Alessio1989
Regular
Nice to know that! Are you talking about all Intel hardware, Gen 7.5 included?Uhh, that's not correct. Intel hardware can do even more than the 64 DWORDS without additional indirection.
Nice to know that! Are you talking about all Intel hardware, Gen 7.5 included?Uhh, that's not correct. Intel hardware can do even more than the 64 DWORDS without additional indirection.
Yes, all hardware. See our earlier presentations - they include big sections on resource binding on both Gen7.5/Gen8 and Gen9.Nice to know that! Are you talking about all Intel hardware, Gen 7.5 included?
On GCN a wave execution starts with the first SGPRs filled with the root arguments. You can use AMD CodeXL to inspect GCN shader microcode. GCN vertex fetch is bindless, and vertex buffer (VBV) uses a resource descriptor like every other shader resource (SRV, UAV, CBV). AMD doesn't have bindless IBVs.MSDN states:
On which hardware?
I also still do not have clear which hardware has a dedicated buffer for root argument versioning and how big is it.
I also still do not have clear which hardware has a dedicated buffer for root argument versioning and how big is it. I guess there are different hardware with different "root buffer" sizes, but none of current hardware has a big one enough to cover the full root signature size allowed by the API (64 DWORDs).
Keep in mind parallelism between multiple kernels on all of the "bigger" architectures. You can possibly have much more than a single root signature loaded on the CU.Uhh, that's not correct. Intel hardware can do even more than the 64 DWORDS without additional indirection.
Pipelining of root argument/signature changes is required on all chips for performance - this *is* the inner loop after all. My statement stands - Gen can easily pipeline 64 DWORDS/draw or more efficiently.Keep in mind parallelism between multiple kernels on all of the "bigger" architectures. You can possibly have much more than a single root signature loaded on the CU.
No, but I installed Intel OpenCL SDK..did you installed the intel media sdk?
„We are a software company!“Latest I've tried were 358.70 which seemed to be working for a few SDK samples I tried.
Latest I've heard from Nvidia, when asking this via email was:
- Starting with the next release of NVIDIA Nsight, Fermi GPUs will no longer be supported.