pub struct CudaPlan { /* private fields */ }Expand description
A graph lowered for one bucket on one GPU.
Implementations§
Source§impl CudaPlan
impl CudaPlan
Sourcepub fn arena_bytes(&self) -> usize
pub fn arena_bytes(&self) -> usize
Arena size in bytes.
Sourcepub fn profile(&mut self)
pub fn profile(&mut self)
Times every step from now on. Each step then waits for the GPU and the graph is not used, so this is for finding where the time goes, not for measuring the total.
Sourcepub fn gemm_timings(&self) -> Vec<((usize, usize, usize), u64, u64)>
pub fn gemm_timings(&self) -> Vec<((usize, usize, usize), u64, u64)>
Time per GEMM shape since CudaPlan::profile: rows, inner size, columns, calls over all
timed runs and nanoseconds, largest first.
Trait Implementations§
Auto Trait Implementations§
impl !Sync for CudaPlan
impl Freeze for CudaPlan
impl RefUnwindSafe for CudaPlan
impl Send for CudaPlan
impl Unpin for CudaPlan
impl UnsafeUnpin for CudaPlan
impl UnwindSafe for CudaPlan
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