Skip to content
New issue

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

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

Already on GitHub? # to your account

XBufVector WIP #10

Open
wants to merge 1 commit into
base: bleve
Choose a base branch
from
Open

XBufVector WIP #10

wants to merge 1 commit into from

Conversation

abhinavdangeti
Copy link
Member

From @steveyen -->

This now passes faiss old tests, but doesn't really exercise any of the new XBufVector functionality yet.

This PR seems to compile, but it does not pass tests -- especially, the ONDISK.test_add fails because I think it's not using an BufIOReader.

make -C build test
Running tests...
Test project /Users/steve.yen/work/membase/dev/tmp/faiss-blevesearch/build
    Start 23: ONDISK.test_add
1/3 Test #23: ONDISK.test_add ...................***Exception: SegFault  0.04 sec
WARNING clustering 1000 points to 40 centroids: please provide at least 1560 training points
Running main() from /Users/steve.yen/dev/tmp/faiss-blevesearch/build/_deps/googletest-src/googletest/src/gtest_main.cc
Note: Google Test filter = ONDISK.test_add
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from ONDISK
[ RUN      ] ONDISK.test_add
From lldb...

[       OK ] ONDISK.make_invlists (954 ms)
[ RUN      ] ONDISK.test_add
resizing /tmp/faiss_tmp_WkfcZL to 64 bytes
resizing /tmp/faiss_tmp_WkfcZL to 128 bytes
resizing /tmp/faiss_tmp_WkfcZL to 256 bytes
resizing /tmp/faiss_tmp_WkfcZL to 512 bytes
resizing /tmp/faiss_tmp_WkfcZL to 1024 bytes
resizing /tmp/faiss_tmp_WkfcZL to 2048 bytes
resizing /tmp/faiss_tmp_WkfcZL to 4096 bytes
resizing /tmp/faiss_tmp_WkfcZL to 8192 bytes
resizing /tmp/faiss_tmp_WkfcZL to 16384 bytes
resizing /tmp/faiss_tmp_WkfcZL to 32768 bytes
resizing /tmp/faiss_tmp_WkfcZL to 65536 bytes
resizing /tmp/faiss_tmp_WkfcZL to 131072 bytes
resizing /tmp/faiss_tmp_WkfcZL to 262144 bytes
Process 8740 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x20)
    frame #0: 0x00000001012f61d0 libfaiss.dylib`faiss::BufIOReader::readPointer(unsigned long) + 16
libfaiss.dylib`faiss::BufIOReader::readPointer:
->  0x1012f61d0 <+16>: ldr    x8, [x9, #0x20]
    0x1012f61d4 <+20>: ldr    x10, [x9, #0x28]
    0x1012f61d8 <+24>: add    x8, x8, x10
    0x1012f61dc <+28>: str    x8, [sp, 

Latest news on this... the old tests now pass. But, that doesn't mean the new codepaths are getting invoked and exercised.

It shows a couple examples of replacing std::vector with XBufVector -- and with replacing calls to READVECTOR() with X_READVECTOR().

"the old tests now pass" ==> correction on this ==> I meant to say the old tests now pass no worse or better than before! On my dev computer, the old tests didn't fully 100% pass before I started working on this... especially, on the ThreadedIndex tests...

        Start  45: ThreadedIndex.SingleException
 45/105 Test  #45: ThreadedIndex.SingleException ...........................................Bus error***Exception:   0.03 sec
        Start  46: ThreadedIndex.MultipleException
 46/105 Test  #46: ThreadedIndex.MultipleException .........................................Bus error***Exception:   0.03 sec

...

98% tests passed, 2 tests failed out of 105

Total Test time (real) = 127.72 sec

The following tests FAILED:
	 45 - ThreadedIndex.SingleException (Bus error)
	 46 - ThreadedIndex.MultipleException (Bus error)
Errors while running CTest

This now passes faiss old tests, but doesn't really
exercise any of the new XBufVector functionality yet.
@abhinavdangeti abhinavdangeti mentioned this pull request Dec 6, 2023
@abhinavdangeti abhinavdangeti added the do not merge Not ready to go in just yet label Apr 26, 2024
@abhinavdangeti abhinavdangeti removed the request for review from moshaad7 October 17, 2024 17:50
# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
do not merge Not ready to go in just yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants