12-12-2018 05:51 AM
I am trying to synthesize the code which contains two large arrays each of size 256. To get initiation interval 1,
#pragma HLS array_partition
is used for those arrays. Initiation interval 1 is achieved, but synthesis is hanging at RTL generation in vivado hls 2018.2 . However, the code is synthesizing successfully in vivado hls 2017.2. Can you suggest a solution for this problem.
12-12-2018 05:54 AM
Tood less info to say anything concrete. But Vivado varies from version to version.
My experience shows Vivado future versions tend to have bugs that older versions don't have.
12-12-2018 06:29 AM
Thanks for the reply... But the answer you gave is very generic that we all know.. any solution to solve or identify the issue that I m facing??
Apart from that we are facing so many intiation interval issues also with data flow pragma too... It's not doing flow properly and finally dusturbimg the interval of the ip where as I m getting interval 1 for 2017.2 tool if we use...
12-13-2018 01:12 AM
12-13-2018 08:07 AM
It's regarding to VivadoHLS CSynth... Still struked with issue and it's not resolved. Please help us ASAP. It became a critical path blocker for us now. Recently only we updated from 2017.2 to 2018.2 , after doing that, regarding to HLS we are facing so many bugs in HLS. Even
#Pragma HLS dataflow
also not working properly ie.., in the top level function where dataflow used , with 2018.2 HLS Synth we are not getting initiation interval 1 where as for the same codes we are getting interval as 1 in 2017.2 synth of HLS. Please address these things and solve ASAP. Please provide us a solution.
12-13-2018 08:24 AM
Okay. I will move this topic to the correct board to get help from HLS community.
Did you try running the HLS Csynth after removing the directive file? Just a small check to see if any pragma is resulting into the issue. Please share the compete log file here.
Also, if there any loops with high iterations, try to reduce the value and re-run it to see if it passes, this will help to understand the hang.
12-13-2018 09:20 PM
12-14-2018 05:44 AM
It seems that you have specified reverse version names for the attached logs. From the logs, it seems that the tool got stuck at the scheduling phase and there is nothing much info on why the tool got stuck. So, can you please give a try in VIvado HLS 2018.3 and if you are still facing the same issue then I would suggest you share the archived project to check this issue at our end.
02-10-2019 09:07 AM
I just wanted to mention that there have been some tool repairs and enhancements in the dataflow processing in the 2018.x software. One of the side-effects of this has been that some forms of code (code organization/structure) that passed prior to this release have been affected. While new bugs can of course be introduced with any code change, in this case it was older HLS software allowing code through the tools that it should not have. In some cases that code worked, in others, it allowed incorrect code to be synthesized which was then caught in CoSim.
For this case, we need to see your code that produces this error. Either your original code, or an example that contains the same code structure which triggers this condition. Can you either post that to this thread, or send an IM to @anusheel or myself?