[GitHub] [lucene] jbellis closed pull request #12303: Address HNSW Searcher performance regression
jbellis closed pull request #12303: Address HNSW Searcher performance regression URL: https://github.com/apache/lucene/pull/12303 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] gsmiller commented on issue #12317: Option for disabling term dictionary compression
gsmiller commented on issue #12317: URL: https://github.com/apache/lucene/issues/12317#issuecomment-1556202535 I'm no expert in this area of our codec, but I'm curious to understand the issue a bit better. From the flame chart you provide, it looks like you're primarily looking at an indexing-related performance issue. Is that correct? When you disabled the patch, did you notice query-time performance changes? Compression isn't only useful for saving disk space; it's useful for keeping index pages hot in the OS cache and getting better data locality, which translates to better query-time performance. If your customer is running a 16gb heap on a machine with 32gb of memory, I wonder if they're benefitting much from the page cache? Is the heap squeezing out available memory for the page cache? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556224675 i made the benchmarks easier to run with something like this: ``` git clone https://github.com/rmuir/vectorbench cd vectorbench mvn verify java -jar target/vectorbench.jar ``` I can confirm everything works on aarch64 and i am experimenting with the unrolling. definitely if we remove the unrolling it gets way slower. so i tried to unroll again (4x instead of 2x), it is only a slight improvement in performance on my skylake: ``` Benchmark (size) Mode Cnt Score Error Units DotProductBenchmark.dotProductNew 1024 thrpt5 9.997 ± 0.999 ops/us DotProductBenchmark.dotProductNewNew1024 thrpt5 11.285 ± 0.161 ops/us DotProductBenchmark.dotProductOld 1024 thrpt5 2.024 ± 0.028 ops/us ``` But on the aarch64 mac additionall unrolling basically doubles throughput again (7.785 -> 14.912). I am experimenting more with this. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] ChrisHegarty commented on pull request #12311: Integrate the Incubating Panama Vector API
ChrisHegarty commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556278132 I didn't get an anywhere with Luceneutil yet! :-( (I haven't been able to run it successfully, getting OOM errors ) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556284488 thanks for sanity checking! i'm still working on the repo and making improvements. would be super-curious if you could 'git pull' and try -psize=1024 on your avx512 machine. hopefully it looks better now there, we were being inefficient. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556285923 With latest commits to that vectorbench I see this on my m1: ``` Benchmark (size) Mode Cnt Score Error Units DotProductBenchmark.dotProductNew 1024 thrpt5 9.463 ± 0.004 ops/us DotProductBenchmark.dotProductNewNew1024 thrpt5 16.106 ± 0.048 ops/us DotProductBenchmark.dotProductOld 1024 thrpt5 3.828 ± 0.003 ops/us ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556287294 thanks, glad it fixes the problem. i am running it across all the sizes we test and seeing how it looks on both my machines. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] ChrisHegarty commented on pull request #12311: Integrate the Incubating Panama Vector API
