找到一款opencl-benchmark 工具
https://github.com/ProjectPhysX/OpenCL-Benchmark
双卡1080ti实测：
yxc@hua-System-Product-Name:~/code/3/OpenCL-Benchmark$ ./make.sh
.-----------------------------------------------------------------------------.
|----------------.------------------------------------------------------------|
| Device ID 0 | NVIDIA GeForce GTX 1080 Ti |
| Device ID 1 | NVIDIA GeForce GTX 1080 Ti |
|----------------'------------------------------------------------------------|
|----------------.------------------------------------------------------------|
| Device ID | 0 |
| Device Name | NVIDIA GeForce GTX 1080 Ti |
| Device Vendor | NVIDIA Corporation |
| Device Driver | 530.41.03 (Linux) |
| OpenCL Version | OpenCL C 1.2 |
| Compute Units | 28 at 1657 MHz (3584 cores, 11.877 TFLOPs/s) |
| Memory, Cache | 11169 MB, 1344 KB global / 48 KB local |
| Buffer Limits | 2792 MB global, 64 KB constant |
|----------------'------------------------------------------------------------|
| Info: OpenCL C code successfully compiled. |
| FP64 compute 0.427 TFLOPs/s (1/32) |
| FP32 compute 12.837 TFLOPs/s ( 1x ) |
| FP16 compute not supported |
| INT64 compute 1.530 TIOPs/s (1/8 ) |
| INT32 compute 4.505 TIOPs/s (1/3 ) |
| INT16 compute 12.740 TIOPs/s ( 1x ) |
| INT8 compute 13.012 TIOPs/s ( 1x ) |
| Memory Bandwidth ( coalesced read ) 372.41 GB/s |
| Memory Bandwidth ( coalesced write) 429.60 GB/s |
| Memory Bandwidth (misaligned read ) 163.06 GB/s |
| Memory Bandwidth (misaligned write) 100.27 GB/s |
| PCIe Bandwidth (send ) 10.41 GB/s |
| PCIe Bandwidth ( receive ) 10.32 GB/s |
| PCIe Bandwidth ( bidirectional) (Gen4 x16) 10.02 GB/s |
|-----------------------------------------------------------------------------|
|----------------.------------------------------------------------------------|
| Device ID | 1 |
| Device Name | NVIDIA GeForce GTX 1080 Ti |
| Device Vendor | NVIDIA Corporation |
| Device Driver | 530.41.03 (Linux) |
| OpenCL Version | OpenCL C 1.2 |
| Compute Units | 28 at 1632 MHz (3584 cores, 11.698 TFLOPs/s) |
| Memory, Cache | 11172 MB, 1344 KB global / 48 KB local |
| Buffer Limits | 2793 MB global, 64 KB constant |
|----------------'------------------------------------------------------------|
| Info: OpenCL C code successfully compiled. |
| FP64 compute 0.431 TFLOPs/s (1/24) |
| FP32 compute 12.888 TFLOPs/s ( 1x ) |
| FP16 compute not supported |
| INT64 compute 1.533 TIOPs/s (1/8 ) |
| INT32 compute 4.533 TIOPs/s (1/3 ) |
| INT16 compute 12.429 TIOPs/s ( 1x ) |
| INT8 compute 13.058 TIOPs/s ( 1x ) |
| Memory Bandwidth ( coalesced read ) 374.60 GB/s |
| Memory Bandwidth ( coalesced write) 429.12 GB/s |
| Memory Bandwidth (misaligned read ) 163.98 GB/s |
| Memory Bandwidth (misaligned write) 100.70 GB/s |
| PCIe Bandwidth (send ) 1.58 GB/s |
| PCIe Bandwidth ( receive ) 1.63 GB/s |
| PCIe Bandwidth ( bidirectional) (Gen1 x16) 1.59 GB/s |
|-----------------------------------------------------------------------------|
|-----------------------------------------------------------------------------|
| Done. Press Enter to exit. |
'-----------------------------------------------------------------------------'