4 ms·
An interesting piece of hardware that I have seen was used in ray tracing where high speed reduced precision FP was used to find polygons that were not visible,
by russell 15y ago
An interesting piece of hardware that I have seen was used in ray tracing where high speed reduced precision FP was used to find polygons that were not visible, those where the low precision hid a change and those that were otherwise visible Conventional high precision processing was needed only for those with changes. The overall performance improvement was 100x.
We are probably close to hitting the wall for general purpose solutions on general purpose hardware, but the future is wide open for application specific hardware.
- Palomides 15y agoare we actually going to see application-specific hardware, or are we going to see something like an FPGA as a coprocessor in consumer hardware?
- Tuna-Fish 15y agoFPGAs really wouldn't help much. The theoretical gains sound good, but frankly, the practical speed of a cpu is measured by how fast it can run crappy code. FPGA in a cpu would just be a pointless waste of silicon that almost nobody would program for. Application-specific hardware is already being added to the cpus -- the newest Intel ones have a hardware RNG and hardware video encoders and decoders.
- bnegreve 15y ago"already" ? GPUs are almost hardware implementations of directX/openGL APIs and it is not new.
- rhino42 15y agoThere's a team at UC-Berkeley which is working on compiling C++ to a combination of CPU/GPU/FPGA code right now, which might change things. I think you need a special architecture to take full advantage of it though.
- russell 15y agoIn the ray tracing example I mentioned above, FPGAs got a 10x improvement. Going to silicon got another 10x.