Align base #3642
Align base #3642
Conversation
|
The first test was only modifying allocating from single buffers. It "only" failed 3 cases from the "Internal Intel CI System/merge/build" test suite. Next going to try only modifying contiguous buffer allocations. |
lyakh
force-pushed
the
align_test
branch
3 times, most recently
from
e068b02 to
de71fae
Compare
|
all tests passed except one - travis-ci/pr, which hasn't completed since almost 24 hours now. Restarting. |
|
SOFCI TEST |
|
This time travis-ci/pr completed successfully. This time several device tests failed, but since these failures weren't there last time, I think they're unrelated. Once this is merged, I'll continue with the remaining more intrusive 2 patches |
lyakh
requested review from
dbaluta,
jajanusz,
lbetlej,
lgirdwood,
libinyang,
mmaka1 and
plbossart
as code owners
|
@zrombel could you please have a look at the quickbuild failures here? Yesterday the test was green, no idea why it decided to re-test and why it's failing now. |
lgirdwood
left a comment
lgirdwood
left a comment
There was a problem hiding this comment.
This all good except path 3, which should be a separate fix in another PR·
src/lib/alloc.c
Outdated
There was a problem hiding this comment.
ehm, they should be using this. Not all memory is equal. This fix should probably be a new PR.
There was a problem hiding this comment.
ok, let's drop / postpone number 3 then, because they're interdependent
|
Not sure why we see module unload fail here. rerun. |
|
SOFCI TEST |
No need for two divisions for an alignment macro, one is enough. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Simplify the align_ptr() function by re-using an existing ALIGN() macro. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
alloc_heap_buffer() is trying to allocate memory from a single heap, no need to check its size on each iteration of an internal loop over maps. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
@lgirdwood failures are so strange. I'd bet they're unrelated, but I don't see many similar failures in other PRs around now. The failure on CNL seems again to be the old problem of IPC contents flipping to 0xffffffff: thesofproject/linux#2498 , as for the other two failures on BSW and BDW I'd really need to debug them. I see a ton of "no bytes to consume" / "no bytes to produce" messages in logs and I don't see how this PR can introduce any functional changes. If anything it should only make alignment calculation slightly faster by avoiding one superfluous division - that's all! I'll split this into 3 PRs just to test... |
|
@lyakh CI looks good now - there has been a back log. |
2 participants
We're taking a new approach to this. Let's merge these 4 patches first, then we'll deal with the two more complex remaining patches from #3607 separately in PRs #3646 and #2647 and #3607 can be closed then.