ChrisHegarty commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556288076 > we were being inefficient. If I understand this correctly, the inefficiency was too many reduceLances, right? You replaced it with addition of the accumulators before reducing. Sounds good. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556288827 yes, i think we have to imagine it as a scalar operation that gets slower as vector size increases. i looked into it and read this answer and changed the code: https://stackoverflow.com/questions/6996764/fastest-way-to-do-horizontal-sse-vector-sum-or-other-reduction/35270026#35270026 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556297297 i pushed one more commit to improve for "unaligned" vectors. the way to think about it, with unrolling, we do 64-at-a-time on avx512. So it isn't good to do worst-case 63 scalar computations just because user had 1023 dimensions or something like that. better to be bounded to 15. It makes things more well-rounded and prevents seeing slowdowns for sizes such as 702 in the test. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556297355 aarch64: ``` Benchmark (size) Mode CntScore Error Units DotProductBenchmark.dotProductNew 1 thrpt5 322.255 ± 0.496 ops/us DotProductBenchmark.dotProductNew 4 thrpt5 247.637 ± 1.027 ops/us DotProductBenchmark.dotProductNew 6 thrpt5 222.331 ± 0.156 ops/us DotProductBenchmark.dotProductNew 8 thrpt5 184.218 ± 0.074 ops/us DotProductBenchmark.dotProductNew 13 thrpt5 128.962 ± 0.017 ops/us DotProductBenchmark.dotProductNew 16 thrpt5 165.321 ± 0.057 ops/us DotProductBenchmark.dotProductNew 25 thrpt5 132.347 ± 0.435 ops/us DotProductBenchmark.dotProductNew 32 thrpt5 145.280 ± 5.810 ops/us DotProductBenchmark.dotProductNew 64 thrpt5 111.581 ± 9.717 ops/us DotProductBenchmark.dotProductNew100 thrpt5 81.562 ± 0.310 ops/us DotProductBenchmark.dotProductNew128 thrpt5 80.488 ± 0.255 ops/us DotProductBenchmark.dotProductNew207 thrpt5 41.804 ± 0.360 ops/us DotProductBenchmark.dotProductNew256 thrpt5 40.422 ± 0.090 ops/us DotProductBenchmark.dotProductNew300 thrpt5 33.164 ± 0.131 ops/us DotProductBenchmark.dotProductNew512 thrpt5 21.133 ± 0.020 ops/us DotProductBenchmark.dotProductNew702 thrpt5 13.383 ± 0.010 ops/us DotProductBenchmark.dotProductNew 1024 thrpt59.402 ± 0.198 ops/us DotProductBenchmark.dotProductNewNew 1 thrpt5 322.162 ± 1.042 ops/us DotProductBenchmark.dotProductNewNew 4 thrpt5 247.393 ± 3.673 ops/us DotProductBenchmark.dotProductNewNew 6 thrpt5 219.753 ± 2.000 ops/us DotProductBenchmark.dotProductNewNew 8 thrpt5 189.224 ± 3.233 ops/us DotProductBenchmark.dotProductNewNew 13 thrpt5 148.712 ± 6.845 ops/us DotProductBenchmark.dotProductNewNew 16 thrpt5 169.608 ± 0.200 ops/us DotProductBenchmark.dotProductNewNew 25 thrpt5 105.866 ± 0.506 ops/us DotProductBenchmark.dotProductNewNew 32 thrpt5 146.394 ± 0.802 ops/us DotProductBenchmark.dotProductNewNew 64 thrpt5 119.317 ± 0.385 ops/us DotProductBenchmark.dotProductNewNew 100 thrpt5 84.921 ± 2.819 ops/us DotProductBenchmark.dotProductNewNew 128 thrpt5 87.055 ± 0.473 ops/us DotProductBenchmark.dotProductNewNew 207 thrpt5 51.933 ± 0.270 ops/us DotProductBenchmark.dotProductNewNew 256 thrpt5 55.509 ± 3.926 ops/us DotProductBenchmark.dotProductNewNew 300 thrpt5 26.955 ± 0.016 ops/us DotProductBenchmark.dotProductNewNew 512 thrpt5 20.205 ± 0.076 ops/us DotProductBenchmark.dotProductNewNew 702 thrpt5 20.980 ± 0.029 ops/us