palaniappan-r's picture
Add markdown sources
d59bb1b verified
# Reducing slew on clock tree with set_max_transition?
Tool: Gate Resizer
Subcategory: Usage question
## Conversation
### oharboe
With `make DESIGN_CONFIG=designs/asap7/sram-64x16/config.mk` I get a lot of slew and I heard about `set_max_transition` and found it used in some designs in ORFS, so I thought I would see what it did...
```
==========================================================================
resizer pre report_checks -path_delay max
--------------------------------------------------------------------------
Startpoint: _358_ (falling edge-triggered flip-flop clocked by io_clk')
Endpoint: io_ram_dout2[43] (output port clocked by io_clk)
Path Group: io_clk
Path Type: max
Fanout Cap Slew Delay Time Description
-----------------------------------------------------------------------------
0.00 0.00 0.00 clock io_clk' (fall edge)
0.00 0.00 clock network delay (ideal)
0.00 0.00 0.00 v _358_/CLK (DFFLQNx1_ASAP7_75t_R)
95.79 123.37 123.37 v _358_/QN (DFFLQNx1_ASAP7_75t_R)
1 6.79 _038_ (net)
96.33 4.05 127.42 v _173_/A (INVx3_ASAP7_75t_R)
45.29 35.23 162.65 ^ _173_/Y (INVx3_ASAP7_75t_R)
1 3.94 io_ram_dout2[43] (net)
45.43 1.45 164.10 ^ io_ram_dout2[43] (out)
164.10 data arrival time
0.00 500.00 500.00 clock io_clk (rise edge)
0.00 500.00 clock network delay (ideal)
0.00 500.00 clock reconvergence pessimism
-100.00 400.00 output external delay
400.00 data required time
-----------------------------------------------------------------------------
400.00 data required time
-164.10 data arrival time
-----------------------------------------------------------------------------
235.90 slack (MET)
```
0.050 ns seems to be a good place to start to bring down the slew from 95.79 above.
When I add `set_max_transition 0.05 [current_design]` to the constraints.sdc, the flow seems to get stuck:
```
Perform port buffering...
[INFO RSZ-0027] Inserted 79 input buffers.
[INFO RSZ-0028] Inserted 128 output buffers.
Perform buffer insertion...
[INFO RSZ-0058] Using max wire length 232um.
[no further output after 10 minutes or so]
```
maybe the units are picoseconds? `set_max_transition 50 [current_design]`, same result, resizing doesn't seem to complete.
Created a draft pull request to check if it completes... https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/pull/1315
### maliberty
You are setting it 0.05ps not ns which is an impossibly fast target.
### oharboe
`set_max_transition 80 [current_design]` is as low as I could get and `make cts` completing in a reasonable amount of time.
Lots of violations:
```
==========================================================================
cts final report_check_types -max_slew -max_cap -max_fanout -violators
--------------------------------------------------------------------------
max slew
Pin Limit Slew Slack
------------------------------------------------------------
_8_44/A 80.00 330.31 -250.31 (VIOLATED)
_8_46/A 80.00 309.18 -229.18 (VIOLATED)
