test_fps.sh 文件内容:
#!/bin/sh
n=0
t1=`date +%s`
count=1000
while [ true ]; do
let n++
if [ $(($n % 10)) -eq 0 ]; then
echo -e ".\c"
fi
if [ $n -gt $count ]; then
break;
fi
cat /dev/zero > /dev/fb0 2>&1
done;
t2=`date +%s`
fps=$(($count/($t2 - $t1)))
echo -e "\ntest result: $fps fps"
widora tiny200s 800x480 测试结果:
# sh test.sh
....................................................................................................
test result: 37 fps
离线
测试方法, 记住开始和结束的时间戳,然后测试1000次写0到fb0,
用 1000除以时间差, 得到粗略的帧率。
离线
这么慢?我记得裸机跑过,有50多fps
离线
这么慢?我记得裸机跑过,有50多fps
这个脚本肯定不怎么准,粗略测试,用来练习一下脚本技术。
离线
这么慢?我记得裸机跑过,有50多fps
RGB模式 可以到60Hz吗?
-------------
-------------
Steward says:
-------------
we use oscilloscope to double check and hw can ouput 60hz
in kernel, I just accumulate and print it
and the result is 40hz
even though boost priority for innterrupt
the result is same as without boos priority
we also check with some engineers who developed f1c500s
they havent checked this when produce this ic
I mean no any verification with interrupt speed
so, I dont know whats the root cause
maybe, kernel or IC
if you know this, please tell me
thanks
离线
这样测试肯定测得不准吧,毕竟本身你运行的这个脚本就不是一直占用CPU的。
离线
#!/bin/sh
n=0
t1=`date +%s`
count=1000
while [ true ]; do
let n++
if [ $(($n % 10)) -eq 0 ]; then
echo -e ".\c"
fi
if [ $n -gt $count ]; then
break;
fi
djpeg -outfile /dev/null /usr/images/hehua.jpg
done;
t2=`date +%s`
fps=$(($count/($t2 - $t1)))
echo -e "\ntest result: $fps fps"
V3s neno指令加速的1280x720 jpeg文件解码粗略测试15fps:
# ./test.sh
....................................................................................................
test result: 15 fps
连续测试同一个文件jpeg解码,解码完成的结果直接扔/dev/null
如果把全志的编解码器搞起来, 至少应该 50fps吧?
离线
貌似发现了更精确的测试方法, 直接解码成 yuv420
libjpeg自带的tjbench:
# tjbench /usr/images/hehua.jpg -yuv -subsamp 420
Testing YUV planar encoding/decoding
>>>>> JPEG 4:2:0 --> BGR (Top-down) <<<<<
Image size: 1280 x 720
Decomp to YUV --> Frame rate: 34.059825 fps
Throughput: 31.389535 Megapixels/sec
YUV Decode --> Frame rate: 71.475664 fps
Throughput: 65.871972 Megapixels/sec
离线