Skip to content

GitLab

Unverified Commit 7fa892e6 authored by Qianfeng's avatar Qianfeng Committed by GitHub
Browse files

Batchnorm-forward implemented using welford method to calculate variance (#403) 



* Update to the batchnorm-forward API and base class

* Fix leeked header including in gridwise_set_buffer_value.hpp

* Add kernels and device file for batchnorm-forward welford supporting both blockwise and multi-block reduction

* Update to the batchnorm-forward example to use the new batchnorm-forward device interface

* Change the batchnorm-forward reference to use sequential welford method

* Change to assign the workspace into four buffers in the host layer

* Use GetReduceCountPerThread functor to replace the initial count for Blockwise and Multiblock welford

* Tiny correction and remove un-used file under example/34_batchnorm

* Renaming in the kernel arguments

* Explicitly use ck::math::sqrt in batchnorm-forward kernels

* Add some comments to some kernels

* Tiny fix

* Generalize the data types in reference_batchnorm_forward_nhwc_c

* Use ck::ignore to mark un-used parameters

* Move GetReduceCountPerThread functor codes from kernel to device

* Remove some un-used codes in device_batchnorm_forward_impl.hpp

* Tiny fix in batchnorm_forward example

* Move GetReduceCountPerThread() to welford_helper.hpp

* Use seperate data type for Scale and Bias

* Renaming in device Op

* Tiny fix in forward example

* Updata to batchnorm-infer (type spliting, renaming)

* Add time and bandwidth measurement to the batchnorm-forward example

* Add support of elementwise operation for batchnorm forward output

* Reduce object copying by passing object as reference type

* Tiny change for performance

* Updates for performance again

* Some Renamings

* Add GetActualVariance template parameter for ThreadwiseWelfordMerge

* Tiny update in reference batchnorm forward nhwc/c

* Move batchnorm multiblock kernel files to grid/batchnorm_multiblock sub-directory

