mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-06-29 19:28:33 +00:00
add preshuffle padding to library
This commit is contained in:
qin letao
@@ -265,6 +265,254 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p3_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p4_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p5_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p6_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
#endif
|
||||
|
||||
#if(defined(CK_ENABLE_BF16) || defined(CK_ENABLE_FP8))
|
||||
@@ -516,6 +764,253 @@ void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma1
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p3_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p4_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p5_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p6_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>&
|
||||
instances);
|
||||
#endif
|
||||
|
||||
template <typename ADataType,
|
||||
@@ -601,6 +1096,46 @@ struct DeviceOperationInstanceFactory<
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p6(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instances_padding(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_p1_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_p2_padding(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p1_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p3_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p4_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p5_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p6_padding(
|
||||
op_ptrs);
|
||||
}
|
||||
}
|
||||
#endif
|
||||
@@ -651,6 +1186,46 @@ struct DeviceOperationInstanceFactory<
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p6(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instances_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instances_padding(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instances_v2_padding(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_p1_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_p2_padding(
|
||||
op_ptrs);
|
||||
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p1_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p2_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p3_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p4_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p5_padding(
|
||||
op_ptrs);
|
||||
add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p6_padding(
|
||||
op_ptrs);
|
||||
}
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -39,6 +39,44 @@ list(APPEND GEMM_MULTIPLY_MULTIPLY_WEIGHT_PRESHUFFLE_INSTANCES
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p4.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p5.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p6.cpp
|
||||
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instance_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instance_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instance_v2_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instance_v2_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_p1_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_p2_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p1_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p2_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p3_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p4_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p5_padding.cpp
|
||||
f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p6_padding.cpp
|
||||
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instance_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instance_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instance_v2_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instance_v2_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_p1_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_p2_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p1_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p2_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p3_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p4_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p5_padding.cpp
|
||||
f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p6_padding.cpp
|
||||
)
|
||||
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
@@ -79,4 +117,42 @@ set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p5.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p6.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_p1_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instance_p2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p1_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p3_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p4_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p5_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_bf16/device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instance_p6_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instance_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instance_v2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_p1_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instance_p2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p1_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p2_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p3_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p4_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p5_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
set_source_files_properties(f8_f8_f16/device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instance_p6_padding.cpp PROPERTIES COMPILE_OPTIONS ";-mllvm;-greedy-reverse-local-assignment=1")
|
||||
|
||||
add_instance_library(device_gemm_multiply_multiply_wp_instance ${GEMM_MULTIPLY_MULTIPLY_WEIGHT_PRESHUFFLE_INSTANCES})
|
||||
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances_p1<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances_p2<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p3_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances_p3<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p4_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances_p4<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p5_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances_p5<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_default_instances_p6(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma16x16_mn_compute_instances_p6<
|
||||
GemmDefault>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_instances_p1<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_compute_instances_p2<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p1_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p2_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p3_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p4_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_bf16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_bf16_mk_mfma_mn_p5_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances_p1<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances_p2<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p3_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances_p3<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p4_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances_p4<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p5_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances_p5<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_default_instances_p6_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma16x16_mn_compute_instances_p6<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_p1_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_instances_p1<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,33 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_default_instances_p2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_compute_instances_p2<
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p1_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p2_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p3_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p4_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instances_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_instances<
|
||||
v1,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -0,0 +1,34 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "device_gemm_multiply_multiply_wp_xdl_f8_f8_f16_mk_mfma_mn.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
void add_device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_default_instances_v2_padding(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitKBPreShuffle<Row,
|
||||
Col,
|
||||
Tuple<Row, Col>,
|
||||
Row,
|
||||
F8,
|
||||
F8,
|
||||
Tuple<F32, F32>,
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
MultiplyMultiply>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_multiply_multiply_weight_preshuffle_xdl_f8_f8_f16_mk_mfma_mn_p5_instances<
|
||||
v2,
|
||||
GemmNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
Reference in New Issue
Block a user