> Which led me to thinking if we even have GPU resources available today in the CI infrastructure to run tests? I was reading through https://issues.apache.org/jira/browse/SOLR-17892 and saw this being mentioned:
Rahul, thank you for bringing this up in our Lucene on GPU meeting today. I just wanted to respond in this thread to point out that we're (NVIDIA) looking at ways we can help with this so that the GPU integration w/ cuvs-lucene can be fully tested each time Apache Solr is released. At the very least, we can make sure we're running the tests frequently and make the results available. On Mon, May 11, 2026 at 1:05 AM Rahul Goswami <[email protected]> wrote: > Hi all, > I recently tried out the cuVS-lucene backed codec in Solr 10.0 for GPU > accelerated vector indexing and I think it is a pretty exciting and > cutting-edge addition to Solr. However at the moment it seems to not work > out of the box and the failure mode is a silent switch to CPU backed vector > indexing (using the lucene native Lucene99HnswVectorsWriter instead of the > cuvs-lucene based writer). > > I submitted a patch for this and validated the same on a GPU machine: > https://github.com/apache/solr/pull/4328 . Also cherry-picked to > branch_10_0 and branch_10x. > > Which led me to thinking if we even have GPU resources available today in > the CI infrastructure to run tests? I was reading through > https://issues.apache.org/jira/browse/SOLR-17892 and saw this being > mentioned: > > - We are also exploring the possibility of contributing GPU resources to > Apache Solr’s CI infrastructure for continuous GPU test coverage (to be > discussed further with the community). > > > Hence checking with the community if there was a conclusion to this (?) > > There is also a mention about benchmarks in the JIRA and would love to see > if the same were also made available at any point? > > Thanks, > Rahul >
