Radeon RX570 detection / OpenCL error

[rfuegen]

there are two RX570 GPUs in my rig, but xmrig seems to detect four:

• ABOUT XMRig/6.8.0-mo1 gcc/4.8.5
• LIBS libuv/1.40.0
• HUGE PAGES supported
• 1GB PAGES unavailable
• CPU AMD Ryzen 9 3900X 12-Core Processor (1) 64-bit AES
  threads:24
• MEMORY 4.1/15.4 GB (27%)
• DONATE 1%
• ASSEMBLY auto:ryzen
• POOL 128tx exploit #1 gulf.moneroocean.stream:10032 algo auto
• COMMANDS 'h' hashrate, 'p' pause, 'r' resume, 's' results, 'c' connection
• ADL press e for health report
• OPENCL #0 AMD Accelerated Parallel Processing/OpenCL 2.1 AMD-APP (3110.6)
• OPENCL GPU #0 01:00.0 Radeon RX 570 Series (Ellesmere) 1366 MHz cu:32 mem:4048/8178 MB
• OPENCL GPU 128tx exploit #1 0a:00.0 Radeon RX 570 Series (Ellesmere) 1366 MHz cu:32 mem:4048/8172 MB
• CUDA disabled
  [2021-03-14 22:18:37.786] benchmk STARTING ALGO PERFORMANCE CALIBRATION (with 20 seconds round)
  [2021-03-14 22:18:37.786] benchmk Algo cn/r Preparation
  [2021-03-14 22:18:37.786] cpu use profile * (24 threads) scratchpad 2048 KB
  [2021-03-14 22:18:37.789] opencl use profile cn/2 (4 threads) scratchpad 2048 KB
  | # | GPU | BUS ID | INTENSITY | WSIZE | MEMORY | NAME
  | 0 | 0 | 01:00.0 | 768 | 8 | 1536 | Radeon RX 570 Series (Ellesmere)
  | 1 | 0 | 01:00.0 | 768 | 8 | 1536 | Radeon RX 570 Series (Ellesmere)
  | 2 | 1 | 0a:00.0 | 768 | 8 | 1536 | Radeon RX 570 Series (Ellesmere)
  | 3 | 1 | 0a:00.0 | 768 | 8 | 1536 | Radeon RX 570 Series (Ellesmere)
  [2021-03-14 22:18:38.343] cpu READY threads 24/24 (24) huge pages 100% 24/24 memory 49152 KB (557 ms)
  [2021-03-14 22:18:38.375] opencl READY threads 4/4 (586 ms)

is this just a display issue or are there 2 threads per GPU? in "OPENCL GPU" lines, two cards are correctly detected.

with rebench-algo: true and cpu: enabled: true, xmrig crashes right after start:

[2021-03-14 22:18:38.343] cpu READY threads 24/24 (24) huge pages 100% 24/24 memory 49152 KB (557 ms)
[2021-03-14 22:18:38.375] opencl READY threads 4/4 (586 ms)
[2021-03-14 22:18:38.407] opencl error CL_INVALID_KERNEL_NAME when calling clCreateKernel for kernel cn1_1000
[2021-03-14 22:18:38.408] opencl thread #3 failed with error CL_INVALID_KERNEL_NAME
[2021-03-14 22:18:38.470] opencl error CL_INVALID_KERNEL_NAME when calling clCreateKernel for kernel cn1_1000
[2021-03-14 22:18:38.470] opencl thread #0 failed with error CL_INVALID_KERNEL_NAME
[2021-03-14 22:18:38.474] opencl error CL_INVALID_KERNEL_NAME when calling clCreateKernel for kernel cn1_1000
[2021-03-14 22:18:38.474] opencl thread #2 failed with error CL_INVALID_KERNEL_NAME
[2021-03-14 22:18:38.552] opencl error CL_INVALID_KERNEL_NAME when calling clCreateKernel for kernel cn1_1000
[2021-03-14 22:18:38.552] opencl thread #1 failed with error CL_INVALID_KERNEL_NAME
[2021-03-14 22:18:38.746] signal Ctrl+C received, exiting
Segmentation fault (core dumped)

OS is Centos8 kernel 4.18.0-240.10.1.el8_3.x86_64
amd-gpu-pro is 20-20-1089974

when mining XHV on CPU+2*GPU, power and fan speed do not match as reported by xmrig (above) and sensors (below):

[2021-03-14 22:40:56.063] opencl #0 01:00.0 0W 0C 0RPM 0/0MHz
[2021-03-14 22:40:56.063] opencl #1 0a:00.0 61W 48C 1154RPM 600/2000MHz
[2021-03-14 22:40:56.063] miner speed 10s/60s/15m 2267.4 2331.2 n/a H/s max 2430.0 H/s

amdgpu-pci-0100
Adapter: PCI adapter
vddgfx: +0.95 V
fan1: 1198 RPM (min = 0 RPM, max = 4500 RPM)
edge: +46.0°C (crit = +94.0°C, hyst = -273.1°C)
power1: 71.24 W (cap = 135.00 W)

amdgpu-pci-0a00
Adapter: PCI adapter
vddgfx: +0.95 V
fan1: 1286 RPM (min = 0 RPM, max = 4500 RPM)
edge: +45.0°C (crit = +94.0°C, hyst = -273.1°C)
power1: 60.19 W (cap = 135.00 W)

I'm not that worried about what is printed on the screen, but would like to run a full rebench for CPU+GPU.

tnx

[SChernykh]

there are two RX570 GPUs in my rig, but xmrig seems to detect four:

It detects two GPUs, uses two threads per GPU. 2*2=4

with rebench-algo: true and cpu: enabled: true, xmrig crashes right after start:

This is MoneroOcean version, not stock XMRig. Report it to MoneroOcean.