search

gaph-pucrs / RS5

RV32I[M][C][V][_Zihpm][_Zkne][_Xosvm]_Zicsr processor
MIT License
9 stars 1 forks source link

Refactor decoding #6

Closed aedalzotto closed 1 year ago

aedalzotto commented 1 year ago

Still needs to test if area and timing are affected by this.

aedalzotto commented 1 year ago

The report below was generated without LUTRAM:

Check reports ![2023-05-22T22:21:27,027643177-03:00](https://github.com/gaph-pucrs/PUC-RS5/assets/12283310/2cccbb50-9126-4a72-a431-62529da3846c) ![2023-05-22T22:17:05,223259735-03:00](https://github.com/gaph-pucrs/PUC-RS5/assets/12283310/5468b509-11be-4275-a813-35b6050b9616) Still not good enough
aedalzotto commented 1 year ago

Modified: 2023-05-23T17:24:38,987675613-03:00 2023-05-23T17:24:19,557470086-03:00

Baseline: 2023-05-23T17:30:57,742056104-03:00 2023-05-23T17:30:45,378612401-03:00

aedalzotto commented 1 year ago

2023-05-23T17:58:44,602278560-03:00 2023-05-23T17:58:33,823266035-03:00

Better now: ~3.7% decrease in LUT utilization A bit of improvement in slack time, but I don't know if anything above 0 is meaningful (check this)

aedalzotto commented 1 year ago

2023-05-23T18:39:32,082707579-03:00 2023-05-23T18:39:19,814491456-03:00

Back to a little bit more LUTs but better timing. I think the next step is changing instruction decoding to one-hot.

marlls1989 commented 1 year ago

How so? The slack is positive, timing is good.

On 22 May 2023, at 22:24, Angelo Elias Dal Zotto @.***> wrote:

https://user-images.githubusercontent.com/12283310/240092500-2cccbb50-9126-4a72-a431-62529da3846c.png https://user-images.githubusercontent.com/12283310/240092502-5468b509-11be-4275-a813-35b6050b9616.png Still not good enough

— Reply to this email directly, view it on GitHub https://github.com/gaph-pucrs/PUC-RS5/pull/6#issuecomment-1558321724, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABE2QN55AM3USVHKS7COG23XHQGUHANCNFSM6AAAAAAYLEFSIQ. You are receiving this because you are subscribed to this thread.