- 14 8月, 2020 15 次提交
-
-
由 Allen 提交于
Our missQueue design means we can not avoid these assertions. We send response before free this missQueueEntry and storeMissQueueEntry. So during this short period, sbuffer may still send down the same block, which is perfectly OK and we should not assert it. LoadMissQueue and StoreMissQueue design may need to be revised.
-
由 Allen 提交于
-
由 Allen 提交于
-
由 Allen 提交于
StoreReq starts to going down.
-
由 William Wang 提交于
-
由 Allen 提交于
-
由 Allen 提交于
Sbuffer may send down blocks with the same idx(but we will block it). Sbuffer should not send down the same block multiple times(this means sbuffer is buggy).
-
由 William Wang 提交于
-
由 Allen 提交于
-
由 Allen 提交于
-
由 Allen 提交于
Do not block replayed reqs. Also, let LoadMissQueue and StoreMissQueue set meta.replay correctly. Initialized replay_resp_ctr to zero in StoreMissEntry.
-
由 Allen 提交于
-
由 Allen 提交于
Now, we can track each req's lifecycle in debug log.
-
由 LinJiawei 提交于
-
由 LinJiawei 提交于
-
- 13 8月, 2020 16 次提交
-
-
由 William Wang 提交于
-
由 Yinan Xu 提交于
-
由 Allen 提交于
-
由 Allen 提交于
-
由 Allen 提交于
to a fully functional IOMSHR. Now, it should support multiple outstanding IO requests. There still exists one problem: will IO tilelink share path with memory tilelink? If so, their transaction id will collide.
-
由 Allen 提交于
-
由 Allen 提交于
**NaiveTLToAXI4 has not been finished yet**. None of them fully support tilelink. FakeTLLLC supports: * AcquireBlock on channel A for permission acquire * ReleaseData on channel C for dcache eviction and writeback It's supposed to work with L1 dcache. Now, we still don't have a fully functional TL to AXI converter, starving for diplomacy!!!
-
由 YikeZhou 提交于
-
由 YikeZhou 提交于
Sbuffer: replace lru
-
由 Yinan Xu 提交于
-
由 Yinan Xu 提交于
-
由 William Wang 提交于
-
由 William Wang 提交于
-
由 William Wang 提交于
-
由 Yinan Xu 提交于
-
由 Allen 提交于
-
- 12 8月, 2020 9 次提交
-
-
由 William Wang 提交于
-
由 William Wang 提交于
-
由 William Wang 提交于
-
由 LinJiawei 提交于
-
由 William Wang 提交于
-
由 Allen 提交于
-
由 Allen 提交于
-
由 Allen 提交于
-
由 Allen 提交于
-