Skip to content
New issue

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

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

Already on GitHub? Sign in to your account

diff memcpy COW optimization #5

Open
wants to merge 73 commits into
base: difftest
Choose a base branch
from
Open

diff memcpy COW optimization #5

wants to merge 73 commits into from

Conversation

tastynoob
Copy link

No description provided.

Moved to catiss repo.
Migrate the original code from OpenXiangShan and re-write it with C++
style classes for the upstream riscv-isa-sim.
Rocket-chip restricts the values of MSTATUS.FS.
rocket-chip sets the xtval to instruction values when exceptions.
However, XiangShan and NutShell simply clear it, which is also
compatible with the RISC-V ISA.
It seems sim_t::addr_to_mem allocates the memory of only PGSIZE bytes
at every time. Therefore, we need to separate memcpy operations.
Different DUTs may have different physical address bits.
Both rocket-chip and NutShell do not implement ASIDs.
We add diff_trace_t class to record Spike's actions including

* instructions executed
* store requests
* load requests
* page-table walker requests

With the above information, we add support for the store commit
diff.
Rocket-chip sets the value to zero.
Load operations may cause exceptions. The is_amo flag must be set
just before the store is processed.
This commit adds the Sdtrig ISA string for the RISC-V Triggers
extension with permission check for the related CSRs.
poemonsense and others added 28 commits July 14, 2023 20:49
This CSR only exists if Sdtrig is enabled.
* misa
* satp
* mimpid
* marchid
Practical hardware designs would have reservation sets of 64B.
However, Spike seems to have only byte-level reservation sets.

I'm not sure whether it would violate the ISA.
If virtual memory has not been enabled, the CPU is not expected
to cache any PTEs and should not behave differently with Spike.
There was a macro MAX_INSN_LENGTH, but it seems not to be used anywhere.
We use it to decode the maximum insn_length.
When skipping instructions, minstret should be incremented.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants