You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I want to optimize memory loads and tried several different instructions to see how it goes. Up to now, however, I got success only with flat_load.
I would like to compile a kernel for RX 4xx/5xx, but if I use global_load_dword** I'm getting "unknown instruction". I suppose that there are some flags for that...
Also, I have a question about "buffer resources". How to access them, and how to use them? I tried several kernel setups but didn't have luck. Is there any table which explains what is where? Who decides what info goes to which register (driver?)? Infos on the net are sparse and most of them are contradictional...
Thanks in advance
The text was updated successfully, but these errors were encountered:
The GLOBAL_* and SCRATCH_* instructions has been introduced in the RX Vega GPU and they are not available in Fiji/Polaris GPU's. The resource buffer was used in old OpenCL implementation and in the first GCN GPU generation (Tahiti, Pitcairn, HD 7xxx). For new OpenCL drivers and newer GPU's the FLAT_* instructions are recommended to access memory.
I want to optimize memory loads and tried several different instructions to see how it goes. Up to now, however, I got success only with flat_load.
I would like to compile a kernel for RX 4xx/5xx, but if I use global_load_dword** I'm getting "unknown instruction". I suppose that there are some flags for that...
Also, I have a question about "buffer resources". How to access them, and how to use them? I tried several kernel setups but didn't have luck. Is there any table which explains what is where? Who decides what info goes to which register (driver?)? Infos on the net are sparse and most of them are contradictional...
Thanks in advance
The text was updated successfully, but these errors were encountered: