* reharden: caravel
~ shift caravel_clocking due to change in size
~ change the pr boundary of caravel_power_routing mag file
~ regenarate lef of caravel_power_routing
* update pdn for `caravel_clocking` & `digital_pll`
* added script to update and generate the power routing views
* ~ run update_power_routing_views from the caravel root with prboundary
* fix output message
* added power routing lef, mag and gds
* fix update_power_routing_views saving wrong cell name
* reharden: caravel
~ incorperate pdn changes
~ re-extract spefs
* fix caravel_power_routing views
* fix abs path in maglef views
* fix abs path in mag views
add substcut layers in gpio_control_block and mgmt_protect
* generate a new chip_io gds
* regenerate gpio_control_block due to mag and gds not in sync
* reharden: caravel
~ change config to pass clean routing
~ use updated views of macros
* lvs clean views
* add caravel top-level generated sdf for all corners
* fix absolute path for mgmt_core_wrapper
Co-authored-by: kareem <kareem.farid@efabless.com>
Co-authored-by: Bassant Hassan <bassant.hassan@efabless.com>
- reimplement using a sparecell
- reimplement using newest open_pdks
!important using openlane pre odb with some local patches which
most if not all are merged in the current head of openlane however
still takes effort to update the interactive script to be latest
openlane compatible
!important override abstract lef generated by openlane. openlane
generates lef and mag that contain def BLOCKAGE layers that cause
congestions during top level routing
- rtl updated
~ add one column to the right to pass placement congestion
~ density adjusted (probably has no effect)
+ manually add isosubstrate layer in mag and gds from older iterations
!important still need to run dynamic simulations
!important depends on some updates to openlane
!important need to be able to recreate using newer openlane versions
- high level changes:
* add larger buffers on output ports
* add buffers on input ports
* adjust sdc file increasing output load and setting a high transition
- detailed changes:
* add interactive script for openlane where the order of events is a bit shuffled
- to add obstruction before pdn
- to manually insert buffers on some ports
- to manually remove buffers inserted by synthesis on for example serial_clock_out
* change openlane config adding extra row and columns to increase the space and fit the
added buffers
* change config to enable buffering
* increase density for better placement?
* change the cell exclude list. some excluded cells didn't make sense
* ef decap cells break dynamic sims?
* add custom pdn script for to duplicate the old pdn
- misc changes:
* fix openlane makefile to properly detect interactive script
!important still need to run dynamic simulations
!important depends on some updates to openlane
update gpio_control_block config for new openlane versions:
- disable `SYNTH_BUFFERING` and `SYNTH_SIZING` to limit the design size
and fit the floorplan
- change `SYNTH_STRATEGY` to `AREA 0` to minimize design cells
- disable `PL_RESIZER_TIMING_OPTIMIZATIONS` and
enable `GLB_RESIZER_TIMING_OPTIMIZATIONS`
- remove `FP_IO_*` and replace them with `FP_DEF_TEMPLATE` for io placement
- set `DECAP_CELL` to not use ef decaps.. i think that was for simulations?
- enable some turned off `QUIT_*` variables
- replace deprecated variables such as `GLB_RT_*`
- customize `pdn.tcl` to force pdn straps to follow the old pattern
- replace `$script_dir` with `$::env(DESIGN_DIR)`
!IMPORTANT - still need to run dynamic simulations
does not pass through an inverter, so that the input can remain
unconnected. Rewired the existing implementation to use an
alternative gate that has an inverting input so that the
user_gpio_out signal can be left undriven when the GPIOs are in
the management enable state. This is a simple logic refactoring
and does not change the logic function. The manual rewiring has
been confirmed by LVS, but at least one GL simulation should be
run to confirm that the logic function remains the same as before.
* (1) Modified the .magicrc file to set a default for PDK if not set in the
environment. (2) Fixed the user ID programming layout to not leave holes
behind when the script moves the vias around (similar to the handling of
the GPIO defaults block). (3) Added substrate isolation to gpio_control_block
and fixed the path references to the standard cells. (4) Fixed the four
missing routes on the Caravan top level. (5) Reinstated the large rendered
labels for the pads on both caravel and caravan. (6) Corrected the top
level gate-level netlist for caravan to add the missing pins to the
management core wrapper. (7) Did the same for the caravan top level RTL.
(8) Created scripts to run full LVS including extracting the management
core wrapper and reading all gate-level verilog submodules. (9) Moved all
of the LVS scripts to the scripts directory.
* Apply automatic changes to Manifest and README.rst
* Made the changes from pull request #73 as they did not get merged
successfully, and if merged now they will generate conflicts with
this pull request in scripts/set_user_id.py. So it's easier to
just manually add them to this pull request.
* Apply automatic changes to Manifest and README.rst
Co-authored-by: RTimothyEdwards <RTimothyEdwards@users.noreply.github.com>
it is exactly overlapping; this resolves issues with LVS. Also:
Corrected the references to the OpenLane/ path in the .mag files
by changing them to the canonical $PDKPATH reference.
view pointing to the contents of ../gds/ so that when the assembled
chip's GDS is generated with "cif *hier write disable", the POR
will continue to have the GDS with the proper hierarchical processing.
verilog for both caravel and caravan. Hand-edited the RTL and GL netlists to
correct this; still need to correct the layouts. The error causes the user1
side clock, load, and reset buffers to drive the user2 side as well as the user1
side, making a huge mess of the routing. Will route this by hand.
got removed from its net by a chang in the LEF view of an I/O pad, and
a lack of declaration of an array to attach to pwr_ctrl_out in the
verilog, which is valid verilog but netgen can't know the bus size
without the no-connect net being declared. The remaining issue has to
do with separation of ground domains in the mgmt_protect block.
Marwan Abbas
Tim Edwards
kareem
Kareem Farid
manarabdelaty