Downloading Page for Benchmarks & Placement Outputs
The contents of each benchmark's .tar.gz
file:
.nodes: contains
the size of each cell, the total number of cells and information of whether a
cell is a FF or not (“FF” after cell size means Flip-Flop, “nFF”
means not Flip-Flop).
.nets: specify the input and out cells of each
nets and total number of nets.
.par: specify the technical parameters for STA, including:
driving resistance (Rd), load capacitance (Cg), unit wire resistance (r), unit
wire capacitance (c) and length unit which explains unit length of the
benchmark is how many micro meters (unit).
.pl: specify the (x. y) coordinates of each cell. There are
two .pl file: one is the placement results of Dragon (has “Dragon” in its
name); one is our incremental placement results (has "Flow" in its name).
TD incremental placement (FlowPLace)
results with 3% WS constraint:
a) Initial placement done with
Dragon without row spacing (Same as above downloadable benchmarks)
|
|
Delay impr. (%) |
WL incr. (%) |
Run time (s) |
|
IBM01 |
15.1 |
9.9 |
203 |
|
IBM02 |
26.3 |
10.1 |
201 |
|
IBM03 |
30.3 |
9.7 |
338 |
|
IBM04 |
21.5 |
9.3 |
484 |
|
IBM05 |
18.0 |
4.3 |
304 |
|
IBM06 |
20.4 |
8.9 |
402 |
|
IBM07 |
14.1 |
6.7 |
729 |
|
IBM08 |
28.4 |
3.8 |
795 |
|
IBM09 |
8.4 |
9.1 |
854 |
|
IBM10 |
15.9 |
8.7 |
811 |
|
IBM11 |
14.1 |
10.2 |
621 |
|
IBM12 |
12.8 |
5.1 |
724 |
|
IBM13 |
18.4 |
8.3 |
916 |
|
IBM14 |
13.2 |
9.7 |
979 |
|
IBM15 |
18.8 |
4.7 |
1142 |
|
IBM16 |
20.5 |
4.3 |
1198 |
|
IBM17 |
25.6 |
2.6 |
1304 |
|
IBM18 |
34.1 |
2.6 |
1451 |
|
Avg. |
19.7 |
7.1 |
747 |
|
DMA |
19.1 |
12.8 |
297 |
|
DSP1 |
21.3 |
10.8 |
294 |
|
DSP2 |
20.8 |
10.2 |
302 |
|
RISC1 |
22.2 |
9.7 |
331 |
|
RISC2 |
19.8 |
10.5 |
387 |
|
Avg.
|
20.6 |
10.9 |
322 |
b) Initial
placement done with Dragon with row spacing = cell height
|
|
Delay impr. (%) |
WL incr. (%) |
Run time (s) |
|
IBM01 |
18.1 |
13.1 |
255 |
|
IBM02 |
27.1 |
11.0 |
302 |
|
IBM03 |
28.4 |
7.3 |
331 |
|
IBM04 |
25.6 |
7.9 |
527 |
|
IBM05 |
17.2 |
7.1 |
304 |
|
IBM06 |
21.2 |
10.2 |
572 |
|
IBM07 |
20.4 |
7.5 |
664 |
|
IBM08 |
24.1 |
5.4 |
714 |
|
IBM09 |
18.9 |
8.0 |
997 |
|
IBM10 |
18.1 |
5.9 |
1005 |
|
IBM11 |
13.8 |
4.6 |
502 |
|
IBM12 |
15.7 |
7.2 |
721 |
|
IBM13 |
17.0 |
7.9 |
851 |
|
IBM14 |
20.7 |
11.2 |
1107 |
|
IBM15 |
22.0 |
6.4 |
1224 |
|
IBM16 |
20.5 |
4.3 |
1141 |
|
IBM17 |
30.2 |
2.7 |
1299 |
|
IBM18 |
30.1 |
1.6 |
1401 |
|
Avg.
|
21.6 |
7.1 |
789 |
|
DMA |
24.3 |
12.9 |
321 |
|
DSP1 |
28.1 |
12.1 |
304 |
|
DSP2 |
27.2 |
13.0 |
307 |
|
RISC1 |
22.7 |
11.0 |
324 |
|
RISC2 |
23.1 |
10.4 |
338 |
|
Avg.
|
25.1 |
11.9 |
319 |
c) Initial
placement done by TD-Dragon without row spacing
|
|
Delay impr. (%) |
WL incr. (%) |
Run time (s) |
|
Matrix |
1.18 |
8.07 |
208 |
|
Mac32 |
4.75 |
2.89 |
319 |
|
Mac64 |
7.51 |
7.91 |
328 |
|
Vp2 |
0.69 |
2.32 |
369 |
|
Avg.
|
3.53 |
5.87 |
281 |
Notes:
(1) It
is not very clear how to run TD-Dragon with row spacing.
(2) At this time, we cannot put TD-Dragon output results since we obtained
the .tlf cell-timing files with a non-distribution agreement. If we are
permitted to put the output results we will do so, as and when
that happens.