BPU clock gating #2733

eastonman · 2024-03-01T08:47:40Z

No description provided.

chenguokai · 2024-03-01T11:30:28Z

CI failed

eastonman · 2024-03-01T12:17:12Z

seems failed for no reason, just re-run for now.

XiangShanRobot · 2024-03-03T15:17:11Z

[Generated by IPC robot]
commit: 6cd72ef

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`6cd72ef`	1.871	0.461	2.093	1.197	2.908	2.178	1.398	0.919	1.430	1.123	3.612	2.656	2.301	3.149

master branch:

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`a42a7ff`	1.878	0.461	2.100	1.213	2.914	2.181	1.395	0.920	1.421	1.126	3.617	2.659	2.302	3.165
`b15e4c0`	1.878	0.461	2.100	1.213	2.895	2.181	1.395	0.919	1.421	1.128	3.617	2.659	2.302	3.165
`3c5d56a`	1.878	0.461	2.100	1.209	2.913	2.182	1.395	0.920	1.421	1.123	3.617	2.660	2.302	3.165
`8abe181`	1.878	0.461	2.100	1.213	2.895	2.179	1.395	0.919	1.421	1.125	3.617	2.666	2.302	3.165
`13156de`	1.878	0.461	2.100	1.213	2.895	2.182	1.397	0.919	1.421	1.126	3.617	2.660	2.302	3.165
`1fcb3bc`	1.878	0.461	2.100	1.211	2.895	2.181	1.395	0.920	1.421	1.123	3.617	2.659	2.302	3.165
`f3c16e1`	1.878	0.461	2.100	1.213	2.895	2.182	1.396	0.920	1.421	1.126	3.617	2.659	2.302	3.165
`45f43e6`	1.878	0.461	2.100	1.211	2.895	2.182	1.395	0.920	1.421	1.127	3.617	2.659	2.302	3.165
`8fae59b`	1.878	0.461	2.100	1.213	2.913	2.182	1.394	0.920	1.421	1.123	3.617	2.647	2.302	3.165
`a61a35e`	1.879	0.460	2.092	1.200	2.895	2.176	1.394	0.917	1.436	1.125	3.617	2.663	2.298	3.134

eastonman · 2024-03-04T03:05:50Z

Behavior change is expected, but I will split the changes affecting performance into another PR.

eastonman · 2024-03-05T03:53:51Z

I believe the performance fluctuation is due to a performance bug fixed in e6b2fa7, which have minor influence.

In 14 microbenchmarks, 6 decreased MPKI, 1 same, and 7 increased. I think the fluctuation is acceptable.

astar.log
copy_and_run.log
coremark.log
gcc.log
gromacs.log
lbm.log
linux.log
mcf.log
microbench.log
milc.log
namd.log
povray.log
wrf.log
xalancbmk.log
(Ubuntu)  manyang@open05 6445 $> cd ../6453
(Ubuntu)  manyang@open05 6453 $> ls | sort |xargs cat | grep BpBRight
[PERF ][time=             2672567] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               633722
[PERF ][time=              882819] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                62733
[PERF ][time=             1503050] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               599581
[PERF ][time=             4178172] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               695115
[PERF ][time=             1719508] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                92771
[PERF ][time=             2295462] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                56652
[PERF ][time=             2961840] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               534243
[PERF ][time=             5443099] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               592129
[PERF ][time=              228655] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                54216
[PERF ][time=             4453785] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               313615
[PERF ][time=             1384306] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               501060
[PERF ][time=             1882376] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               456550
[PERF ][time=             2172840] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               254440
[PERF ][time=             1587779] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,              1103260
(Ubuntu)  manyang@open05 6453 $> cd ../6445
(Ubuntu)  manyang@open05 6445 $> ls | sort |xargs cat | grep BpBRight
[PERF ][time=             2661985] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               633704
[PERF ][time=              882163] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                62734
[PERF ][time=             1498152] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               599656
[PERF ][time=             4123637] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               695197
[PERF ][time=             1715795] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                92805
[PERF ][time=             2292792] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                56652
[PERF ][time=             2967776] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               534057
[PERF ][time=             5435407] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               591821
[PERF ][time=              230140] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,                54165
[PERF ][time=             4441659] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               313613
[PERF ][time=             1382280] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               501060
[PERF ][time=             1880573] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               456503
[PERF ][time=             2172362] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,               254455
[PERF ][time=             1579787] TOP.SimTop.l_soc.core_with_l2.core.frontend.ftq: BpBRight,              1103272

eastonman · 2024-03-05T06:15:24Z

re-running CI with LFSR changes reverted.

