huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 18:48:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

Optimize for the contiguous case.

Browse Source
This commit is contained in:
laurent
2023-06-23 11:23:49 +01:00
parent 1a90f9d3a6
commit 4ffdeb4e23
1 changed files with 19 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
src/cuda_backend.rs
View File

@ -396,14 +396,22 @@ impl CudaStorage {
let ds = dev.htod_copy([dims, src_stride].concat())?;
match (&self.slice, &mut dst.slice) {
(CudaStorageSlice::F32(src), CudaStorageSlice::F32(dst)) => {
let func = dev.get_or_load_func("ucopy_f32", kernels::UNARY)?;
let mut dst = dst.slice_mut(dst_offset..);
if src_shape.is_contiguous(src_stride) {
dev.dtod_copy(src, &mut dst)?
} else {
let func = dev.get_or_load_func("ucopy_f32", kernels::UNARY)?;
// SAFETY: Set later by running the kernel.
let params = (el_count, dims.len(), &ds, src, &mut dst);
// SAFETY: ffi.
unsafe { func.launch(cfg, params) }?
}
}
(CudaStorageSlice::F64(src), CudaStorageSlice::F64(dst)) => {
let mut dst = dst.slice_mut(dst_offset..);
if src_shape.is_contiguous(src_stride) {
dev.dtod_copy(src, &mut dst)?
} else {
let func = dev.get_or_load_func("ucopy_64", kernels::UNARY)?;
let mut dst = dst.slice_mut(dst_offset..);
// SAFETY: Set later by running the kernel.
@ -411,6 +419,7 @@ impl CudaStorage {
// SAFETY: ffi.
unsafe { func.launch(cfg, params) }?;
}
}
_ => {
return Err(CudaError::InternalError(
"dtype mismatch in copy_strided op",