root@renegade:~# 7zr b 7-Zip (a) [64] 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21 p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,4 CPUs LE) LE CPU Freq: 932 1381 1381 1382 1381 1381 1381 1382 1380 RAM size: 3924 MB, # CPU hardware threads: 4 RAM usage: 882 MB, # Benchmark threads: 4 Compressing | Decompressing Dict Speed Usage R/U Rating | Speed Usage R/U Rating KiB/s % MIPS MIPS | KiB/s % MIPS MIPS 22: 2225 312 695 2165 | 55372 361 1309 4724 23: 2185 317 703 2227 | 57714 382 1306 4994 24: 2127 323 708 2287 | 58544 396 1296 5139 25: 2139 338 723 2443 | 56202 397 1258 5002 ---------------------------------- | ------------------------------ Avr: 322 707 2281 | 384 1292 4965 Tot: 353 1000 3623 root@renegade:~/tinymembench# ./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 : 1700.1 MB/s (1.6%) C copy backwards (32 byte blocks) : 1743.9 MB/s (2.3%) C copy backwards (64 byte blocks) : 1721.9 MB/s (2.6%) C copy : 1707.6 MB/s (1.0%) C copy prefetched (32 bytes step) : 1548.5 MB/s (0.2%) C copy prefetched (64 bytes step) : 1827.9 MB/s (10.0%) C 2-pass copy : 1804.7 MB/s (5.0%) C 2-pass copy prefetched (32 bytes step) : 1360.1 MB/s (3.0%) C 2-pass copy prefetched (64 bytes step) : 1293.9 MB/s (0.3%) C fill : 7072.5 MB/s (1.2%) C fill (shuffle within 16 byte blocks) : 7016.5 MB/s (7.4%) C fill (shuffle within 32 byte blocks) : 6955.3 MB/s (0.4%) C fill (shuffle within 64 byte blocks) : 6931.6 MB/s (2.2%) --- standard memcpy : 1615.7 MB/s (4.0%) standard memset : 7486.8 MB/s (4.5%) --- NEON LDP/STP copy : 1911.9 MB/s (1.6%) NEON LDP/STP copy pldl2strm (32 bytes step) : 1573.4 MB/s (0.6%) NEON LDP/STP copy pldl2strm (64 bytes step) : 1827.3 MB/s NEON LDP/STP copy pldl1keep (32 bytes step) : 1969.8 MB/s (5.8%) NEON LDP/STP copy pldl1keep (64 bytes step) : 1974.2 MB/s NEON LD1/ST1 copy : 1876.4 MB/s (2.3%) NEON STP fill : 7485.5 MB/s NEON STNP fill : 2580.4 MB/s (1.3%) ARM LDP/STP copy : 1910.5 MB/s ARM STP fill : 7485.3 MB/s ARM STNP fill : 2522.3 MB/s (0.5%) ========================================================================== == Framebuffer read tests. == == == == Many ARM devices use a part of the system memory as the framebuffer, == == typically mapped as uncached but with write-combining enabled. == == Writes to such framebuffers are quite fast, but reads are much == == slower and very sensitive to the alignment and the selection of == == CPU instructions which are used for accessing memory. == == == == Many x86 systems allocate the framebuffer in the GPU memory, == == accessible for the CPU via a relatively slow PCI-E bus. Moreover, == == PCI-E is asymmetric and handles reads a lot worse than writes. == == == == If uncached framebuffer reads are reasonably fast (at least 100 MB/s == == or preferably >300 MB/s), then using the shadow framebuffer layer == == is not necessary in Xorg DDX drivers, resulting in a nice overall == == performance improvement. For example, the xf86-video-fbturbo DDX == == uses this trick. == ========================================================================== NEON LDP/STP copy (from framebuffer) : 359.5 MB/s NEON LDP/STP 2-pass copy (from framebuffer) : 356.2 MB/s NEON LD1/ST1 copy (from framebuffer) : 99.0 MB/s (0.5%) NEON LD1/ST1 2-pass copy (from framebuffer) : 98.8 MB/s (3.8%) ARM LDP/STP copy (from framebuffer) : 190.9 MB/s ARM LDP/STP 2-pass copy (from framebuffer) : 190.1 MB/s (0.4%) ========================================================================== == 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 : 0.1 ns / 0.1 ns 65536 : 5.0 ns / 8.5 ns 131072 : 7.6 ns / 11.7 ns 262144 : 10.4 ns / 15.7 ns 524288 : 57.6 ns / 90.9 ns 1048576 : 86.2 ns / 122.3 ns 2097152 : 101.6 ns / 135.6 ns 4194304 : 114.2 ns / 146.7 ns 8388608 : 121.3 ns / 152.9 ns 16777216 : 126.4 ns / 157.5 ns 33554432 : 130.0 ns / 160.8 ns 67108864 : 141.8 ns / 181.8 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 : 0.1 ns / 0.1 ns 65536 : 4.9 ns / 8.4 ns 131072 : 7.6 ns / 11.9 ns 262144 : 10.4 ns / 15.7 ns 524288 : 57.6 ns / 90.8 ns 1048576 : 86.2 ns / 122.3 ns 2097152 : 100.6 ns / 134.3 ns 4194304 : 107.7 ns / 139.3 ns 8388608 : 111.2 ns / 141.5 ns 16777216 : 112.9 ns / 142.5 ns 33554432 : 113.6 ns / 143.0 ns 67108864 : 113.9 ns / 143.3 ns root@renegade:~# sysbench --test=cpu --cpu-max-prime=20000 run --num-threads=$(grep -c '^processor' /proc/cpuinfo) sysbench 0.4.12: multi-threaded system evaluation benchmark Running the test with following options: Number of threads: 4 Doing CPU performance benchmark Threads started! Done. Maximum prime number checked in CPU test: 20000 Test execution summary: total time: 6.5578s total number of events: 10000 total time taken by event execution: 26.2197 per-request statistics: min: 2.61ms avg: 2.62ms max: 4.64ms approx. 95 percentile: 2.62ms Threads fairness: events (avg/stddev): 2500.0000/4.36 execution time (avg/stddev): 6.5549/0.00