Skip to content

fix: CUDA bitpacked sliced output allocation #8622

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
0ax1 merged 1 commit into from
Jun 29, 2026
Merged

fix: CUDA bitpacked sliced output allocation #8622

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 13 additions & 5 deletions vortex-cuda/src/kernel/encodings/bitpacked.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -177,14 +177,16 @@ where
// Get CUDA view of input
let input_view = device_input.cuda_view::<A::Physical>()?;

let output_len = offset + len;

// Allocate output buffer
let output_slice = ctx.device_alloc::<A>(len.next_multiple_of(1024))?;
let output_slice = ctx.device_alloc::<A>(output_len.next_multiple_of(1024))?;
let output_buf = CudaDeviceBuffer::new(output_slice);
let output_view = output_buf.as_view::<A>();

let output_width = size_of::<A>() * 8;
let cuda_function = bitpacked_cuda_kernel(bit_width, output_width, ctx)?;
let config = bitpacked_cuda_launch_config(output_width, len)?;
let config = bitpacked_cuda_launch_config(output_width, output_len)?;

// We hold this here to keep the device buffers alive.
let device_patches = if let Some(patches) = patches {
Expand All @@ -199,7 +201,7 @@ where

let patches_arg = build_gpu_patches(device_patches.as_ref())?;

ctx.launch_kernel_config(&cuda_function, config, len, |args| {
ctx.launch_kernel_config(&cuda_function, config, output_len, |args| {
args.arg(&input_view)
.arg(&output_view)
.arg(&reference)
Expand Down Expand Up @@ -552,8 +554,14 @@ mod tests {
Ok(())
}

#[rstest]
#[case(67, 3969)]
#[case(1, 1025)]
#[crate::test]
fn test_cuda_bitunpack_sliced() -> VortexResult<()> {
fn test_cuda_bitunpack_sliced(
#[case] slice_start: usize,
#[case] slice_end: usize,
) -> VortexResult<()> {
let mut ctx = vortex_array::array_session().create_execution_ctx();
let bit_width = 32;
let mut cuda_ctx = CudaSession::create_execution_ctx(&crate::cuda_session())
Expand All @@ -570,7 +578,7 @@ mod tests {

let bitpacked_array = BitPacked::encode(&primitive_array.into_array(), bit_width, &mut ctx)
.vortex_expect("operation should succeed in test");
let sliced_array = bitpacked_array.into_array().slice(67..3969)?;
let sliced_array = bitpacked_array.into_array().slice(slice_start..slice_end)?;
let gpu_result = block_on(async {
BitPackedExecutor
.execute(sliced_array.clone(), &mut cuda_ctx)
Expand Down
Loading