Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

There is a contradiction between 'lane sequencer to functional units' and 'lane sequencer to operand requester'. #259

Open
TimLee-123 opened this issue Nov 7, 2023 · 1 comment
Labels
bug Something isn't working

Comments

@TimLee-123
Copy link

TimLee-123 commented Nov 7, 2023

Issue

We identified a contradiction during verification when the lane sequencer communicates with both the functional unit and the
operand requester.

The main reason for this is that when an instruction is sent to the lane sequencer and vl is 0, it does not proceed to the subsequent functional unit. However, at the same time, the lane sequencer instructs the operand requester to fetch operands for an instruction that is no longer needed for computation.

For example, there are two instructions. The first one is a 'vmv' instruction, and its vl parameter is set to 0. The second instruction is a 'vadd' instruction. From the waveform below, It can be observed that the 'vmv' instruction is notably absent within the instructions transmitted from the lane sequencer to the VALU. However, the lane sequencer still instructs the operand requester to initiate a request for this instruction. This ultimately results in the vs1 of the 'vmv' instruction being treated as the vs1 of the 'vadd' instruction, leading to an error.

From the waveform, it can be seen that both v8 and v24 have values of 0. However, due to the aforementioned reasons, the vs1 of vadd is changed to come from v0, ultimately resulting in a non-zero value being written back to v4.

By suggestion, when the valid signal from the lane sequencer to functional units is 0, the valid signal from the lane sequencer to the operand requester should also be set to 0.

Supplementary Pictures

image

image

image

image

image

@mp-17
Copy link
Collaborator

mp-17 commented Feb 26, 2024

Hello @TimLee-123 ,

Thank you very much for all your contributions. We are planning to solve all the issues in bundle with a targeted verification effort.

@mp-17 mp-17 added the bug Something isn't working label Jun 15, 2024
# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

2 participants