Re: [I] Significant drop in recall for int8 scalar quantization using maximum_inner_product [lucene]

2024-05-18 Thread via GitHub
naveentatikonda commented on issue #13350: URL: https://github.com/apache/lucene/issues/13350#issuecomment-2119034285 > Scalar quantization in Lucene is by default 7 bits. Meaning the range of values is actually 0-127. For 7 bits, it makes sense. But, the example I was referring to is

Re: [PR] Reduce the overhead of `IndexInput#prefetch` when data is cached in RAM. [lucene]

2024-05-18 Thread via GitHub
rmuir commented on code in PR #13381: URL: https://github.com/apache/lucene/pull/13381#discussion_r1605828738 ## lucene/core/src/java21/org/apache/lucene/store/MemorySegmentIndexInput.java: ## @@ -57,6 +58,7 @@ abstract class MemorySegmentIndexInput extends IndexInput implement

Re: [PR] Reduce the overhead of `IndexInput#prefetch` when data is cached in RAM. [lucene]

2024-05-18 Thread via GitHub
rmuir commented on code in PR #13381: URL: https://github.com/apache/lucene/pull/13381#discussion_r1605825907 ## lucene/core/src/java21/org/apache/lucene/store/MemorySegmentIndexInput.java: ## @@ -57,6 +58,7 @@ abstract class MemorySegmentIndexInput extends IndexInput implement

Re: [PR] Reduce the overhead of `IndexInput#prefetch` when data is cached in RAM. [lucene]

2024-05-18 Thread via GitHub
rmuir commented on code in PR #13381: URL: https://github.com/apache/lucene/pull/13381#discussion_r1605825529 ## lucene/core/src/java/org/apache/lucene/util/BitUtil.java: ## @@ -303,4 +303,12 @@ public static int zigZagDecode(int i) { public static long zigZagDecode(long l) {

Re: [PR] Convert more classes to record classes [lucene]

2024-05-18 Thread via GitHub
uschindler commented on PR #13328: URL: https://github.com/apache/lucene/pull/13328#issuecomment-2118884788 Hi @shubhamvishu, can you check my comments? The changes here look fine, please fix the remaining issues. Uwe -- This is an automated message from the Apache Git Service. To

Re: [PR] Add a separate option to allow running Panama Vectorization for all tests with suitable C2 defaults [lucene]

2024-05-18 Thread via GitHub
uschindler commented on PR #13351: URL: https://github.com/apache/lucene/pull/13351#issuecomment-2118883258 Thanks! -- 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 unsub

Re: [PR] Add a MemorySegment Vector scorer - for scoring without copying on-heap [lucene]

2024-05-18 Thread via GitHub
uschindler commented on PR #13339: URL: https://github.com/apache/lucene/pull/13339#issuecomment-2118880465 Back from vacation. Will look into that till Tuesday! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL

Re: [PR] Reduce the overhead of `IndexInput#prefetch` when data is cached in RAM. [lucene]

2024-05-18 Thread via GitHub
uschindler commented on PR #13381: URL: https://github.com/apache/lucene/pull/13381#issuecomment-2118878096 P.S.: Actually when looking at the code, the `MemorySegment#load()` method calls `madvise(MADV_WILLNEED)`. So we could also implement prefetch using `load()`. The only problem

Re: [PR] Reduce the overhead of `IndexInput#prefetch` when data is cached in RAM. [lucene]

2024-05-18 Thread via GitHub
uschindler commented on code in PR #13381: URL: https://github.com/apache/lucene/pull/13381#discussion_r1605815826 ## lucene/core/src/java21/org/apache/lucene/store/MemorySegmentIndexInput.java: ## @@ -344,7 +354,11 @@ public void prefetch(long offset, long length) throws IOExc

Re: [PR] Reduce the overhead of `IndexInput#prefetch` when data is cached in RAM. [lucene]

2024-05-18 Thread via GitHub
rmuir commented on code in PR #13381: URL: https://github.com/apache/lucene/pull/13381#discussion_r1605799608 ## lucene/core/src/java21/org/apache/lucene/store/PosixNativeAccess.java: ## @@ -165,6 +181,44 @@ private Integer mapReadAdvice(ReadAdvice readAdvice) { }; } +

Re: [PR] Replace Map by primitive IntObjectHashMap. [lucene]

2024-05-18 Thread via GitHub
bruno-roustant merged PR #13368: URL: https://github.com/apache/lucene/pull/13368 -- 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...@luc