DotProductBenchmark.dotProductNewNew1024 thrpt5 16.244 ± 0.172 ops/us DotProductBenchmark.dotProductOld 1 thrpt5 347.324 ± 5.857 ops/us DotProductBenchmark.dotProductOld 4 thrpt5 247.653 ± 0.285 ops/us DotProductBenchmark.dotProductOld 6 thrpt5 262.647 ± 1.425 ops/us DotProductBenchmark.dotProductOld 8 thrpt5 227.780 ± 0.477 ops/us DotProductBenchmark.dotProductOld 13 thrpt5 154.562 ± 2.007 ops/us DotProductBenchmark.dotProductOld 16 thrpt5 154.300 ± 0.460 ops/us DotProductBenchmark.dotProductOld 25 thrpt5 92.981 ± 0.083 ops/us DotProductBenchmark.dotProductOld 32 thrpt5 89.480 ± 0.211 ops/us DotProductBenchmark.dotProductOld 64 thrpt5 50.378 ± 0.064 ops/us DotProductBenchmark.dotProductOld100 thrpt5 38.107 ± 0.115 ops/us DotProductBenchmark.dotProductOld128 thrpt5 27.906 ± 0.030 ops/us DotProductBenchmark.dotProductOld207 thrpt5 21.094 ± 0.016 ops/us DotProductBenchmark.dotProductOld256 thrpt5 14.728 ± 0.004 ops/us DotProductBenchmark.dotProductOld300 thrpt5 14.860 ± 0.025 ops/us DotProductBenchmark.dotProductOld512 thrpt57.496 ± 0.041 ops/us DotProductBenchmark.dotProductOld702 thrpt56.369 ± 0.015 ops/us DotProductBenchmark.dotProductOld 1024 thrpt53.826 ± 0.002 ops/us ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556297561 skylake: ``` Benchmark (size) Mode CntScore Error Units DotProductBenchmark.dotProductNew 1 thrpt5 153.702 ± 2.576 ops/us DotProductBenchmark.dotProductNew 4 thrpt5 95.861 ± 1.403 ops/us DotProductBenchmark.dotProductNew 6 thrpt5 93.582 ± 1.640 ops/us DotProductBenchmark.dotProductNew 8 thrpt5 81.923 ± 1.045 ops/us DotProductBenchmark.dotProductNew 13 thrpt5 66.178 ± 0.789 ops/us DotProductBenchmark.dotProductNew 16 thrpt5 62.173 ± 1.191 ops/us DotProductBenchmark.dotProductNew 25 thrpt5 40.726 ± 0.455 ops/us DotProductBenchmark.dotProductNew 32 thrpt5 59.063 ± 6.797 ops/us DotProductBenchmark.dotProductNew 64 thrpt5 51.108 ± 1.368 ops/us DotProductBenchmark.dotProductNew100 thrpt5 35.460 ± 0.310 ops/us DotProductBenchmark.dotProductNew128 thrpt5 39.522 ± 0.356 ops/us DotProductBenchmark.dotProductNew207 thrpt5 21.369 ± 0.182 ops/us DotProductBenchmark.dotProductNew256 thrpt5 26.010 ± 0.112 ops/us DotProductBenchmark.dotProductNew300 thrpt5 19.118 ± 0.389 ops/us DotProductBenchmark.dotProductNew512 thrpt5 17.368 ± 0.755 ops/us DotProductBenchmark.dotProductNew702 thrpt5 11.338 ± 0.143 ops/us DotProductBenchmark.dotProductNew 1024 thrpt5 10.073 ± 0.113 ops/us DotProductBenchmark.dotProductNewNew 1 thrpt5 152.223 ± 0.854 ops/us DotProductBenchmark.dotProductNewNew 4 thrpt5 114.786 ± 1.555 ops/us DotProductBenchmark.dotProductNewNew 6 thrpt5 91.451 ± 0.874 ops/us DotProductBenchmark.dotProductNewNew 8 thrpt5 81.767 ± 0.345 ops/us DotProductBenchmark.dotProductNewNew 13 thrpt5 67.915 ± 0.889 ops/us DotProductBenchmark.dotProductNewNew 16 thrpt5 64.509 ± 1.064 ops/us DotProductBenchmark.dotProductNewNew 25 thrpt5 53.764 ± 1.037 ops/us DotProductBenchmark.dotProductNewNew 32 thrpt5 62.759 ± 0.942 ops/us DotProductBenchmark.dotProductNewNew 64 thrpt5 55.151 ± 0.396 ops/us DotProductBenchmark.dotProductNewNew 100 thrpt5 37.558 ± 0.996 ops/us DotProductBenchmark.dotProductNewNew 128 thrpt5 46.005 ± 0.733 ops/us DotProductBenchmark.dotProductNewNew 207 thrpt5 26.135 ± 0.780 ops/us DotProductBenchmark.dotProductNewNew 256 thrpt5 30.208 ± 0.115 ops/us DotProductBenchmark.dotProductNewNew 300 thrpt5 22.830 ± 1.903 ops/us DotProductBenchmark.dotProductNewNew 512 thrpt5 17.916 ± 0.216 ops/us DotProductBenchmark.dotProductNewNew 702 thrpt5 12.854 ± 1.727 ops/us DotProductBenchmark.dotProductNewNew1024 thrpt5 11.620 ± 0.291 ops/us DotProductBenchmark.dotProductOld 1 thrpt5 162.477 ± 3.116 ops/us DotProductBenchmark.dotProductOld 4 thrpt5 120.188 ± 2.748 ops/us DotProductBenchmark.dotProductOld 6 thrpt5 120.427 ± 1.619 ops/us DotProductBenchmark.dotProductOld 8 thrpt5 98.704 ± 2.279 ops/us DotProductBenchmark.dotProductOld 13 thrpt5 76.331 ± 1.940 ops/us DotProductBenchmark.dotProductOld 16 thrpt5 67.417 ± 1.456 ops/us DotProductBenchmark.dotProductOld 25 thrpt5 47.443 ± 0.513 ops/us DotProductBenchmark.dotProductOld 32 thrpt5 43.270 ± 4.112 ops/us DotProductBenchmark.dotProductOld 64 thrpt5 26.506 ± 0.826 ops/us DotProductBenchmark.dotProductOld100 thrpt5 16.793 ± 0.163 ops/us DotProductBenchmark.dotProductOld128 thrpt5 14.332 ± 0.207 ops/us DotProductBenchmark.dotProductOld207 thrpt59.032 ± 0.155 ops/us DotProductBenchmark.dotProductOld256 thrpt57.853 ± 0.115 ops/us DotProductBenchmark.dotProductOld300 thrpt56.331 ± 0.025 ops/us DotProductBenchmark.dotProductOld512 thrpt54.027 ± 0.023 ops/us DotProductBenchmark.dotProductOld702 thrpt52.762 ± 0.041 ops/us DotProductBenchmark.dotProductOld 1024 thrpt52.003 ± 0.020 ops/us ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556403183
I pushed a new benchmark to https://github.com/rmuir/vectorbench for the
binary dot product.
Basically this has to act like:
```
int sum = 0;
for (...) {
short product = (short) (a[i] * b[i]);
sum += (int) product;
}
```
So it is tricky to do with totally generic implementation (just using
SPECIES_PREFERRED). for avx256, it means you read byte vector of length 32 and
then work each half as short (2 short vectors of length 16), and then the same
thing again for each half as int (4 int vectors of length 8). This generic
approach only gives me a 2x speedup which is a little disappointing.
but this is a stupid approach if you have 256-bit vectors. You can just use
ByteVector.SPECIES_64, ShortVector.SPECIES_128, and IntVector.SPECIES_256 and
the whole thing is much faster.
on my skylake (has avx 256 and gets the optimized 256-bit impl)
```
Benchmark(size) Mode CntScore Error
Units
BinaryDotProductBenchmark.dotProductNew 1 thrpt5 159.476 ± 8.177
ops/us
BinaryDotProductBenchmark.dotProductNew 128 thrpt5 41.759 ± 0.267
ops/us
BinaryDotProductBenchmark.dotProductNew 207 thrpt5 25.094 ± 0.107
ops/us
BinaryDotProductBenchmark.dotProductNew 256 thrpt5 24.841 ± 0.124
ops/us
BinaryDotProductBenchmark.dotProductNew 300 thrpt5 19.624 ± 0.891
ops/us
BinaryDotProductBenchmark.dotProductNew 512 thrpt5 13.763 ± 0.171
ops/us
BinaryDotProductBenchmark.dotProductNew 702 thrpt59.792 ± 0.388
ops/us
BinaryDotProductBenchmark.dotProductNew1024 thrpt56.878 ± 0.834
ops/us
BinaryDotProductBenchmark.dotProductOld 1 thrpt5 160.423 ± 6.845
ops/us
BinaryDotProductBenchmark.dotProductOld 128 thrpt5 13.300 ± 0.159
ops/us
BinaryDotProductBenchmark.dotProductOld 207 thrpt58.678 ± 0.293
ops/us
BinaryDotProductBenchmark.dotProductOld 256 thrpt56.892 ± 0.331
ops/us
BinaryDotProductBenchmark.dotProductOld 300 thrpt56.008 ± 0.438
ops/us
BinaryDotProductBenchmark.dotProductOld 512 thrpt53.613 ± 0.192
ops/us
BinaryDotProductBenchmark.dotProductOld 702 thrpt52.710 ± 0.167
ops/us
BinaryDotProductBenchmark.dotProductOld1024 thrpt51.825 ± 0.125
ops/us
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
