Installed #AMD's #OpenCL implementation (minus kernel DKMS stuff) from their #Debian repository (something ROCm-3.8 or similar). The `clinfo` it bundled in /opt worked great, but `/usr/bin/clinfo` would not find the AMD #GPU, and so neither most other OpenCL applications. Solution was to create an #ICD text file in `/etc` pointing to the relevant shared library:
$ cat /etc/OpenCL/vendors/amdocl64.icd
A generic #x86_64 C++ build on Intel Core2Duo runs about half the speed as the POCL CPU OpenCL implementation. Maybe I can drop my attempts at #SIMD (which benefit strongly from non-portable -march=native, as I haven't figured out runtime CPU detection and compiling multiple versions) and punt that to the #OpenCL runtime compiler(s). A fallback in case of no OpenCL might still be handy though...
Turns out I just needed to `make SIMD=0` with no code changes necessary. Now build from clean takes ~15mins, which is still long but most rebuilds during development hopefully don't need to recompile all of it.
Welcome to post.lurk.org, an instance for discussions around cultural freedom, experimental, new media art, net and computational culture, and things like that.