XiangShanRobot · 2024-03-05T10:31:21Z

[Generated by IPC robot]
commit: 37153f8

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`37153f8`	1.878	0.461	2.100	1.210	2.895	2.181	1.394	0.919	1.421	1.123	3.617	2.628	2.302	3.165

master branch:

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`a42a7ff`	1.878	0.461	2.100	1.213	2.914	2.181	1.395	0.920	1.421	1.126	3.617	2.659	2.302	3.165
`b15e4c0`	1.878	0.461	2.100	1.213	2.895	2.181	1.395	0.919	1.421	1.128	3.617	2.659	2.302	3.165
`3c5d56a`	1.878	0.461	2.100	1.209	2.913	2.182	1.395	0.920	1.421	1.123	3.617	2.660	2.302	3.165
`8abe181`	1.878	0.461	2.100	1.213	2.895	2.179	1.395	0.919	1.421	1.125	3.617	2.666	2.302	3.165
`13156de`	1.878	0.461	2.100	1.213	2.895	2.182	1.397	0.919	1.421	1.126	3.617	2.660	2.302	3.165
`1fcb3bc`	1.878	0.461	2.100	1.211	2.895	2.181	1.395	0.920	1.421	1.123	3.617	2.659	2.302	3.165
`f3c16e1`	1.878	0.461	2.100	1.213	2.895	2.182	1.396	0.920	1.421	1.126	3.617	2.659	2.302	3.165
`45f43e6`	1.878	0.461	2.100	1.211	2.895	2.182	1.395	0.920	1.421	1.127	3.617	2.659	2.302	3.165
`8fae59b`	1.878	0.461	2.100	1.213	2.913	2.182	1.394	0.920	1.421	1.123	3.617	2.647	2.302	3.165
`a61a35e`	1.879	0.460	2.092	1.200	2.895	2.176	1.394	0.917	1.436	1.125	3.617	2.663	2.298	3.134

XiangShanRobot · 2024-03-05T17:46:59Z

[Generated by IPC robot]
commit: 5e916e7

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`5e916e7`	1.878	0.461	2.100	1.211	2.913	2.179	1.395	0.919	1.421	1.123	3.617	2.662	2.302	3.165

master branch:

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`a42a7ff`	1.878	0.461	2.100	1.213	2.914	2.181	1.395	0.920	1.421	1.126	3.617	2.659	2.302	3.165
`b15e4c0`	1.878	0.461	2.100	1.213	2.895	2.181	1.395	0.919	1.421	1.128	3.617	2.659	2.302	3.165
`3c5d56a`	1.878	0.461	2.100	1.209	2.913	2.182	1.395	0.920	1.421	1.123	3.617	2.660	2.302	3.165
`8abe181`	1.878	0.461	2.100	1.213	2.895	2.179	1.395	0.919	1.421	1.125	3.617	2.666	2.302	3.165
`13156de`	1.878	0.461	2.100	1.213	2.895	2.182	1.397	0.919	1.421	1.126	3.617	2.660	2.302	3.165
`1fcb3bc`	1.878	0.461	2.100	1.211	2.895	2.181	1.395	0.920	1.421	1.123	3.617	2.659	2.302	3.165
`f3c16e1`	1.878	0.461	2.100	1.213	2.895	2.182	1.396	0.920	1.421	1.126	3.617	2.659	2.302	3.165
`45f43e6`	1.878	0.461	2.100	1.211	2.895	2.182	1.395	0.920	1.421	1.127	3.617	2.659	2.302	3.165
`8fae59b`	1.878	0.461	2.100	1.213	2.913	2.182	1.394	0.920	1.421	1.123	3.617	2.647	2.302	3.165
`a61a35e`	1.879	0.460	2.092	1.200	2.895	2.176	1.394	0.917	1.436	1.125	3.617	2.663	2.298	3.134

eastonman · 2024-03-06T10:31:17Z

I suggest these clock-gating PRs should be documented with comments before merging.
#2733 #2745 #2734

this commit ports clock gating optimization from nanhu if applicable, most ungated register is opted Co-authored-by: Liang Sen <liangsen20z@ict.ac.cn>

some control signal inside Tage is using data signals from io this commit add io control to these signals

XiangShanRobot · 2024-04-16T19:31:56Z

[Generated by IPC robot]
commit: 7f85bd1

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`7f85bd1`	1.707	0.449	2.112	1.153	1.643	1.183	2.331	0.908	1.398	0.832	2.349	2.136	1.783	2.911

master branch:

