<reg>:53 "operation not supported"

[kmodexc]

I try to mine on a NVIDIA Jetson Nano. I have build the xmrgi-cuda and xmrig from source. There were no problems. NVML should be installed (ubuntu package cuda-nvml-dev-10-2) so i dont know why it has problems loading it.

The problem is that the mining thread stops with thread #0 failed with error <reg>:53 "operation not supported" and then does nothing.

  * ABOUT        XMRig/6.8.1 gcc/7.5.0
 * LIBS         libuv/1.40.1-dev OpenSSL/1.1.1 hwloc/1.11.9
 * HUGE PAGES   supported
 * 1GB PAGES    unavailable
 * CPU          ARM Cortex-A57 (1) 64-bit AES
                L2:2.0 MB L3:0.0 MB 4C/4T NUMA:1
 * MEMORY       1.2/3.9 GB (32%)
 * DONATE       5%
 * POOL #1      xmrpool.eu:9999 algo auto
 * COMMANDS     hashrate, pause, resume, results, connection
 * OPENCL       disabled
 * CUDA         10.2/10.2/6.5.0
 * NVML         disabled (failed to load NVML)
 * CUDA GPU     #0 00:00.0 NVIDIA Tegra X1 921/12 MHz smx:1 arch:53 mem:2687/3964 MB
[2021-02-06 23:16:29.159]  net      use pool xmrpool.eu:9999 TLSv1.2 54.37.7.208
[2021-02-06 23:16:29.159]  net      fingerprint (SHA-256): "719f838fa8ae7844b7f25917c91d32185c594c4d36a3ed5bab233a433ecf9910"
[2021-02-06 23:16:29.160]  net      new job from xmrpool.eu:9999 diff 50000 algo rx/0 height 2291178
[2021-02-06 23:16:29.160]  cpu      use argon2 implementation default
[2021-02-06 23:16:29.161]  randomx  init dataset algo rx/0 (4 threads) seed 6c5c1c954231ee43...
[2021-02-06 23:16:29.163]  randomx  allocated 2336 MB (2080+256) huge pages 0% 0/1168 +JIT (2 ms)
[2021-02-06 23:16:55.436]  randomx  dataset ready (26273 ms)
[2021-02-06 23:16:55.437]  nvidia   use profile  rx  (1 thread) scratchpad 2048 KB
|  # | GPU |  BUS ID | INTENSITY | THREADS | BLOCKS | BF |  BS | MEMORY | NAME
|  0 |   0 | 00:00.0 |        64 |      32 |      2 |  0 |   0 |    128 | NVIDIA Tegra X1
[2021-02-06 23:16:55.832]  nvidia   READY threads 1/1 (395 ms)
[2021-02-06 23:16:55.849]  nvidia   thread #0 failed with error <reg>:53 "operation not supported"
[2021-02-06 23:16:58.161]  net      new job from xmrpool.eu:9999 diff 50000 algo rx/0 height 2291178
[2021-02-06 23:17:36.027]  net      new job from xmrpool.eu:9999 diff 50000 algo rx/0 height 2291179
[2021-02-06 23:17:36.842]  net      new job from xmrpool.eu:9999 diff 25000 algo rx/0 height 2291179
[2021-02-06 23:17:55.548]  nvidia   #0 00:00.0   0W  0C
[2021-02-06 23:17:55.549]  miner    speed 10s/60s/15m n/a n/a n/a H/s max n/a H/s

[N304LX]

The same here, but, it was caused for my antivirus, turn it off or add as exeption the cuda plugin, may be it work's for you too

[Spudz76]

Not enough memory for RandomX

RandomX is a CPU algorithm not intended for GPUs

Possibly NVML needs to run as root depending how the Linux Distro is configured (assuming this isn't regular Ubuntu).

[gavan1]

@einsteinmaster Did you find a fix? I recommend you rename the issue to thread #0 failed with error <reg>:53 "operation not supported"

@Spudz76 could you point me to the algos that use the gpu or maybe point to what needs to be in the config file to isolate the cuda core ?

I've disabled the cpu and net jobs are picked up by xmrig but I get the

[2021-06-06 23:17:55.548]  nvidia   #0 00:00.0   0W  0C
[2021-04-06 23:17:55.549]  miner    speed 10s/60s/15m n/a n/a n/a H/s max n/a H/s

Ubuntu Bionic / Jetson Nano devkit 4GB.

Thanks

[Spudz76]

RandomX probably not going to work and is not a GPU algorithm so even if it did it would be silly slow. Try other algos, use MoneroOcean for cross-mining, like would be done for any other sort of GPU to earn XMR (without processing RandomX).

[kmodexc]

To me it looks like an architecture problem. Jetson nano runs on arm64 where there is some binary code build for another architecture. I had no time tracking that problem, but i would probably search for pre-build dependencys and find out if they can be build from source.

@Spudz76 the GPU has 4GB RAM as my log shows so i guess this should not be a problem. In fact the jetsons GPU memory is shared with the CPU RAM. https://forums.developer.nvidia.com/t/jetson-tx2-gpu-memory/48994

@gavan1 had the same results.

[sigkill]

This appears to be related to threads to block count from everything I've read on the nvidia forum. Not had luck patching, but if there were flags to change the blocks / threads for xmrig, that would make figuring this out much easier.

[sigkill]

"too many resources requested for launch" is common as well.

[Spudz76]

Edit blocks/threads in config.json

[g0lg0th7]

thread #0 failed with error :59 "unspecified launch failure

Folllowing this opened gpu issue, any update ?