Skip to content

Fix load_q_impl tensor templates #20

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
ivandobskygithub merged 1 commit into from
Nov 25, 2025
Merged

Fix load_q_impl tensor templates #20

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
9 changes: 5 additions & 4 deletions hopper/mainloop_fwd_sm90_tma_gmma_ws.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,6 +119,7 @@ struct CollectiveMainloopFwdSm90 {
static constexpr int NumMmaThreadsQK = size(TiledMmaQK{});
static constexpr int NumMmaThreads = size(TiledMmaPV{});
static constexpr int NumProducerThreads = !Transpose_V && Use_TMA_KV && Use_TMA_Q ? cutlass::NumThreadsPerWarp : cutlass::NumThreadsPerWarpGroup;
static constexpr bool SingleProducerWarp = NumProducerThreads == cutlass::NumThreadsPerWarp;
static_assert(NumMmaThreadsQK % cutlass::NumThreadsPerWarpGroup == 0);
static_assert(NumMmaThreads % cutlass::NumThreadsPerWarpGroup == 0);
static constexpr int NumMmaWarpGroups = NumMmaThreads / cutlass::NumThreadsPerWarpGroup;
Expand Down Expand Up @@ -646,11 +647,11 @@ struct CollectiveMainloopFwdSm90 {
}
}

template <typename SharedStorage>
template <typename SharedStorage, typename TensorQ, typename TensorQv>
CUTLASS_DEVICE void load_q_impl(
std::true_type /*UseTmaQ*/, Params const& params, SharedStorage &shared_storage,
SeqlenInfo_t const& seqlen_info, cute::tuple<int32_t, int32_t, int32_t, int32_t> block_coord,
int thread_idx, bool is_varlen_q, int warp_idx_in_warpgroup, Tensor sQ, Tensor sQv) {
int thread_idx, bool is_varlen_q, int warp_idx_in_warpgroup, TensorQ sQ, TensorQv sQv) {

int const m_block = get<0>(block_coord);
int const bidh = get<1>(block_coord);
Expand Down Expand Up @@ -694,11 +695,11 @@ struct CollectiveMainloopFwdSm90 {
}
}

template <typename SharedStorage>
template <typename SharedStorage, typename TensorQ, typename TensorQv>
CUTLASS_DEVICE void load_q_impl(
std::false_type /*UseTmaQ*/, Params const& params, SharedStorage &shared_storage,
SeqlenInfo_t const& seqlen_info, cute::tuple<int32_t, int32_t, int32_t, int32_t> block_coord,
int thread_idx, bool is_varlen_q, int /*warp_idx_in_warpgroup*/, Tensor sQ, Tensor sQv) {
int thread_idx, bool is_varlen_q, int /*warp_idx_in_warpgroup*/, TensorQ sQ, TensorQv sQv) {

int const m_block = get<0>(block_coord);
int const bidh = get<1>(block_coord);
Expand Down