Repository navigation
Run the nightly and cache regressions again, on the riscv-arch-test ELFs - #1963
Merged
jordancarlin merged 5 commits intoOct 12, 2026
Merged
Conversation
The testbench asks make for an ELF's .objdump.addr only. The recipe moved
.addr into place first, so a parallel simulation of the same ELF could
find .addr up to date and open .lab before it existed ("Could not open
the label maps"), and the ELF halted without reporting to tohost.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Signed-off-by: David Harris <David_Harris@hmc.edu>
The derivative configurations and the cache tests still named the tests.vh suites, so --nightly and --cache were refused. Each derivative configuration now runs the ACT ELFs of the configuration it is derived from, by leaf directory or by a glob within one where it lacks an extension the rest of the directory needs (addActTests). noS and noU run the ELFs of rv32i, rv64i and rv32imc, because the rv32gc and rv64gc ELFs set up S mode as they boot. The cache tests run the rv64gc ACT directories with the cache loggers; CacheSim.py already empties its model at each reset the loggers record, so a run may hold several ELFs. --nightly also builds config/deriv, and the embench and coremark checks get the grepstr their TestCase now requires. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Signed-off-by: David Harris <David_Harris@hmc.edu>
… hour --testfloat and --nightly ran the testfloat tests without building their vectors, so every one failed to open its .tv file on a fresh checkout. Under --testfloat's 30-minute limit the fma tests of the largest FPUs (fdh, fdq, fdqh) also timed out; they get the hour nightly allows. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Signed-off-by: David Harris <David_Harris@hmc.edu>
Since the move to ELF directories, sim_run simulated only bd_speedopt_speed, so benchmark_speed.py found no results for bd_sizeopt_speed and make run failed. Both builds now run, in parallel because each takes about half an hour. wsim names its ELF list after the run (--name) instead of the testbench mode, so the two embench runs no longer overwrite each other's list. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Signed-off-by: David Harris <David_Harris@hmc.edu>
The benchmark checks had not run since nightly was turned off, and the RTL merged since then shifted some stall, cache and cycle counts. CoreMark/MHz and CPI are unchanged; two Embench speed scores moved by 0.01. The new values are from this branch, and a second CoreMark sweep reproduced them exactly. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Signed-off-by: David Harris <David_Harris@hmc.edu>
jordancarlin
approved these changes
Oct 12, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
regression-wally --nightlyand--cachewere refused after #1895 because the derivative configurations and cache tests still named tests.vh suites. Each derivative configuration now runs the riscv-arch-test ELFs of the configuration it is derived from, by leaf directory or by a glob within one where it lacks an extension the rest of the directory needs; noS and noU use the rv32i, rv64i and rv32imc ELFs, since the rv32gc and rv64gc ELFs set up S mode as they boot, so they cover less than before until the derivatives get ACT configurations of their own. Getting the full nightly to run also needed: building config/deriv and the testfloat vectors, a required field for the embench and coremark checks, simulating embench's size-optimized build again (with wsim naming its ELF lists after--nameso parallel runs of one mode don't collide), an hour for--testfloat's largest fma tests, refreshed CoreMark and Embench expected results, and a testbench/Makefile fix for a race in which a simulation could open an ELF's label maps before both existed. Nightly then passes except for rv32imc'spriv/ExceptionsSmon Questa and VCS, which hangs at a misaligned fetch from a nonexistent address (Verilator passes) and #1968 fixes, and the two RV32 Q configurations'ZfaD-fmvp.d.x, which #1962 fixes.🤖 Generated with Claude Code