pub struct Glm5NextDsaWorkspace { /* private fields */ }Expand description
2026-09-25: Scratch reused across decode_k calls; each layer owns one.
Implementations§
Source§impl Glm5NextDsaWorkspace
impl Glm5NextDsaWorkspace
Sourcepub fn new(
gpu: &dyn GpuBackend,
cfg: &Glm5NextDsaConfig,
max_rows: usize,
) -> Result<Self>
pub fn new( gpu: &dyn GpuBackend, cfg: &Glm5NextDsaConfig, max_rows: usize, ) -> Result<Self>
2026-09-25: max_rows (at least 1) is the largest k this workspace serves. Sized by
it: the projection outputs, attn_out, sl, the batched-selector buffers, and the
selection scratch, which is planned at super::super::state::max_dsa_context tokens and
max_rows query rows. The staging rows and bt do not depend on it.
Auto Trait Implementations§
impl Freeze for Glm5NextDsaWorkspace
impl RefUnwindSafe for Glm5NextDsaWorkspace
impl Send for Glm5NextDsaWorkspace
impl Sync for Glm5NextDsaWorkspace
impl Unpin for Glm5NextDsaWorkspace
impl UnwindSafe for Glm5NextDsaWorkspace
Blanket Implementations§
Source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere
T: ?Sized,
Source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Mutably borrows from an owned value. Read more