Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- root@nanopct4:~/tinymembench# taskset -c 5 ./tinymembench ; taskset -c 3 ./tinymembench
- tinymembench v0.4.9 (simple benchmark for memory throughput and latency)
- ==========================================================================
- == Memory bandwidth tests ==
- == ==
- == Note 1: 1MB = 1000000 bytes ==
- == Note 2: Results for 'copy' tests show how many bytes can be ==
- == copied per second (adding together read and writen ==
- == bytes would have provided twice higher numbers) ==
- == Note 3: 2-pass copy means that we are using a small temporary buffer ==
- == to first fetch data into it, and only then write it to the ==
- == destination (source -> L1 cache, L1 cache -> destination) ==
- == Note 4: If sample standard deviation exceeds 0.1%, it is shown in ==
- == brackets ==
- ==========================================================================
- C copy backwards : 4087.7 MB/s
- C copy backwards (32 byte blocks) : 4089.3 MB/s
- C copy backwards (64 byte blocks) : 4089.2 MB/s
- C copy : 4091.0 MB/s
- C copy prefetched (32 bytes step) : 4067.9 MB/s
- C copy prefetched (64 bytes step) : 4065.3 MB/s
- C 2-pass copy : 3595.6 MB/s
- C 2-pass copy prefetched (32 bytes step) : 3771.6 MB/s
- C 2-pass copy prefetched (64 bytes step) : 3762.5 MB/s
- ^C
- root@nanopct4:~/tinymembench# cpufreq-set -g performance
- root@nanopct4:~/tinymembench# taskset -c 5 ./tinymembench ; taskset -c 3 ./tinymembench
- tinymembench v0.4.9 (simple benchmark for memory throughput and latency)
- ==========================================================================
- == Memory bandwidth tests ==
- == ==
- == Note 1: 1MB = 1000000 bytes ==
- == Note 2: Results for 'copy' tests show how many bytes can be ==
- == copied per second (adding together read and writen ==
- == bytes would have provided twice higher numbers) ==
- == Note 3: 2-pass copy means that we are using a small temporary buffer ==
- == to first fetch data into it, and only then write it to the ==
- == destination (source -> L1 cache, L1 cache -> destination) ==
- == Note 4: If sample standard deviation exceeds 0.1%, it is shown in ==
- == brackets ==
- ==========================================================================
- C copy backwards : 4089.0 MB/s
- C copy backwards (32 byte blocks) : 4089.9 MB/s
- C copy backwards (64 byte blocks) : 4091.0 MB/s
- C copy : 4094.3 MB/s
- C copy prefetched (32 bytes step) : 4070.8 MB/s
- C copy prefetched (64 bytes step) : 4068.3 MB/s
- C 2-pass copy : 3597.1 MB/s
- C 2-pass copy prefetched (32 bytes step) : 3776.2 MB/s
- C 2-pass copy prefetched (64 bytes step) : 3764.9 MB/s
- C fill : 9035.1 MB/s (0.7%)
- C fill (shuffle within 16 byte blocks) : 9073.4 MB/s (0.2%)
- C fill (shuffle within 32 byte blocks) : 9079.7 MB/s
- C fill (shuffle within 64 byte blocks) : 9079.3 MB/s
- ---
- standard memcpy : 4094.2 MB/s
- standard memset : 9033.6 MB/s (0.7%)
- ---
- NEON LDP/STP copy : 4093.2 MB/s
- NEON LDP/STP copy pldl2strm (32 bytes step) : 4134.2 MB/s
- NEON LDP/STP copy pldl2strm (64 bytes step) : 4134.0 MB/s
- NEON LDP/STP copy pldl1keep (32 bytes step) : 4064.0 MB/s
- NEON LDP/STP copy pldl1keep (64 bytes step) : 4061.0 MB/s
- NEON LD1/ST1 copy : 4092.6 MB/s
- NEON STP fill : 9028.1 MB/s (0.7%)
- NEON STNP fill : 9035.0 MB/s
- ARM LDP/STP copy : 4093.3 MB/s
- ARM STP fill : 9031.8 MB/s (0.7%)
