search

fireice-uk / xmr-stak

Free Monero RandomX Miner and unified CryptoNight miner
GNU General Public License v3.0
4.05k stars 1.79k forks source link

nvrtcCompileProgram failed: NVRTC_ERROR_INVALID_OPTION --gpu-architecture #2488

Closed o2genum closed 5 years ago

o2genum commented 5 years ago

Which operating system do you use? Windows 10

To Reproduce I've compiled latest version 2.10.6.

When starting, I get these errors:

[2019-07-23 18:59:48] : Mining coin: cryptonight_r
[2019-07-23 18:59:49] : Found AMD platform index id = 1, name = Advanced Micro Devices, Inc.
[2019-07-23 18:59:49] : Found OpenCL GPU Baffin.
[2019-07-23 18:59:49] : AMD: GPU (OpenCL) configuration stored in file 'amd.txt'
[2019-07-23 18:59:49] : Compiling code and initializing GPUs. This will take a while...
[2019-07-23 18:59:49] : Device 0 work size 8 / 32.
[2019-07-23 18:59:49] : OpenCL device 0 - Load precompiled code from file C:\Users\o2gen\AppData\Local/.openclcache/4417dc6ed2e28e13b104ef0e2ae2fd2c47f295ce5b23ea58b2d0928ba72f14a1.openclbin
[2019-07-23 18:59:49] : Device 0 work size 8 / 32.
[2019-07-23 18:59:49] : OpenCL device 0 - Load precompiled code from file C:\Users\o2gen\AppData\Local/.openclcache/4417dc6ed2e28e13b104ef0e2ae2fd2c47f295ce5b23ea58b2d0928ba72f14a1.openclbin
[2019-07-23 18:59:49] : Starting AMD GPU (OpenCL) thread 0, no affinity.
[2019-07-23 18:59:49] : Starting AMD GPU (OpenCL) thread 1, no affinity.
[2019-07-23 18:59:49] : NVIDIA: try to load library 'xmrstak_cuda_backend'
CUDA [10.1/9.0] GPU#0, device architecture 75: "GeForce GTX 1660"...
device init succeeded
[2019-07-23 18:59:49] : NVIDIA: GPU configuration stored in file 'nvidia.txt'
NVIDIA: found 1 potential device's
[2019-07-23 18:59:49] : Starting NVIDIA GPU thread 0, no affinity.
CUDA [10.1/9.0] GPU#0, device architecture 75: "GeForce GTX 1660"...
device init succeeded
[2019-07-23 18:59:49] : NVIDIA: use library 'xmrstak_cuda_backend'
[2019-07-23 18:59:49] : CPU configuration stored in file 'cpu.txt'
[2019-07-23 18:59:49] : Starting 1x thread, affinity: 0.
[2019-07-23 18:59:49] : hwloc: memory pinned
[2019-07-23 18:59:49] : Starting 1x thread, affinity: 1.
[2019-07-23 18:59:49] : hwloc: memory pinned
[2019-07-23 18:59:49] : Starting 1x thread, affinity: 2.
[2019-07-23 18:59:49] : hwloc: memory pinned
[2019-07-23 18:59:49] : Fast-connecting to mypool.org:14444 pool ...
[2019-07-23 18:59:49] : Pool mypool.org:14444 connected. Logging in...
[2019-07-23 18:59:50] : Difficulty changed. Now: 120001.
[2019-07-23 18:59:50] : Pool logged in.
[2019-07-23 18:59:50] : enable cryptonight_r asm 'intel_avx' cpu's
[2019-07-23 18:59:50] : enable cryptonight_r asm 'intel_avx' cpu's
[2019-07-23 18:59:50] : nvrtcCompileProgram failed: NVRTC_ERROR_INVALID_OPTION
[2019-07-23 18:59:50] : Program compile log: nvrtc: error: invalid value for --gpu-architecture (-arch)

[CUDA] Error gpu 0: <cryptonight_core_cpu_hash>:1015 "device kernel image is invalid[2019-07-23 18:59:50] : enable cryptonight_r asm 'intel_avx' cpu's
"

What may be causing these errors?

psychocrypt commented 5 years ago

install cuda 10.1 You gpu is to new for old cuda 9

Andrey Moiseev notifications@github.com schrieb am Di., 23. Juli 2019, 18:15:

Describe the bug A clear and concise description of what the bug is.

Which operating system do you use? Windows 10

To Reproduce I've compiled latest version 2.10.6.