-
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] rmuir commented on pull request #12311: Integrate the Incubating Panama Vector API
rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556405150 and here's the results on my aarch64 mac, which has only 128-bit vectors and gets that disappointing generic impl: ``` Benchmark(size) Mode CntScore Error Units BinaryDotProductBenchmark.dotProductNew 1 thrpt5 334.839 ± 0.368 ops/us BinaryDotProductBenchmark.dotProductNew 128 thrpt5 34.097 ± 0.026 ops/us BinaryDotProductBenchmark.dotProductNew 207 thrpt5 22.045 ± 0.063 ops/us BinaryDotProductBenchmark.dotProductNew 256 thrpt5 18.782 ± 0.603 ops/us BinaryDotProductBenchmark.dotProductNew 300 thrpt5 15.932 ± 0.049 ops/us BinaryDotProductBenchmark.dotProductNew 512 thrpt59.985 ± 0.010 ops/us BinaryDotProductBenchmark.dotProductNew 702 thrpt57.322 ± 0.001 ops/us BinaryDotProductBenchmark.dotProductNew1024 thrpt55.146 ± 0.002 ops/us BinaryDotProductBenchmark.dotProductOld 1 thrpt5 343.722 ± 0.640 ops/us BinaryDotProductBenchmark.dotProductOld 128 thrpt5 24.981 ± 0.102 ops/us BinaryDotProductBenchmark.dotProductOld 207 thrpt5 14.944 ± 0.128 ops/us BinaryDotProductBenchmark.dotProductOld 256 thrpt5 12.541 ± 0.006 ops/us BinaryDotProductBenchmark.dotProductOld 300 thrpt5 10.663 ± 0.005 ops/us BinaryDotProductBenchmark.dotProductOld 512 thrpt56.198 ± 0.019 ops/us BinaryDotProductBenchmark.dotProductOld 702 thrpt54.531 ± 0.062 ops/us BinaryDotProductBenchmark.dotProductOld1024 thrpt53.108 ± 0.003 ops/us ``` I'd be curious how we could implement this with better performance, especially if we can just have a single generic impl like the float one. I feel like I must be doing it wrong :) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] tang-hi opened a new pull request, #12322: NeighborQueue set incomplemete false when call clear
tang-hi opened a new pull request, #12322: URL: https://github.com/apache/lucene/pull/12322 ### Description solve the bug that @msokolov mentioned in [PR](https://github.com/apache/lucene/pull/12255#issuecomment-1553088549) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] zhaih merged pull request #12257: Add multi-thread searchability to OnHeapHnswGraph
zhaih merged PR #12257: URL: https://github.com/apache/lucene/pull/12257 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org - To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
[GitHub] [lucene] zhaih commented on a diff in pull request #12246: Set word2vec getSynonyms method synchronized
zhaih commented on code in PR #12246:
URL: https://github.com/apache/lucene/pull/12246#discussion_r1199956128
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/synonym/word2vec/Word2VecSynonymProvider.java:
##
@@ -42,6 +42,7 @@ public class Word2VecSynonymProvider {
private static final VectorSimilarityFunction SIMILARITY_FUNCTION =
VectorSimilarityFunction.DOT_PRODUCT;
private static final VectorEncoding VECTOR_ENCODING = VectorEncoding.FLOAT32;
+ private static final int NO_LIMIT_ON_VISITED_NODES = Integer.MAX_VALUE;
private final Word2VecModel word2VecModel;
private final HnswGraph hnswGraph;
Review Comment:
I merged https://github.com/apache/lucene/pull/12257, but I think we still
need to make a change here to declare it as `OnHeapHnswGraph` to make it work.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
-
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
