#GPU OPENCL BENCHMARK LIST HOW TO#

You can figure this out with some checks for the device vendor+name, which will reveal the microarchitecture: string name = trim(cl_device.getInfo()) // device name 128 (Nvidia Maxwell/Pascal/Ampere/Hopper/Ada, AMD RDNA/RDNA2).This depends on the GPU microarchitecture and can be: There is one remaining unknown: the number of cores per CU. Int ipc = ipc = cl_device.getInfo()=CL_DEVICE_TYPE_GPU ? 2 : 32 // IPC (instructions per cycle) is 2 for GPUs and 32 for most modern CPUsįloat tflops = 1E-6f*(float)cores*(float)ipc*(float)clock_frequency Int clock_frequency = (uint)cl_device.getInfo() // in MHz But: You get the number of CUs CL_DEVICE_MAX_COMPUTE_UNITS, the peak core clock speed according to the data sheet CL_DEVICE_MAX_CLOCK_FREQUENCY, and the instructions per cycle (ipc), which can give you at least a good estimate for FP32 TFlops: int compute_units = (uint)cl_device.getInfo() // compute units (CUs) can contain multiple cores depending on the microarchitecture OpenCL clGetDeviceInfo provides neither of these. memory bandwidth (in some cases also cache bandwidth).FP32 TFlops (in some cases also FP64/FP16 TFlops).There is 2 specs that mainly determine the performance of a GPU: I want to estimate the computing power automatically by the program just from information provided by clGetDeviceInfo() But this is some vendor specific information which I have to search as a user form external sources. Related: OpenCL: Confused by CL_DEVICE_MAX_COMPUTE_UNITSīy reading some materials (wiki, nVidia documentation) I can find that each compute unit has 128 shader cores (or CUDA cores). I tried to print all possible divice parameters provided by clGetDeviceInfo() but none seems to give me required information: OpenCL platform count 2ĭEVICE: pthread-Intel(R) Core(TM) i5-10400F CPU 2.90GHzĭEVICE_VERSION: OpenCL 1.2 pocl HSTR: pthread-x86_64-pc-linux-gnu-skylake GFLOPS = CL_DEVICE_MAX_CLOCK_FREQUENCY * CL_DEVICE_MAX_COMPUTE_UNITS * CL_DEVICE_MAX_WORK_GROUP_SIZEīut this does not fit of Intel CPU where I have CL_DEVICE_MAX_WORK_GROUP_SIZE = 4096 which would make it more powerfull than GPU A work-group executes on a single compute unit.

The number of parallel compute units on the OpenCL device. Looking in the documentation: CL_DEVICE_MAX_COMPUTE_UNITS While clock frequnecy 1777 MHz seems reasonable, the 28 comoute units seems way to low from my NVIDIA GeForce RTX 3060 which have Core Config: 3584 112:48:28:112 according to wikipedia Now I can get CL_DEVICE_MAX_CLOCK_FREQUENCY and CL_DEVICE_MAX_COMPUTE_UNITS form clGetDeviceInfo() But still it is some rough estimate of crude computing capabilities.

* yes, I know this is extremely crude, as each operation takes different number of clock cycles on each architecture, and there are huge inefficiencies due to cash-misses etc. What I wanted to do is to go over all devices and evaluate total theoretical computing power in GFLOPS simply by multiplying:

#GPU OPENCL BENCHMARK LIST HOW TO#

I'm trying to make automatic decision how to distribute workload between CPU and GPU.

Gpu opencl benchmark list