XSIM libraries are not compiled into VHDL_LIBS

[dbee-lumacron]

When running XSIM on latest vivado tools 2025.2 with:

source <path-to-OsvvmLibraries>/OsvvmLibraries/Scripts/StartXSIM.tcl
build  ../OsvvmLibraries
build  ../OsvvmLibraries/RunDemoTests.pro

I can see that VHDL_LIBS/XSIM-2025.2 is created, but it's never populated.

Instead, everything goes into xsim.dir. I assume this means that the library directory scripting isn't working:

• SetLibraryDirectory [LibraryDirectory]
  • Set the directory in which the libraries will be created to LibraryDirectory.
  • If LibraryDirectory is not specified, use the CurrentSimulationDirectory.
  • By default, libraries are created in <LibraryDirectory>/VHDL_LIBS/<tool version>/.

I also observe that we get folders in xsim.dir for each testcase e.g. osvvm_TbAxi4.TbAxi4_DemoMemoryReadWrite1.

Do the same set of steps for Questa results in expected behaviour. VHDL_LIBS/Base-2025.1 and doesn't include osvvm_TbAxi4.TbAxi4_DemoMemoryReadWrite1.

P.S. I only got two failures running xsim with RunDemoTests, both were in AxiStream and related to log files.

%% 3410000 ps Alert ERROR in Default, diff OsvvmTemp_XSIM/TbStream_SendGetDemo1.log C:/Users/dbee/repos/OsvvmLibraries/AXI4/AxiStream/testbench/../ValidatedResults//TbStream_SendGetDemo1.log, File miscompare on line 1

%% 1240000 ps Alert ERROR in Default, diff OsvvmTemp_XSIM/TbStream_SendGetPacketBurst1.log C:/Users/dbee/repos/OsvvmLibraries/AXI4/AxiStream/testbench/../ValidatedResults//TbStream_SendGetPacketBurst1.log, File miscompare on line 2

Is it perhaps a windows-specific issue? Let me know if I should create a separate issue.

[JimLewis]

Hi Daniel,
I have not yet figured out a way to get XSIM to use OSVVM's library directory. It has been some time since I checked to see if there are any updates regarding this. If anyone has successfully got xsim to take their library, I would appreciate input.

Miscompares are due to XSIM not understanding type time. If you print and tell it 1 ns, it prints with ns units. If you tell it to print using 1000 ps, it prints in ps units. If your simulator resolution is 1 ps, and you store 1 ns into a time variable, and then use that variable to control the time resolution for printing, it seems to see it as 1000 ps. At one point I had an issue in to them that is indirectly related to this. It would not hurt to submit another issue to them. They have made much progress since the last issues I submitted. And I submitted them in priority order and with held low ranked ones. I am not sure if this one made the cut or not since it was not a full flat out failure.

It is not bad for what it is. If you have not tried NVC though you should.

Best Regards,
Jim