Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
lib.sh: perf_analyze(): drop --skip-to-first-trace
The --skip-to-first-trace option was added as a precaution to exclude log pollution from before the start of a test but how much pollution it excludes was never assessed; we just don't know. Hopefully not much? Replace, see below. mtrace timestamps used to return to zero on every DSP boot, which --skip-to-first-trace depends on. After SOF commit 76b0979c6453 ("sync time between host and fw for logging"), mtrace timestamps still return to zero (briefly) but only on "cold" DSP boots, not on when booting from IMR. Booting from IMR has now been enabled by default in most configurations. That combination breaks --skip-to-first-trace. The only practical short-term solution is to turn off --skip-to-first-trace and take the risk of old logs polluting the test results, which is what this commit does. If results pollution proves to be a problem, we could in the longer term try to disable IMR boot just before running this test using sof_debug as discussed in thesofproject#1153 This would also measure from a possibly cleaner DSP state. New, DSP "debug ops" being proposed in thesofproject/linux#4726 might also help. Fixes: thesofproject#1153 Signed-off-by: Baofeng Tian <[email protected]> Signed-off-by: Chao Song <[email protected]>
- Loading branch information