* Fuse mean and bias in the normalization calculation
Co-authored-by: default avatarroot <root@dc-smc-18.amd.com>
Co-authored-by: default avatarrocking5566 <ChunYu.Lai@amd.com>
parent 337642a4
Show whitespace changes
Inline Side-by-side
Showing with 2631 additions and 665 deletions
example/34_batchnorm/batchnorm_common.hpp
View file @ 7fa892e6
......@@ -10,131 +10,17 @@
#include "ck/utility/data_type.hpp"
// binary operation used to calculate invVariance from mean and meansquare
struct InvVariance
{
InvVariance(double epsilon) : epsilon_(epsilon){};
template <typename T>
__host__ __device__ constexpr void operator()(T& y, const T& mean, const T& meansquare) const
{
static_assert(std::is_same<T, float>::value || std::is_same<T, double>::value,
"Data type is not supported by this operation!");
using ck::type_convert;
using ck::math::sqrt;
T tmp_epsilon = type_convert<T>(epsilon_);
y = meansquare - mean * mean;
y = 1.0f / sqrt(tmp_epsilon + y);
};
double epsilon_;
};
// (4-in, 2-out) element-wise operation used to update the moving average of mean and variance
struct MovingAverage
{
MovingAverage(double factor) : factor_(factor){};
template <typename T>
__host__ __device__ constexpr void operator()(T& y0,
T& y1,
const T& mean,
const T& runningMean,
const T& meansquare,
const T& runningVariance) const
{
static_assert(std::is_same<T, float>::value || std::is_same<T, double>::value,
"Data type is not supported by this operation!");
using ck::type_convert;
T tmp_factor = type_convert<T>(factor_);
T variance = meansquare - mean * mean;
y0 = runningMean * (type_convert<T>(1.0f) - tmp_factor) + mean * tmp_factor;
y1 = runningVariance * (type_convert<T>(1.0f) - tmp_factor) + variance * tmp_factor;
};
double factor_;
};
struct MovingAverageAndInvVariance
{
MovingAverageAndInvVariance(double epsilon, double factor)
: epsilon_(epsilon), factor_(factor){};
template <typename T>
__host__ __device__ constexpr void operator()(T& y0, // resultRunningMean
T& y1, // resultRunningVariance
T& y2, // saveInvVariance
const T& mean,
const T& runningMean,
const T& meansquare,
const T& runningVariance) const
{
static_assert(std::is_same<T, float>::value || std::is_same<T, double>::value,
"Data type is not supported by this operation!");
using ck::type_convert;
using ck::math::sqrt;
T tmp_epsilon = type_convert<T>(epsilon_);
T tmp_factor = type_convert<T>(factor_);
T variance = meansquare - mean * mean;
y0 = runningMean * (type_convert<T>(1.0f) - tmp_factor) + mean * tmp_factor;
y1 = runningVariance * (type_convert<T>(1.0f) - tmp_factor) + variance * tmp_factor;
y2 = 1.0f / sqrt(tmp_epsilon + variance);
};
double epsilon_;
double factor_;
};
struct NormalizeInInfer
{
NormalizeInInfer(double epsilon = 1e-4) : epsilon_(epsilon) {}
template <typename T1, typename T2>
template <typename T1, typename T2, typename T3, typename T4>
__host__ __device__ constexpr void operator()(T1& y,
const T1& x,
const T2& mean,
const T2& variance,
const T2& gamma,
const T2& beta) const
{
static_assert(std::is_same<T2, float>::value || std::is_same<T2, double>::value,
"Data type is not supported by this operation!");
using ck::type_convert;
using ck::math::sqrt;
T2 tmp_x, tmp_y;
tmp_x = type_convert<T2>(x);
tmp_y = ((tmp_x - mean) / sqrt(variance + type_convert<T2>(epsilon_))) * gamma + beta;
y = type_convert<T1>(tmp_y);
};
double epsilon_;
};
struct NormalizeInForward
{
NormalizeInForward(double epsilon = 1e-4) : epsilon_(epsilon) {}
template <typename T1, typename T2>
__host__ __device__ constexpr void operator()(T1& y,
const T1& x,
const T2& mean,
const T2& meansquare,
const T2& gamma,
const T2& beta) const
const T3& gamma,
const T4& beta) const
{
static_assert(std::is_same<T2, float>::value || std::is_same<T2, double>::value,
"Data type is not supported by this operation!");
......@@ -143,11 +29,12 @@ struct NormalizeInForward
using ck::math::sqrt;
T2 tmp_x, tmp_y;
T2 variance = meansquare - mean * mean;
tmp_x = type_convert<T2>(x);
tmp_y = ((tmp_x - mean) / sqrt(variance + type_convert<T2>(epsilon_))) * gamma + beta;
tmp_y = ((tmp_x - mean) / sqrt(variance + type_convert<T2>(epsilon_))) *
type_convert<T2>(gamma) +
type_convert<T2>(beta);
y = type_convert<T1>(tmp_y);
};
......
example/34_batchnorm/batchnorm_forward_impl.hpp deleted 100644 → 0
View file @ 337642a4
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <cassert>
#include <vector>
#include "ck/ck.hpp"
#include "ck/utility/reduction_operator.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_multiple_reduce_multiblock.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_elementwise.hpp"
#include "batchnorm_common.hpp"
template <typename InOutDataType,
typename AccDataType,
ck::index_t Rank,
ck::index_t NumBatchNormReduceDim,
bool fastest_dim_is_reduced = false>
int bnorm_fwd(bool time_kernel,
bool updateMovingAverage,
bool saveMeanAndInvVariance,
const std::array<int, NumBatchNormReduceDim> reduceDims,
const std::array<ck::index_t, Rank> xyLengths,
const std::array<ck::index_t, Rank> xStrides,
const std::array<ck::index_t, Rank> yStrides,
const std::array<ck::index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarLengths,
const std::array<ck::index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarStrides,
const void* p_x,
const void* p_scale,
const void* p_bias,
void* p_y,
double exponentialAverageFactor,
void* p_runningMean,
void* p_runningVariance,
double epsilon,
void* p_saveMean,
void* p_saveInvVariance,
void* p_tmp_mean,
void* p_tmp_meansquare)
{
static_assert(NumBatchNormReduceDim < Rank,
"Invalid number of reduced dimensions for batchnorm!");
constexpr ck::index_t NumScaleBiasMeanVarDim = Rank - NumBatchNormReduceDim;
using InElementwiseOperation_Mean = ck::tensor_operation::element_wise::PassThrough;
using AccElementwiseOperation_Mean = ck::tensor_operation::element_wise::UnaryDivide;
using InElementwiseOperation_Meansquare = ck::tensor_operation::element_wise::UnarySquare;
using AccElementwiseOperation_Meansquare = ck::tensor_operation::element_wise::UnaryDivide;
using DeviceMeanAndMeansquareInstance =
ck::tensor_operation::device::DeviceMultipleReduceMultiBlock<
2,
InOutDataType,
AccDataType,
ck::Tuple<AccDataType, AccDataType>,
Rank,
NumBatchNormReduceDim,
ck::reduce::Add,
ck::Tuple<InElementwiseOperation_Mean, InElementwiseOperation_Meansquare>,
ck::Tuple<AccElementwiseOperation_Mean, AccElementwiseOperation_Meansquare>,
ck::InMemoryDataOperationEnum::Set,
false, // PropagateNan
256,
16,
16,
1,
1,
fastest_dim_is_reduced ? 1 : 0,
1,
ck::Sequence<1, 1>>;
using DeviceNormalizeInstance = ck::tensor_operation::device::DeviceElementwise<
ck::Tuple<InOutDataType, AccDataType, AccDataType, AccDataType, AccDataType>, // x, mean,
// meansquare,
// scale, bias
ck::Tuple<InOutDataType>, // y
NormalizeInForward,
Rank,
2, // MPerthread
ck::Sequence<1, 1, 1, 1, 1>, // scalarPerVector: x, mean, meansquare, scale, bias
ck::Sequence<1>>; // scalarPerVector: y
using DeviceInvVarianceInstance = ck::tensor_operation::device::DeviceElementwise<
ck::Tuple<AccDataType, AccDataType>, // mean, meansquare
ck::Tuple<AccDataType>, // invVariance
InvVariance,
NumScaleBiasMeanVarDim,
2, // MPerthread
ck::Sequence<1, 1>, // scalarPerVector: mean, meansquare
ck::Sequence<1>>; // scalarPerVector: invVariance
using DeviceMovingAverageInstance = ck::tensor_operation::device::DeviceElementwise<
ck::Tuple<AccDataType, AccDataType, AccDataType, AccDataType>, // old moving mean, new mean,
// old moving variance, new
// meansquare
ck::Tuple<AccDataType, AccDataType>, // updated moving mean, updated moving variance
MovingAverage,
NumScaleBiasMeanVarDim,
4, // MPerthread
ck::Sequence<1, 1, 1, 1>, // scalarPerVector: old moving mean, new mean, old moving
// variance, new meansquare
ck::Sequence<1, 1>>; // scalarPerVector: updated moving mean, updated moving variance
using DeviceMovingAverageAndInvVarianceInstance =
ck::tensor_operation::device::DeviceElementwise<
ck::Tuple<AccDataType, AccDataType, AccDataType, AccDataType>, // old moving mean, new
// mean, old moving
// variance, new
// meansquare
ck::Tuple<AccDataType, AccDataType, AccDataType>, // updated moving mean, updated moving
// variancem, invVariance
MovingAverageAndInvVariance,
NumScaleBiasMeanVarDim,
4, // MPerthread
ck::Sequence<1, 1, 1, 1>, // scalarPerVector: old moving mean, new mean, old moving
// variance, new meansquare
ck::Sequence<1, 1, 1>>; // scalarPerVector: updated moving mean, updated moving variance
auto invariantDims = get_invariant_dims<Rank, NumBatchNormReduceDim>(reduceDims);
std::array<ck::index_t, Rank> aligned_scaleBiasMeanVarStrides{0};
int i = 0;
for(auto dim : invariantDims)
{
assert(xyLengths[dim] == bnScaleBiasMeanVarLengths[i]);
aligned_scaleBiasMeanVarStrides[dim] = bnScaleBiasMeanVarStrides[i];
i++;
};
int32_t reduceLength = 1;
for(auto dim : reduceDims)
reduceLength *= xyLengths[dim];
int32_t invariantLength = 1;
for(auto dim : invariantDims)
invariantLength *= xyLengths[dim];
size_t total_length = static_cast<size_t>(invariantLength) * reduceLength;
float avg_time = 0.0f;
std::size_t num_bytes = 0;
auto dev_mean_and_meansquare = DeviceMeanAndMeansquareInstance{};
void* p_mean = saveMeanAndInvVariance ? p_saveMean : p_tmp_mean;
const AccDataType alpha = ck::type_convert<AccDataType>(1.0f);
const AccDataType beta = ck::type_convert<AccDataType>(0.0f);
auto argument_ptr1 = dev_mean_and_meansquare.MakeArgumentPointer(
xyLengths,
xStrides,
bnScaleBiasMeanVarLengths,
{bnScaleBiasMeanVarStrides, bnScaleBiasMeanVarStrides},
reduceDims,
{&alpha, &alpha},
{&beta, &beta},
p_x,
{p_mean, p_tmp_meansquare},
ck::make_tuple(InElementwiseOperation_Mean{}, InElementwiseOperation_Meansquare{}),
ck::make_tuple(AccElementwiseOperation_Mean{reduceLength},
AccElementwiseOperation_Meansquare{reduceLength}));
auto dev_normalize = DeviceNormalizeInstance{};
auto argument_ptr2 =
dev_normalize.MakeArgumentPointer(xyLengths,
{xStrides,
aligned_scaleBiasMeanVarStrides,
aligned_scaleBiasMeanVarStrides,
aligned_scaleBiasMeanVarStrides,
aligned_scaleBiasMeanVarStrides},
{yStrides},
{p_x, p_mean, p_tmp_meansquare, p_scale, p_bias},
{p_y},
NormalizeInForward{epsilon});
if(!dev_mean_and_meansquare.IsSupportedArgument(argument_ptr1.get()) ||
!dev_normalize.IsSupportedArgument(argument_ptr2.get()))
{
std::cout << "The runtime parameters seems not supported by the Devic, exiting!"
<< std::endl;
return (-1);
};
auto invoker_ptr1 = dev_mean_and_meansquare.MakeInvokerPointer();
auto invoker_ptr2 = dev_normalize.MakeInvokerPointer();
avg_time += invoker_ptr1->Run(argument_ptr1.get(), StreamConfig{nullptr, time_kernel});
avg_time += invoker_ptr2->Run(argument_ptr2.get(), StreamConfig{nullptr, time_kernel});
num_bytes +=
(total_length * sizeof(InOutDataType) + invariantLength * 2 * sizeof(AccDataType)) + // No.1
(total_length * (1 * sizeof(InOutDataType) + 4 * sizeof(AccDataType)) +
total_length * sizeof(InOutDataType)); // No.2
if(saveMeanAndInvVariance && updateMovingAverage)
{
auto dev_moving_average_inv_variance = DeviceMovingAverageAndInvVarianceInstance{};
auto argument_ptr3 = dev_moving_average_inv_variance.MakeArgumentPointer(
bnScaleBiasMeanVarLengths,
{bnScaleBiasMeanVarStrides,
bnScaleBiasMeanVarStrides,
bnScaleBiasMeanVarStrides,
bnScaleBiasMeanVarStrides},
{bnScaleBiasMeanVarStrides, bnScaleBiasMeanVarStrides, bnScaleBiasMeanVarStrides},
{p_mean, p_runningMean, p_tmp_meansquare, p_runningVariance},
{p_runningMean, p_runningVariance, p_saveInvVariance},
MovingAverageAndInvVariance{epsilon, exponentialAverageFactor});
if(!dev_moving_average_inv_variance.IsSupportedArgument(argument_ptr3.get()))
{
std::cout << "Runtime parameters not supported by the Device, exiting!" << std::endl;
return (-1);
};
auto invoker_ptr3 = dev_moving_average_inv_variance.MakeInvokerPointer();
avg_time += invoker_ptr3->Run(argument_ptr3.get(), StreamConfig{nullptr, time_kernel});
num_bytes += invariantLength * (4 + 3) * sizeof(AccDataType) * 2; // No.5
}
else if(saveMeanAndInvVariance)
{
auto dev_inv_variance = DeviceInvVarianceInstance{};
auto argument_ptr3 = dev_inv_variance.MakeArgumentPointer(
bnScaleBiasMeanVarLengths,
{bnScaleBiasMeanVarStrides, bnScaleBiasMeanVarStrides},
{bnScaleBiasMeanVarStrides},
{p_mean, p_tmp_meansquare},
{p_saveInvVariance},
InvVariance{epsilon});
if(!dev_inv_variance.IsSupportedArgument(argument_ptr3.get()))
{
std::cout << "Runtime parameters not supported by the Device, exiting!" << std::endl;
return (-1);
};
auto invoker_ptr3 = dev_inv_variance.MakeInvokerPointer();
