7.2. 视频编解码性能 

视频编解码性能
- BM1684 性能测试
- BM1684X 性能测试

7.2.1. BM1684 性能测试 

本节我们列举了BM1684芯片对于视频编码、解码、转码的性能数据，此外，我们附上了每一个性能测试的测试命令，您可以按照表格中的测试命令进行验证。

编解码能力性能测试

芯片

场景

ffmpeg命令行

1684

decode

#!/bin/bash
rm -rf *.log *.txt
video=test_video/jellyfish-3-mbps-hd-h264.mkv
tpus=1
thread_num=8
for ((i=0; i<$tpus; i++))
do
{
    if [[ $tpus -gt `cat /proc/bmsophon/chip_num` ]];then
        echo "The number of tpu entered is greater than the actual number."
        return -1
    fi
    for ((j=0; j<$thread_num; j++))
    do
    {
        nohup ffmpeg -benchmark -output_format 101 -zero_copy 1 -sophon_idx ${i} -c:v h264_bm -i $video -f null /dev/null >> ${video}_thread_${j}_tpu_${i}.log 2>&1 &
    } &
    done
}
done
echo "Start Wait End..."
sleep 30
echo "Start Calc fps..."
for ((i=0; i<$tpus; i++))
do
    for ((j=0; j<$thread_num; j++))
    do
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1| awk -F '=' '{print $3}'|tr -d 'a-z'|tr -d ' ' >> fps.txt    #fps
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1| awk -F '=' '{print $2}'|tr -d 'a-z'|tr -d ' ' >> frame.txt    #frames
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 5|head -n 1| awk -F '=' '{print $4}'|tr -d 'a-z'|tr -d ' ' >> time.txt    #frames
        cat fps.txt | awk '{sum+=$1} END {print sum}' > fpssum.txt
        cat frame.txt | awk '{sum+=$1} END {print sum}' > framesum.txt
        cat time.txt | awk '{sum+=$1} END {print sum}' > timesum.txt
    done
done
    echo "chip_num: $tpus"
        echo "total_frames: $(<framesum.txt)"
        cat timesum.txt | awk '{task_num = "'$tpus'"*"'$thread_num'"} {print $1/task_num}' > timeavg.txt
        echo "avg_time: $(<timeavg.txt)"
        #echo "fps: $(<fpssum.txt)"
        f_sum=$(<framesum.txt)
        time_avg=$(<timeavg.txt)
        awk 'BEGIN{printf "%.2f\n",'$f_sum'/'$time_avg'}' > speed.txt
        speed=$(<speed.txt)
    echo "speed: ${speed}"
        rm -rf f*.txt
        rm -rf t*.txt
        rm -rf *.log

编解码能力性能测试

芯片

场景

ffmpeg命令行

1684

encode

#!/bin/bash
rm -rf *.log f*.txt t*.txt
video=test_video/jellyfish-3-mbps-hd-h264.mkv
tpus=1         #chip num: one SC5+ chip_num=3
thread_num=8   #8 cores
for ((i=0; i<$tpus; i++))
do
{
    if [[ $tpus -gt `cat /proc/bmsophon/chip_num` ]];then
        echo "The number of tpu entered is greater than the actual number."
        return -1
    fi
    for ((j=0; j<$thread_num; j++))
    do
    {
    nohup ffmpeg -benchmark -hwaccel bmcodec -hwaccel_device $i -c:v h264_bm -i $video  -c:v h264_bm -g 256 -b:v 3MB -enc-params "gop_preset=4:mb_rc=1:delta_qp=3:min_qp=20:max_qp=33" -f null /dev/null >> ${video}_thread_${j}_tpu_${i}.log 2>&1 &
    } &
    done
}
done
echo "Start Wait End..."
sleep 105
echo "Start Calc fps..."
for ((i=0; i<$tpus; i++))
do
    for ((j=0; j<$thread_num; j++))
    do
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1|awk -F "=" '{print $3}'|tr -d "a-z"|tr -d " " >> fpsall.txt
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1|awk -F "=" '{print $2}'|tr -d " "|tr -d "a-z" >> frameall.txt
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 5|head -n 1|awk -F "=" '{print $4}'|tr -d "a-z"|tr -d " " >> timeall.txt
        cat fpsall.txt | awk '{sum+=$1} END {print sum}' > fpssum.txt
        cat frameall.txt | awk '{sum+=$1} END {print sum}' > framesum.txt
        cat timeall.txt | awk '{sum+=$1} END {print sum}' > timesum.txt
    done
done
        echo "chip_num: $tpus"
        echo "total_frames: $(<framesum.txt)"
        cat timesum.txt | awk '{task_num = "'$tpus'"*"'$thread_num'"} {print $1/task_num}' > timeavg.txt
        echo "avg_time: $(<timeavg.txt)"
        #echo "fps: $(<fpssum.txt)"
    f_sum=$(<framesum.txt)
    time_avg=$(<timeavg.txt)
    awk 'BEGIN{printf "%.2f\n",'$f_sum'/'$time_avg'}' > speed.txt
    speed=$(<speed.txt)
    echo "speed: ${speed}"
        rm -rf f*.txt
        rm -rf t*.txt
        rm -rf *.log

