Skip to content

Support alpha beta scaling for GEMM #78

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
asroy merged 8 commits into from
Feb 11, 2022
Merged

Support alpha beta scaling for GEMM #78

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
fdbfa96
[What] Add 2d version of bias, prepare to implement alpha / beta scaling
Feb 7, 2022
6bed25a
Add alpha / beta functor
Feb 7, 2022
0372d81
Refine parameter of example
Feb 8, 2022
9bafd93
[What] Use real type instead of template
Feb 8, 2022
8a896c4
Rename parameter for general operator
Feb 10, 2022
10560a7
Remove redundant comment
Feb 10, 2022
0c043d7
Fix compile error
Feb 11, 2022
02d8ce0
Merge remote-tracking branch 'origin/develop' into gemm_alpha_beta
Feb 11, 2022
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
39 changes: 39 additions & 0 deletions composable_kernel/include/tensor_operation/element_wise_operation.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,45 @@ struct PassThrough
__host__ __device__ void operator()(half_t& y, const half_t& x) const { y = x; }
};

struct Add
{
__host__ __device__ constexpr void
operator()(float& dst, const float& src_y, const float& bias) const
Comment thread
asroy marked this conversation as resolved.
Outdated
{
// FIXME - Use float (acc type) bias in the future.
Comment thread
rocking5566 marked this conversation as resolved.
Outdated
dst = src_y + bias;
}

__host__ __device__ constexpr void
operator()(half_t& dst, const half_t& src_y, const half_t& bias) const
Comment thread
asroy marked this conversation as resolved.
Outdated
{
// FIXME - Use float (acc type) bias in the future.
dst = src_y + bias;
}
};

struct AlphaBetaAdd
{
AlphaBetaAdd(float alpha, float beta) : alpha_(alpha), beta_(beta) {}

__host__ __device__ constexpr void
operator()(float& dst, const float& src_y, const float& bias) const
{
dst = alpha_ * src_y + beta_ * bias;
}

__host__ __device__ constexpr void
operator()(half_t& dst, const half_t& src_y, const half_t& bias) const
{
// FIXME - Let src_y be acc type
dst = static_cast<half_t>(alpha_ * static_cast<float>(src_y) +
beta_ * static_cast<float>(bias));
}

float alpha_;
float beta_;
};

struct AddRelu
{
__host__ __device__ constexpr void operator()(float& y, const float& x0, const float& x1) const
Expand Down
29 changes: 29 additions & 0 deletions device_operation/include/device_gemm.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,35 @@ namespace ck {
namespace tensor_operation {
namespace device {

template <typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation>
struct DeviceGemmBias : public BaseOperator
{
virtual std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
const void* p_bias,
void* p_c,
ck::index_t M,
ck::index_t N,
ck::index_t K,
ck::index_t StrideA,
ck::index_t StrideB,
ck::index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op) = 0;

virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
};

template <typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation>
using DeviceGemmBiasPtr = std::unique_ptr<
DeviceGemmBias<AElementwiseOperation, BElementwiseOperation, CElementwiseOperation>>;

template <typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation>
Expand Down
Loading