Results 65 comments of valeriob01

I suspect a problem with your amdkfd, you could try to reinstall the drivers.

I'm sorry about this error, the command should be `dmesg | grep amd` , and it must show if your gpus have been initialized.

That's tricky, your gpus are initialized on numbers 0 and 1, that is good for gpuowl, -device 0 and -device 1. But for mfakto the numbers start at 1, so...

There may be some issue with the driver, but the most important part is that you got it working. Good number crunching!

BTW, there is a patched version which compiles on Linux, I use this version: https://github.com/preda/mfakto

I said it in the forum already but just in case. To optimize mfakto performance: mfakto -d \ --perftest 1 It will write new parameters in the mfakto.ini file so...

As a test I have downgraded the kernel to 4.19 and ROCm to 3.1 and to gcc-8. Result: ``` make gcc-8 -m64 -Wall -O3 -funroll-loops -ffast-math -finline-functions -frerun-loop-opt -fgcse-sm -fgcse-las...

Even before there was a huge amount of warnings on Linux but mfakto compiled and ran well. Now even the old code (from preda/mfakto) is failing on perftest. My systems...

That does not sound ok as I am trying to compile for OpenCL .

I hope this helps, compiler trace: ``` make gcc -m64 -Wall -O3 -funroll-loops -ffast-math -finline-functions -frerun-loop-opt -fgcse-sm -fgcse-las -flto -I/opt/rocm-3.3.0/opencl/include -DBUILD_OPENCL -funroll-all-loops -funsafe-loop-optimizations -fira-region=all -fsched-spec-load -fsched-stalled-insns=10 -fsched-stalled-insns-dep=10 -fno-align-labels -c sieve.c...