1 lscpu
2 架构：                           x86_64
3 CPU 运行模式：32-bit, 64-bit
4 字节序：                         Little Endian
5 Address sizes:                   43 bits physical, 48 bits virtual
6 CPU:                             128
7 在线 CPU 列表：0-127
8 每个核的线程数：                 2
9 每个座的核数：                   32
10 座：                             2
11 NUMA 节点：2
12 厂商 ID：HygonGenuine
13 CPU 系列：24
14 型号：                           1
15 型号名称：                       Hygon C86 7280 32-core Processor
16 步进：                           1
17CPU MHz：2108.234
18 BogoMIPS：                       3999.45
19 虚拟化：                         AMD-V
20 L1d 缓存：2 MiB
21 L1i 缓存：4 MiB
22 L2 缓存：32 MiB
23 L3 缓存：128 MiB
24 //注意这里和真实物理架构不一致，bios配置了Die Interleaving Enable
25 //表示每路内多个Die内存交织分配，这样整个一路就是一个大Die
26 NUMA 节点0 CPU：0-31,64-95  
27 NUMA 节点1 CPU：32-63,96-127
28 //enable die interleaving 后继续streaming测试
29 //最终测试结果表现就是7/15/23/31 core性能一致，因为默认一个numa内内存交织分配
30 //可以看到同一路下的四个die内存交织访问，所以4个node内存延时一样了（被平均），都不如就近快
31
32
33 time for i in $(seq 7 8 64); do echo $i; numactl -C $i -m 0 ./bin/stream -W 5 -N 5 -M 64M; done
34 7
35 STREAM copy latency: 1.48 nanoseconds
36 STREAM copy bandwidth: 10782.58 MB/sec
37 STREAM scale latency: 1.20 nanoseconds
38 STREAM scale bandwidth: 13364.38 MB/sec
39 STREAM add latency: 1.46 nanoseconds
40 STREAM add bandwidth: 16408.32 MB/sec
41 STREAM triad latency: 1.53 nanoseconds
42 STREAM triad bandwidth: 15696.00 MB/sec
43 15
44 STREAM copy latency: 1.51 nanoseconds
45 STREAM copy bandwidth: 10601.25 MB/sec
46 STREAM scale latency: 1.24 nanoseconds
47 STREAM scale bandwidth: 12855.87 MB/sec
48 STREAM add latency: 1.46 nanoseconds
49 STREAM add bandwidth: 16382.42 MB/sec
50 STREAM triad latency: 1.53 nanoseconds
51 STREAM triad bandwidth: 15691.48 MB/sec
52 23
53 STREAM copy latency: 1.50 nanoseconds
54 STREAM copy bandwidth: 10700.61 MB/sec
55 STREAM scale latency: 1.27 nanoseconds
56 STREAM scale bandwidth: 12634.63 MB/sec
57 STREAM add latency: 1.47 nanoseconds
58 STREAM add bandwidth: 16370.67 MB/sec
59 STREAM triad latency: 1.55 nanoseconds
60 STREAM triad bandwidth: 15455.75 MB/sec
61 31
62 STREAM copy latency: 1.50 nanoseconds
63 STREAM copy bandwidth: 10637.39 MB/sec
64 STREAM scale latency: 1.25 nanoseconds
65 STREAM scale bandwidth: 12778.99 MB/sec
66 STREAM add latency: 1.46 nanoseconds
67 STREAM add bandwidth: 16420.65 MB/sec
68 STREAM triad latency: 1.61 nanoseconds
69 STREAM triad bandwidth: 14946.80 MB/sec
70 39
71 STREAM copy latency: 2.35 nanoseconds
72 STREAM copy bandwidth: 6807.09 MB/sec
73 STREAM scale latency: 2.32 nanoseconds
74 STREAM scale bandwidth: 6906.93 MB/sec
75 STREAM add latency: 1.63 nanoseconds
76 STREAM add bandwidth: 14729.23 MB/sec
77 STREAM triad latency: 3.36 nanoseconds
78 STREAM triad bandwidth: 7151.67 MB/sec
79 47
80 STREAM copy latency: 2.31 nanoseconds
81 STREAM copy bandwidth: 6938.47 MB/sec