  • Latest dependencies from the releases page
  • CUDA SDK 9.0
  • OCL SDK

When starting, I get these errors:

[2019-07-23 18:59:48] : Mining coin: cryptonight_r [2019-07-23 18:59:49] : Found AMD platform index id = 1, name = Advanced Micro Devices, Inc. [2019-07-23 18:59:49] : Found OpenCL GPU Baffin. [2019-07-23 18:59:49] : AMD: GPU (OpenCL) configuration stored in file 'amd.txt' [2019-07-23 18:59:49] : Compiling code and initializing GPUs. This will take a while... [2019-07-23 18:59:49] : Device 0 work size 8 / 32. [2019-07-23 18:59:49] : OpenCL device 0 - Load precompiled code from file C:\Users\o2gen\AppData\Local/.openclcache/4417dc6ed2e28e13b104ef0e2ae2fd2c47f295ce5b23ea58b2d0928ba72f14a1.openclbin [2019-07-23 18:59:49] : Device 0 work size 8 / 32. [2019-07-23 18:59:49] : OpenCL device 0 - Load precompiled code from file C:\Users\o2gen\AppData\Local/.openclcache/4417dc6ed2e28e13b104ef0e2ae2fd2c47f295ce5b23ea58b2d0928ba72f14a1.openclbin [2019-07-23 18:59:49] : Starting AMD GPU (OpenCL) thread 0, no affinity. [2019-07-23 18:59:49] : Starting AMD GPU (OpenCL) thread 1, no affinity. [2019-07-23 18:59:49] : NVIDIA: try to load library 'xmrstak_cuda_backend' CUDA [10.1/9.0] GPU#0, device architecture 75: "GeForce GTX 1660"... device init succeeded [2019-07-23 18:59:49] : NVIDIA: GPU configuration stored in file 'nvidia.txt' NVIDIA: found 1 potential device's [2019-07-23 18:59:49] : Starting NVIDIA GPU thread 0, no affinity. CUDA [10.1/9.0] GPU#0, device architecture 75: "GeForce GTX 1660"... device init succeeded [2019-07-23 18:59:49] : NVIDIA: use library 'xmrstak_cuda_backend' [2019-07-23 18:59:49] : CPU configuration stored in file 'cpu.txt' [2019-07-23 18:59:49] : Starting 1x thread, affinity: 0. [2019-07-23 18:59:49] : hwloc: memory pinned [2019-07-23 18:59:49] : Starting 1x thread, affinity: 1. [2019-07-23 18:59:49] : hwloc: memory pinned [2019-07-23 18:59:49] : Starting 1x thread, affinity: 2. [2019-07-23 18:59:49] : hwloc: memory pinned [2019-07-23 18:59:49] : Fast-connecting to mypool.org:14444 pool ... [2019-07-23 18:59:49] : Pool mypool.org:14444 connected. Logging in... [2019-07-23 18:59:50] : Difficulty changed. Now: 120001. [2019-07-23 18:59:50] : Pool logged in. [2019-07-23 18:59:50] : enable cryptonight_r asm 'intel_avx' cpu's [2019-07-23 18:59:50] : enable cryptonight_r asm 'intel_avx' cpu's [2019-07-23 18:59:50] : nvrtcCompileProgram failed: NVRTC_ERROR_INVALID_OPTION [2019-07-23 18:59:50] : Program compile log: nvrtc: error: invalid value for --gpu-architecture (-arch)

[CUDA] Error gpu 0: :1015 "device kernel image is invalid[2019-07-23 18:59:50] : enable cryptonight_r asm 'intel_avx' cpu's "

What may be causing these errors?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/fireice-uk/xmr-stak/issues/2488?email_source=notifications&email_token=AGFTDNTUFCTIN7XIBRVRERTQA4VARA5CNFSM4IGGHQFKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HA66HMQ, or mute the thread https://github.com/notifications/unsubscribe-auth/AGFTDNTRRSPWUXTQJXOREBDQA4VARANCNFSM4IGGHQFA .

psychocrypt commented 5 years ago

please remove the build folder before you compile with cuda 10

o2genum commented 5 years ago

@psychocrypt OK, it worked.

If I'm distributing the miner to users, what is the best way to support both CUDA 9.0 and CUDA 10.0?

I'm still seeing both xmrstak_cuda_backend_cuda10_0.dll and xmrstak_cuda_backend.dll in the released ZIP, is this for some king of fallback from one CUDA version to another?

psychocrypt commented 5 years ago

yes this will allow us to ship the miner wi4h optimizarion for different cuda version and therefore gpus.