mole99
commented
1 year ago Hi @d-m-bailey, I have now tried the following:
- "FP_PDN_CHECK_NODES": 1
- Specifying the leading
\inFP_PDN_MACRO_HOOKS
"FP_PDN_MACRO_HOOKS": [
"\\memory[0].sram vccd1 vssd1 vccd1 vssd1,",
"\\memory[1].sram vccd1 vssd1 vccd1 vssd1,",
"\\memory[2].sram vccd1 vssd1 vccd1 vssd1,",
"\\memory[3].sram vccd1 vssd1 vccd1 vssd1,",
"sram vccd1 vssd1 vccd1 vssd1"
],Thanks for the tip with runs/tag/config.tcl, that seems to be correct now at least:
set ::env(FP_PDN_MACRO_HOOKS) {\memory[0].sram vccd1 vssd1 vccd1 vssd1, \memory[1].sram vccd1 vssd1 vccd1 vssd1, \memory[2].sram vccd1 vssd1 vccd1 vssd1, \memory[3].sram vccd1 vssd1 vccd1 vssd1, sram vccd1 vssd1 vccd1 vssd1}But the actual issue still persists, the SRAM macros instantiated inside the generate do not have any power ports in the powered verilog (tmp/signoff/25-user_project_wrapper.pnl.v) and thus are not connected to power.
For example one of the SRAMs with the generate keyword:
sky130_sram_2kbyte_1rw1r_32x512_8 \memory[2].sram (.csb0(la_data_in[0]),
.csb1(la_data_in[47]),
.web0(la_data_in[1]),
.clk0(wb_clk_i),
.clk1(wb_clk_i),
.addr0({la_data_in[14],
la_data_in[13],
la_data_in[12],
la_data_in[11],
...The SRAM instantiated in the normal way, on the other hand, has the power ports as expected:
sky130_sram_2kbyte_1rw1r_32x512_8 sram (.csb0(la_data_in[0]),
.csb1(la_data_in[47]),
.web0(la_data_in[1]),
.clk0(wb_clk_i),
.clk1(wb_clk_i),
.vccd1(vccd1),
.vssd1(vssd1),
.addr0({la_data_in[14],
la_data_in[13],
la_data_in[12],
la_data_in[11],
la_data_in[10],
...
donn
kareefardi
Description
When instantiating multiple macros via the Verilog
generatestatement the power connections are lost somewhere in the flow and thus the macros are not connected to power.I had a look at the resulting file
tmp/signoff/25-user_project_wrapper.pnl.v. The macro I had instantiated normally is connected tovccd1/vssd1, the other four instances via thegeneratestatement are not connected tovccd1/vssd1.Example code snippet:
Expected Behavior
The macros should be connected to power.
Environment report
Reproduction material
No reproducible was packaged because the flow did not fail.
I created a repository, ol-generate-bug, based on
caravel_user_projectto easily reproduce the issue. Please note that I installedsky130Anext to the project folder underdependencies/pdks, so that the sram can be found.After
OPENLANE_ROOT,PDK_ROOTandPDKare exported, you can run the flow viamake user_project_wrapper.Relevant log output