[cub] Add cooperative cached high-bin histograms - #10568

Open

robobryce wants to merge 24 commits into

NVIDIA:mainfrom

robobryce:pr/histocache/gmem-cooperative-baseline

Commits on Sep 13, 2026

[cub] Add cooperative global-memory histogram baseline
brycelelbach
committed
d854c06
[cub] Add shared-memory cache for high-bin histograms
brycelelbach
committed
fc30ef0
[cub] Address cooperative histogram review
brycelelbach
committed
b9f346c
[cub] Harden cooperative histogram dispatch
brycelelbach
committed
aeab26b
[cub] Fix cooperative launch compatibility
brycelelbach
committed
0f05451
[cub] Keep histogram stream operators inline
brycelelbach
committed
6abdfbd
[cub] Port production high-bin histogram optimizations
brycelelbach
committed
d29d80f
[cub] Handle full-width histogram ranges
brycelelbach
committed
0f181eb
[cub] Fix full-width histogram range validation
brycelelbach
committed
2fad4a7
[cub] Isolate high-bin histogram optimizations
brycelelbach
committed
1438850
[cub] Isolate high-bin histogram dispatch
brycelelbach
committed
f228b10
[cub] Match raw high-bin cache occupancy
brycelelbach
committed
f9b7c35
[cub] Use explicit high-bin launch tuning
brycelelbach
committed
dae1495
[cub] Match raw high-bin launch and hot paths
brycelelbach
committed
02c85bd
[cub] Match raw histogram fallback and hot path
brycelelbach
committed
db8eeb3
[cub] Match raw RANGE cache capacity
brycelelbach
committed
933a731
[cub] Match raw cooperative grid sizing
brycelelbach
committed
e534e4d
[cub] Tune high-bin work-grid size
brycelelbach
committed
2b1d16b
[cub] Fix high-bin policy validation
brycelelbach
committed
fc6cf61
[cub] Match raw multi-channel RANGE occupancy
brycelelbach
committed
6800bb3
[cub] Preserve multi-channel EVEN SMEM range
brycelelbach
committed
6f4f11c
[cub] Fix cooperative histogram CI portability
brycelelbach
committed
99394a7
[cub] Fix cooperative histogram MSVC builds
brycelelbach
committed
c846360
[cub] Refactor cooperative histogram policies
brycelelbach
committed
1e359ac

Navigation Menu

[cub] Add cooperative cached high-bin histograms - #10568

[cub] Add cooperative cached high-bin histograms#10568
robobryce wants to merge 24 commits into
NVIDIA:mainfrom
robobryce:pr/histocache/gmem-cooperative-baseline

Commits on Sep 13, 2026

[cub] Add cooperative global-memory histogram baseline

[cub] Add shared-memory cache for high-bin histograms

[cub] Address cooperative histogram review

[cub] Harden cooperative histogram dispatch

[cub] Fix cooperative launch compatibility

[cub] Keep histogram stream operators inline

[cub] Port production high-bin histogram optimizations

[cub] Handle full-width histogram ranges

[cub] Fix full-width histogram range validation

[cub] Isolate high-bin histogram optimizations

[cub] Isolate high-bin histogram dispatch

[cub] Match raw high-bin cache occupancy

[cub] Use explicit high-bin launch tuning

[cub] Match raw high-bin launch and hot paths

[cub] Match raw histogram fallback and hot path

[cub] Match raw RANGE cache capacity

[cub] Match raw cooperative grid sizing

[cub] Tune high-bin work-grid size

[cub] Fix high-bin policy validation

[cub] Match raw multi-channel RANGE occupancy

[cub] Preserve multi-channel EVEN SMEM range

[cub] Fix cooperative histogram CI portability

[cub] Fix cooperative histogram MSVC builds

[cub] Refactor cooperative histogram policies