Skip to content

Remove no longer needed const_cast #3325

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

Open
zcbenz wants to merge 1 commit into
base: main
Choose a base branch
from
+12 −19
Open

Remove no longer needed const_cast #3325

Show file tree
Hide file tree
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
14 changes: 3 additions & 11 deletions mlx/backend/cuda/reduce/all_reduce.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -98,9 +98,7 @@ void all_reduce(
size_t block_step;
size_t insize = in.size();
Dtype dt = in.dtype();

// Cub doesn't like const pointers for load (sigh).
void* indata = const_cast<void*>(gpu_ptr<void>(in));
void* indata = gpu_ptr<void>(in);

// Large array so allocate an intermediate and accumulate there
std::tie(blocks, threads, block_step) = get_args(insize, N_READS);
Expand All @@ -120,7 +118,7 @@ void all_reduce(
kernel,
blocks,
threads,
static_cast<T*>(indata),
indata,
gpu_ptr<U>(intermediate),
block_step,
insize);
Expand All @@ -143,13 +141,7 @@ void all_reduce(
using U = typename cu::ReduceResult<OP, T>::type;
auto kernel = cu::all_reduce<T, U, OP, N_READS>;
encoder.add_kernel_node(
kernel,
blocks,
threads,
static_cast<T*>(indata),
gpu_ptr<U>(out),
block_step,
insize);
kernel, blocks, threads, indata, gpu_ptr<U>(out), block_step, insize);
});
});
}
Expand Down
8 changes: 2 additions & 6 deletions mlx/backend/cuda/reduce/col_reduce.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -282,8 +282,6 @@ void col_reduce_looped(
using OP = MLX_GET_TYPE(reduce_type_tag);
using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
using U = typename cu::ReduceResult<OP, T>::type;
// Cub doesn't like const pointers for vectorized loads. (sigh)
T* indata = const_cast<T*>(gpu_ptr<T>(in));

constexpr int N_READS = 4;
constexpr int BM = 32;
Expand All @@ -296,7 +294,7 @@ void col_reduce_looped(
kernel,
grid,
blocks,
indata,
gpu_ptr<T>(in),
gpu_ptr<U>(out),
static_cast<cu::ColReduceArgs>(args),
out.size() / args.reduction_stride);
Expand Down Expand Up @@ -389,8 +387,6 @@ void col_reduce_two_pass(
using OP = MLX_GET_TYPE(reduce_type_tag);
using T = cuda_type_t<MLX_GET_TYPE(type_tag)>;
using U = typename cu::ReduceResult<OP, T>::type;
// Cub doesn't like const pointers for vectorized loads. (sigh)
T* indata = const_cast<T*>(gpu_ptr<T>(in));

constexpr int N_READS = 4;
constexpr int BM = 32;
Expand All @@ -403,7 +399,7 @@ void col_reduce_two_pass(
kernel,
grid,
blocks,
indata,
gpu_ptr<T>(in),
gpu_ptr<U>(intermediate),
static_cast<cu::ColReduceArgs>(args),
out.size() / args.reduction_stride);
Expand Down
9 changes: 7 additions & 2 deletions mlx/backend/cuda/reduce/row_reduce.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -268,10 +268,15 @@ void row_reduce_simple(
kernel = cu::row_reduce_simple<T, U, OP, N_READS, 2>;
}

T* indata = const_cast<T*>(gpu_ptr<T>(in));
int size = plan.shape.back();
encoder.add_kernel_node(
kernel, grid, block, indata, gpu_ptr<U>(out), out.size(), size);
kernel,
grid,
block,
gpu_ptr<T>(in),
gpu_ptr<U>(out),
out.size(),
size);
});
});
}
Expand Down
Loading