Repository navigation
Conversation
fefb201 to
aa62381
Compare
Merging this PR will regress 3 benchmarks4 warnings, 1 note
Warning Please fix the performance issues or acknowledge them on CodSpeed. Performance Changes
Tip Investigate this regression by commenting Comparing Footnotes
|
aa62381 to
c1c91ba
Compare
c1c91ba to
0affd29
Compare
3367172 to
f783c34
Compare
661d07c to
4c93a25
Compare
|
@claude review |
|
Claude finished @danking's task in 2m 28s —— View job PR #10828 review: selective PCO take
The sparse/grouped design holds up. Requests are sorted, repeated rows are resolved once, each page is decoded once, and results are scattered into an initialized output. Bounds checks run before any allocation, and legacy metadata uses a single ranked validity mask instead of a per-index scan. The kernel registration matches how ALP, BitPacked and the others do it. I found one test-module compile error that will break CI, and one large block of code that can never run. 🔴 Blocking1. 🟠 Should address2.
3. The CodSpeed regression in
If scalar-at scans end up in this kernel, that fixed overhead per call is a plausible cause. A cheap early exit for 4. The module docs still show the old registration. 🟡 Minor / style
✅ Looks good
ChecksI did not run any builds or tests; this was a read-only review. Item 1 comes from reading the import list, not from a compile. CI "Linters and Tests" was still queued when I checked. |
4c93a25 to
ce5758f
Compare
d4b2869 to
73dbbe4
Compare
e348dea to
fbe9df9
Compare
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
…stry guards Signed-off-by: Daniel King <dan@spiraldb.com>
Signed-off-by: Daniel King <dan@spiraldb.com>
e2ba944 to
76e27b4
Compare
|
Addressed the remaining applicable take feedback in 76e27b4: scatter directly from each decoded primitive page using checked reads, eliminating the per-page temporary index array, second take dispatch, and selected-value array. The public example now calls initialize, basic and new test sessions exercise that registration, and vortex-file initializes PCO outside its retained array-registry guard. The older Run/Output/contiguous planner, BTreeSet page estimate, legacy selective path, and codec zero-fill changes are absent from the current diff, as requested by Daniel. Global semantic prefix validation and corrupt-index conformance tests remain intentionally excluded under his explicit scope/validation policy; this does not promise every malformed index returns Err rather than panicking. The checked page-value reads preserve the previous gather's local bounds failure. Local independent source review passed; no builds/tests or timing reruns were performed, and no CodSpeed regression was acknowledged. |
|
This PR has been marked as stale because it has been open for 14 days with no activity. Please comment or remove the stale label if you wish to keep it active, otherwise it will be closed in 7 days |
Use the logical-row page index to decompress only pages containing requested non-null values for primitive-index take. Group requests by page, reuse chunk decoder setup, and gather values in the requested order while preserving duplicates and logical nulls. Files without an index retain the existing fallback; compact range take and slice handling are introduced separately.