Parallel `findClkSkew`

[kbieganski]

Runs findClkSkew in parallel.

Requires this thread safety fix (PR #219).

findClkSkew basically retrieves the absolute maximum skew per clock. We can do it in separate threads and then reduce over the partial results from each thread.

Multi-threaded performance

Compared against thread-safe-network (from PR #219), as master often crashes with -threads 8.

Global routing

Ibex

Command	Mean [s]	Min [s]	Max [s]	Relative
ibex / 5_1_grt / thread-safe-network / 8 threads	118.088 ± 0.341	117.563	118.504	3.58 ± 0.02
ibex / 5_1_grt / parallel-find-clk-skew / 8 threads	32.994 ± 0.102	32.837	33.152	1.00

BlackParrot

Command	Mean [s]	Min [s]	Max [s]	Relative
black_parrot / 5_1_grt / thread-safe-network / 8 threads	766.336 ± 3.214	762.248	773.285	1.34 ± 0.01
black_parrot / 5_1_grt / parallel-find-clk-skew / 8 threads	573.284 ± 2.252	570.765	577.828	1.00

Flow from synthesis to 5_1_grt

Ibex

Command	Mean [s]	Min [s]	Max [s]	Relative
ibex / 1_1_synth to 5_1_grt / thread-safe-network / 8 threads	380.836 ± 0.602	380.027	381.910	1.63 ± 0.01
ibex / 1_1_synth to 5_1_grt / parallel-find-clk-skew / 8 threads	234.039 ± 0.684	232.489	234.902	1.00

BlackParrot

Command	Mean [s]	Min [s]	Max [s]	Relative
black_parrot / 1_1_synth to 5_1_grt / thread-safe-network / 8 threads	3533.167 ± 8.196	3520.372	3543.953	1.10 ± 0.01
black_parrot / 1_1_synth to 5_1_grt / parallel-find-clk-skew / 8 threads	3215.166 ± 13.543	3200.957	3239.229	1.00

Single-threaded performance

This is a performance regression check.

Global routing

Ibex

Command	Mean [s]	Min [s]	Max [s]	Relative
ibex / 5_1_grt / master / 1 thread	125.214 ± 0.776	123.797	126.203	1.02 ± 0.01
ibex / 5_1_grt / thread-safe-network / 1 thread	122.379 ± 0.670	121.476	123.630	1.00
ibex / 5_1_grt / parallel-find-clk-skew / 1 thread	123.304 ± 0.497	122.479	124.033	1.01 ± 0.01

Command	Mean [s]	Min [s]	Max [s]	Relative
ibex / 5_1_grt / master / 1 thread	124.140 ± 3.096	118.442	127.865	1.01 ± 0.03
ibex / 5_1_grt / thread-safe-network / 1 thread	122.610 ± 0.516	121.760	123.394	1.00
ibex / 5_1_grt / parallel-find-clk-skew / 1 thread	122.836 ± 0.563	122.244	123.900	1.00 ± 0.01

BlackParrot

Command	Mean [s]	Min [s]	Max [s]	Relative
black_parrot / 5_1_grt / master / 1 thread	834.350 ± 7.627	817.349	842.018	1.02 ± 0.01
black_parrot / 5_1_grt / thread-safe-network / 1 thread	828.409 ± 6.578	816.466	837.040	1.01 ± 0.01
black_parrot / 5_1_grt / parallel-find-clk-skew / 1 thread	818.827 ± 4.998	811.516	825.794	1.00

Flow from synthesis to 5_1_grt (Ibex only)

Command	Mean [s]	Min [s]	Max [s]	Relative
ibex / 1_1_synth to 5_1_grt / master / 1 thread	406.818 ± 1.115	405.531	408.481	1.00
ibex / 1_1_synth to 5_1_grt / parallel-find-clk-skew / 1 thread	408.049 ± 1.234	405.287	409.360	1.00 ± 0.00