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
{{ message }}
This repository has been archived by the owner on Jan 26, 2024. It is now read-only.
Even compiling the simple square example, this obviously needs to use hipMemcpy.
However, I was kind of surprised that internally in hipMemcpy the BlitProgram is used which (in an OpenCL fashion) seems to be built on first use using the compiler manager, and thus requiring effectively a full llvm dependency. In particular the device libs need to be present (which was what made me even aware of this issue)
Doesn't this undermine the whole point of doing binary builds for each ISA in hipamd in the first place?
Even compiling the simple square example, this obviously needs to use hipMemcpy.
However, I was kind of surprised that internally in hipMemcpy the BlitProgram is used which (in an OpenCL fashion) seems to be built on first use using the compiler manager, and thus requiring effectively a full llvm dependency. In particular the device libs need to be present (which was what made me even aware of this issue)
Doesn't this undermine the whole point of doing binary builds for each ISA in hipamd in the first place?
To reproduce the issue run
Ideally this should only happen if hiprtc is used.
The text was updated successfully, but these errors were encountered: