cpufp：测试X86 CPU峰值性能的测试工具

Intel Alder Lake i7-1280p
For single Golden Cove(big) core:
$ ./cpufp --thread_pool=[0]
Number Threads: 1
Thread Pool Binding: 0
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| AVX_VNNI        | INT8      | 590.31 GOPS      |
| AVX_VNNI        | INT16     | 295.06 GOPS      |
| FMA             | FP32      | 149.87 GFLOPS    |
| FMA             | FP64      | 74.931 GFLOPS    |
| AVX             | FP32      | 112.39 GFLOPS    |
| AVX             | FP64      | 56.203 GFLOPS    |
| SSE             | FP32      | 56.054 GFLOPS    |
| SSE             | FP64      | 28.001 GFLOPS    |
--------------------------------------------------
For multiple big cores:
$ ./cpufp --thread_pool=[0,2,4,6,8,10]
Number Threads: 6
Thread Pool Binding: 0 2 4 6 8 10
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| AVX_VNNI        | INT8      | 2636.8 GOPS      |
| AVX_VNNI        | INT16     | 1319.1 GOPS      |
| FMA             | FP32      | 670.05 GFLOPS    |
| FMA             | FP64      | 335 GFLOPS       |
| AVX             | FP32      | 502.4 GFLOPS     |
| AVX             | FP64      | 251.2 GFLOPS     |
| SSE             | FP32      | 250.42 GFLOPS    |
| SSE             | FP64      | 125.16 GFLOPS    |
--------------------------------------------------
For single Gracemont(little) core:
$ ./cpufp --thread_pool=[12]
Number Threads: 1
Thread Pool Binding: 12
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| AVX_VNNI        | INT8      | 114.89 GOPS      |
| AVX_VNNI        | INT16     | 57.445 GOPS      |
| FMA             | FP32      | 57.444 GFLOPS    |
| FMA             | FP64      | 28.723 GFLOPS    |
| AVX             | FP32      | 28.723 GFLOPS    |
| AVX             | FP64      | 14.362 GFLOPS    |
| SSE             | FP32      | 28.312 GFLOPS    |
| SSE             | FP64      | 14.361 GFLOPS    |
--------------------------------------------------
For multiple little cores:
$ ./cpufp --thread_pool=[12-19]
Number Threads: 8
Thread Pool Binding: 12 13 14 15 16 17 18 19
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| AVX_VNNI        | INT8      | 867.99 GOPS      |
| AVX_VNNI        | INT16     | 434 GOPS         |
| FMA             | FP32      | 434 GFLOPS       |
| FMA             | FP64      | 217 GFLOPS       |
| AVX             | FP32      | 217.01 GFLOPS    |
| AVX             | FP64      | 108.5 GFLOPS     |
| SSE             | FP32      | 216.39 GFLOPS    |
| SSE             | FP64      | 108.5 GFLOPS     |
--------------------------------------------------
AMD Ryzen9 6900HX(Zen3+)
For single core:
$ ./cpufp --thread_pool=[0]
Number Threads: 1
Thread Pool Binding: 0
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| FMA             | FP32      | 156.18 GFLOPS    |
| FMA             | FP64      | 78.371 GFLOPS    |
| AVX             | FP32      | 156.55 GFLOPS    |
| AVX             | FP64      | 78.256 GFLOPS    |
| SSE             | FP32      | 78.219 GFLOPS    |
| SSE             | FP64      | 38.99 GFLOPS     |
--------------------------------------------------
For multi-cores:
$ ./cpufp --thread_pool=[0,2,4,6,8,10,12,14]
Number Threads: 8
Thread Pool Binding: 0 2 4 6 8 10 12 14
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| FMA             | FP32      | 1151.2 GFLOPS    |
| FMA             | FP64      | 569.89 GFLOPS    |
| AVX             | FP32      | 1088.7 GFLOPS    |
| AVX             | FP64      | 536.37 GFLOPS    |
| SSE             | FP32      | 541.35 GFLOPS    |
| SSE             | FP64      | 269.56 GFLOPS    |
--------------------------------------------------
Intel Celeron N5105(Jasper Lake)
For single core:
$ ./cpufp --thread_pool=[0]
Number Threads: 1
Thread Pool Binding: 0
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| SSE             | FP32      | 23.102 GFLOPS    |
| SSE             | FP64      | 11.564 GFLOPS    |
--------------------------------------------------
For multi_cores:
$ ./cpufp --thread_pool=[0-3]
Number Threads: 4
Thread Pool Binding: 0 1 2 3
--------------------------------------------------
| Instruction Set | Data Type | Peak Performance |
| SSE             | FP32      | 89.327 GFLOPS    |
| SSE             | FP64      | 44.664 GFLOPS    |
--------------------------------------------------
目前还没有ARM版本的，可以深度做一下其他芯片版本的测试，还有Cache啊，访存延迟什么的，带宽什么，如果更加全面就更好了。