rmuir commented on issue #12023:
URL: https://github.com/apache/lucene/issues/12023#issuecomment-1353025015
determinization has already been removed here. that is the problem.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub
rmuir closed issue #12023: Mechanism to interrupt long-running/resource
intensive queries
URL: https://github.com/apache/lucene/issues/12023
--
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
Bukhtawar opened a new issue, #12023:
URL: https://github.com/apache/lucene/issues/12023
### Description
As a part of https://github.com/opensearch-project/OpenSearch/issues/687 we
detected that regex queries can run into tight loops for quite long. Below is
the stack trace of the
rmuir commented on issue #12021:
URL: https://github.com/apache/lucene/issues/12021#issuecomment-1352977757
This looks like a bug in solr code (SolrDocumentFetcher) so I'd recommend
opening a bug over at https://github.com/apache/solr
--
This is an automated message from the Apache Git
iverase commented on code in PR #12022:
URL: https://github.com/apache/lucene/pull/12022#discussion_r1049521314
##
lucene/CHANGES.txt:
##
@@ -68,6 +68,8 @@ Bug Fixes
* LUCENE-10599: LogMergePolicy is more likely to keep merging segments until
they reach the maximum merge
iverase commented on code in PR #12022:
URL: https://github.com/apache/lucene/pull/12022#discussion_r1049521314
##
lucene/CHANGES.txt:
##
@@ -68,6 +68,8 @@ Bug Fixes
* LUCENE-10599: LogMergePolicy is more likely to keep merging segments until
they reach the maximum merge
craigtaverner opened a new pull request, #12022:
URL: https://github.com/apache/lucene/pull/12022
Fixes https://github.com/apache/lucene/issues/12020
### Description
When performing a search using a shape geometry query of relation type
`QueryRelation.CONTAINS`, it is possible
nosvalds opened a new issue, #12021:
URL: https://github.com/apache/lucene/issues/12021
### Description
## Issue
For fields using `large="true"`, large fields (which is what they are
intended for) can be truncated in v9+ of Lucene.
Example fieldtype definition:
```
craigtaverner opened a new issue, #12020:
URL: https://github.com/apache/lucene/issues/12020
### Description
When performing a search using a shape geometry query of relation type
`QueryRelation.CONTAINS`, it is possible to get a false positive when two
geometries intersect, but
LuXugang commented on code in PR #12017:
URL: https://github.com/apache/lucene/pull/12017#discussion_r1049162986
##
lucene/core/src/test/org/apache/lucene/search/TestBooleanQuery.java:
##
@@ -1015,6 +1015,80 @@ public void testDisjunctionRandomClausesMatchesCount()
throws
LuXugang commented on code in PR #12017:
URL: https://github.com/apache/lucene/pull/12017#discussion_r1049163040
##
lucene/core/src/test/org/apache/lucene/search/TestBooleanQuery.java:
##
@@ -1015,6 +1015,80 @@ public void testDisjunctionRandomClausesMatchesCount()
throws
LuXugang commented on code in PR #12017:
URL: https://github.com/apache/lucene/pull/12017#discussion_r1049162892
##
lucene/core/src/java/org/apache/lucene/search/BooleanWeight.java:
##
@@ -470,14 +470,19 @@ private int reqCount(LeafReaderContext context) throws
IOException {
uschindler commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352328979
> I will check it out and inspect the coverage report from .js tests and see
if there are any holes. If i find them I will push more tests. I am just really
paranoid about some of
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352322297
I will check it out and inspect the coverage report from .js tests and see
if there are any holes. If i find them I will push more tests. I am just really
paranoid about some of the slow
uschindler commented on code in PR #12016:
URL: https://github.com/apache/lucene/pull/12016#discussion_r1049026102
##
lucene/expressions/src/java/module-info.java:
##
@@ -19,7 +19,7 @@
module org.apache.lucene.expressions {
Review Comment:
Ok. No problem. I just noticed
reta commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352291207
> sorry i didnt do the other tests, I gotta run for now. i can do them later
tonight or tomorrow if you need but just wanted to prototype it out
@rmuir tests have been migrated,
reta commented on code in PR #12016:
URL: https://github.com/apache/lucene/pull/12016#discussion_r1049019777
##
lucene/expressions/src/java/module-info.java:
##
@@ -19,7 +19,7 @@
module org.apache.lucene.expressions {
Review Comment:
ah ... it is still automatic :(
rmuir commented on code in PR #12016:
URL: https://github.com/apache/lucene/pull/12016#discussion_r1049016403
##
lucene/expressions/src/java/org/apache/lucene/expressions/js/JavascriptCompilerSettings.java:
##
@@ -0,0 +1,51 @@
+/*
+ * Licensed to the Apache Software Foundation
uschindler commented on code in PR #12016:
URL: https://github.com/apache/lucene/pull/12016#discussion_r1049015663
##
lucene/expressions/src/java/org/apache/lucene/expressions/js/JavascriptCompilerSettings.java:
##
@@ -0,0 +1,51 @@
+/*
+ * Licensed to the Apache Software
reta commented on code in PR #12016:
URL: https://github.com/apache/lucene/pull/12016#discussion_r1049014391
##
lucene/expressions/src/java/org/apache/lucene/expressions/js/JavascriptCompilerSettings.java:
##
@@ -0,0 +1,51 @@
+/*
+ * Licensed to the Apache Software Foundation
uschindler commented on code in PR #12016:
URL: https://github.com/apache/lucene/pull/12016#discussion_r1049010145
##
lucene/expressions/src/java/org/apache/lucene/expressions/js/JavascriptCompilerSettings.java:
##
@@ -0,0 +1,51 @@
+/*
+ * Licensed to the Apache Software
uschindler commented on code in PR #12016:
URL: https://github.com/apache/lucene/pull/12016#discussion_r1049005864
##
lucene/expressions/src/java/module-info.java:
##
@@ -19,7 +19,7 @@
module org.apache.lucene.expressions {
Review Comment:
Can we now remove this
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352259233
sorry i didnt do the other tests, I gotta run for now. i can do them later
tonight or tomorrow if you need but just wanted to prototype it out
--
This is an automated message from the
reta commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352258413
> I pushed a proposal to your branch (only fixing one of the tests in .js to
use it). If you are really against it, just revert the commit. But i think it
keeps tests simpler.
:+1:
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352248407
I pushed a proposal to your branch (only fixing one of the tests in .js to
use it). If you are really against it, just revert the commit. But i think it
keeps tests simpler.
--
This is
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352177919
we don't need to parameterize the pickiness IMO, we can just turn it on in
these tests. Thanks for getting this hooked up. I will look at your branch and
try to play with it.
--
This
reta commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1352084948
@rmuir first pass over pickiness, all tests are now run w/o diagnostics
listener (picky / not picky mode), the rough observations so far are
encouraging, there are no significant changes
benwtrent opened a new pull request, #12019:
URL: https://github.com/apache/lucene/pull/12019
I noticed that the Lucene94 backward-codec was still allowing KNN fields
writer. I have removed it, and updated the related tests similarly to other KNN
backward-codec changes.
There are
benwtrent commented on code in PR #11946:
URL: https://github.com/apache/lucene/pull/11946#discussion_r1048856364
##
lucene/core/src/java/org/apache/lucene/search/KnnVectorQuery.java:
##
@@ -76,12 +91,29 @@ public KnnVectorQuery(String field, float[] target, int k) {
*
msokolov commented on code in PR #11946:
URL: https://github.com/apache/lucene/pull/11946#discussion_r1048846385
##
lucene/core/src/java/org/apache/lucene/search/KnnVectorQuery.java:
##
@@ -76,12 +91,29 @@ public KnnVectorQuery(String field, float[] target, int k) {
*
rmuir commented on issue #11963:
URL: https://github.com/apache/lucene/issues/11963#issuecomment-1351946133
I would tend to expect some organization like this:
indexing:
* ByteVectorField
* FloatVectorField
searching:
* ByteVectorField.newQuery
*
benwtrent commented on issue #11963:
URL: https://github.com/apache/lucene/issues/11963#issuecomment-1351900658
OK, step one of this is done with the `KnnByteVectorQuery`. I am next
approaching a new `KnnByteVectorField` and this will cause some refactoring to
`LeafReader` or
AlmostFamiliar commented on issue #9231:
URL: https://github.com/apache/lucene/issues/9231#issuecomment-1351830837
Are there still plans to merge this?
I would also be very interested in this feature.
--
This is an automated message from the Apache Git Service.
To respond to the
benwtrent commented on code in PR #11946:
URL: https://github.com/apache/lucene/pull/11946#discussion_r1048764283
##
lucene/core/src/java/org/apache/lucene/search/KnnVectorQuery.java:
##
@@ -76,12 +91,29 @@ public KnnVectorQuery(String field, float[] target, int k) {
*
benwtrent commented on code in PR #11946:
URL: https://github.com/apache/lucene/pull/11946#discussion_r1048763428
##
lucene/core/src/java/org/apache/lucene/search/KnnVectorQuery.java:
##
@@ -76,12 +91,29 @@ public KnnVectorQuery(String field, float[] target, int k) {
*
jdconrad commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351772233
> javacc?
I was under the impression this was EOL? If it's still well supported I'm
not familiar enough with the code generation to know if this would avoid the
pitfalls ANTLR
msokolov commented on PR #11946:
URL: https://github.com/apache/lucene/pull/11946#issuecomment-1351747886
It seems there are conflicts due to a recent refactor of this query - would
you mind merging from main and resolving those please?
--
This is an automated message from the Apache Git
uschindler commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351743460
> One other possibility would be to hand-roll the expression lexer/parser.
This would get rid of the need for any additional dependencies and generated
code. From what I can tell the
uschindler commented on PR #12014:
URL: https://github.com/apache/lucene/pull/12014#issuecomment-1351734347
+1
--
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
rmuir commented on PR #12014:
URL: https://github.com/apache/lucene/pull/12014#issuecomment-1351680479
I looked at what e.g. glibc does here as a fallback out of curiousity, for
floats it is very simple (using Dekker algorithm), but requires changing the FP
rounding mode, which you cant do
jdconrad commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351666949
One other possibility would be to hand-roll the expression lexer/parser.
This would get rid of the need for any additional dependencies and generated
code. From what I can tell the API
dweiss commented on PR #12014:
URL: https://github.com/apache/lucene/pull/12014#issuecomment-1351660443
I honestly don't know who can use this method without any provided cpuid
check... We actually use fma in our code but do so by detecting the performance
difference between a naive
benwtrent commented on PR #12014:
URL: https://github.com/apache/lucene/pull/12014#issuecomment-1351560532
Holy crap, creating `BigDecimal` and then multiplying & adding is crazy.
This is a completely unacceptable fallback calculation for this method.
+1 on banning its use in the
jpountz merged PR #12018:
URL: https://github.com/apache/lucene/pull/12018
--
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:
benwtrent commented on PR #12018:
URL: https://github.com/apache/lucene/pull/12018#issuecomment-1351392230
@jpountz Here is the backport.
--
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
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351378582
another way to say it: shading is a terrible, TERRIBLE idea and you only
hear about it in java, because java is the only language with developers that
are bad enough to consider it.
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351374308
I feel that if we publish a shaded jar we become guilty of "contributing to
the delinquency" of terrible supply chain management, by hiding third party
dependencies and their versions
uschindler commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351368145
> > Can we specify our dependencies in a different way (e.g. exact version)
in our maven stuff so this won't happen? e.g. you can do this in python, and
specify that you depend on
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351366346
Yes, java showed what a disaster it is around supply chains with the log4j
vulnerability. Shading should not even be considered as an option.
--
This is an automated message from the
benwtrent opened a new pull request, #12018:
URL: https://github.com/apache/lucene/pull/12018
Backport of #12004
This is the first commit of a much larger refactor. The overall goal is to
separate the concerns of byte vectors and float vectors. Making their usage and
APIs clearer
dweiss commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351355396
> Can we specify our dependencies in a different way (e.g. exact version) in
our maven stuff so this won't happen? e.g. you can do this in python, and
specify that you depend on antlr ==
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351345609
I also dont understand the issue where ppl think they can modify arbitrary
versions of lucene dependencies.
Can we specify our dependencies in a different way (e.g. exact version)
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351332547
personally i am against the shading. I think it is a huge antipattern, it
hides third-party artifacts completely. think about someone trying to do
security or license audit and they have
benwtrent commented on PR #12004:
URL: https://github.com/apache/lucene/pull/12004#issuecomment-1351211685
@uschindler 100%! I can get the backport PR done asap
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
jpountz commented on code in PR #12017:
URL: https://github.com/apache/lucene/pull/12017#discussion_r1048383541
##
lucene/core/src/java/org/apache/lucene/search/BooleanWeight.java:
##
@@ -470,14 +470,19 @@ private int reqCount(LeafReaderContext context) throws
IOException {
dweiss commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1351155914
> With ASM this was hard to do but with Grade it is quite easy. Just add
another configuration with those two dependencies and use the "Gradle Shadow
Plugin"
LuXugang opened a new pull request, #12017:
URL: https://github.com/apache/lucene/pull/12017
When BooleanQuery is pure disjunction, if at least one clause could match
all docs, then we could get right `count` even though there was other clause
whose `count` is unknown.
--
This is
fcofdez commented on PR #11997:
URL: https://github.com/apache/lucene/pull/11997#issuecomment-1351000741
@jpountz thanks for looking into this, unfortunately I didn't have time to
go through the last round of review comments. The change makes sense to me.
--
This is an automated message
jpountz commented on PR #11997:
URL: https://github.com/apache/lucene/pull/11997#issuecomment-1350964750
@fcofdez I hope you don't mind but I checked out your branch to look into
the above issue and ended up pushing my changes. These new fields not only
support indexing sorted numeric doc
uschindler commented on PR #12004:
URL: https://github.com/apache/lucene/pull/12004#issuecomment-1350722742
> @benwtrent Would you mind helping with the backport by creating a PR?
There are a few conflicts and I could use some help with resolving them.
There are also Java 17 switch
jpountz commented on PR #12004:
URL: https://github.com/apache/lucene/pull/12004#issuecomment-1350702865
@benwtrent Would you mind helping with the backport by creating a PR? There
are a few conflicts and I could use some help with resolving them.
--
This is an automated message from the
iverase commented on PR #12006:
URL: https://github.com/apache/lucene/pull/12006#issuecomment-1350678308
> If it gives a good speedup, we could think about applying the same trick
to other queries too, if we can do it without making things too ugly. For
example LatLonPoint, IntPoint, etc
uschindler commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1350666129
Hi, I agree with all Robert say. I would also like to make another
suggestion (that could also be applied when this has been fixed). To me it
looks like a bad decission of antlr,
jpountz merged PR #12004:
URL: https://github.com/apache/lucene/pull/12004
--
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:
gf2121 commented on PR #12006:
URL: https://github.com/apache/lucene/pull/12006#issuecomment-1350639546
This brings a small jump for Distance Filter LatLonPoint task.
http://people.apache.org/~mikemccand/geobench.html
--
This is an automated message from the Apache Git Service.
To
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1350455204
Simple package-private `static` method to turn on the pickiness should do it.
I don't want to see 100 new constructors/abstractions added with booleans,
just because antlr made a
jdconrad commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1350071474
I agree with @rmuir that having an ambiguity check for tests similar to
Painless would be great for expressions. I'm a bit surprised this change didn't
require much additionally to the
reta commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349804323
Thanks for encouraging @rmuir ! I will be working on the matter this week
and share my findings, thank you!
--
This is an automated message from the Apache Git Service.
To respond to
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349778653
In general, sorry if i discouraged before, it is really just a frustrating
situation
If you get stuck, just leave the PR open. I will try to dig into this too, I
have been through
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349610775
cc: @jdconrad who might remember a lot more about this than me
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349606909
here is coverage report using the current antlr. I guess i dont know why so
much is missing here:
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349604029
As far as inspecting coverage, I suspect it is pretty good. But there is
instructions in https://github.com/apache/lucene/blob/main/help/tests.txt on
how to generate reports.
--
This
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349602516
@reta I remember doing this adds overhead, that's why it is a boolean there.
so it really just needs to be something we do from tests. for example it could
be a package-private setter or
reta commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349601503
> Looks like this in painless:
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349599904
Looks like this in painless:
rmuir commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349597878
the only way i know to prevent the traps is to do like painless and "enable
picky mode" which fails test instead of doing slow things. and to have 100%
test coverage of grammar!
--
reta commented on PR #12016:
URL: https://github.com/apache/lucene/pull/12016#issuecomment-1349213966
@rmuir @uschindler what kind of (performance? jmh?) testing would help to
discard / prove that moving to 4.11.x makes / does not make sense. You have
definitely seen traps in the past, I
reta opened a new pull request, #12016:
URL: https://github.com/apache/lucene/pull/12016
Signed-off-by: Andriy Redko
### Description
The Apache Lucene is using quite old version of ANTLR 4.5.1-1. By itself, it
is not a showstopper, but more profound issue is that some ANTLR
rmuir closed issue #12012: Spotless runs before javac in `gradle check` which
results in less-than-helpful errors on compilation problems
URL: https://github.com/apache/lucene/issues/12012
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on
rmuir commented on PR #12013:
URL: https://github.com/apache/lucene/pull/12013#issuecomment-1348842317
Yeah, I'm highly suspicious of this cache. I dug into this and found that
previously stored fields (!) were used for this lookup. So no surprise there
was a cache around it, since this is
rmuir commented on PR #12015:
URL: https://github.com/apache/lucene/pull/12015#issuecomment-1348790082
Thanks again, it really helps make the backporting process less painful for
me.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to
dweiss commented on PR #12015:
URL: https://github.com/apache/lucene/pull/12015#issuecomment-1348772038
Sorry - I see it now. I've pushed to my local fork at gh. Didn't notice it.
I've just pushed to 9x on apache as well.
--
This is an automated message from the Apache Git Service.
To
dweiss commented on PR #12015:
URL: https://github.com/apache/lucene/pull/12015#issuecomment-1348767456
You have?! Well, that's strange - I've pushed it to 9x... Can you reproduce
it with the head?
--
This is an automated message from the Apache Git Service.
To respond to the message,
rmuir commented on PR #12015:
URL: https://github.com/apache/lucene/pull/12015#issuecomment-1348669494
@dweiss do you mind pushing to 9.x, otherwise i will do it for you. I just
hit this again in 9.x :)
--
This is an automated message from the Apache Git Service.
To respond to the
rmuir merged PR #11998:
URL: https://github.com/apache/lucene/pull/11998
--
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:
rmuir commented on PR #12015:
URL: https://github.com/apache/lucene/pull/12015#issuecomment-1348488213
thank you @dweiss
--
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.
iverase merged PR #11988:
URL: https://github.com/apache/lucene/pull/11988
--
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:
iverase closed issue #11986: Polygons failing to tessellate
URL: https://github.com/apache/lucene/issues/11986
--
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,
dweiss commented on issue #12012:
URL: https://github.com/apache/lucene/issues/12012#issuecomment-1347874082
I've applied the patch to 9x and main, btw.
--
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
dweiss merged PR #12015:
URL: https://github.com/apache/lucene/pull/12015
--
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:
dweiss commented on issue #12012:
URL: https://github.com/apache/lucene/issues/12012#issuecomment-1347872711
I filed a PR #12015 . Indeed the 'core' formatting task seems to be named
'spotlessJava' (spotless + format name), the check and apply are just attached
to it in a fancy manner via
dweiss commented on issue #12012:
URL: https://github.com/apache/lucene/issues/12012#issuecomment-1347844975
> I think we are not slowing down the spotlessJava task which is the one
actually failing for me?
Eh. I don't know what the relationships between those spotless tasks are.
rmuir merged PR #12010:
URL: https://github.com/apache/lucene/pull/12010
--
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:
rmuir closed issue #12009: Find (and fix) places where we treat a long as a
double without an explicit cast
URL: https://github.com/apache/lucene/issues/12009
--
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
rmuir commented on PR #12014:
URL: https://github.com/apache/lucene/pull/12014#issuecomment-1347620183
Yeah, I think if the fallback java code was 2x, 4x, or 8x slower (like you
would expect from these intrinsics), we wouldn't be having this conversation :)
--
This is an automated
gsmiller commented on PR #12014:
URL: https://github.com/apache/lucene/pull/12014#issuecomment-1347617300
+1, seems reasonable to me. We can always remove this ban in the future if
there's a good reason, but seems reasonable to put this in place to prevent it
sneaking in for now.
--
rmuir opened a new pull request, #12014:
URL: https://github.com/apache/lucene/pull/12014
When FMA is not supported by the hardware, these methods fall back to
`BigDecimal` usage [1] which causes them to be 2500x slower [2].
While most hardware in the last 10 years may have the
uschindler commented on PR #12013:
URL: https://github.com/apache/lucene/pull/12013#issuecomment-1347501489
Please remove the cache. Totally useless. It hurts more because the escape
analysis can't work.
When tiered compilation stepped in, this is just a useless extra
My rule:
rmuir commented on PR #12013:
URL: https://github.com/apache/lucene/pull/12013#issuecomment-1347439475
Also (summoning my inner @uschindler), curious what the perf impact is if we
simply remove these caches completely. Maybe they are not relevant to the
performance anymore due to faceting
rmuir commented on issue #12012:
URL: https://github.com/apache/lucene/issues/12012#issuecomment-1347423692
I ran this patch on the mac m1 and also got `BUILD SUCCESSFUL in 3m 33s`.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to
501 - 600 of 27689 matches
Mail list logo