~linaro-toolchain-dev/cortex-strings/trunk

Added some scripts to run the benchmarks and to plot the results.

added added

removed removed

#!/bin/bash

# Benchmark all variants of all functions

# memcpy: 9.905206 s for 10000000 calls to memcpy of 3328 bytes. ~3204.202 MB/s

# To run for 10 s, transfer 30 GB. So loops is 30e9 / size

for t in memcpy memset memchr strcpy strlen strcmp strchr bounce; do

for variant in try-all try-bionic try-csl try-glibc try-newlib try-none try-plain; do

#for size in 256 512 1024 2048 4096 8192 16384 $(seq 4 4 128); do

for size in $(seq 4 4 128); do

if [[ $size -le 128 ]]; then

loops=$(( 250000000 - 1000000 * $size ))

else

loops=$(( 30000000000 / $size ))

./$variant -t $t -c $size -l $loops

done

250000000 - 1000000*size