clkbuf_4_4_0_io_clk/Y 80.00 294.55 -214.55 (VIOLATED)
SRAM2RW16x32_2/CE1 80.00 249.22 -169.22 (VIOLATED)
SRAM2RW16x32_2/CE2 80.00 249.17 -169.17 (VIOLATED)
SRAM2RW16x32/CE1 80.00 224.86 -144.86 (VIOLATED)
SRAM2RW16x32/CE2 80.00 224.83 -144.83 (VIOLATED)
_58_73/A 80.00 193.42 -113.42 (VIOLATED)
_58_74/A 80.00 193.40 -113.40 (VIOLATED)
_8_52/A 80.00 193.35 -113.35 (VIOLATED)
_8_51/A 80.00 193.32 -113.32 (VIOLATED)
_8_48/A 80.00 193.25 -113.25 (VIOLATED)
_8_53/A 80.00 193.22 -113.22 (VIOLATED)
_8_50/A 80.00 193.03 -113.03 (VIOLATED)
_58_66/A 80.00 192.96 -112.96 (VIOLATED)
_58/A 80.00 192.91 -112.91 (VIOLATED)
_58_71/A 80.00 192.76 -112.76 (VIOLATED)
_58_60/A 80.00 192.65 -112.65 (VIOLATED)
_58_67/A 80.00 192.64 -112.64 (VIOLATED)
_8_17/A 80.00 192.28 -112.28 (VIOLATED)
_8_18/A 80.00 192.23 -112.23 (VIOLATED)
clkbuf_4_11_0_io_clk/Y 80.00 191.85 -111.85 (VIOLATED)
_8_55/A 80.00 154.73 -74.73 (VIOLATED)
_8_24/A 80.00 154.72 -74.72 (VIOLATED)
_58_70/A 80.00 154.72 -74.72 (VIOLATED)
_58_77/A 80.00 154.71 -74.71 (VIOLATED)
_8_54/A 80.00 154.67 -74.67 (VIOLATED)
_8_49/A 80.00 154.66 -74.66 (VIOLATED)
_8_57/A 80.00 154.55 -74.55 (VIOLATED)
_58_75/A 80.00 154.42 -74.42 (VIOLATED)
_58_63/A 80.00 154.36 -74.36 (VIOLATED)
clkbuf_4_10_0_io_clk/Y 80.00 153.74 -73.74 (VIOLATED)
SRAM2RW16x32_3/CE1 80.00 148.30 -68.30 (VIOLATED)
_8_19/A 80.00 142.96 -62.96 (VIOLATED)
_8_36/A 80.00 142.96 -62.96 (VIOLATED)
clkbuf_4_2_0_io_clk/Y 80.00 142.96 -62.96 (VIOLATED)
clkbuf_4_9_0_io_clk/A 80.00 140.76 -60.76 (VIOLATED)
clkbuf_4_8_0_io_clk/A 80.00 140.75 -60.75 (VIOLATED)
clkbuf_4_10_0_io_clk/A 80.00 139.69 -59.69 (VIOLATED)
clkbuf_4_4_0_io_clk/A 80.00 139.66 -59.66 (VIOLATED)
clkbuf_4_5_0_io_clk/A 80.00 139.65 -59.65 (VIOLATED)
clkbuf_4_11_0_io_clk/A 80.00 139.64 -59.64 (VIOLATED)
clkbuf_4_6_0_io_clk/A 80.00 139.22 -59.22 (VIOLATED)
clkbuf_4_7_0_io_clk/A 80.00 139.19 -59.19 (VIOLATED)
clkbuf_4_0_0_io_clk/A 80.00 138.58 -58.58 (VIOLATED)
clkbuf_4_1_0_io_clk/A 80.00 138.12 -58.12 (VIOLATED)
clkbuf_4_15_0_io_clk/A 80.00 137.02 -57.02 (VIOLATED)
clkbuf_4_14_0_io_clk/A 80.00 137.02 -57.02 (VIOLATED)
clkbuf_4_13_0_io_clk/A 80.00 137.01 -57.01 (VIOLATED)
clkbuf_4_12_0_io_clk/A 80.00 136.99 -56.99 (VIOLATED)
clkbuf_4_3_0_io_clk/A 80.00 136.86 -56.86 (VIOLATED)
clkbuf_4_2_0_io_clk/A 80.00 136.22 -56.22 (VIOLATED)
clkbuf_0_io_clk/Y 80.00 132.32 -52.32 (VIOLATED)
SRAM2RW16x32_2/O2[23] 0.23 45.42 -45.19 (VIOLATED)
SRAM2RW16x32_2/O2[24] 0.23 45.26 -45.04 (VIOLATED)
SRAM2RW16x32_2/O2[16] 0.23 41.77 -41.54 (VIOLATED)
SRAM2RW16x32_3/CE2 80.00 120.37 -40.37 (VIOLATED)
SRAM2RW16x32_2/O2[30] 0.23 40.24 -40.01 (VIOLATED)
SRAM2RW16x32_2/O2[26] 0.23 38.87 -38.65 (VIOLATED)
SRAM2RW16x32_2/O2[29] 0.23 38.54 -38.31 (VIOLATED)
SRAM2RW16x32_2/O2[12] 0.23 38.16 -37.93 (VIOLATED)
SRAM2RW16x32_2/O2[17] 0.23 36.78 -36.55 (VIOLATED)
SRAM2RW16x32_2/O2[7] 0.23 36.69 -36.46 (VIOLATED)
```
### oharboe
This PR uses set_max_transition. I have an example to study now, so closing.
https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/pull/1311