Dev cuda#86

Draft

suzannejin wants to merge 104 commits intodevfrom

dev-cuda

Commits on Jan 29, 2026

Added nvtx for baseline
zalbanob
committed
minor change
zalbanob
committed
more nvtx
zalbanob
committed
Initial impl
zalbanob
committed
HOpefully fixed?

zalbanob
authored and
zalbanob
committed
Atleast generates bindings
zalbanob
committed
Fixed binding issue due to namespace generation
zalbanob
committed
Major fix to the api strucutre
zalbanob
committed
Fix
zalbanob
committed
CPU backend functional
zalbanob
committed
GPU backend working
zalbanob
committed
Cmake fix tests

zalbanob
authored and
zalbanob
committed
Add nvbench as a submodule under bench/nvbench

zalbanob
authored and
zalbanob
committed
Added catch as submodule

zalbanob
authored and
zalbanob
committed
Tests link

zalbanob
authored and
zalbanob
committed
Unfinalized tests

zalbanob
authored and
zalbanob
committed
Wish I had known Rcpp cannot be used without R runtime (https://stackoverflow.com/questions/47074438/compiling-rcpp-code-gives-segmentation-fault)
zalbanob
committed
Added testall cpp interface to CMAKE

zalbanob
authored and
zalbanob
committed
Made Testthat work through the cpp interface

zalbanob
authored and
zalbanob
committed
Fix
zalbanob
committed
Hindsight is 20/20

zalbanob
authored and
zalbanob
committed
Fixed bindings
zalbanob
committed
Initial test integration works
zalbanob
committed
Fixed a couple problems with the scripts
zalbanob
committed
Fixed the problem with alignemnet
zalbanob
committed
Added Ctz,comparision kernels, fixed utils copy

zalbanob
authored and
zalbanob
committed
Added thrust to cmake

zalbanob
authored and
zalbanob
committed
Minor fixes to cmake, ctz not working
zalbanob
committed
Fixed ctz
zalbanob
committed
Fixed the bindings
zalbanob
committed
inlining dispatchers
zalbanob
committed
un-inlining dispatchers
zalbanob
committed
implemented graflex::getFDR cuda
zalbanob
committed
added tests for GPU getFDR and count
zalbanob
committed
inital implementation of odd ratio computation
zalbanob
committed
Fixed deadlock issue
zalbanob
committed
merged init and compute into one kernel
zalbanob
committed
Fixed race
zalbanob
committed
added more kernels to graflex
zalbanob
committed
Added non-optimal graflex cuda dispatch impl
zalbanob
committed
Fixed minor allocation issues
zalbanob
committed
minor change
zalbanob
committed
Minor fixes to prevent potential overflow
zalbanob
committed
minor Style fix
zalbanob
committed
Fixed a bug in lrm alpha unweighted
zalbanob
committed
intial impl of cutlass - syncing for ide
zalbanob
committed
Cutlass omega

zalbanob
authored and
zalbanob
committed
Fix
zalbanob
committed
Save
zalbanob
committed
Working Omega cutlass kernel but without overflow handleing
zalbanob
committed
local save
zalbanob
committed
Update omega kernel
zalbanob
committed
Cleanup
zalbanob
committed
Cleanup
zalbanob
committed
Fix ctz
zalbanob
committed
Fix lrv
zalbanob
committed
Inital impl of backend
zalbanob
committed
further backend kernels
zalbanob
committed
deprecated cutlass
zalbanob
committed
More backend kernels and fix copyToNumericVector
zalbanob
committed
added alr
zalbanob
committed
more backend impl and tests
zalbanob
committed
Minor fixes
zalbanob
committed
Fixes backend cpu code, implements more backend kernels (as well as fixing them), implements col major varients of kernels which proved to be more performant on microbenchmark
zalbanob
committed
impl linRcpp
zalbanob
committed
Kernel Fixes
zalbanob
committed
cleanup
zalbanob
committed
implemented phiRcpp but we deadlock (as expected on large matrices)
zalbanob
committed
added symRcpp kernel
zalbanob
committed
Added nvtx for baseline
zalbanob
committed
added launch bounds to better inform the compiler
zalbanob
committed
intial movement into traits
zalbanob
committed
more cleanup
zalbanob
committed
made nvtx optional by setting propr.enable_nvtx to true/false
zalbanob
committed
Minor refactor
zalbanob
committed
add singularity def to here
zalbanob
committed
Added CI finally
zalbanob
committed
Added CI finally and Fixed lg2 to lg_e
zalbanob
committed
More clear test message
zalbanob
committed
Fix numerical issue
zalbanob
committed
Sync
zalbanob
committed
Added cub load store
zalbanob
committed
cosmetics
zalbanob
committed
cosmetics 2
zalbanob
committed
Added CC detection
zalbanob
committed
CPU Harmonic
zalbanob
committed
Updated R files to 60bd9da
zalbanob
committed
More optimizations
zalbanob
committed
Added Kernel Traits
zalbanob
committed
W-LRM: arithmetic mean to harmonic mean and fixes
zalbanob
committed
W-LRV moved to harmonic mean
zalbanob
committed
Fixed CPU backend
zalbanob
committed
Added fix from dev/3ecdb11
zalbanob
committed
Added lrv and lrm tests based on real input
zalbanob
committed
split phase phi
zalbanob
committed
Added nvtx markers
zalbanob
committed
Fixes
zalbanob
committed
Fixes
zalbanob
committed
Added support for integration with nvbenchr
zalbanob
committed
Added benchmarking and profiling code
zalbanob
committed
removed singularity
zalbanob
committed
removed catch2
zalbanob
committed

Commits

Commits on Jan 29, 2026

Commits on Feb 5, 2026