Relative paths in run_modelsim.py and other changes to make Modelsim work on non-Utah machines

[nachiket]

---

name: Pull request
about: Push a change to this project

Motivate of the pull request

• To address an existing issue. If so, please provide a link to the issue. Standard-cell flow #225 (later comments on vvp hanging)
• Breaking new feature. If so, please decribe details in the description part.

Describe the technical details

• Modelsim simulation backend isn’t operational.

1. Removed the end_flow_with_test line from end of task.conf file
2. Modified run_modelsim.py to use relative paths for task_dir instead of assuming OPENFPGA_ROOT contains all task files
3. Tweaked openfpga shell script to adjust the simulation TCL file path generated by the modelsim backend.
4. Need to add —modelsim_ini runtime flag to customize modelsim path to your own machines installation. This should be explained in the documentation.

Which part of the code base require a change

In general, modification on existing submodules are not acceptable. You should push changes to upstream.

• VPR
• OpenFPGA libraries
• FPGA-Verilog
• FPGA-Bitstream
• FPGA-SDC
• FPGA-SPICE
• Flow scripts
• Architecture library
• Cell library

Checklist of the pull request

• Require code changes.
• Require new tests to be added
• Require an update on documentation

Impact of the pull request

• Require a change on Quality of Results (QoR)
• Break back-compatibility. If so, please list who may be influenced.

[ganeshgore]

LGTM, these features are not tested on CI. @tangxifan You can merge if everything is good.

[nachiket]

You should check why the Modelsim tcl files were being generated in a sub folder SimulationDeck for the write_verilog_testbench command. I couldn’t figure out how to make run_modelsim pick them up from that sub folder without hard coding in the paths so I just modified the openfpga shell script instead to drop that folder path.

…

On Feb 19, 2021, at 10:05 PM, ganeshgore ***@***.***> wrote:  LGTM, these features are not tested on CI. @tangxifan You can merge if everything is good. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[tangxifan]

You should check why the Modelsim tcl files were being generated in a sub folder SimulationDeck for the write_verilog_testbench command. I couldn’t figure out how to make run_modelsim pick them up from that sub folder without hard coding in the paths so I just modified the openfpga shell script instead to drop that folder path.
…
On Feb 19, 2021, at 10:05 PM, ganeshgore @.***> wrote:  LGTM, these features are not tested on CI. @tangxifan You can merge if everything is good. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

That is a valid solution in my opinion.

[tangxifan]

Can you avoid to modifying existing script templates? We can create a new template script with this feature.

[nachiket]

I agree. Is there a way to use run_modelsim.py with this script unmodified? Id rather not have to delete the simulationdeck folder path at all but can’t see how else to use the Modelsim flow.

…

On Feb 19, 2021, at 10:44 PM, tangxifan ***@***.***> wrote:  @tangxifan requested changes on this pull request. In openfpga_flow/openfpga_shell_scripts/fix_device_route_chan_width_example_script.openfpga: > @@ -55,7 +55,7 @@ write_fabric_verilog --file ./SRC --explicit_port_mapping --include_timing --pri # - Enable top-level testbench which is a full verification including programming circuit and core logic of FPGA # - Enable pre-configured top-level testbench which is a fast verification skipping programming phase # - Simulation ini file is optional and is needed only when you need to interface different HDL simulators using openfpga flow-run scripts -write_verilog_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --print_top_testbench --print_preconfig_top_testbench --print_simulation_ini ./SimulationDeck/simulation_deck.ini --include_signal_init --support_icarus_simulator --explicit_port_mapping +write_verilog_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --print_top_testbench --print_preconfig_top_testbench --print_simulation_ini ./simulation_deck.ini --include_signal_init --explicit_port_mapping Can you avoid to modifying existing script templates? We can create a new template script with this feature. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[tangxifan]

I agree. Is there a way to use run_modelsim.py with this script unmodified? Id rather not have to delete the simulationdeck folder path at all but can’t see how else to use the Modelsim flow.
…
On Feb 19, 2021, at 10:44 PM, tangxifan @.***> wrote:  @tangxifan requested changes on this pull request. In openfpga_flow/openfpga_shell_scripts/fix_device_route_chan_width_example_script.openfpga: > @@ -55,7 +55,7 @@ write_fabric_verilog --file ./SRC --explicit_port_mapping --include_timing --pri # - Enable top-level testbench which is a full verification including programming circuit and core logic of FPGA # - Enable pre-configured top-level testbench which is a fast verification skipping programming phase # - Simulation ini file is optional and is needed only when you need to interface different HDL simulators using openfpga flow-run scripts -write_verilog_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --print_top_testbench --print_preconfig_top_testbench --print_simulation_ini ./SimulationDeck/simulation_deck.ini --include_signal_init --support_icarus_simulator --explicit_port_mapping +write_verilog_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --print_top_testbench --print_preconfig_top_testbench --print_simulation_ini ./simulation_deck.ini --include_signal_init --explicit_port_mapping Can you avoid to modifying existing script templates? We can create a new template script with this feature. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

@nachiket You can define variables in the openfpga shell script and then define the value of the variables in task configuration file. For example, you can define ${OPENFPGA_SIMULATION_DECK_INI_FILE} in the openfpga shell script.
We have an example here.

