Open lomotos10 opened 1 year ago
I update my xrt and the shell of u280 solved this problem
@xtfan Can you share your xbmgmt examine
details? I've updated my XRT and DPTs (both Deployment and Development) to 2022.1 but running host
fails... Thanks in advance!
xbmgmt examine
:
$ xbmgmt examine
System Configuration
OS Name : Linux
Release : 4.15.0-192-generic
Version : #203-Ubuntu SMP Wed Aug 10 17:40:03 UTC 2022
Machine : x86_64
CPU Cores : 12
Memory : 32156 MB
Distribution : Ubuntu 18.04.6 LTS
GLIBC : 2.27
Model : To Be Filled By O.E.M.
XRT
Version : 2.13.479
Branch : 2022.1
Hash : 5e92a513c6950e79638b1a879ddb882da34fc683
Hash Date : 2022-06-25 02:08:32
XOCL : 2.13.479, 5e92a513c6950e79638b1a879ddb882da34fc683
XCLMGMT : 2.13.479, 5e92a513c6950e79638b1a879ddb882da34fc683
Devices present
BDF : Shell Platform UUID Device ID Device Ready*
[0000:2b:00.0] : xilinx_u280_gen3x16_xdma_base_1 283BAB8F-654D-8674-968F-4DA57F7FA5D7 mgmt(inst=11008) Yes
* Devices that are not ready will have reduced functionality when using XRT tools
Run result:
$ ./host ../build_dir.hw.xilinx_u280_gen3x16_xdma_1_202211_1/network.xclbin
Found Platform
Platform Name: Xilinx
INFO: Reading ../build_dir.hw.xilinx_u280_gen3x16_xdma_1_202211_1/network.xclbin
Loading: '../build_dir.hw.xilinx_u280_gen3x16_xdma_1_202211_1/network.xclbin'
Trying to program device[0]: xilinx_u280_gen3x16_xdma_base_1
Device[0]: program successful!
local_IP:a01d498, boardNum:0
XRT build version: 2.13.479
Build hash: 5e92a513c6950e79638b1a879ddb882da34fc683
Build date: 2022-06-25 02:08:32
Git branch: 2022.1
PID: 18860
UID: 1000
[Tue Sep 13 07:12:05 2022 GMT]
HOST: [hostname]
EXE: [base_dir]/host/host
[XRT] ERROR: std::bad_alloc
host/iperf_krnl/host.cpp:159 Error calling err = network_kernel.setArg(4, buffer_r2), error code is: -5
Hi,I also encounter this error. I change the size of the buffer in the host.cpp to get rid of this error.
@xtfan Thanks! May I ask exactly how you changed the size? I tried increasing the buffer size by 2x or more, but this didn't help the situation.
@lomotos10 When I change the DATA_SIZE to be 6250000, the error is gone. But I don't know the negative impact of this change.
@xtfan Thank you, now ./host
works. To be clear, does the application still work for you? (e.g. iperf for iperf_krnl)
@lomotos10 I have no any 100G NIC or switch for experiment so I just connect two U280 directly that run the same iperf_krnl. It seems that each FPGA card can send and receive packages from each other according to the debug ila. But I can't determine whether each of them is working correctly. Actually, I have tried corundum project (opensource) as a 100G NIC. However, the corundum failed to establish link with the 100G cmac in this project. Until now, I still can't run iperf on the server to test the iperf_krnl due to the lack of 100G NIC.
I'm using U250 with Vitis 2022.2 and I encounter this same error. Is there a way to fix this error. Using U280 give the same error.
When I contacted the developers by mail I was redirected to the https://github.com/fpgasystems/fpga-network-stack project because that project is being maintained, unlike this one.
I tried building the project on the new
vitis_2022_1
branch targeting U280, and it resulted in the following error onmake all
:There seems to be a mismatch between the Vitis and DTP versions specified in the README. (Vitis 2022.1, DTP xilinx_u280_xdma_201920_3)
Maybe the README needs an update, or is there a way that I can make the build work?