Skip to content
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

Fixing batched half precision complex GEMM #298

Merged
merged 1 commit into from
Oct 17, 2022
Merged

Fixing batched half precision complex GEMM #298

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
20 changes: 15 additions & 5 deletions include/matx/transforms/matmul.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,6 +343,10 @@ class matxMatMulHandle_t {
cublasLtMatrixLayoutDestroy(Adesc);
cublasLtMatmulDescDestroy(operationDesc);
}

matxFree(a_hp);
matxFree(b_hp);
matxFree(c_hp);
}

/**
Expand Down Expand Up @@ -412,6 +416,9 @@ class matxMatMulHandle_t {
cublasLtMatrixLayout_t BtransformDesc = nullptr;
cublasLtMatrixLayout_t CtransformDesc = nullptr;
cublasLtMatmulHeuristicResult_t heuristicResult = {};
void *c_hp = nullptr; // Make these void since they only work on complex types
void *a_hp = nullptr;
void *b_hp = nullptr;
size_t workspaceSize = 1 << 25UL; // 16MB buffer suggested by cuBLAS team
void *workspace = nullptr;
detail::MatMulParams_t params_;
Expand Down Expand Up @@ -634,21 +641,24 @@ class matxMatMulHandle_t {

auto a_shape = a.Shape();
*(a_shape.begin() + a.Rank() - 2) = a.Size(a.Rank() - 2) * 2;
auto a_planar = make_tensor<typename T1::value_type>(a_shape, MATX_ASYNC_DEVICE_MEMORY, stream);
matxAlloc(&a_hp, a.Bytes(), MATX_ASYNC_DEVICE_MEMORY, stream);
auto a_planar = make_tensor<typename T2::value_type>(reinterpret_cast<typename T2::value_type*>(a_hp), a_shape, false);

auto b_shape = b.Shape();
*(b_shape.begin() + b.Rank() - 2) = b.Size(b.Rank() - 2) * 2;
auto b_planar = make_tensor<typename T2::value_type>(b_shape, MATX_ASYNC_DEVICE_MEMORY, stream);
matxAlloc(&b_hp, b.Bytes(), MATX_ASYNC_DEVICE_MEMORY, stream);
auto b_planar = make_tensor<typename T3::value_type>(reinterpret_cast<typename T3::value_type*>(b_hp), b_shape, false);

auto c_shape = c.Shape();
*(c_shape.begin() + c.Rank() - 2) = c.Size(c.Rank() - 2) * 2;
auto c_planar = make_tensor<typename T3::value_type>(c_shape, MATX_ASYNC_DEVICE_MEMORY, stream);
matxAlloc(&c_hp, c.Bytes(), MATX_ASYNC_DEVICE_MEMORY, stream);
auto c_planar = make_tensor<typename T1::value_type>(reinterpret_cast<typename T1::value_type*>(c_hp), c_shape, false);

// Convert A/B to planar layout
(a_planar = planar(a)).run(stream);
(b_planar = planar(b)).run(stream);

// update poitners to planar data.
// update pointers to planar data.
// must use Reset because types for planar are different
a_adj.Reset(reinterpret_cast<T1 *>(a_planar.Data()));
b_adj.Reset(reinterpret_cast<T2 *>(b_planar.Data()));
Expand Down Expand Up @@ -701,11 +711,11 @@ class matxMatMulHandle_t {
}
else {
for (size_t iter = 0; iter < total_iter; iter++) {

// Get pointers into A/B/C for this round
auto ap = std::apply([&a_adj](auto... param) { return a_adj.GetPointer(param...); }, idx);
auto bp = std::apply([&b_adj](auto... param) { return b_adj.GetPointer(param...); }, idx);
auto cp = std::apply([&c_adj](auto... param) { return c_adj.GetPointer(param...); }, idx);

auto res = cublasLtMatmul(
ltHandle, operationDesc, &salpha, (void *)ap,
Adesc, (void *)bp, Bdesc, &sbeta,
Expand Down