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
`23761fd`
`0c00289`	1.707	0.449	2.112	1.155	1.643	1.183	2.332	0.908	1.383	0.832	2.349	2.142	1.783	2.911
`eef81af`	1.862	0.460	2.087	1.204	2.903	2.177	2.332	0.929	1.427	1.138	3.610	2.663	2.297	3.183
`875ae3b`	1.862	0.460	2.087	1.204	2.903	2.177	2.332	0.928	1.427	1.138	3.610	2.659	2.297	3.183
`f6916db`	1.862	0.460	2.087	1.204	2.903	2.177	2.332	0.929	1.427	1.138	3.610	2.659	2.297	3.183
`4d931b7`	1.862	0.460	2.087	1.204	2.903	2.177	2.332	0.928	1.427	1.138	3.610	2.657	2.297	3.183
`9afa8a4`	1.862	0.460	2.087	1.204	2.903	2.177	2.332	0.928	1.427	1.138	3.610	2.657	2.297	3.183
`ef6723f`	1.862	0.460	2.087	1.204	2.903	2.177	2.331	0.929	1.427	1.138	3.610	2.648	2.297	3.183
`58a9a40`	1.862	0.460	2.087	1.204	2.903	2.177	2.332	0.929	1.427	1.138	3.610	2.657	2.297	3.183
`8f62644`	1.862	0.460	2.087	1.204	2.903	2.177	2.331	0.929	1.427	1.138	3.610	2.645	2.297	3.183

eastonman · 2024-04-17T02:21:52Z

@Tang-Haojin I believe this can be merged?

Tang-Haojin · 2024-04-17T03:27:42Z

Need squash or rebase?

eastonman · 2024-04-17T03:43:33Z

@Tang-Haojin prefer squash, not big PR.

eastonman requested a review from chenguokai as a code owner March 1, 2024 08:47

chenguokai approved these changes Mar 1, 2024

View reviewed changes

eastonman added the do not merge Do not merge this pull request label Mar 4, 2024

eastonman removed the do not merge Do not merge this pull request label Mar 5, 2024

eastonman added the do not merge Do not merge this pull request label Mar 6, 2024

eastonman removed the do not merge Do not merge this pull request label Apr 1, 2024

bpu: port clock gating opt from nanhu

f90826d

this commit ports clock gating optimization from nanhu if applicable, most ungated register is opted Co-authored-by: Liang Sen <liangsen20z@ict.ac.cn>

eastonman force-pushed the bpu-clock-gating branch from 5e916e7 to 38bc65b Compare April 16, 2024 11:23

eastonman added 4 commits April 16, 2024 19:23

bpu: fix missing control sig

54a35d4

some control signal inside Tage is using data signals from io this commit add io control to these signals

bpu: use RegNextWithEnable in Valid[T]

90b7f1d

bpu: remove LFSR clock gating

c5bfc76

bpu: revert RegEnable on reset_u

7f85bd1

eastonman force-pushed the bpu-clock-gating branch from 38bc65b to 7f85bd1 Compare April 16, 2024 11:30

Tang-Haojin merged commit 7af6acb into master Apr 17, 2024
4 checks passed

Tang-Haojin deleted the bpu-clock-gating branch April 17, 2024 03:43

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

BPU clock gating #2733

BPU clock gating #2733

eastonman commented Mar 1, 2024

chenguokai commented Mar 1, 2024

eastonman commented Mar 1, 2024

XiangShanRobot commented Mar 3, 2024

eastonman commented Mar 4, 2024

eastonman commented Mar 5, 2024 •

edited

Loading

eastonman commented Mar 5, 2024

XiangShanRobot commented Mar 5, 2024

XiangShanRobot commented Mar 5, 2024

eastonman commented Mar 6, 2024

XiangShanRobot commented Apr 16, 2024

eastonman commented Apr 17, 2024

Tang-Haojin commented Apr 17, 2024

eastonman commented Apr 17, 2024

BPU clock gating #2733

BPU clock gating #2733

Conversation

eastonman commented Mar 1, 2024

chenguokai commented Mar 1, 2024

eastonman commented Mar 1, 2024

XiangShanRobot commented Mar 3, 2024

eastonman commented Mar 4, 2024

eastonman commented Mar 5, 2024 • edited Loading

eastonman commented Mar 5, 2024

XiangShanRobot commented Mar 5, 2024

XiangShanRobot commented Mar 5, 2024

eastonman commented Mar 6, 2024

XiangShanRobot commented Apr 16, 2024

eastonman commented Apr 17, 2024

Tang-Haojin commented Apr 17, 2024

eastonman commented Apr 17, 2024

eastonman commented Mar 5, 2024 •

edited

Loading