- ARM STNP fill : 9031.8 MB/s
- ==========================================================================
- == Memory latency test ==
- == ==
- == Average time is measured for random memory accesses in the buffers ==
- == of different sizes. The larger is the buffer, the more significant ==
- == are relative contributions of TLB, L1/L2 cache misses and SDRAM ==
- == accesses. For extremely large buffer sizes we are expecting to see ==
- == page table walk with several requests to SDRAM for almost every ==
- == memory access (though 64MiB is not nearly large enough to experience ==
- == this effect to its fullest). ==
- == ==
- == Note 1: All the numbers are representing extra time, which needs to ==
- == be added to L1 cache latency. The cycle timings for L1 cache ==
- == latency can be usually found in the processor documentation. ==
- == Note 2: Dual random read means that we are simultaneously performing ==
- == two independent memory accesses at a time. In the case if ==
- == the memory subsystem can't handle multiple outstanding ==
- == requests, dual random read has the same timings as two ==
- == single reads performed one after another. ==
- ==========================================================================
- block size : single random read / dual random read, [MADV_NOHUGEPAGE]
- 1024 : 0.0 ns / 0.0 ns
- 2048 : 0.0 ns / 0.0 ns
- 4096 : 0.0 ns / 0.0 ns
- 8192 : 0.0 ns / 0.0 ns
- 16384 : 0.0 ns / 0.0 ns
- 32768 : 0.0 ns / 0.0 ns
- 65536 : 4.5 ns / 7.1 ns
- 131072 : 6.8 ns / 9.6 ns
- 262144 : 9.8 ns / 12.8 ns
- 524288 : 11.3 ns / 14.6 ns
- 1048576 : 16.2 ns / 23.1 ns
- 2097152 : 93.2 ns / 140.0 ns
- 4194304 : 130.7 ns / 174.1 ns
- 8388608 : 155.1 ns / 194.7 ns
- 16777216 : 166.9 ns / 199.3 ns
- 33554432 : 173.6 ns / 206.7 ns
- 67108864 : 185.0 ns / 225.6 ns
- block size : single random read / dual random read, [MADV_HUGEPAGE]
- 1024 : 0.0 ns / 0.0 ns
- 2048 : 0.0 ns / 0.0 ns
- 4096 : 0.0 ns / 0.0 ns
- 8192 : 0.0 ns / 0.0 ns
- 16384 : 0.0 ns / 0.0 ns
- 32768 : 0.0 ns / 0.0 ns
- 65536 : 4.5 ns / 7.1 ns
- 131072 : 6.7 ns / 9.6 ns
- 262144 : 7.9 ns / 10.5 ns
- 524288 : 8.5 ns / 10.9 ns
- 1048576 : 12.4 ns / 17.8 ns
- 2097152 : 85.7 ns / 130.5 ns
- 4194304 : 122.5 ns / 163.8 ns
- 8388608 : 144.1 ns / 182.2 ns
- 16777216 : 152.9 ns / 185.3 ns
- 33554432 : 156.9 ns / 185.1 ns
- 67108864 : 161.6 ns / 187.2 ns
- tinymembench v0.4.9 (simple benchmark for memory throughput and latency)
- ==========================================================================
- == Memory bandwidth tests ==
- == ==
- == Note 1: 1MB = 1000000 bytes ==
- == Note 2: Results for 'copy' tests show how many bytes can be ==
- == copied per second (adding together read and writen ==
- == bytes would have provided twice higher numbers) ==
- == Note 3: 2-pass copy means that we are using a small temporary buffer ==
- == to first fetch data into it, and only then write it to the ==
- == destination (source -> L1 cache, L1 cache -> destination) ==
- == Note 4: If sample standard deviation exceeds 0.1%, it is shown in ==
- == brackets ==
- ==========================================================================
- C copy backwards : 1886.1 MB/s (1.6%)
- C copy backwards (32 byte blocks) : 1897.8 MB/s (3.4%)
- C copy backwards (64 byte blocks) : 1886.0 MB/s (2.8%)
- C copy : 2012.7 MB/s (0.7%)
- C copy prefetched (32 bytes step) : 1410.2 MB/s
