Phoneix-Bench / issues /FPGA-MAFIA_fpga_mafia_pr479
OvertureMMXXI's picture
Add 511-case data manifest (testbench_verilator, issues, snapshots) (#2)
123e810
Raw
History Blame
1.83 kB
# core_rrv - create a dedicated test for the timer interrupt during a illegal instruction trap
- Source URL: https://github.com/FPGA-MAFIA/fpga_mafia/issues/477
Create a test that has both "illegal instruction trap" and "timer interrupts".
I am hoping to see hit this case:
- illegal instruction trap exception handler is in progress (MIE set to 0 by HW)
- Timer interrupt time condition met -> but MIE is 0, so no exception is taken yet
- illegal instruction handler return with MRET (MIE back to 1)
- Timer interrupt condition is still set so the timer interrupt exception handler takes place
- We return from the timer interrupts and continue execution
PR Title: add test timer exception after illegal instruction trap + debug csr's
Status: Merged on Feb 14, 2024
Description: No description was provided in the pull request.
Commits (7 total):
1. "add clear and set csr macros"
2. "add debud csrs. dpc, dcsr, dscratch0, dscratch1"
3. "add test for checking alive illegal before timer interrupt"
4. "disable mie to avoid timer exception while printing. may cause misses"
5. "Merge branch 'main' into core_rrv"
6. "modify regrees file"
7. "add debug csrs, custom_mtime, custom_mtimecmp and custom_lfsr"
Review Comments:
amichai-bd (reviewer/approver):
- Regarding cycle/instret counter synchronization: "The plan is to artificially match the ref model performance IPC with the DUT performance so the 2 tests can be compared correctly."
- On LFSR updates: Expected hardware LFSR updates "only when there is a read/write from the LFSR CSR" by moving logic into read/write cases
roman012285 (author) response to LFSR comment: Indicated the fix hadn't been implemented yet ("ืขื•ื“ ืœื ื”ื’ืขืชื™ ืœื–ื”")
Linked Issue:
Issue #477: "create a dedicated test for the timer interrupt during a illegal instruction trap"