编解码能力性能测试
芯片	场景	测试结果
1684	decode
1684	encode

编解码能力性能测试

芯片

场景

ffmpeg命令行

1684

transcode

转 32K 低码流：

ffmpeg -hwaccel bmcodec -hwaccel_device 0 -c:v h264_bm -output_format 101 -i test.264 -vf "scale_bm=352:288" -c:v h264_bm -g 256 -b:v 32K -y result.264

转 1Mbps 高码流：

ffmpeg -hwaccel bmcodec -hwaccel_device 0 -c:v h264_bm -output_format 101 -i test.264 -vf "scale_bm=352:288" -c:v h264_bm -g 256 -b:v 1M -y result.ts

-hwaccel 使用硬件API
-hwaccel_device 指定设备，使用bm-smi查看
-c:v 解码器
-output_format 输出数据的格式。设为0，则输出线性排列的未压缩数据；设为101，则输出压缩数据。缺省值为0。推荐设置为101，输出压缩数据。可以节省内存、节省带宽。输出的压缩数据，可以调用scale_bm filter解压缩成正常的YUV数据。
-i 输入
-vf ffmpeg过滤器，scale_bm将frame resize为目标宽高
-c:v 编码器
-g 关键帧间隔控制
-b:v 输出码流
-y 输出时覆盖原文件result.ts 输出文件

7.2.2. BM1684X 性能测试 

本节我们列举了BM1684X芯片对于视频编码、解码、转码的性能数据，此外，我们附上了每一个性能测试的测试命令，您可以按照表格中的测试命令进行验证。

编解码能力性能测试

芯片

场景

ffmpeg命令行

1684X

decode

#!/bin/bash
rm -rf *.log *.txt
video=test_video/jellyfish-3-mbps-hd-h264.mkv
tpus=1
thread_num=8
for ((i=0; i<$tpus; i++))
do
{
    if [[ $tpus -gt `cat /proc/bmsophon/chip_num` ]];then
        echo "The number of tpu entered is greater than the actual number."
        return -1
    fi
    for ((j=0; j<$thread_num; j++))
    do
    {
        nohup ffmpeg -benchmark -output_format 101 -zero_copy 1 -sophon_idx ${i} -c:v h264_bm -i $video -f null /dev/null >> ${video}_thread_${j}_tpu_${i}.log 2>&1 &
    } &
    done
}
done
echo "Start Wait End..."
sleep 30
echo "Start Calc fps..."
for ((i=0; i<$tpus; i++))
do
    for ((j=0; j<$thread_num; j++))
    do
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1| awk -F '=' '{print $3}'|tr -d 'a-z'|tr -d ' ' >> fps.txt    #fps
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1| awk -F '=' '{print $2}'|tr -d 'a-z'|tr -d ' ' >> frame.txt    #frames
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 5|head -n 1| awk -F '=' '{print $4}'|tr -d 'a-z'|tr -d ' ' >> time.txt    #frames
        cat fps.txt | awk '{sum+=$1} END {print sum}' > fpssum.txt
        cat frame.txt | awk '{sum+=$1} END {print sum}' > framesum.txt
        cat time.txt | awk '{sum+=$1} END {print sum}' > timesum.txt
    done
done
    echo "chip_num: $tpus"
        echo "total_frames: $(<framesum.txt)"
        cat timesum.txt | awk '{task_num = "'$tpus'"*"'$thread_num'"} {print $1/task_num}' > timeavg.txt
        echo "avg_time: $(<timeavg.txt)"
        #echo "fps: $(<fpssum.txt)"
        f_sum=$(<framesum.txt)
        time_avg=$(<timeavg.txt)
        awk 'BEGIN{printf "%.2f\n",'$f_sum'/'$time_avg'}' > speed.txt
        speed=$(<speed.txt)
    echo "speed: ${speed}"
        rm -rf f*.txt
        rm -rf t*.txt
        rm -rf *.log

编解码能力性能测试

芯片

场景

ffmpeg命令行

1684X

encode

#!/bin/bash
rm -rf *.log f*.txt t*.txt
video=test_video/jellyfish-3-mbps-hd-h264.mkv
tpus=1         #chip num: one SC5+ chip_num=3
thread_num=8   #8 cores
for ((i=0; i<$tpus; i++))
do
{
    if [[ $tpus -gt `cat /proc/bmsophon/chip_num` ]];then
        echo "The number of tpu entered is greater than the actual number."
        return -1
    fi
    for ((j=0; j<$thread_num; j++))
    do
    {
    nohup ffmpeg -benchmark -hwaccel bmcodec -hwaccel_device $i -c:v h264_bm -i $video  -c:v h264_bm -g 256 -b:v 3MB -enc-params "gop_preset=4:mb_rc=1:delta_qp=3:min_qp=20:max_qp=33" -f null /dev/null >> ${video}_thread_${j}_tpu_${i}.log 2>&1 &
    } &
    done
}
done
echo "Start Wait End..."
sleep 105
echo "Start Calc fps..."
for ((i=0; i<$tpus; i++))
do
    for ((j=0; j<$thread_num; j++))
    do
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1|awk -F "=" '{print $3}'|tr -d "a-z"|tr -d " " >> fpsall.txt
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 7|head -n 1|awk -F "=" '{print $2}'|tr -d " "|tr -d "a-z" >> frameall.txt
        cat ${video}_thread_${j}_tpu_${i}.log | tail -n 5|head -n 1|awk -F "=" '{print $4}'|tr -d "a-z"|tr -d " " >> timeall.txt
        cat fpsall.txt | awk '{sum+=$1} END {print sum}' > fpssum.txt
        cat frameall.txt | awk '{sum+=$1} END {print sum}' > framesum.txt
        cat timeall.txt | awk '{sum+=$1} END {print sum}' > timesum.txt
    done
done
        echo "chip_num: $tpus"
        echo "total_frames: $(<framesum.txt)"
        cat timesum.txt | awk '{task_num = "'$tpus'"*"'$thread_num'"} {print $1/task_num}' > timeavg.txt
        echo "avg_time: $(<timeavg.txt)"
        #echo "fps: $(<fpssum.txt)"
    f_sum=$(<framesum.txt)
    time_avg=$(<timeavg.txt)
    awk 'BEGIN{printf "%.2f\n",'$f_sum'/'$time_avg'}' > speed.txt
    speed=$(<speed.txt)
    echo "speed: ${speed}"
        rm -rf f*.txt
        rm -rf t*.txt
        rm -rf *.log

编解码能力性能测试
芯片	场景	测试结果
1684X	decode
1684X	encode

编解码能力性能测试

芯片

场景

ffmpeg命令行

1684X

transcode

转 32K 低码流：

ffmpeg -hwaccel bmcodec -hwaccel_device 0 -c:v h264_bm -output_format 101 -i test.264 -vf "scale_bm=352:288" -c:v h264_bm -g 256 -b:v 32K -y result.264

转 1Mbps 高码流：

ffmpeg -hwaccel bmcodec -hwaccel_device 0 -c:v h264_bm -output_format 101 -i test.264 -vf "scale_bm=352:288" -c:v h264_bm -g 256 -b:v 1M -y result.ts

-hwaccel 使用硬件API
-hwaccel_device 指定设备，使用bm-smi查看
-c:v 解码器
-output_format 输出数据的格式。设为0，则输出线性排列的未压缩数据；设为101，则输出压缩数据。缺省值为0。推荐设置为101，输出压缩数据。可以节省内存、节省带宽。输出的压缩数据，可以调用scale_bm filter解压缩成正常的YUV数据。
-i 输入
-vf ffmpeg过滤器，scale_bm将frame resize为目标宽高
-c:v 编码器
-g 关键帧间隔控制
-b:v 输出码流
-y 输出时覆盖原文件result.ts 输出文件

7.2. 视频编解码性能

7.2.1. BM1684 性能测试

7.2.2. BM1684X 性能测试

7.2. 视频编解码性能 

7.2.1. BM1684 性能测试 

7.2.2. BM1684X 性能测试 