Cory Bloor

Hmm... the existence of xnack on much older GPUs confuses me. I was under the impression that there was progressive development on this feature from GCN to CDNA 1 to...

@Admyfast, I believe that the ROCm minimum system requirements include PCIe 3.0 or greater. I'm not sure if that's the cause of your problem, but that's one thing that stood...

The HIP Runtime works on gfx1031 (the RX 6700 XT's ISA), but none of the math libraries in the HIP SDK are built for gfx1031. The rocFFT/hipFFT libraries use hiprtc...

I'm a bit surprised that you're having trouble with ROCm 5.4.1 on the 7900 XTX, as that architecture is gfx1100 and most of the AMD-provided binaries for ROCm 5.4.1 contain...

The `PREFIX` option is not used anymore, so the problem is further down. I should talk with @lawruble13 about this, since I don't fully understand the purpose of `ROCMInstallTargets` vs....

I'm looking at my schedule for the next month and I'm not sure when I'm going to get a chance to really dig into this, so I'm marking this as...

@pfultz2, @lawruble13, how do we handle the failing OSX builds on a mergeback like this? Do we need to cherry-pick b04260e1c8a78689e2ce18b17c79648b56b5546b into the 6.1 release-staging branch first?

> It looks like `DEPENDS_EXPLICIT_ONLY` was made available in cmake 3.27 which isn't available on all of the supported rocm platforms e.g. ubuntu 20 is 3.16. Currently, we have cmake_minimum_required...

> So yea, [.debug_addr is not yet supported on Debian tooling](https://manpages.debian.org/testing/dwz/dwz.1.en.html). Which currently prevents debugging HIP host-side. Is that **[Bug 27375 - [dwz, dwarf5] .debug_addr support](https://sourceware.org/bugzilla/show_bug.cgi?id=27375)**? A comment from September...

You might find [this script](https://gist.github.com/cgmb/7cd9a481c42ce132b5d6420380becef3) useful. Not everything in it will install to the right directory with DCMAKE_INSTALL_PREFIX specified, but it builds most of the components in ROCm 4.5 from...