avg_time += invoker_ptr3->Run(argument_ptr3.get(), StreamConfig{nullptr, time_kernel});
num_bytes += invariantLength * (2 + 1) * sizeof(AccDataType);
}
else if(updateMovingAverage)
{
auto dev_moving_average = DeviceMovingAverageInstance{};
auto argument_ptr3 = dev_moving_average.MakeArgumentPointer(
bnScaleBiasMeanVarLengths,
{bnScaleBiasMeanVarStrides,
bnScaleBiasMeanVarStrides,
bnScaleBiasMeanVarStrides,
bnScaleBiasMeanVarStrides},
{bnScaleBiasMeanVarStrides, bnScaleBiasMeanVarStrides},
{p_mean, p_runningMean, p_tmp_meansquare, p_runningVariance},
{p_runningMean, p_runningVariance},
MovingAverage{exponentialAverageFactor});
if(!dev_moving_average.IsSupportedArgument(argument_ptr3.get()))
{
std::cout << "Runtime parameters not supported by the Device, exiting!" << std::endl;
return (-1);
};
auto invoker_ptr3 = dev_moving_average.MakeInvokerPointer();
avg_time += invoker_ptr3->Run(argument_ptr3.get(), StreamConfig{nullptr, time_kernel});
num_bytes += invariantLength * (4 + 2) * sizeof(AccDataType) * 2; // No.5
};
if(time_kernel)
{
float gb_per_sec = num_bytes / 1.E6 / avg_time;
std::cout << "Perf: " << avg_time << " ms, " << gb_per_sec << " GB/s" << std::endl;
};
return (0);
};
example/34_batchnorm/batchnorm_forward_nhwc.cpp
View file @ 7fa892e6
......@@ -15,13 +15,9 @@
#include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/utility/host_common_util.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_batchnorm_forward_nhwc_c.hpp"
#include "batchnorm_forward_impl.hpp"
template <typename InOutDataType, typename AccDataType>
using ReferenceBatchNormFwdInstance =
ck::tensor_operation::host::ReferenceBatchNormFwd_Input_N_H_W_C_Output_C<InOutDataType,
AccDataType>;
#include "ck/tensor_operation/gpu/device/impl/device_batchnorm_forward_impl.hpp"
#include "ck/library/utility/host_common_util.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
static struct option long_options[] = {{"inOutLengths", required_argument, nullptr, 'D'},
{"verify", required_argument, nullptr, 'v'},
......@@ -44,6 +40,7 @@ class BatchNormFwdArg
int data_type = 0;
int init_method = 2;
bool time_kernel = false;
bool use_multiblock_welford = false;
public:
void show_usage(const char* cmd)
......@@ -68,6 +65,7 @@ class BatchNormFwdArg
"value, 3=decimal value)"
<< std::endl;
std::cout << "Arg5: time kernel (0=no, 1=yes)" << std::endl;
std::cout << "Arg6: use multi-block welford (0=n0, 1=yes)" << std::endl;
};
int processArgs(int argc, char* argv[])
......@@ -110,14 +108,15 @@ class BatchNormFwdArg
};
};
if(optind + 5 > argc)
if(optind + 6 > argc)
throw std::runtime_error("Invalid cmd-line arguments, more argumetns are needed!");
data_type = std::atoi(argv[optind++]);
updateMovingAverage = std::atoi(argv[optind++]);
saveMeanAndInvVariance = std::atoi(argv[optind++]);
init_method = std::atoi(argv[optind++]);
time_kernel = static_cast<bool>(std::atoi(argv[optind]));
time_kernel = static_cast<bool>(std::atoi(argv[optind++]));
use_multiblock_welford = static_cast<bool>(std::atoi(argv[optind]));
if(data_type != 0 && data_type != 1 && data_type != 3 && data_type != 5 && data_type != 6)
return (-1);
......@@ -128,7 +127,7 @@ class BatchNormFwdArg
using namespace ck;
template <typename InOutDataType, typename AccDataType>
template <typename InOutDataType, typename AccDataType, bool UseMultiblockInK>
bool bnorm_fwd_nhwc_test(bool do_verification,
int init_method,
bool time_kernel,
......@@ -273,73 +272,140 @@ bool bnorm_fwd_nhwc_test(bool do_verification,
scaleBiasMeanVarStrides.end(),
i_scaleBiasMeanVarStrides.begin());
int result = 0;
// used for saving meansquare
DeviceMem workspace(sizeof(AccDataType) * 2 * resultSaveMean_ref.mDesc.GetElementSpaceSize() +
128);
using PassThroughOp = ck::tensor_operation::element_wise::PassThrough;
using DeviceBatchNormFwdInstance =
ck::tensor_operation::device::DeviceBatchNormFwdImpl<InOutDataType,
InOutDataType,
AccDataType,
AccDataType, // ScaleDataType
AccDataType, // BiasDataType
AccDataType, // MeanVarDataType
PassThroughOp, // YElementwiseOp
Rank,
NumReduceDim,
UseMultiblockInK,
256,
16,
16,
1,
2,
0,
1,
1,
1,
1,
1>;
void* p_tmp_mean = workspace.GetDeviceBuffer();
void* p_tmp_meansquare =
static_cast<char*>(p_tmp_mean) +
(sizeof(AccDataType) * resultSaveMean_ref.mDesc.GetElementSpaceSize() + 63) / 64 * 64;
auto batchnorm_fwd = DeviceBatchNormFwdInstance{};
result = bnorm_fwd<InOutDataType, AccDataType, Rank, NumReduceDim, false>(
time_kernel,
updateMovingAverage,
saveMeanAndInvVariance,
{0, 1, 2},
auto argument_ptr = batchnorm_fwd.MakeArgumentPointer(
i_inOutLengths,
i_inOutStrides,
i_inOutStrides,
{0, 1, 2},
i_scaleBiasMeanVarLengths,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
x_dev.GetDeviceBuffer(),
bnScale_dev.GetDeviceBuffer(),
bnBias_dev.GetDeviceBuffer(),
y_dev.GetDeviceBuffer(),
averageFactor,
updateMovingAverage ? resultRunningMean_dev.GetDeviceBuffer() : nullptr,
updateMovingAverage ? resultRunningVariance_dev.GetDeviceBuffer() : nullptr,
epsilon,
PassThroughOp{},
y_dev.GetDeviceBuffer(),
saveMeanAndInvVariance ? resultSaveMean_dev.GetDeviceBuffer() : nullptr,
saveMeanAndInvVariance ? resultSaveInvVariance_dev.GetDeviceBuffer() : nullptr,
p_tmp_mean,
p_tmp_meansquare);
averageFactor,
updateMovingAverage ? resultRunningMean_dev.GetDeviceBuffer() : nullptr,
updateMovingAverage ? resultRunningVariance_dev.GetDeviceBuffer() : nullptr);
if(result < 0)
if(!batchnorm_fwd.IsSupportedArgument(argument_ptr.get()))
{
std::cout << "The runtime parameters seems not supported by the BatchNorm device instance, "
"exiting!"
<< std::endl;
return (false);
};
size_t workspace_sz = batchnorm_fwd.GetWorkSpaceSize(argument_ptr.get());
DeviceMem workspace_dev(workspace_sz);
batchnorm_fwd.SetWorkSpacePointer(argument_ptr.get(), workspace_dev.GetDeviceBuffer());
auto invoker_ptr = batchnorm_fwd.MakeInvokerPointer();
if(time_kernel)
{
float avg_time = 0.0f;
size_t num_bytes = 0;
size_t total_length = inOutLengths[0] * inOutLengths[1] * inOutLengths[2] * inOutLengths[3];
size_t invariant_length = inOutLengths[3];
avg_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, time_kernel});
// inputing of x, scale, bias, outputing of y
num_bytes +=
total_length * sizeof(InOutDataType) * 2 + invariant_length * sizeof(AccDataType) * 2;
// outputing of mean, inv-variance
num_bytes += saveMeanAndInvVariance ? invariant_length * sizeof(AccDataType) * 2 : 0;
// updating of moving mean, variance
num_bytes += updateMovingAverage ? invariant_length * sizeof(AccDataType) * 4 : 0;
float gb_per_sec = num_bytes / 1.E6 / avg_time;
std::cout << "Perf: " << avg_time << " ms, " << gb_per_sec << " GB/s" << std::endl;
}
else
(void)invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, time_kernel});
bool pass = true;
if(do_verification)
{
auto batchNormFwd_ref = ReferenceBatchNormFwdInstance<InOutDataType, AccDataType>{};
using ReferenceBatchNormFwdInstance =
ck::tensor_operation::host::ReferenceBatchNormFwd_Input_N_H_W_C_Output_C<InOutDataType,
InOutDataType,
AccDataType,
AccDataType,
AccDataType,
AccDataType,
PassThroughOp>;
auto batchNormFwd_ref = ReferenceBatchNormFwdInstance{};
auto argument_ptr_ref = batchNormFwd_ref.MakeArgumentPointer(
i_inOutLengths,
i_inOutStrides,
i_inOutStrides,
{0, 1, 2},
i_scaleBiasMeanVarLengths,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
x.mData.data(),
bnScale.mData.data(),
bnBias.mData.data(),
y_ref.mData.data(),
0.1, // exponentialAverageFactor
updateMovingAverage ? resultRunningMean_ref.mData.data() : nullptr, // resultRunningMean
updateMovingAverage ? resultRunningVariance_ref.mData.data()
: nullptr, // resultRunningVariance
epsilon,
PassThroughOp{},
y_ref.mData.data(),
saveMeanAndInvVariance ? resultSaveMean_ref.mData.data() : nullptr,
saveMeanAndInvVariance ? resultSaveInvVariance_ref.mData.data() : nullptr);
saveMeanAndInvVariance ? resultSaveInvVariance_ref.mData.data() : nullptr,
averageFactor,
updateMovingAverage ? resultRunningMean_ref.mData.data() : nullptr,
updateMovingAverage ? resultRunningVariance_ref.mData.data() : nullptr);
if(!batchNormFwd_ref.IsSupportedArgument(argument_ptr_ref.get()))
{
std::cout
<< "The runtime parameters seems not supported by the BatchNorm instance, exiting!"
std::cout << "The runtime parameters seems not supported by the BatchNorm reference "
"instance, exiting!"
<< std::endl;
return (-2);
return (false);
};
auto invoker_ptr_ref = batchNormFwd_ref.MakeInvokerPointer();
......@@ -365,6 +431,8 @@ bool bnorm_fwd_nhwc_test(bool do_verification,
if(saveMeanAndInvVariance)
{
using ck::host_common::dumpBufferToFile;
Tensor<AccDataType> resultSaveMean(scaleBiasMeanVarLengths);
Tensor<AccDataType> resultSaveInvVariance(scaleBiasMeanVarLengths);
......@@ -396,7 +464,17 @@ int main(int argc, char* argv[])
if(arg.data_type == 0)
{
pass = bnorm_fwd_nhwc_test<ck::half_t, float>(arg.do_verification,
if(arg.use_multiblock_welford)
pass = bnorm_fwd_nhwc_test<ck::half_t, float, true>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
arg.updateMovingAverage,
arg.saveMeanAndInvVariance,
averageFactor,
epsilon);
else
pass = bnorm_fwd_nhwc_test<ck::half_t, float, false>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
......@@ -407,7 +485,17 @@ int main(int argc, char* argv[])
}
else if(arg.data_type == 1)
{
pass = bnorm_fwd_nhwc_test<float, float>(arg.do_verification,
if(arg.use_multiblock_welford)
pass = bnorm_fwd_nhwc_test<float, float, true>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
arg.updateMovingAverage,
arg.saveMeanAndInvVariance,
averageFactor,
epsilon);
else
pass = bnorm_fwd_nhwc_test<float, float, false>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
......@@ -418,7 +506,17 @@ int main(int argc, char* argv[])
}
else if(arg.data_type == 3)
{
pass = bnorm_fwd_nhwc_test<int8_t, float>(arg.do_verification,
if(arg.use_multiblock_welford)
pass = bnorm_fwd_nhwc_test<int8_t, float, true>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
arg.updateMovingAverage,
arg.saveMeanAndInvVariance,
averageFactor,
epsilon);
else
pass = bnorm_fwd_nhwc_test<int8_t, float, false>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
......@@ -429,7 +527,17 @@ int main(int argc, char* argv[])
}
else if(arg.data_type == 5)
{
pass = bnorm_fwd_nhwc_test<ck::bhalf_t, float>(arg.do_verification,
if(arg.use_multiblock_welford)
pass = bnorm_fwd_nhwc_test<ck::bhalf_t, float, true>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
arg.updateMovingAverage,
arg.saveMeanAndInvVariance,
averageFactor,
epsilon);
else
pass = bnorm_fwd_nhwc_test<ck::bhalf_t, float, false>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
......@@ -440,7 +548,17 @@ int main(int argc, char* argv[])
}
else if(arg.data_type == 6)
{
pass = bnorm_fwd_nhwc_test<double, double>(arg.do_verification,
if(arg.use_multiblock_welford)
pass = bnorm_fwd_nhwc_test<double, double, true>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
arg.updateMovingAverage,
arg.saveMeanAndInvVariance,
averageFactor,
epsilon);
else
pass = bnorm_fwd_nhwc_test<double, double, false>(arg.do_verification,
arg.init_method,
arg.time_kernel,
arg.inOutLengths,
......@@ -452,12 +570,21 @@ int main(int argc, char* argv[])
}
else
{
pass = bnorm_fwd_nhwc_test<ck::half_t, float>(true,
pass = bnorm_fwd_nhwc_test<ck::half_t, float, true>(true,
2,
false, // don't time kernel
{128, 16, 16, 1024},
{128, 16, 6, 512},
true,
true,
averageFactor,
epsilon);
pass = pass && bnorm_fwd_nhwc_test<ck::half_t, float, false>(true,
2,
false, // don't time kernel
{128, 16, 3, 1024},
true,
true,
false,
averageFactor,
epsilon);
};
......
example/34_batchnorm/batchnorm_infer_impl.hpp
View file @ 7fa892e6
......@@ -14,8 +14,12 @@
#include "batchnorm_common.hpp"
template <typename InOutDataType,
template <typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType,
ck::index_t Rank,
ck::index_t NumBatchNormReduceDim,
bool fastest_dim_is_reduced = false>
......@@ -26,7 +30,9 @@ int bnorm_infer(
const std::array<ck::index_t, Rank> xStrides,
const std::array<ck::index_t, Rank> yStrides,
const std::array<ck::index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarLengths,
const std::array<ck::index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarStrides,
const std::array<ck::index_t, Rank - NumBatchNormReduceDim> bnScaleStrides,
const std::array<ck::index_t, Rank - NumBatchNormReduceDim> bnBiasStrides,
const std::array<ck::index_t, Rank - NumBatchNormReduceDim> bnMeanVarStrides,
const void* p_x,
const void* p_scale,
const void* p_bias,
......@@ -41,11 +47,11 @@ int bnorm_infer(
"Invalid number of reduced dimensions for batchnorm!");
using DeviceNormalizeInstance = ck::tensor_operation::device::DeviceElementwise<
ck::Tuple<InOutDataType, AccDataType, AccDataType, AccDataType, AccDataType>, // x, mean,
ck::Tuple<XDataType, AccDataType, AccDataType, AccDataType, AccDataType>, // x, mean,
// variance,
// scale,
// bias,
ck::Tuple<InOutDataType>, // y
ck::Tuple<YDataType>, // y
NormalizeInInfer,
Rank,
2, // MPerthread
......@@ -53,14 +59,18 @@ int bnorm_infer(
ck::Sequence<1>>; // scalarPerVector: y
auto invariantDims = get_invariant_dims<Rank, NumBatchNormReduceDim>(reduceDims);
std::array<ck::index_t, Rank> aligned_scaleBiasMeanVarStrides{0};
std::array<ck::index_t, Rank> aligned_bnScaleStrides{0};
std::array<ck::index_t, Rank> aligned_bnBiasStrides{0};
std::array<ck::index_t, Rank> aligned_bnMeanVarStrides{0};
int i = 0;
for(auto dim : invariantDims)
{
assert(xyLengths[dim] == bnScaleBiasMeanVarLengths[i]);
aligned_scaleBiasMeanVarStrides[dim] = bnScaleBiasMeanVarStrides[i];
aligned_bnScaleStrides[dim] = bnScaleStrides[i];
aligned_bnBiasStrides[dim] = bnBiasStrides[i];
aligned_bnMeanVarStrides[dim] = bnMeanVarStrides[i];
i++;
};
......@@ -84,10 +94,10 @@ int bnorm_infer(
auto argument_ptr1 = dev_normalize.MakeArgumentPointer(
xyLengths,
{xStrides,
aligned_scaleBiasMeanVarStrides,
aligned_scaleBiasMeanVarStrides,
aligned_scaleBiasMeanVarStrides,
aligned_scaleBiasMeanVarStrides},
aligned_bnMeanVarStrides,
aligned_bnMeanVarStrides,
aligned_bnScaleStrides,
aligned_bnBiasStrides},
{yStrides},
{p_x, p_estimatedMean, p_estimatedVariance, p_scale, p_bias},
{p_y},
......@@ -105,8 +115,10 @@ int bnorm_infer(
avg_time += invoker_ptr1->Run(argument_ptr1.get(), StreamConfig{nullptr, time_kernel});
num_bytes += (total_length * (1 * sizeof(InOutDataType) + 4 * sizeof(AccDataType)) +
total_length * sizeof(InOutDataType));
num_bytes += total_length * sizeof(XDataType) +
invariantLength *
(sizeof(ScaleDataType) + sizeof(BiasDataType) + 2 * sizeof(MeanVarDataType)) +
total_length * sizeof(YDataType);
if(time_kernel)
{
......
example/34_batchnorm/batchnorm_infer_nhwc.cpp
View file @ 7fa892e6
......@@ -18,11 +18,6 @@
#include "batchnorm_infer_impl.hpp"
template <typename InOutDataType, typename AccDataType>
using ReferenceBatchNormInferInstance =
ck::tensor_operation::host::ReferenceBatchNormInfer_Input_N_H_W_C_Output_C<InOutDataType,
AccDataType>;
static struct option long_options[] = {{"inOutLengths", required_argument, nullptr, 'D'},
{"verify", required_argument, nullptr, 'v'},
{"help", no_argument, nullptr, '?'},
......@@ -236,14 +231,23 @@ bool bnorm_infer_nhwc_test(bool do_verification,
int result = 0;
result = bnorm_infer<InOutDataType, AccDataType, Rank, NumReduceDim, false>(
time_kernel,
result = bnorm_infer<InOutDataType,
InOutDataType,
AccDataType,
AccDataType,
AccDataType,
AccDataType,
Rank,
NumReduceDim,
false>(time_kernel,
{0, 1, 2},
i_inOutLengths,
i_inOutStrides,
i_inOutStrides,
i_scaleBiasMeanVarLengths,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
x_dev.GetDeviceBuffer(),
bnScale_dev.GetDeviceBuffer(),
bnBias_dev.GetDeviceBuffer(),
......@@ -259,7 +263,15 @@ bool bnorm_infer_nhwc_test(bool do_verification,
if(do_verification)
{
auto batchNormInfer_ref = ReferenceBatchNormInferInstance<InOutDataType, AccDataType>{};
using ReferenceBatchNormInferInstance =
ck::tensor_operation::host::ReferenceBatchNormInfer_Input_N_H_W_C_Output_C<
InOutDataType,
InOutDataType,
AccDataType,
AccDataType,
AccDataType,
AccDataType>;
auto batchNormInfer_ref = ReferenceBatchNormInferInstance{};
auto argument_ptr_ref =
batchNormInfer_ref.MakeArgumentPointer(i_inOutLengths,
......@@ -267,6 +279,8 @@ bool bnorm_infer_nhwc_test(bool do_verification,
i_inOutStrides,
i_scaleBiasMeanVarLengths,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
i_scaleBiasMeanVarStrides,
x.mData.data(),
bnScale.mData.data(),
bnBias.mData.data(),
......
include/ck/tensor_operation/gpu/device/device_batchnorm_forward.hpp
View file @ 7fa892e6
......@@ -13,31 +13,36 @@ namespace ck {
namespace tensor_operation {
namespace device {
template <index_t Rank, index_t NumBatchNormReduceDim>
template <index_t Rank, index_t NumBatchNormReduceDim, typename YElementwiseOp>
struct DeviceBatchNormFwd : public BaseOperator
{
virtual std::unique_ptr<BaseArgument> MakeArgumentPointer(
const std::array<index_t, Rank> xyLengths,
const std::array<index_t, Rank> xStrides,
const std::array<index_t, Rank> yStrides,
const std::array<int, NumBatchNormReduceDim> reduceDims,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarLengths,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnBiasStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnMeanVarStrides,
const void* p_x,
const void* bnScale,
const void* bnBias,
double epsilon,
const YElementwiseOp y_elementwise_op,
void* p_y,
void* resultSaveMean,
void* resultSaveInvVariance,
double exponentialAverageFactor,
void* resultRunningMean,
void* resultRunningVariance,
double epsilon,
void* resultSaveMean,
void* resultSaveInvVariance) = 0;
void* resultRunningVariance) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
};
template <index_t Rank, index_t NumBatchNormReduceDim>
using DeviceBatchNormFwdPtr = std::unique_ptr<DeviceBatchNormFwd<Rank, NumBatchNormReduceDim>>;
template <index_t Rank, index_t NumBatchNormReduceDim, typename YElementwiseOp>
using DeviceBatchNormFwdPtr =
std::unique_ptr<DeviceBatchNormFwd<Rank, NumBatchNormReduceDim, YElementwiseOp>>;
} // namespace device
} // namespace tensor_operation
......
include/ck/tensor_operation/gpu/device/device_batchnorm_infer.hpp
View file @ 7fa892e6
......@@ -21,7 +21,9 @@ struct DeviceBatchNormInfer : public BaseOperator
const std::array<index_t, Rank> xStrides,
const std::array<index_t, Rank> yStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarLengths,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnBiasStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnMeanVarStrides,
const void* p_x,
const void* bnScale,
const void* bnBias,
......
include/ck/tensor_operation/gpu/device/impl/device_batchnorm_forward_impl.hpp 0 → 100644
View file @ 7fa892e6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/reduction_operator.hpp"
#include "ck/tensor_operation/gpu/device/device_batchnorm_forward.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/device/welford_helper.hpp"
#include "ck/tensor_operation/gpu/grid/batchnorm_multiblock/gridwise_multiblock_welford_first_half.hpp"
#include "ck/tensor_operation/gpu/grid/batchnorm_multiblock/gridwise_multiblock_welford_second_half_batchnorm_forward_final.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batchnorm_forward_blockwise_welford.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType,
typename YElementwiseOp,
index_t Rank,
index_t NumBatchNormReduceDim,
bool UseMultiblockInK,
index_t BlockSize,
index_t MThreadClusterSize,
index_t KThreadClusterSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t XSrcYDstVectorDim,
index_t XSrcVectorSize,
index_t YDstVectorSize,
index_t ScaleSrcVectorSize,
index_t BiasSrcVectorSize,
index_t MeanVarSrcDstVectorSize>
struct DeviceBatchNormFwdImpl
: public DeviceBatchNormFwd<Rank, NumBatchNormReduceDim, YElementwiseOp>
{
static_assert(Rank <= 6, "Bigger Rank size is not supported!");
static_assert(BlockSize == MThreadClusterSize * KThreadClusterSize,
"Invalid thread cluster size assignments!");
static_assert((XSrcYDstVectorDim == 0 && MThreadSliceSize % XSrcVectorSize == 0) ||
(XSrcYDstVectorDim == 1 && KThreadSliceSize % XSrcVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static constexpr index_t NumInvariantDim = Rank - NumBatchNormReduceDim;
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
static auto MakeXY2dDescriptor(const std::array<index_t, Rank>& xyLengths,
const std::array<index_t, Rank>& xyStrides,
int blkGroupSize,
int numBlockTileIteration)
{
const auto tupleXYLengths =
generate_tuple([&](auto I) { return xyLengths[I]; }, Number<Rank>{});
const auto tupleXYStrides =
generate_tuple([&](auto I) { return xyStrides[I]; }, Number<Rank>{});
const auto raw_grid_desc = make_naive_tensor_descriptor(tupleXYLengths, tupleXYStrides);
const auto grid_desc_m_k = [&]() {
using InvariantDims = typename arithmetic_sequence_gen<0, NumInvariantDim, 1>::type;
using ReduceDims = typename arithmetic_sequence_gen<NumInvariantDim, Rank, 1>::type;
const auto reduceDimLengths =
generate_tuple([&](auto I) { return xyLengths[NumInvariantDim + I]; },
Number<NumBatchNormReduceDim>{});
const auto invariantDimLengths =
generate_tuple([&](auto I) { return xyLengths[I]; }, Number<NumInvariantDim>{});
return transform_tensor_descriptor(raw_grid_desc,
make_tuple(make_merge_transform(invariantDimLengths),
make_merge_transform(reduceDimLengths)),
make_tuple(InvariantDims{}, ReduceDims{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}();
const auto invariantLength = grid_desc_m_k.GetLength(Number<0>{});
const auto reduceLength = grid_desc_m_k.GetLength(Number<1>{});
const int workSizePerBlock = K_BlockTileSize * numBlockTileIteration;
const auto mPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
const auto kPad = workSizePerBlock * blkGroupSize - reduceLength;
auto grid_desc_m_k_padded =
transform_tensor_descriptor(grid_desc_m_k,
make_tuple(make_right_pad_transform(invariantLength, mPad),
make_right_pad_transform(reduceLength, kPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (grid_desc_m_k_padded);
};
static auto MakeMeanVarCountOutputMG2dDescriptor(int invariantLength, int blkGroupSize)
{
const auto grid_desc_m_g =
make_naive_tensor_descriptor_packed(make_tuple(invariantLength, blkGroupSize));
const auto mPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
auto grid_desc_m_g_padded =
transform_tensor_descriptor(grid_desc_m_g,
make_tuple(make_right_pad_transform(invariantLength, mPad),
make_pass_through_transform(blkGroupSize)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (grid_desc_m_g_padded);
};
static auto MakeMeanVarCountInputMK2dDescriptor(int invariantLength, int blkGroupSize)
{
const auto reduceLength = blkGroupSize;
const auto grid_desc_m_k =
make_naive_tensor_descriptor_packed(make_tuple(invariantLength, reduceLength));
const auto mPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
const auto kPad =
math::integer_least_multiple(reduceLength, KThreadClusterSize) - reduceLength;
auto grid_desc_m_k_padded =
transform_tensor_descriptor(grid_desc_m_k,
make_tuple(make_right_pad_transform(invariantLength, mPad),
make_right_pad_transform(reduceLength, kPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (grid_desc_m_k_padded);
};
static auto
MakeScaleBiasMeanVar1dDescriptor(const std::array<index_t, NumInvariantDim>& lengths,
const std::array<index_t, NumInvariantDim>& strides)
{
const auto tupleLengths =
generate_tuple([&](auto I) { return lengths[I]; }, Number<NumInvariantDim>{});
const auto tupleStrides =
generate_tuple([&](auto I) { return strides[I]; }, Number<NumInvariantDim>{});
auto raw_grid_desc = make_naive_tensor_descriptor(tupleLengths, tupleStrides);
auto grid_desc_m = transform_tensor_descriptor(
raw_grid_desc,
make_tuple(make_merge_transform(tupleLengths)),
make_tuple(typename arithmetic_sequence_gen<0, NumInvariantDim, 1>::type{}),
make_tuple(Sequence<0>{}));
const auto invariantLength = grid_desc_m.GetLength(Number<0>{});
const auto mPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
auto grid_desc_m_padded =
transform_tensor_descriptor(grid_desc_m,
make_tuple(make_right_pad_transform(invariantLength, mPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return (grid_desc_m_padded);
};
using XYGridDesc_M_K = decltype(MakeXY2dDescriptor({1}, {1}, 1, 1));
using ScaleBiasMeanVarGridDesc_M = decltype(MakeScaleBiasMeanVar1dDescriptor({1}, {1}));
struct Argument : public BaseArgument
{
Argument(const std::array<index_t, Rank> xyLengths,
const std::array<index_t, Rank> xStrides,
const std::array<index_t, Rank> yStrides,
const std::array<int, NumBatchNormReduceDim> reduceDims,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarLengths,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnBiasStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnMeanVarStrides,
const XDataType* p_x,
const ScaleDataType* p_scale,
const BiasDataType* p_bias,
const YElementwiseOp y_elementwise_op,
double epsilon,
YDataType* p_y,
MeanVarDataType* resultSaveMean,
MeanVarDataType* resultSaveInvVariance,
double averageFactor,
MeanVarDataType* resultRunningMean,
MeanVarDataType* resultRunningVariance)
: bnScaleBiasMeanVarLengths_(bnScaleBiasMeanVarLengths),
bnScaleStrides_(bnScaleStrides),
bnBiasStrides_(bnBiasStrides),
bnMeanVarStrides_(bnMeanVarStrides),
p_x_(p_x),
p_scale_(p_scale),
p_bias_(p_bias),
y_elementwise_op_(y_elementwise_op),
p_y_(p_y),
resultSaveMean_(resultSaveMean),
resultSaveInvVariance_(resultSaveInvVariance),
resultRunningMean_(resultRunningMean),
resultRunningVariance_(resultRunningVariance)
{
xyLengths_ =
shuffle_tensor_dimensions<Rank, NumBatchNormReduceDim>(xyLengths, reduceDims);
xStrides_ =
shuffle_tensor_dimensions<Rank, NumBatchNormReduceDim>(xStrides, reduceDims);
yStrides_ =
shuffle_tensor_dimensions<Rank, NumBatchNormReduceDim>(yStrides, reduceDims);
std::tie(invariant_length_, reduce_length_) =
get_2d_lengths<Rank, NumBatchNormReduceDim>(xyLengths_);
epsilon_ = type_convert<AccDataType>(epsilon);
averageFactor_ = type_convert<AccDataType>(averageFactor);
updateMovingAverage_ =
(resultRunningMean != nullptr && resultRunningVariance != nullptr);
saveMeanInvVariance_ = (resultSaveMean != nullptr && resultSaveInvVariance_ != nullptr);
if(UseMultiblockInK)
{
int iterations = 1;
while(true)
{
int testBlkGroupSize = (reduce_length_ + (K_BlockTileSize * iterations) - 1) /
(K_BlockTileSize * iterations);
// we want the blkGroupSize be not more than 128
if(testBlkGroupSize <= 128)
break;
iterations++;
};
blkGroupSize_ = (reduce_length_ + (K_BlockTileSize * iterations) - 1) /
(K_BlockTileSize * iterations);
numBlockTileIteration_ = iterations;
}
else
{
blkGroupSize_ = 1;
numBlockTileIteration_ = (reduce_length_ + K_BlockTileSize - 1) / K_BlockTileSize;
};
gridSize_ = (invariant_length_ + M_BlockTileSize - 1) / M_BlockTileSize * blkGroupSize_;
x_grid_desc_m_k_ =
MakeXY2dDescriptor(xyLengths_, xStrides_, blkGroupSize_, numBlockTileIteration_);
y_grid_desc_m_k_ =
MakeXY2dDescriptor(xyLengths_, yStrides_, blkGroupSize_, numBlockTileIteration_);
scale_grid_desc_m_ =
MakeScaleBiasMeanVar1dDescriptor(bnScaleBiasMeanVarLengths, bnScaleStrides_);
bias_grid_desc_m_ =
MakeScaleBiasMeanVar1dDescriptor(bnScaleBiasMeanVarLengths, bnBiasStrides_);
mean_var_grid_desc_m_ =
MakeScaleBiasMeanVar1dDescriptor(bnScaleBiasMeanVarLengths, bnMeanVarStrides_);
}
AccDataType epsilon_;
AccDataType averageFactor_;
bool updateMovingAverage_;
bool saveMeanInvVariance_;
std::array<index_t, Rank> xyLengths_;
std::array<index_t, Rank> xStrides_;
std::array<index_t, Rank> yStrides_;
std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarLengths_;
std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleStrides_;
std::array<index_t, Rank - NumBatchNormReduceDim> bnBiasStrides_;
std::array<index_t, Rank - NumBatchNormReduceDim> bnMeanVarStrides_;
const XDataType* p_x_;
const ScaleDataType* p_scale_;
const BiasDataType* p_bias_;
const YElementwiseOp y_elementwise_op_;
YDataType* p_y_;
MeanVarDataType* resultSaveMean_;
MeanVarDataType* resultSaveInvVariance_;
MeanVarDataType* resultRunningMean_;
MeanVarDataType* resultRunningVariance_;
long_index_t invariant_length_;
long_index_t reduce_length_;
int blkGroupSize_;
int numBlockTileIteration_;
size_t gridSize_;
XYGridDesc_M_K x_grid_desc_m_k_;
XYGridDesc_M_K y_grid_desc_m_k_;
ScaleBiasMeanVarGridDesc_M scale_grid_desc_m_;
ScaleBiasMeanVarGridDesc_M bias_grid_desc_m_;
ScaleBiasMeanVarGridDesc_M mean_var_grid_desc_m_;
void* workspace_mean_;
void* workspace_variance_;
void* workspace_count_;
};
size_t GetWorkSpaceSize(const BaseArgument* pArg) const override
{
const Argument* pArg_ = dynamic_cast<const Argument*>(pArg);
size_t workspace_size = 0;
if(UseMultiblockInK && pArg_->blkGroupSize_ > 1)
{
// workspace for welford intermediate mean
workspace_size +=
pArg_->invariant_length_ * pArg_->blkGroupSize_ * sizeof(MeanVarDataType) + 64;
// workspace for welford intermediate variance
workspace_size +=
pArg_->invariant_length_ * pArg_->blkGroupSize_ * sizeof(MeanVarDataType) + 64;
// workspace for welford intermediate count
workspace_size +=
pArg_->invariant_length_ * pArg_->blkGroupSize_ * sizeof(int32_t) + 64;
}
return (workspace_size);
};
void SetWorkSpacePointer(BaseArgument* pArg, void* p_workspace) const override
{
Argument* pArg_ = dynamic_cast<Argument*>(pArg);
pArg_->p_workspace_ = p_workspace;
if(UseMultiblockInK && pArg_->blkGroupSize_ > 1)
{
// setup buffer used for intermediate welford mean
pArg_->workspace_mean_ = static_cast<char*>(pArg_->p_workspace_);
index_t mean_space_sz =
pArg_->invariant_length_ * pArg_->blkGroupSize_ * sizeof(MeanVarDataType);
mean_space_sz = math::integer_least_multiple(mean_space_sz, 64);
// setup buffer used for intermediate welford varirance
pArg_->workspace_variance_ =
reinterpret_cast<char*>(pArg_->workspace_mean_) + mean_space_sz;
index_t variance_space_sz =
pArg_->invariant_length_ * pArg_->blkGroupSize_ * sizeof(MeanVarDataType);
variance_space_sz = math::integer_least_multiple(variance_space_sz, 64);
// setup buffer used for intermediate welfor count
pArg_->workspace_count_ =
reinterpret_cast<char*>(pArg_->workspace_variance_) + variance_space_sz;
};
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
float avg_time = 0;
if(UseMultiblockInK && arg.blkGroupSize_ > 1)
{
using GetReduceCountPerThreadFunctor =
GetReduceCountPerThreadForMultiblockWelford<K_BlockTileSize, KThreadSliceSize>;
GetReduceCountPerThreadFunctor get_reduce_count_per_thread(
arg.blkGroupSize_, arg.numBlockTileIteration_, arg.reduce_length_);
const auto mean_var_count_grid_desc_m_g =
DeviceBatchNormFwdImpl::MakeMeanVarCountOutputMG2dDescriptor(
arg.invariant_length_, arg.blkGroupSize_);
const auto mean_var_count_grid_desc_m_k =
DeviceBatchNormFwdImpl::MakeMeanVarCountInputMK2dDescriptor(
arg.invariant_length_, arg.blkGroupSize_);
using MeanVarCountGridDesc_M_G = decltype(mean_var_count_grid_desc_m_g);
using MeanVarCountGridDesc_M_K = decltype(mean_var_count_grid_desc_m_k);
using GridwiseMultiblockWelfordFirstHalf_ =
GridwiseMultiblockWelfordFirstHalf<XDataType,
AccDataType,
MeanVarDataType,
XYGridDesc_M_K,
MeanVarCountGridDesc_M_G,
GetReduceCountPerThreadFunctor,
BlockSize,
MThreadClusterSize,
KThreadClusterSize,
MThreadSliceSize,
KThreadSliceSize,
XSrcYDstVectorDim,
XSrcVectorSize>;
using GridwiseWelfordSecondHalfBatchNormForwardFinal_ =
GridwiseWelfordSecondHalfBatchNormForwardFinal<XDataType,
YDataType,
AccDataType,
ScaleDataType,
BiasDataType,
MeanVarDataType,
YElementwiseOp,
XYGridDesc_M_K,
MeanVarCountGridDesc_M_K,
ScaleBiasMeanVarGridDesc_M,
ScaleBiasMeanVarGridDesc_M,
BlockSize,
MThreadClusterSize,
KThreadClusterSize,
MThreadSliceSize,
KThreadSliceSize,
XSrcYDstVectorDim,
XSrcVectorSize,
YDstVectorSize,
ScaleSrcVectorSize,
BiasSrcVectorSize,
MeanVarSrcDstVectorSize>;
index_t numMeanVarCountBlockTileIteration =
(arg.blkGroupSize_ + KThreadClusterSize - 1) / KThreadClusterSize;
const auto kern_multiblock_welford_first_half =
kernel_multiblock_welford_first_half<GridwiseMultiblockWelfordFirstHalf_,
XDataType,
MeanVarDataType,
XYGridDesc_M_K,
MeanVarCountGridDesc_M_G,
GetReduceCountPerThreadFunctor>;
const auto kern_welford_second_half_batchnorm_forward_final =
kernel_welford_second_half_batchnorm_forward_final<
GridwiseWelfordSecondHalfBatchNormForwardFinal_,
XDataType,
YDataType,
AccDataType,
ScaleDataType,
BiasDataType,
MeanVarDataType,
YElementwiseOp,
XYGridDesc_M_K,
MeanVarCountGridDesc_M_K,
ScaleBiasMeanVarGridDesc_M,
ScaleBiasMeanVarGridDesc_M>;
avg_time +=
launch_and_time_kernel(stream_config,
kern_multiblock_welford_first_half,
dim3(arg.gridSize_),
dim3(BlockSize),
0,
arg.x_grid_desc_m_k_,
mean_var_count_grid_desc_m_g,
get_reduce_count_per_thread,
arg.numBlockTileIteration_,
arg.p_x_,
static_cast<MeanVarDataType*>(arg.workspace_mean_),
static_cast<MeanVarDataType*>(arg.workspace_variance_),
static_cast<int32_t*>(arg.workspace_count_));
avg_time +=
launch_and_time_kernel(stream_config,
kern_welford_second_half_batchnorm_forward_final,
dim3(arg.gridSize_),
dim3(BlockSize),
0,
arg.x_grid_desc_m_k_,
arg.y_grid_desc_m_k_,
mean_var_count_grid_desc_m_k,
arg.scale_grid_desc_m_,
arg.bias_grid_desc_m_,
arg.mean_var_grid_desc_m_,
arg.blkGroupSize_,
arg.numBlockTileIteration_,
numMeanVarCountBlockTileIteration,
arg.epsilon_,
static_cast<MeanVarDataType*>(arg.workspace_mean_),
static_cast<MeanVarDataType*>(arg.workspace_variance_),
static_cast<int32_t*>(arg.workspace_count_),
arg.p_x_,
arg.p_scale_,
arg.p_bias_,
arg.y_elementwise_op_,
arg.p_y_,
arg.updateMovingAverage_,
arg.averageFactor_,
arg.resultRunningMean_,
arg.resultRunningVariance_,
arg.saveMeanInvVariance_,
arg.resultSaveMean_,
arg.resultSaveInvVariance_);
}
else
{
using GetReduceCountPerThreadFunctor =
GetReduceCountPerThreadForBlockwiseWelford<K_BlockTileSize, KThreadSliceSize>;
GetReduceCountPerThreadFunctor get_reduce_count_per_thread(
arg.numBlockTileIteration_, arg.reduce_length_);
using GridwiseBatchNormForwardWithBlockwiseWelford_ =
GridwiseBatchNormForwardWithBlockwiseWelford<XDataType,
YDataType,
AccDataType,
ScaleDataType,
BiasDataType,
MeanVarDataType,
YElementwiseOp,
XYGridDesc_M_K,
ScaleBiasMeanVarGridDesc_M,
ScaleBiasMeanVarGridDesc_M,
GetReduceCountPerThreadFunctor,
BlockSize,
MThreadClusterSize,
KThreadClusterSize,
MThreadSliceSize,
KThreadSliceSize,
XSrcYDstVectorDim,
XSrcVectorSize,
YDstVectorSize,
ScaleSrcVectorSize,
BiasSrcVectorSize,
MeanVarSrcDstVectorSize>;
const auto kern_batchnorm_fwd = kernel_batchnorm_forward_with_blockwise_welford<
GridwiseBatchNormForwardWithBlockwiseWelford_,
XDataType,
YDataType,
AccDataType,
ScaleDataType,
BiasDataType,
MeanVarDataType,
YElementwiseOp,
XYGridDesc_M_K,
ScaleBiasMeanVarGridDesc_M,
ScaleBiasMeanVarGridDesc_M,
GetReduceCountPerThreadFunctor>;
avg_time += launch_and_time_kernel(stream_config,
kern_batchnorm_fwd,
dim3(arg.gridSize_),
dim3(BlockSize),
0,
arg.x_grid_desc_m_k_,
arg.y_grid_desc_m_k_,
arg.scale_grid_desc_m_,
arg.bias_grid_desc_m_,
arg.mean_var_grid_desc_m_,
get_reduce_count_per_thread,
arg.numBlockTileIteration_,
arg.epsilon_,
arg.p_x_,
arg.p_scale_,
arg.p_bias_,
arg.y_elementwise_op_,
arg.p_y_,
arg.updateMovingAverage_, // true or false
arg.averageFactor_,
arg.resultRunningMean_,
arg.resultRunningVariance_,
arg.saveMeanInvVariance_, // true or false
arg.resultSaveMean_,
arg.resultSaveInvVariance_);
};
return (avg_time);
};
float Run(const BaseArgument* pArg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(pArg), stream_config);
};
};
bool IsSupportedArgument(const BaseArgument* pArg) override
{
const Argument* pArg_ = dynamic_cast<const Argument*>(pArg);
if constexpr(XSrcYDstVectorDim == 0)
{
if(pArg_->xStrides_[NumInvariantDim - 1] != 1 ||
pArg_->yStrides_[NumInvariantDim - 1] != 1)
return false;
if(pArg_->xyLengths_[NumInvariantDim - 1] % XSrcVectorSize != 0 ||
pArg_->xyLengths_[NumInvariantDim - 1] % YDstVectorSize != 0)
return false;
}
else
{
if(pArg_->xStrides_[Rank - 1] != 1 || pArg_->yStrides_[Rank - 1] != 1)
return false;
if(pArg_->xyLengths_[Rank - 1] % XSrcVectorSize != 0 ||
pArg_->xyLengths_[Rank - 1] % YDstVectorSize != 0)
return false;
};
if(pArg_->bnScaleStrides_[NumInvariantDim - 1] != 1 && ScaleSrcVectorSize != 1)
return false;
if(pArg_->bnBiasStrides_[NumInvariantDim - 1] != 1 && BiasSrcVectorSize != 1)
return false;
if(pArg_->bnScaleBiasMeanVarLengths_[NumInvariantDim - 1] % ScaleSrcVectorSize != 0)
return false;
if(pArg_->bnScaleBiasMeanVarLengths_[NumInvariantDim - 1] % BiasSrcVectorSize != 0)
return false;
if(pArg_->bnMeanVarStrides_[NumInvariantDim - 1] != 1 && MeanVarSrcDstVectorSize != 1)
return false;
if(pArg_->bnScaleBiasMeanVarLengths_[NumInvariantDim - 1] % MeanVarSrcDstVectorSize != 0)
return false;
bool is_valid = true;
static_for<0, NumInvariantDim, 1>{}([&](auto I) {
if(pArg_->xyLengths_[I] != pArg_->bnScaleBiasMeanVarLengths_[I])
is_valid = false;
});
if(!is_valid)
return false;
return true;
};
std::unique_ptr<BaseArgument> MakeArgumentPointer(
const std::array<index_t, Rank> xyLengths,
const std::array<index_t, Rank> xStrides,
const std::array<index_t, Rank> yStrides,
const std::array<int, NumBatchNormReduceDim> reduceDims,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleBiasMeanVarLengths,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnScaleStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnBiasStrides,
const std::array<index_t, Rank - NumBatchNormReduceDim> bnMeanVarStrides,
const void* p_x,
const void* p_scale,
const void* p_bias,
double epsilon,
const YElementwiseOp y_elementwise_op,
void* p_y,
void* resultSaveMean,
void* resultSaveInvVariance,
double averageFactor,
void* resultRunningMean,
void* resultRunningVariance) override
{
return std::make_unique<Argument>(xyLengths,
xStrides,
yStrides,
reduceDims,
bnScaleBiasMeanVarLengths,
bnScaleStrides,
bnBiasStrides,
bnMeanVarStrides,
static_cast<const XDataType*>(p_x),
static_cast<const ScaleDataType*>(p_scale),
static_cast<const BiasDataType*>(p_bias),
y_elementwise_op,
epsilon,
static_cast<YDataType*>(p_y),
static_cast<MeanVarDataType*>(resultSaveMean),
static_cast<MeanVarDataType*>(resultSaveInvVariance),
averageFactor,
static_cast<MeanVarDataType*>(resultRunningMean),
static_cast<MeanVarDataType*>(resultRunningVariance));
};
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>();
};
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchNormFwdImpl<" << BlockSize << ",";
str << "M_C" << MThreadClusterSize << "_S" << MThreadSliceSize << ",";
str << "K_C" << KThreadClusterSize << "_S" << KThreadSliceSize << ",";
str << "XSrcYDstVectorDim_" << XSrcYDstVectorDim << ",";
str << "VectorSize_X" << XSrcVectorSize << "_scale_" << ScaleSrcVectorSize << "_bias_" << BiasSrcVectorSize << "_mean_var_" << MeanVarSrcDstVectorSize << "_Y" << YDstVectorSize << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/welford_helper.hpp 0 → 100644
View file @ 7fa892e6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
namespace ck {
namespace tensor_operation {
namespace device {
template <index_t K_BlockTileSize, index_t KThreadSliceSize>
struct GetReduceCountPerThreadForBlockwiseWelford
{
GetReduceCountPerThreadForBlockwiseWelford(index_t numBlockTileIteration,
long_index_t reduce_length)
: numBlockTileIteration_{numBlockTileIteration}
{
count_in_last_tile_ = reduce_length % K_BlockTileSize;
};
__device__ index_t operator()(index_t thread_k_cluster_id) const
{
if(count_in_last_tile_ == 0)
return (KThreadSliceSize * numBlockTileIteration_);
else
{
index_t num_complete_slice = count_in_last_tile_ / KThreadSliceSize;
index_t count_in_last_slice = count_in_last_tile_ % KThreadSliceSize;
if(thread_k_cluster_id < num_complete_slice)
return (KThreadSliceSize * numBlockTileIteration_);
else if(thread_k_cluster_id == num_complete_slice)
return (KThreadSliceSize * (numBlockTileIteration_ - 1) + count_in_last_slice);
else
return (KThreadSliceSize * (numBlockTileIteration_ - 1));
};
};
index_t numBlockTileIteration_;
index_t count_in_last_tile_;
};
template <index_t K_BlockTileSize, index_t KThreadSliceSize>
struct GetReduceCountPerThreadForMultiblockWelford
{
GetReduceCountPerThreadForMultiblockWelford(index_t blkGroupSize,
index_t numBlockTileIteration,
long_index_t reduce_length)
: blkGroupSize_(blkGroupSize), numBlockTileIteration_{numBlockTileIteration}
{
last_block_reduce_length_ =
reduce_length - K_BlockTileSize * numBlockTileIteration_ * (blkGroupSize_ - 1);
numBlockTileIterationByLastBlock_ =
(last_block_reduce_length_ + K_BlockTileSize - 1) / K_BlockTileSize;
};
__device__ index_t operator()(index_t block_local_id, index_t thread_k_cluster_id) const
{
if(last_block_reduce_length_ == K_BlockTileSize * numBlockTileIteration_ ||
block_local_id < blkGroupSize_ - 1)
return (KThreadSliceSize * numBlockTileIteration_);
index_t count_in_last_tile = last_block_reduce_length_ % K_BlockTileSize;
if(count_in_last_tile == 0)
return (KThreadSliceSize * numBlockTileIterationByLastBlock_);
else
{
index_t num_complete_slice = count_in_last_tile / KThreadSliceSize;
if(thread_k_cluster_id < num_complete_slice)
return (KThreadSliceSize * numBlockTileIterationByLastBlock_);
else if(thread_k_cluster_id == num_complete_slice)
return (KThreadSliceSize * (numBlockTileIterationByLastBlock_ - 1) +
count_in_last_tile);
else
return (KThreadSliceSize * (numBlockTileIterationByLastBlock_ - 1));
};
};
index_t blkGroupSize_;
index_t numBlockTileIteration_;
index_t last_block_reduce_length_;
index_t numBlockTileIterationByLastBlock_;
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/grid/batchnorm_multiblock/gridwise_multiblock_welford_first_half.hpp 0 → 100644
View file @ 7fa892e6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck/utility/data_type.hpp"
#include "ck/utility/math.hpp"
#include "ck/tensor_operation/gpu/block/blockwise_welford.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_welford.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
namespace ck {
template <typename GridwiseMultiblockWelfordFirstHalf_,
typename XDataType,
typename MeanVarDataType,
typename XGridDesc_M_K,
typename MeanVarCountGridDesc_M_G,
typename GetReduceCountPerThreadFunctor>
__global__ void kernel_multiblock_welford_first_half(
const XGridDesc_M_K x_grid_desc_m_k,
const MeanVarCountGridDesc_M_G mean_var_count_grid_desc_m_g,
const GetReduceCountPerThreadFunctor get_reduce_count_per_thread,
index_t num_k_block_tile_iteration,
const XDataType* const __restrict__ p_x,
MeanVarDataType* const p_welford_mean,
MeanVarDataType* const p_welford_variance,
int32_t* const p_welford_count)
{
GridwiseMultiblockWelfordFirstHalf_::Run(x_grid_desc_m_k,
mean_var_count_grid_desc_m_g,
get_reduce_count_per_thread,
num_k_block_tile_iteration,
p_x,
p_welford_mean,
p_welford_variance,
p_welford_count);
};
template <typename XDataType,
typename AccDataType,
typename MeanVarDataType,
typename XGridDesc_M_K,
typename MeanVarCountGridDesc_M_G,
typename GetReduceCountPerThreadFunctor,
index_t BlockSize,
index_t MThreadClusterSize,
index_t KThreadClusterSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t XSrcCountSrcVectorDim,
index_t XSrcCountSrcVectorSize>
struct GridwiseMultiblockWelfordFirstHalf
{
static_assert((XSrcCountSrcVectorDim == 0 && MThreadSliceSize % XSrcCountSrcVectorSize == 0) ||
(XSrcCountSrcVectorDim == 1 &&
KThreadSliceSize % XSrcCountSrcVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static constexpr bool reorder_thread_cluster = (XSrcCountSrcVectorDim == 0);
using ThreadClusterLengths_M_K = Sequence<MThreadClusterSize, KThreadClusterSize>;
using ThreadBufferDimAccessOrder =
typename conditional<reorder_thread_cluster, Sequence<1, 0>, Sequence<0, 1>>::type;
using ThreadClusterArrangeOrder =
typename conditional<reorder_thread_cluster, Sequence<1, 0>, Sequence<0, 1>>::type;
static constexpr auto thread_cluster_desc =
make_cluster_descriptor(ThreadClusterLengths_M_K{}, ThreadClusterArrangeOrder{});
using ThreadReduceSrcDesc_M_K = decltype(make_naive_tensor_descriptor_packed(
make_tuple(Number<MThreadSliceSize>{}, Number<KThreadSliceSize>{})));
using ThreadReduceDstDesc_M =
decltype(make_naive_tensor_descriptor_packed(make_tuple(Number<MThreadSliceSize>{})));
using ThreadwiseWelford =
ThreadwiseWelford<AccDataType, ThreadReduceSrcDesc_M_K, ThreadReduceDstDesc_M>;
using BlockwiseWelford = BlockwiseWelford<AccDataType,
BlockSize,
ThreadClusterLengths_M_K,
ThreadClusterArrangeOrder,
false>;
using PassThroughOp = tensor_operation::element_wise::PassThrough;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
__device__ static void Run(const XGridDesc_M_K& x_grid_desc_m_k,
const MeanVarCountGridDesc_M_G& mean_var_count_grid_desc_m_g,
const GetReduceCountPerThreadFunctor& get_reduce_count_per_thread,
index_t num_k_block_tile_iteration,
const XDataType* const __restrict__ p_x,
MeanVarDataType* const p_welford_mean,
MeanVarDataType* const p_welford_variance,
int32_t* const p_welford_count)
{
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize * KThreadSliceSize, true>
x_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
welford_mean_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
welford_var_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, int32_t, MThreadSliceSize, true>
welford_count_thread_buf;
const index_t blkgroup_size = mean_var_count_grid_desc_m_g.GetLength(I1);
const index_t thread_local_id = get_thread_local_1d_id();
const index_t block_global_id = get_block_1d_id();
const index_t blkgroup_id = block_global_id / blkgroup_size;
const index_t block_local_id = block_global_id % blkgroup_size;
const auto thread_cluster_idx =
thread_cluster_desc.CalculateBottomIndex(make_multi_index(thread_local_id));
const auto thread_m_cluster_id = thread_cluster_idx[I0];
const auto thread_k_cluster_id = thread_cluster_idx[I1];
using ThreadBufferLengths_M_K = Sequence<MThreadSliceSize, KThreadSliceSize>;
using ThreadBufferLengths_M_1 = Sequence<MThreadSliceSize, 1>;
constexpr auto thread_buffer_desc_m_k = make_naive_tensor_descriptor_packed(
make_tuple(Number<MThreadSliceSize>{}, Number<KThreadSliceSize>{}));
constexpr auto thread_buffer_desc_m_1 = make_naive_tensor_descriptor_packed(
make_tuple(Number<MThreadSliceSize>{}, Number<1>{}));
const index_t reduceSizePerBlock = K_BlockTileSize * num_k_block_tile_iteration;
auto threadwise_x_load = ThreadwiseTensorSliceTransfer_v2<XDataType,
AccDataType,
XGridDesc_M_K,
decltype(thread_buffer_desc_m_k),
ThreadBufferLengths_M_K,
ThreadBufferDimAccessOrder,
XSrcCountSrcVectorDim,
XSrcCountSrcVectorSize,
1,
true>(
x_grid_desc_m_k,
make_multi_index(blkgroup_id * M_BlockTileSize + thread_m_cluster_id * MThreadSliceSize,
block_local_id * reduceSizePerBlock +
thread_k_cluster_id * KThreadSliceSize));
auto threadwise_welford_mean_var_store =
ThreadwiseTensorSliceTransfer_v1r3<AccDataType,
MeanVarDataType,
decltype(thread_buffer_desc_m_1),
MeanVarCountGridDesc_M_G,
PassThroughOp,
ThreadBufferLengths_M_1,
Sequence<0, 1>,
1,
1,
InMemoryDataOperationEnum::Set,
1,
true>(
mean_var_count_grid_desc_m_g,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize,
block_local_id),
PassThroughOp{});
auto threadwise_welford_count_store =
ThreadwiseTensorSliceTransfer_v1r3<int32_t,
int32_t,
decltype(thread_buffer_desc_m_1),
MeanVarCountGridDesc_M_G,
PassThroughOp,
ThreadBufferLengths_M_1,
Sequence<0, 1>,
1,
1,
InMemoryDataOperationEnum::Set,
1,
true>(
mean_var_count_grid_desc_m_g,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize,
block_local_id),
PassThroughOp{});
constexpr auto thread_copy_fwd_step_m_k = make_multi_index(0, K_BlockTileSize);
const auto x_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_x, x_grid_desc_m_k.GetElementSpaceSize());
auto welford_mean_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_welford_mean, mean_var_count_grid_desc_m_g.GetElementSpaceSize());
auto welford_var_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_welford_variance, mean_var_count_grid_desc_m_g.GetElementSpaceSize());
auto welford_count_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_welford_count, mean_var_count_grid_desc_m_g.GetElementSpaceSize());
auto threadwise_welford = ThreadwiseWelford();
threadwise_welford.max_count_ =
get_reduce_count_per_thread(block_local_id, thread_k_cluster_id);
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
welford_mean_thread_buf(I) = type_convert<AccDataType>(0.0f);
welford_var_thread_buf(I) = type_convert<AccDataType>(0.0f);
});
for(index_t reducedTiles = 0; reducedTiles < num_k_block_tile_iteration; ++reducedTiles)
{
threadwise_x_load.Run(x_grid_desc_m_k,
x_global_val_buf,
thread_buffer_desc_m_k,
make_tuple(I0, I0),
x_thread_buf);
threadwise_x_load.MoveSrcSliceWindow(x_grid_desc_m_k, thread_copy_fwd_step_m_k);
threadwise_welford.Run(x_thread_buf, welford_mean_thread_buf, welford_var_thread_buf);
}
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
if constexpr(I > 0)
block_sync_lds();
welford_count_thread_buf(I) = threadwise_welford.cur_count_;
BlockwiseWelford::Run(
welford_mean_thread_buf(I), welford_var_thread_buf(I), welford_count_thread_buf(I));
});
if(thread_k_cluster_id == 0)
{
threadwise_welford_mean_var_store.Run(thread_buffer_desc_m_1,
make_tuple(I0, I0),
welford_mean_thread_buf,
mean_var_count_grid_desc_m_g,
welford_mean_global_val_buf);
threadwise_welford_mean_var_store.Run(thread_buffer_desc_m_1,
make_tuple(I0, I0),
welford_var_thread_buf,
mean_var_count_grid_desc_m_g,
welford_var_global_val_buf);
threadwise_welford_count_store.Run(thread_buffer_desc_m_1,
make_tuple(I0, I0),
welford_count_thread_buf,
mean_var_count_grid_desc_m_g,
welford_count_global_val_buf);
};
}
};
} // namespace ck
include/ck/tensor_operation/gpu/grid/batchnorm_multiblock/gridwise_multiblock_welford_second_half_batchnorm_forward_final.hpp 0 → 100644
View file @ 7fa892e6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck/utility/data_type.hpp"
#include "ck/utility/math_v2.hpp"
#include "ck/tensor_operation/gpu/block/blockwise_welford.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_welford.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
