Skip to content

GitLab

Commit 860433ea authored by root's avatar root
Browse files

Remove unused mgroup

parent ebb5522c
Expand all Show whitespace changes
Inline Side-by-side
Showing with 85 additions and 149 deletions
example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16_comp_bf8_fp8.cpp
View file @ 860433ea
......@@ -69,7 +69,6 @@ using DeviceConvBwdWeightInstance =
2, // CBlockTransferScalarPerVector_NWaveNPerXdl
ck::BlockGemmPipelineScheduler::Intrawave, // BlkGemmPipeSched
ck::BlockGemmPipelineVersion::v1, // BlkGemmPipelineVer
1, // NumGroupsToMerge
ComputeTypeA, // ComputeTypeA
ComputeTypeB>; // ComputeTypeB
......
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_xdl_cshuffle.hpp
View file @ 860433ea
......@@ -37,7 +37,6 @@ template <typename GridwiseGemm,
typename BGridDesc_BK0_N_K1,
typename CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
typename ComputePtrOffsetOfBatch,
index_t NumGroupsToMerge,
bool HasMainKBlockLoop,
InMemoryDataOperationEnum CGlobalMemoryDataOperation,
index_t MinimumOccupancy = 1,
......@@ -57,7 +56,7 @@ __global__ void
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || \
defined(__gfx94__))
const index_t g_idx = __builtin_amdgcn_readfirstlane(blockIdx.z * NumGroupsToMerge);
const index_t g_idx = __builtin_amdgcn_readfirstlane(blockIdx.z);
const index_t k_idx = __builtin_amdgcn_readfirstlane(blockIdx.y * num_k_per_block);
const long_index_t a_batch_offset =
......@@ -92,7 +91,6 @@ template <typename GridwiseGemm,
typename BGridDesc_BK0_N_K1,
typename CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
typename ComputePtrOffsetOfBatch,
index_t NumGroupsToMerge,
bool HasMainKBlockLoop,
InMemoryDataOperationEnum CGlobalMemoryDataOperation,
index_t MinimumOccupancy = 1,
......@@ -113,7 +111,7 @@ __global__ void
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || \
defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__))
// offset base pointer for each work-group
const index_t g_idx = __builtin_amdgcn_readfirstlane(blockIdx.z * NumGroupsToMerge);
const index_t g_idx = __builtin_amdgcn_readfirstlane(blockIdx.z);
const index_t k_idx = __builtin_amdgcn_readfirstlane(blockIdx.y * num_k_per_block);
const long_index_t a_batch_offset =
......@@ -190,7 +188,6 @@ template <ck::index_t NDimSpatial,
index_t CBlockTransferScalarPerVector_NWaveNPerXdl,
BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave,
BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1,
index_t NumGroupsToMerge = 1,
typename ComputeTypeA = InDataType,
typename ComputeTypeB = ComputeTypeA>
struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
......@@ -240,7 +237,7 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
NPerBlock,
K1Number,
K0PerBlock / K1Number,
NumGroupsToMerge,
1 /*NumGroupsToMerge*/,
ConvBackwardWeightSpecialization>{};
template <ck::index_t NDim, typename ck::enable_if<NDim == 1, bool>::type = false>
......@@ -544,7 +541,7 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
index_t gdx, gdy, gdz;
std::tie(gdx, gdy, gdz) = GridwiseGemm::CalculateGridSize(
gemm_arg.M, gemm_arg.N, gemm_arg.KBatch, arg.Conv_G_ / NumGroupsToMerge);
gemm_arg.M, gemm_arg.N, gemm_arg.KBatch, arg.Conv_G_);
float ave_time = 0;
......@@ -621,7 +618,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy>;
......@@ -636,7 +632,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy>;
......@@ -657,7 +652,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -674,7 +668,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -693,7 +686,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -714,7 +706,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -735,7 +726,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -756,7 +746,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -776,7 +765,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -797,7 +785,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -817,7 +804,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -834,7 +820,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -853,7 +838,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -874,7 +858,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -895,7 +878,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -916,7 +898,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -936,7 +917,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -957,7 +937,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -982,7 +961,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -998,7 +976,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -1017,7 +994,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -1033,7 +1009,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -1055,7 +1030,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -1071,7 +1045,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy,
......@@ -1090,7 +1063,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -1106,7 +1078,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
true,
InMemoryDataOperationEnum::Set,
minimum_occupancy,
......@@ -1130,7 +1101,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
false,
InMemoryDataOperationEnum::AtomicAdd,
minimum_occupancy>;
......@@ -1145,7 +1115,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
NumGroupsToMerge,
false,
InMemoryDataOperationEnum::Set,
minimum_occupancy>;
......@@ -1235,23 +1204,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
}
}
if constexpr(NumGroupsToMerge > 1)
{
// support only if whole M and N can be proccessed on one block
if(!(GemmM <= MPerBlock && GemmN <= NPerBlock))
{
return false;
}
if(!(arg.Conv_C_ == 1 && arg.Conv_K_ == 1))
{
return false;
}
if(arg.Conv_G_ % NumGroupsToMerge != 0)
{
return false;
}
}
if(!(arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0 &&
arg.Conv_K_ % ABlockTransferSrcScalarPerVector == 0))
{
......@@ -1400,21 +1352,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
return str.str();
}
void SetWorkSpacePointer(BaseArgument* p_arg,
void* p_workspace,
const StreamConfig& = StreamConfig{}) const override
{
auto p_arg_ = dynamic_cast<Argument*>(p_arg);
if(p_arg_)
{
p_arg_->p_workspace_ = p_workspace;
}
else
throw std::runtime_error(
"The argument pointer is not an object of "
"DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle::Argument structure!");
}
};
} // namespace device
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment