atlas200 AI的PCIE试玩

2 篇文章 0 订阅
1 篇文章 0 订阅

最近开始玩atlas200了,先上一张美图。

今天先看看pcie的速度吧。

基本信息:

1、atlas200作为rc,FPGA 690T作为ep端;

2、atlas200和690T都支持PCIE3.0 4X,但是目前我们只用到2X;

3、EP的驱动使用XDMA,简单方便。

速度测试:

1、h2c(RC到EP)

./dma_to_device -v 1 -d /dev/xdma0_h2c_0 -f ./datafile_32M.bin -s 0x100000 -c 10
打印如下:
dev /dev/xdma0_h2c_0, addr 0x0, aperture 0x0, size 0x100000, offset 0x0, count 10
host buffer 0x101000 = 0xe7ffcc880000
host buffer 0x101000 = 0xe7ffcc880000
#0: CLOCK_MONOTONIC 0.000920677 sec. write 1048576 bytes
#1: CLOCK_MONOTONIC 0.000834687 sec. write 1048576 bytes
#2: CLOCK_MONOTONIC 0.000833386 sec. write 1048576 bytes
#3: CLOCK_MONOTONIC 0.000830729 sec. write 1048576 bytes
#4: CLOCK_MONOTONIC 0.000834063 sec. write 1048576 bytes
#5: CLOCK_MONOTONIC 0.000830521 sec. write 1048576 bytes
#6: CLOCK_MONOTONIC 0.000828489 sec. write 1048576 bytes
#7: CLOCK_MONOTONIC 0.000848438 sec. write 1048576 bytes
#8: CLOCK_MONOTONIC 0.000823958 sec. write 1048576 bytes
#9: CLOCK_MONOTONIC 0.000829688 sec. write 1048576 bytes
** Avg time device /dev/xdma0_h2c_0, total time 8414636 nsec, avg_time = 841463.625000, size = 1048576, BW = 1246.133545 
/dev/xdma0_h2c_0 ** Average BW = 1048576, 1246.133545

2、c2h(EP到RC)

./dma_from_device -v 1 -d /dev/xdma0_c2h_0 -f output_datafile_4K.bin -s 0x100000 -c 10
打印如下:
dma0_c2h_0 -f output_datafile_4K.bin -s 0x100000 -c 10from_device -v 1 -d /dev/xd
dev /dev/xdma0_c2h_0, addr 0x0, aperture 0x0, size 0x100000, offset 0x0, count 10
host buffer 0x101000, 0xe7ffd6b65000.
#0: CLOCK_MONOTONIC 0.001695312 sec. read 1048576/1048576 bytes
#1: CLOCK_MONOTONIC 0.000915104 sec. read 1048576/1048576 bytes
#2: CLOCK_MONOTONIC 0.000908333 sec. read 1048576/1048576 bytes
#3: CLOCK_MONOTONIC 0.000935625 sec. read 1048576/1048576 bytes
#4: CLOCK_MONOTONIC 0.000925052 sec. read 1048576/1048576 bytes
#5: CLOCK_MONOTONIC 0.000914531 sec. read 1048576/1048576 bytes
#6: CLOCK_MONOTONIC 0.000902656 sec. read 1048576/1048576 bytes
#7: CLOCK_MONOTONIC 0.000913854 sec. read 1048576/1048576 bytes
#8: CLOCK_MONOTONIC 0.000918334 sec. read 1048576/1048576 bytes
#9: CLOCK_MONOTONIC 0.000920104 sec. read 1048576/1048576 bytes
** Avg time device /dev/xdma0_c2h_0, total time 9948905 nsec, avg_time = 994890.500000, size = 1048576, BW = 1053.961182 
/dev/xdma0_c2h_0 ** Average BW = 1048576, 1053.961182

总结:

fpga到cpu的速度是1000MB,pcie2.0 4x的理论速度为2000MB;

cpu到fpga的速度是1250MB,pcie2.0 4x的理论速度为2000MB。

速度有待提升,具体可参考:HI3559AV100和FPGA 7K690T的PCIE接口调试记录-续_hi3559 + fpga-CSDN博客

  • 21
    点赞
  • 20
    收藏
    觉得还不错? 一键收藏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值