namespace ck {
template <typename GridwiseWelfordSecondHalfBatchNormForwardFinal_,
typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType,
typename YElementwiseOp,
typename XYGridDesc_M_K,
typename MeanVarCountGridDesc_M_K,
typename ScaleBiasGridDesc_M,
typename MeanVarGridDesc_M>
__global__ void kernel_welford_second_half_batchnorm_forward_final(
const XYGridDesc_M_K x_grid_desc_m_k,
const XYGridDesc_M_K y_grid_desc_m_k,
const MeanVarCountGridDesc_M_K mean_var_count_grid_desc_m_k,
const ScaleBiasGridDesc_M scale_grid_desc_m,
const ScaleBiasGridDesc_M bias_grid_desc_m,
const MeanVarGridDesc_M mean_var_grid_desc_m,
index_t blkgroup_size,
index_t num_xy_k_block_tile_iteration,
index_t num_mean_var_count_k_block_tile_iteration,
AccDataType epsilon,
const MeanVarDataType* const __restrict__ p_in_welford_mean,
const MeanVarDataType* const __restrict__ p_in_welford_variance,
const int32_t* const __restrict__ p_in_welford_count,
const XDataType* const __restrict__ p_x,
const ScaleDataType* const __restrict__ p_scale,
const BiasDataType* const __restrict__ p_bias,
const YElementwiseOp y_elementwise_op,
YDataType* const __restrict__ p_y,
bool updateMovingAverage,
AccDataType averageFactor,
MeanVarDataType* const __restrict__ resultRunningMean,
MeanVarDataType* const __restrict__ resultRunningVariance,
bool saveMeanInvVariance,
MeanVarDataType* const __restrict__ resultSaveMean,
MeanVarDataType* const __restrict__ resultSaveInvVariance)
{
GridwiseWelfordSecondHalfBatchNormForwardFinal_::Run(x_grid_desc_m_k,
y_grid_desc_m_k,
mean_var_count_grid_desc_m_k,
scale_grid_desc_m,
bias_grid_desc_m,
mean_var_grid_desc_m,
blkgroup_size,
num_xy_k_block_tile_iteration,
num_mean_var_count_k_block_tile_iteration,
epsilon,
p_in_welford_mean,
p_in_welford_variance,
p_in_welford_count,
p_x,
p_scale,
p_bias,
y_elementwise_op,
p_y,
updateMovingAverage,
averageFactor,
resultRunningMean,
resultRunningVariance,
saveMeanInvVariance,
resultSaveMean,
resultSaveInvVariance);
};
template <typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType,
typename YElementwiseOp,
typename XYGridDesc_M_K,
typename MeanVarCountGridDesc_M_K,
typename ScaleBiasGridDesc_M,
typename MeanVarGridDesc_M,
index_t BlockSize,
index_t MThreadClusterSize,
index_t KThreadClusterSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t XSrcYDstVectorDim,
index_t XSrcVectorSize,
index_t YDstVectorSize,
index_t ScaleSrcVectorSize,
index_t BiasSrcVectorSize,
index_t MeanVarSrcDstVectorSize>
struct GridwiseWelfordSecondHalfBatchNormForwardFinal
{
static_assert((XSrcYDstVectorDim == 0 && MThreadSliceSize % XSrcVectorSize == 0) ||
(XSrcYDstVectorDim == 1 && KThreadSliceSize % XSrcVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static_assert((XSrcYDstVectorDim == 0 && MThreadSliceSize % YDstVectorSize == 0) ||
(XSrcYDstVectorDim == 1 && KThreadSliceSize % YDstVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static constexpr bool reorder_thread_cluster = (XSrcYDstVectorDim == 0);
using ThreadClusterLengths_M_K = Sequence<MThreadClusterSize, KThreadClusterSize>;
using ThreadBufferDimAccessOrder =
typename conditional<reorder_thread_cluster, Sequence<1, 0>, Sequence<0, 1>>::type;
using ThreadClusterArrangeOrder =
typename conditional<reorder_thread_cluster, Sequence<1, 0>, Sequence<0, 1>>::type;
static constexpr auto thread_cluster_desc =
make_cluster_descriptor(ThreadClusterLengths_M_K{}, ThreadClusterArrangeOrder{});
using ThreadReduceSrcDesc_M_1 = decltype(
make_naive_tensor_descriptor_packed(make_tuple(Number<MThreadSliceSize>{}, Number<1>{})));
using ThreadReduceDstDesc_M =
decltype(make_naive_tensor_descriptor_packed(make_tuple(Number<MThreadSliceSize>{})));
using ThreadwiseWelford =
ThreadwiseWelfordMerge<AccDataType, ThreadReduceSrcDesc_M_1, ThreadReduceDstDesc_M>;
using BlockwiseWelford = BlockwiseWelford<AccDataType,
BlockSize,
ThreadClusterLengths_M_K,
ThreadClusterArrangeOrder>;
using PassThroughOp = tensor_operation::element_wise::PassThrough;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
__device__ static void Run(const XYGridDesc_M_K& x_grid_desc_m_k,
const XYGridDesc_M_K& y_grid_desc_m_k,
const MeanVarCountGridDesc_M_K& mean_var_count_grid_desc_m_k,
const ScaleBiasGridDesc_M& scale_grid_desc_m,
const ScaleBiasGridDesc_M& bias_grid_desc_m,
const MeanVarGridDesc_M& mean_var_grid_desc_m,
index_t blkgroup_size,
index_t num_xy_k_block_tile_iteration,
index_t num_mean_var_count_k_block_tile_iteration,
AccDataType epsilon,
const MeanVarDataType* const __restrict__ p_in_welford_mean,
const MeanVarDataType* const __restrict__ p_in_welford_variance,
const int32_t* const __restrict__ p_in_welford_count,
const XDataType* const __restrict__ p_x,
const ScaleDataType* const __restrict__ p_scale,
const BiasDataType* const __restrict__ p_bias,
const YElementwiseOp y_elementwise_op,
YDataType* const __restrict__ p_y,
bool updateMovingAverage,
AccDataType averageFactor,
MeanVarDataType* const __restrict__ resultRunningMean,
MeanVarDataType* const __restrict__ resultRunningVariance,
bool saveMeanInvVariance,
MeanVarDataType* const __restrict__ resultSaveMean,
MeanVarDataType* const __restrict__ resultSaveInvVariance)
{
using ck::math::sqrt;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize * 1, true>
in_welford_mean_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize * 1, true>
in_welford_var_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, int32_t, MThreadSliceSize * 1, true>
in_welford_count_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
welford_mean_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
welford_var_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, int32_t, MThreadSliceSize, true>
welford_count_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize * KThreadSliceSize, true>
x_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize * KThreadSliceSize, true>
y_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true> scale_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true> bias_thread_buf;
const index_t thread_local_id = get_thread_local_1d_id();
const index_t block_global_id = get_block_1d_id();
const index_t blkgroup_id = block_global_id / blkgroup_size;
const index_t block_local_id = block_global_id % blkgroup_size;
const auto thread_cluster_idx =
thread_cluster_desc.CalculateBottomIndex(make_multi_index(thread_local_id));
const auto thread_m_cluster_id = thread_cluster_idx[I0];
const auto thread_k_cluster_id = thread_cluster_idx[I1];
using ThreadBufferLengths_M_K = Sequence<MThreadSliceSize, KThreadSliceSize>;
using ThreadBufferLengths_M = Sequence<MThreadSliceSize>;
using ThreadBufferLengths_M_1 = Sequence<MThreadSliceSize, 1>;
constexpr auto thread_buffer_desc_m_k = make_naive_tensor_descriptor_packed(
make_tuple(Number<MThreadSliceSize>{}, Number<KThreadSliceSize>{}));
constexpr auto thread_buffer_desc_m =
make_naive_tensor_descriptor_packed(make_tuple(Number<MThreadSliceSize>{}));
constexpr auto thread_buffer_desc_m_1 = make_naive_tensor_descriptor_packed(
make_tuple(Number<MThreadSliceSize>{}, Number<1>{}));
auto threadwise_mean_var_load_m_k =
ThreadwiseTensorSliceTransfer_v2<MeanVarDataType,
AccDataType,
MeanVarCountGridDesc_M_K,
decltype(thread_buffer_desc_m_1),
ThreadBufferLengths_M_1,
Sequence<0, 1>,
1,
1,
1,
true>(
mean_var_count_grid_desc_m_k,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize,
thread_k_cluster_id * 1));
auto threadwise_count_load_m_k =
ThreadwiseTensorSliceTransfer_v2<int32_t,
int32_t,
MeanVarCountGridDesc_M_K,
decltype(thread_buffer_desc_m_1),
ThreadBufferLengths_M_1,
Sequence<0, 1>,
1,
1,
1,
true>(
mean_var_count_grid_desc_m_k,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize,
thread_k_cluster_id * 1));
const auto welford_mean_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_in_welford_mean, mean_var_count_grid_desc_m_k.GetElementSpaceSize());
const auto welford_var_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_in_welford_variance, mean_var_count_grid_desc_m_k.GetElementSpaceSize());
const auto welford_count_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_in_welford_count, mean_var_count_grid_desc_m_k.GetElementSpaceSize());
constexpr auto mean_var_count_thread_copy_step_m_k =
make_multi_index(0, KThreadClusterSize * 1);
// Step 1: do final welford reduction to get mean and variance
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
welford_mean_thread_buf(I) = type_convert<AccDataType>(0.0f);
welford_var_thread_buf(I) = type_convert<AccDataType>(0.0f);
welford_count_thread_buf(I) = 0;
});
for(index_t reducedTiles = 0; reducedTiles < num_mean_var_count_k_block_tile_iteration;
++reducedTiles)
{
threadwise_mean_var_load_m_k.Run(mean_var_count_grid_desc_m_k,
welford_mean_global_val_buf,
thread_buffer_desc_m_1,
make_tuple(I0, I0),
in_welford_mean_thread_buf);
threadwise_mean_var_load_m_k.Run(mean_var_count_grid_desc_m_k,
welford_var_global_val_buf,
thread_buffer_desc_m_1,
make_tuple(I0, I0),
in_welford_var_thread_buf);
threadwise_count_load_m_k.Run(mean_var_count_grid_desc_m_k,
welford_count_global_val_buf,
thread_buffer_desc_m_1,
make_tuple(I0, I0),
in_welford_count_thread_buf);
ThreadwiseWelford::Run(in_welford_mean_thread_buf,
in_welford_var_thread_buf,
in_welford_count_thread_buf,
welford_mean_thread_buf,
welford_var_thread_buf,
welford_count_thread_buf);
threadwise_mean_var_load_m_k.MoveSrcSliceWindow(mean_var_count_grid_desc_m_k,
mean_var_count_thread_copy_step_m_k);
threadwise_count_load_m_k.MoveSrcSliceWindow(mean_var_count_grid_desc_m_k,
mean_var_count_thread_copy_step_m_k);
}
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
if constexpr(I > 0)
block_sync_lds();
BlockwiseWelford::Run(
welford_mean_thread_buf(I), welford_var_thread_buf(I), welford_count_thread_buf(I));
});
// Step 2: do normalization and output y
const index_t workSizePerBlock = K_BlockTileSize * num_xy_k_block_tile_iteration;
auto threadwise_x_load = ThreadwiseTensorSliceTransfer_v2<XDataType,
AccDataType,
XYGridDesc_M_K,
decltype(thread_buffer_desc_m_k),
ThreadBufferLengths_M_K,
ThreadBufferDimAccessOrder,
XSrcYDstVectorDim,
XSrcVectorSize,
1,
true>(
x_grid_desc_m_k,
make_multi_index(blkgroup_id * M_BlockTileSize + thread_m_cluster_id * MThreadSliceSize,
workSizePerBlock * block_local_id +
thread_k_cluster_id * KThreadSliceSize));
auto threadwise_y_store =
ThreadwiseTensorSliceTransfer_v1r3<AccDataType,
YDataType,
decltype(thread_buffer_desc_m_k),
XYGridDesc_M_K,
YElementwiseOp,
ThreadBufferLengths_M_K,
ThreadBufferDimAccessOrder,
XSrcYDstVectorDim,
YDstVectorSize,
InMemoryDataOperationEnum::Set,
1,
true>(
y_grid_desc_m_k,
make_multi_index(
blkgroup_id * M_BlockTileSize + thread_m_cluster_id * MThreadSliceSize,
workSizePerBlock * block_local_id + thread_k_cluster_id * KThreadSliceSize),
y_elementwise_op);
auto threadwise_scale_load =
ThreadwiseTensorSliceTransfer_v2<ScaleDataType,
AccDataType,
ScaleBiasGridDesc_M,
decltype(thread_buffer_desc_m),
ThreadBufferLengths_M,
Sequence<0>,
0,
ScaleSrcVectorSize,
1,
true>(
scale_grid_desc_m,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize));
auto threadwise_bias_load = ThreadwiseTensorSliceTransfer_v2<BiasDataType,
AccDataType,
ScaleBiasGridDesc_M,
decltype(thread_buffer_desc_m),
ThreadBufferLengths_M,
Sequence<0>,
0,
BiasSrcVectorSize,
1,
true>(
bias_grid_desc_m,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize));
const auto x_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_x, x_grid_desc_m_k.GetElementSpaceSize());
const auto scale_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_scale, scale_grid_desc_m.GetElementSpaceSize());
const auto bias_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_bias, bias_grid_desc_m.GetElementSpaceSize());
auto y_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_y, y_grid_desc_m_k.GetElementSpaceSize());
threadwise_scale_load.Run(scale_grid_desc_m,
scale_global_val_buf,
thread_buffer_desc_m,
make_tuple(I0),
scale_thread_buf);
threadwise_bias_load.Run(bias_grid_desc_m,
bias_global_val_buf,
thread_buffer_desc_m,
make_tuple(I0),
bias_thread_buf);
constexpr auto xy_thread_copy_step_m_k = make_multi_index(0, K_BlockTileSize);
for(index_t workTiles = 0; workTiles < num_xy_k_block_tile_iteration; ++workTiles)
{
threadwise_x_load.Run(x_grid_desc_m_k,
x_global_val_buf,
thread_buffer_desc_m_k,
make_tuple(I0, I0),
x_thread_buf);
static_for<0, MThreadSliceSize, 1>{}([&](auto iM) {
AccDataType multiplier =
scale_thread_buf[iM] / sqrt(welford_var_thread_buf[iM] + epsilon);