- C copy prefetched (64 bytes step) : 1624.7 MB/s
- C 2-pass copy : 1615.8 MB/s
- C 2-pass copy prefetched (32 bytes step) : 1163.1 MB/s
- C 2-pass copy prefetched (64 bytes step) : 1142.3 MB/s
- C fill : 8427.2 MB/s (0.1%)
- C fill (shuffle within 16 byte blocks) : 8425.9 MB/s
- C fill (shuffle within 32 byte blocks) : 8424.7 MB/s
- C fill (shuffle within 64 byte blocks) : 8424.9 MB/s
- ---
- standard memcpy : 2031.3 MB/s
- standard memset : 8449.6 MB/s (0.2%)
- ---
- NEON LDP/STP copy : 2051.3 MB/s (0.2%)
- NEON LDP/STP copy pldl2strm (32 bytes step) : 1358.5 MB/s (1.3%)
- NEON LDP/STP copy pldl2strm (64 bytes step) : 1716.7 MB/s
- NEON LDP/STP copy pldl1keep (32 bytes step) : 2230.4 MB/s
- NEON LDP/STP copy pldl1keep (64 bytes step) : 2233.0 MB/s
- NEON LD1/ST1 copy : 2022.6 MB/s (2.1%)
- NEON STP fill : 8450.9 MB/s (0.2%)
- NEON STNP fill : 3514.7 MB/s (3.8%)
- ARM LDP/STP copy : 2048.3 MB/s (0.3%)
- ARM STP fill : 8449.6 MB/s (0.2%)
- ARM STNP fill : 3365.4 MB/s (2.9%)
- ==========================================================================
- == Memory latency test ==
- == ==
- == Average time is measured for random memory accesses in the buffers ==
- == of different sizes. The larger is the buffer, the more significant ==
- == are relative contributions of TLB, L1/L2 cache misses and SDRAM ==
- == accesses. For extremely large buffer sizes we are expecting to see ==
- == page table walk with several requests to SDRAM for almost every ==
- == memory access (though 64MiB is not nearly large enough to experience ==
- == this effect to its fullest). ==
- == ==
- == Note 1: All the numbers are representing extra time, which needs to ==
- == be added to L1 cache latency. The cycle timings for L1 cache ==
- == latency can be usually found in the processor documentation. ==
- == Note 2: Dual random read means that we are simultaneously performing ==
- == two independent memory accesses at a time. In the case if ==
- == the memory subsystem can't handle multiple outstanding ==
- == requests, dual random read has the same timings as two ==
- == single reads performed one after another. ==
- ==========================================================================
- block size : single random read / dual random read, [MADV_NOHUGEPAGE]
- 1024 : 0.0 ns / 0.0 ns
- 2048 : 0.0 ns / 0.0 ns
- 4096 : 0.0 ns / 0.0 ns
- 8192 : 0.0 ns / 0.0 ns
- 16384 : 0.0 ns / 0.0 ns
- 32768 : 0.0 ns / 0.0 ns
- 65536 : 4.8 ns / 8.2 ns
- 131072 : 7.4 ns / 11.1 ns
- 262144 : 8.7 ns / 12.2 ns
- 524288 : 9.8 ns / 13.8 ns
- 1048576 : 79.1 ns / 122.1 ns
- 2097152 : 116.9 ns / 157.6 ns
- 4194304 : 141.7 ns / 177.4 ns
- 8388608 : 155.1 ns / 187.1 ns
- 16777216 : 163.2 ns / 193.6 ns
- 33554432 : 167.3 ns / 197.8 ns
- 67108864 : 169.3 ns / 201.7 ns
- block size : single random read / dual random read, [MADV_HUGEPAGE]
- 1024 : 0.0 ns / 0.0 ns
- 2048 : 0.0 ns / 0.0 ns
- 4096 : 0.0 ns / 0.0 ns
- 8192 : 0.0 ns / 0.0 ns
- 16384 : 0.0 ns / 0.0 ns
- 32768 : 0.0 ns / 0.0 ns
- 65536 : 4.8 ns / 8.0 ns
- 131072 : 7.4 ns / 11.1 ns
- 262144 : 8.7 ns / 12.2 ns
- 524288 : 9.8 ns / 13.6 ns
- 1048576 : 79.1 ns / 122.0 ns
- 2097152 : 116.7 ns / 157.4 ns
- 4194304 : 135.5 ns / 169.1 ns
- 8388608 : 144.7 ns / 173.2 ns
- 16777216 : 149.6 ns / 174.7 ns
- 33554432 : 152.0 ns / 175.4 ns
- 67108864 : 153.3 ns / 175.6 ns
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement