Vamana build optimization set and fp16 support - #2264

Open

bkarsin wants to merge 37 commits into

NVIDIA:mainfrom

bkarsin:vamana-build-opt

Commits on Jun 15, 2026

Initial multi-warp greedy search optimization
bkarsin
committed
a4d82e1
Add fp16 support for Vamana build/serialize
bkarsin
committed
05cc8cc
Fix recall issue with multi-warp optimization
bkarsin
committed
5bb5e6c
Speed up fp16 vamana build with promoted query vectors and optimized L2 comparators.
bkarsin
committed
3dca1b0
perf(vamana): P1_cache_candidate_vector — Cache the accepted candidate vector in shared memory in the RobustPrune occlusion loop
bkarsin
committed
02b5638
perf(vamana): O1_parallel_scan_buffer_pool — Replace single-thread prefix sum with cub scan and hoist per-batch reverse-edge allocations
bkarsin
committed
9ec9b99
perf(vamana): N3_multiwarp_robust_prune — Parallelize RobustPrune occlusion across multiple warps per query (raise occupancy)
bkarsin
committed
696d3d2
perf(vamana): N7_reuse_search_distances_in_prune — Reuse GreedySearch query->existing-edge distances in the RobustPrune merge (avoid recompute)
bkarsin
committed
24f2ed7

Commits on Jun 16, 2026

Commits on Jun 24, 2026

Clean up and remove dead code
bkarsin
committed
4a0c179
pre-commit fixes
bkarsin
committed
cb91c99
Fix bug with odd dimension data for some cases
bkarsin
committed
2dc5fa9
Merge branch 'NVIDIA:main' into vamana-build-opt
bkarsin
authored
a9b8a8d

Commits on Jul 15, 2026

Merge branch 'main' into vamana-build-opt
cjnolet
authored
5da19e0

Commits on Jul 17, 2026

Merge branch 'main' into vamana-build-opt
bkarsin
authored
c055c60
pre-commit fix
bkarsin
committed
df13b42

Commits on Jul 22, 2026

Merge branch 'main' into vamana-build-opt
cjnolet
authored
adf1827
Fix launch bounds for older arch builds
bkarsin
committed
e008a09
Add fp16 tests
bkarsin
committed
350ed89
Merge branch 'main' into vamana-build-opt
bkarsin
authored
74caba9
Add missing test file
bkarsin
committed
b8238f3

Commits on Jul 24, 2026

Change distribution of fp16 vamana unit tests
bkarsin
committed
70164ef

Commits on Jul 25, 2026

Commits on Jul 28, 2026

replace kernel with map_offset and add raft directives
bkarsin
committed
03a29e4
Merge branch 'main' into vamana-build-opt
bkarsin
authored
3cc19c4

Commits on Jul 29, 2026

Replace CUDA_ARCH with raft::resource::get_device_properties to choose
bkarsin
committed
38d7f1a

Commits on Jul 30, 2026

Fix bug with RAFT directives
bkarsin
committed
6cac7b1
Merge branch 'main' into vamana-build-opt
bkarsin
authored
1db9190

Commits on Aug 19, 2026

Merge branch 'main' into vamana-build-opt
bkarsin
authored
ecc61d1
Replace thrust vector with raft
bkarsin
committed
e1e0b7e

Commits on Aug 20, 2026

Fix build error for some architectures
bkarsin
committed
f0579f9

Commits on Aug 25, 2026

Address review comments
bkarsin
committed
c4d4d6c