search

KastnerRG / riffa

The RIFFA development repository
https://riffa.ucsd.edu
Other
766 stars 313 forks source link

Xilinx tools not correctly implementing asynchronous FIFO. #13

Closed drichmond closed 8 years ago

drichmond commented 8 years ago

This causes a missing first-dword in about 1-in-10 transfers.

Does not seem to affect Altera (unlikely it's a logic bug in our code)

cjmraziii commented 7 years ago

What was the resolution of this issue?

drichmond commented 7 years ago

Wasn't Xilinx specific. In the end, the state machine in tx_port_monitor had a bug in it, that was caused by an earlier fix on a related issue.

Why do you ask?

cjmraziii commented 7 years ago

I am seeing an issue that sounds similar where sometimes the 1st two DWORDS are coming through wrong. It seems to be much less prevalent using the user_clk output of the RIFFA module vs my internal 100MHz asynchronous clock. Was the issue resolved in 2.2.2?

drichmond commented 7 years ago

Yes

On Nov 18, 2016, at 11:11 AM, cjmraziii notifications@github.com wrote:

I am seeing an issue that sounds similar where sometimes the 1st two DWORDS are coming through wrong. It seems to be much less prevalent using the user_clk output of the RIFFA module vs my internal 100MHz asynchronous clock. Was the issue resolved in 2.2.2?

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/KastnerRG/riffa/issues/13#issuecomment-261615673, or mute the thread https://github.com/notifications/unsubscribe-auth/AGFUdJKQGAnExhBbZtW78AyJLIc8Ogm7ks5q_fhlgaJpZM4JgRDp.