back in vogue as part of a many-core processing array.

        Originally posted by jakubo View Post
        would there be any use setting up a system like folding@home has been some time ago to provide simulation and processing power to developers via internet? like some huge cluster of linux users using cpu power, and when matured openCL to help the linux community?
        is it even possible? security issues? other "things"..?
        Security issues would probably be a showstopper unless we hosted the system internally, then it would only be a serious challenge. Simulators do seem to be getting better at scaling across a lot of cores, although I don't know if the scaling is at the point where anyone with less patience than an archaeologist could live with the performance.

        Comment

        Working...
        X