[nachiket]

Hmm, I don’t immediately see it in the linked example. Also, from what I can tell, the run_modelsim.py script needs to have access to this variable for picking up the correct sim folder.

…

On Feb 22, 2021, at 11:06 AM, tangxifan ***@***.***> wrote:  I agree. Is there a way to use run_modelsim.py with this script unmodified? Id rather not have to delete the simulationdeck folder path at all but can’t see how else to use the Modelsim flow. … On Feb 19, 2021, at 10:44 PM, tangxifan @.***> wrote:  @tangxifan requested changes on this pull request. In openfpga_flow/openfpga_shell_scripts/fix_device_route_chan_width_example_script.openfpga: > @@ -55,7 +55,7 @@ write_fabric_verilog --file ./SRC --explicit_port_mapping --include_timing --pri # - Enable top-level testbench which is a full verification including programming circuit and core logic of FPGA # - Enable pre-configured top-level testbench which is a fast verification skipping programming phase # - Simulation ini file is optional and is needed only when you need to interface different HDL simulators using openfpga flow-run scripts -write_verilog_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --print_top_testbench --print_preconfig_top_testbench --print_simulation_ini ./SimulationDeck/simulation_deck.ini --include_signal_init --support_icarus_simulator --explicit_port_mapping +write_verilog_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --print_top_testbench --print_preconfig_top_testbench --print_simulation_ini ./simulation_deck.ini --include_signal_init --explicit_port_mapping Can you avoid to modifying existing script templates? We can create a new template script with this feature. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe. @nachiket You can define variables in the openfpga shell script and then define the value of the variables in task configuration file. For example, you can define ${OPENFPGA_SIMULATION_DECK_INI_FILE} in the openfpga shell script. We have an example here. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[tangxifan]

@nachiket The example shows the use of variable ${OPENFPGA_VPR_ROUTE_CHAN_WIDTH}. You can create new variable for your own usage.

In the openfpga shell script, you can define a variable

OpenFPGA/openfpga_flow/openfpga_shell_scripts/generate_bitstream_fix_device_example_script.openfpga

Line 3 in 4315660

vpr ${VPR_ARCH_FILE} ${VPR_TESTBENCH_BLIF} --clock_modeling route --route_chan_width ${OPENFPGA_VPR_ROUTE_CHAN_WIDTH} --device ${OPENFPGA_VPR_DEVICE_LAYOUT}

In the task configuration file, you can use it by

OpenFPGA/openfpga_flow/tasks/fpga_bitstream/generate_bitstream/device_96x96/config/task.conf

Line 23 in 4315660

openfpga_vpr_route_chan_width=100

According to what I know, the variable is not visible to the run_modelsim.py script.

I am a bit confused by what you need exactly here. If you can provide more details, I can help.

[nachiket]

Ok that example of variable usage makes sense. I was looking for the SIM variable directly. @ganeshgore can chime in how the run_modelsim.py file is supposed to be used. But it appears that it only looks for the tcl script in the root folder of the generated run. The OpenFPGA shell script seems to create a folder SimulationDeck/ inside the run. So to make this work correctly I had to drop the SimulationDeck/ path from the OpenFPGA shell script. A better way may be to allow the Modelsim script to accept a variable defining the folder containing the tcl file.

…

On Feb 22, 2021, at 12:11 PM, tangxifan ***@***.***> wrote:  @nachiket The example shows the use of variable ${OPENFPGA_VPR_ROUTE_CHAN_WIDTH}. You can create new variable for your own usage. In the openfpga shell script, you can define a variable https://github.com/lnis-uofu/OpenFPGA/blob/4315660bf1bf20b264883b277f28a02f35851d55/openfpga_flow/openfpga_shell_scripts/generate_bitstream_fix_device_example_script.openfpga#L3 In the task configuration file, you can use it by https://github.com/lnis-uofu/OpenFPGA/blob/4315660bf1bf20b264883b277f28a02f35851d55/openfpga_flow/tasks/fpga_bitstream/generate_bitstream/device_96x96/config/task.conf#L23 According to what I know, the variable is not visible to the run_modelsim.py script. I am a bit confused by what you need exactly here. If you can provide more details, I can help. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.