AccDataType fused_mean_bias =
bias_thread_buf[iM] - welford_mean_thread_buf[iM] * multiplier;
static_for<0, KThreadSliceSize, 1>{}([&](auto iK) {
constexpr auto offset =
thread_buffer_desc_m_k.CalculateOffset(make_tuple(iM, iK));
y_thread_buf(Number<offset>{}) =
x_thread_buf[Number<offset>{}] * multiplier + fused_mean_bias;
});
});
threadwise_y_store.Run(thread_buffer_desc_m_k,
make_tuple(I0, I0),
y_thread_buf,
y_grid_desc_m_k,
y_global_val_buf);
threadwise_x_load.MoveSrcSliceWindow(x_grid_desc_m_k, xy_thread_copy_step_m_k);
threadwise_y_store.MoveDstSliceWindow(y_grid_desc_m_k, xy_thread_copy_step_m_k);
}
// Step 3: update the moving average of mean and variance (optional)
if(updateMovingAverage && block_local_id == 0 && thread_k_cluster_id == 0)
{
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
running_mean_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
running_var_thread_buf;
auto running_mean_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultRunningMean, mean_var_grid_desc_m.GetElementSpaceSize());
auto running_var_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultRunningVariance, mean_var_grid_desc_m.GetElementSpaceSize());
auto threadwise_mean_var_load_m =
ThreadwiseTensorSliceTransfer_v2<MeanVarDataType,
AccDataType,
MeanVarGridDesc_M,
decltype(thread_buffer_desc_m),
ThreadBufferLengths_M,
Sequence<0>,
0,
MeanVarSrcDstVectorSize,
1,
true>(
mean_var_grid_desc_m,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize));
threadwise_mean_var_load_m.Run(mean_var_grid_desc_m,
running_mean_global_buf,
thread_buffer_desc_m,
make_tuple(I0),
running_mean_thread_buf);
threadwise_mean_var_load_m.Run(mean_var_grid_desc_m,
running_var_global_buf,
thread_buffer_desc_m,
make_tuple(I0),
running_var_thread_buf);
AccDataType oneMinusAverageFactor = type_convert<AccDataType>(1.0) - averageFactor;
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
running_mean_thread_buf(I) = running_mean_thread_buf[I] * oneMinusAverageFactor +
welford_mean_thread_buf[I] * averageFactor;
running_var_thread_buf(I) = running_var_thread_buf[I] * oneMinusAverageFactor +
welford_var_thread_buf[I] * averageFactor;
});
auto threadwise_mean_var_store =
ThreadwiseTensorSliceTransfer_v1r3<AccDataType,
MeanVarDataType,
decltype(thread_buffer_desc_m),
MeanVarGridDesc_M,
PassThroughOp,
ThreadBufferLengths_M,
Sequence<0>,
0,
MeanVarSrcDstVectorSize,
InMemoryDataOperationEnum::Set,
1,
true>(
mean_var_grid_desc_m,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize),
PassThroughOp{});
threadwise_mean_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
running_mean_thread_buf,
mean_var_grid_desc_m,
running_mean_global_buf);
threadwise_mean_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
running_var_thread_buf,
mean_var_grid_desc_m,
running_var_global_buf);
};
// Step 4: save mean and inv-variance (optional)
if(saveMeanInvVariance && block_local_id == 0 && thread_k_cluster_id == 0)
{
auto result_mean_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultSaveMean, mean_var_grid_desc_m.GetElementSpaceSize());
auto result_inv_var_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultSaveInvVariance, mean_var_grid_desc_m.GetElementSpaceSize());
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
welford_var_thread_buf(I) =
type_convert<AccDataType>(1.0f) / sqrt(epsilon + welford_var_thread_buf[I]);
});
auto threadwise_mean_inv_var_store =
ThreadwiseTensorSliceTransfer_v1r3<AccDataType,
MeanVarDataType,
decltype(thread_buffer_desc_m),
MeanVarGridDesc_M,
PassThroughOp,
ThreadBufferLengths_M,
Sequence<0>,
0,
MeanVarSrcDstVectorSize,
InMemoryDataOperationEnum::Set,
1,
true>(
mean_var_grid_desc_m,
make_multi_index(blkgroup_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize),
PassThroughOp{});
threadwise_mean_inv_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
welford_mean_thread_buf,
mean_var_grid_desc_m,
result_mean_global_buf);
threadwise_mean_inv_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
welford_var_thread_buf,
mean_var_grid_desc_m,
result_inv_var_global_buf);
};
}
};
} // namespace ck
include/ck/tensor_operation/gpu/grid/gridwise_batchnorm_forward_blockwise_welford.hpp 0 → 100644
View file @ 7fa892e6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck/utility/data_type.hpp"
#include "ck/utility/math_v2.hpp"
#include "ck/tensor_operation/gpu/block/blockwise_welford.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_welford.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
namespace ck {
template <typename GridwiseBatchrNormForwardWithBlockwiseWelford_,
typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType,
typename YElementwiseOp,
typename XYGridDesc_M_K,
typename ScaleBiasGridDesc_M,
typename MeanVarGridDesc_M,
typename GetReduceCountPerThreadFunctor>
__global__ void kernel_batchnorm_forward_with_blockwise_welford(
const XYGridDesc_M_K x_grid_desc_m_k,
const XYGridDesc_M_K y_grid_desc_m_k,
const ScaleBiasGridDesc_M scale_grid_desc_m,
const ScaleBiasGridDesc_M bias_grid_desc_m,
const MeanVarGridDesc_M mean_var_grid_desc_m,
const GetReduceCountPerThreadFunctor get_reduce_count_per_thread,
index_t num_k_block_tile_iteration,
AccDataType epsilon,
const XDataType* const __restrict__ p_x,
const ScaleDataType* const __restrict__ p_scale,
const BiasDataType* const __restrict__ p_bias,
const YElementwiseOp y_elementwise_op,
YDataType* const __restrict__ p_y,
bool updateMovingAverage,
AccDataType averageFactor,
MeanVarDataType* const __restrict__ resultRunningMean,
MeanVarDataType* const __restrict__ resultRunningVariance,
bool saveMeanInvVariance,
MeanVarDataType* const __restrict__ resultSaveMean,
MeanVarDataType* const __restrict__ resultSaveInvVariance)
{
GridwiseBatchrNormForwardWithBlockwiseWelford_::Run(x_grid_desc_m_k,
y_grid_desc_m_k,
scale_grid_desc_m,
bias_grid_desc_m,
mean_var_grid_desc_m,
get_reduce_count_per_thread,
num_k_block_tile_iteration,
epsilon,
p_x,
p_scale,
p_bias,
y_elementwise_op,
p_y,
updateMovingAverage,
averageFactor,
resultRunningMean,
resultRunningVariance,
saveMeanInvVariance,
resultSaveMean,
resultSaveInvVariance);
};
template <typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType,
typename YElementwiseOp,
typename XYGridDesc_M_K,
typename ScaleBiasGridDesc_M,
typename MeanVarGridDesc_M,
typename GetReduceCountPerThreadFunctor,
index_t BlockSize,
index_t MThreadClusterSize,
index_t KThreadClusterSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t XSrcYDstVectorDim,
index_t XSrcVectorSize,
index_t YDstVectorSize,
index_t ScaleSrcVectorSize,
index_t BiasSrcVectorSize,
index_t MeanVarSrcDstVectorSize>
struct GridwiseBatchNormForwardWithBlockwiseWelford
{
static_assert((XSrcYDstVectorDim == 0 && MThreadSliceSize % XSrcVectorSize == 0) ||
(XSrcYDstVectorDim == 1 && KThreadSliceSize % XSrcVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static_assert((XSrcYDstVectorDim == 0 && MThreadSliceSize % YDstVectorSize == 0) ||
(XSrcYDstVectorDim == 1 && KThreadSliceSize % YDstVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static constexpr bool reorder_thread_cluster = (XSrcYDstVectorDim == 0);
using ThreadClusterLengths_M_K = Sequence<MThreadClusterSize, KThreadClusterSize>;
using ThreadBufferDimAccessOrder =
typename conditional<reorder_thread_cluster, Sequence<1, 0>, Sequence<0, 1>>::type;
using ThreadClusterArrangeOrder =
typename conditional<reorder_thread_cluster, Sequence<1, 0>, Sequence<0, 1>>::type;
static constexpr auto thread_cluster_desc =
make_cluster_descriptor(ThreadClusterLengths_M_K{}, ThreadClusterArrangeOrder{});
using ThreadReduceSrcDesc_M_K = decltype(make_naive_tensor_descriptor_packed(
make_tuple(Number<MThreadSliceSize>{}, Number<KThreadSliceSize>{})));
using ThreadReduceDstDesc_M =
decltype(make_naive_tensor_descriptor_packed(make_tuple(Number<MThreadSliceSize>{})));
using ThreadwiseWelford =
ThreadwiseWelford<AccDataType, ThreadReduceSrcDesc_M_K, ThreadReduceDstDesc_M>;
using BlockwiseWelford = BlockwiseWelford<AccDataType,
BlockSize,
ThreadClusterLengths_M_K,
ThreadClusterArrangeOrder>;
using PassThroughOp = tensor_operation::element_wise::PassThrough;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
__device__ static void Run(const XYGridDesc_M_K& x_grid_desc_m_k,
const XYGridDesc_M_K& y_grid_desc_m_k,
const ScaleBiasGridDesc_M& scale_grid_desc_m,
const ScaleBiasGridDesc_M& bias_grid_desc_m,
const MeanVarGridDesc_M& mean_var_grid_desc_m,
const GetReduceCountPerThreadFunctor& get_reduce_count_per_thread,
index_t num_k_block_tile_iteration,
AccDataType epsilon,
const XDataType* const __restrict__ p_x,
const ScaleDataType* const __restrict__ p_scale,
const BiasDataType* const __restrict__ p_bias,
const YElementwiseOp y_elementwise_op,
YDataType* const __restrict__ p_y,
bool updateMovingAverage,
AccDataType averageFactor,
MeanVarDataType* const __restrict__ resultRunningMean,
MeanVarDataType* const __restrict__ resultRunningVariance,
bool saveMeanInvVariance,
MeanVarDataType* const __restrict__ resultSaveMean,
MeanVarDataType* const __restrict__ resultSaveInvVariance)
{
using ck::math::sqrt;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize * KThreadSliceSize, true>
x_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true> scale_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true> bias_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize * KThreadSliceSize, true>
y_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true> mean_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true> var_thread_buf;
const index_t thread_local_id = get_thread_local_1d_id();
const index_t block_global_id = get_block_1d_id();
const auto thread_cluster_idx =
thread_cluster_desc.CalculateBottomIndex(make_multi_index(thread_local_id));
const auto thread_m_cluster_id = thread_cluster_idx[I0];
const auto thread_k_cluster_id = thread_cluster_idx[I1];
using ThreadBufferLengths_M_K = Sequence<MThreadSliceSize, KThreadSliceSize>;
using ThreadBufferLengths_M = Sequence<MThreadSliceSize>;
constexpr auto thread_buffer_desc_m_k = make_naive_tensor_descriptor_packed(
make_tuple(Number<MThreadSliceSize>{}, Number<KThreadSliceSize>{}));
constexpr auto thread_buffer_desc_m =
make_naive_tensor_descriptor_packed(make_tuple(Number<MThreadSliceSize>{}));
auto threadwise_x_load = ThreadwiseTensorSliceTransfer_v2<XDataType,
AccDataType,
XYGridDesc_M_K,
decltype(thread_buffer_desc_m_k),
ThreadBufferLengths_M_K,
ThreadBufferDimAccessOrder,
XSrcYDstVectorDim,
XSrcVectorSize,
1,
true>(
x_grid_desc_m_k,
make_multi_index(block_global_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize,
thread_k_cluster_id * KThreadSliceSize));
auto threadwise_y_store =
ThreadwiseTensorSliceTransfer_v1r3<AccDataType,
YDataType,
decltype(thread_buffer_desc_m_k),
XYGridDesc_M_K,
YElementwiseOp,
ThreadBufferLengths_M_K,
ThreadBufferDimAccessOrder,
XSrcYDstVectorDim,
YDstVectorSize,
InMemoryDataOperationEnum::Set,
1,
true>(
y_grid_desc_m_k,
make_multi_index(block_global_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize,
thread_k_cluster_id * KThreadSliceSize),
y_elementwise_op);
auto threadwise_scale_load =
ThreadwiseTensorSliceTransfer_v2<ScaleDataType,
AccDataType,
ScaleBiasGridDesc_M,
decltype(thread_buffer_desc_m),
ThreadBufferLengths_M,
Sequence<0>,
0,
ScaleSrcVectorSize,
1,
true>(
scale_grid_desc_m,
make_multi_index(block_global_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize));
auto threadwise_bias_load = ThreadwiseTensorSliceTransfer_v2<BiasDataType,
AccDataType,
ScaleBiasGridDesc_M,
decltype(thread_buffer_desc_m),
ThreadBufferLengths_M,
Sequence<0>,
0,
BiasSrcVectorSize,
1,
true>(
bias_grid_desc_m,
make_multi_index(block_global_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize));
constexpr auto thread_copy_fwd_step_m_k = make_multi_index(0, K_BlockTileSize);
constexpr auto thread_copy_bwd_step_m_k = make_multi_index(0, -K_BlockTileSize);
const auto x_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_x, x_grid_desc_m_k.GetElementSpaceSize());
const auto scale_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_scale, scale_grid_desc_m.GetElementSpaceSize());
const auto bias_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_bias, bias_grid_desc_m.GetElementSpaceSize());
auto y_global_val_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_y, y_grid_desc_m_k.GetElementSpaceSize());
// Step 1: do welford reduction to get mean and variance
auto threadwise_welford = ThreadwiseWelford();
threadwise_welford.max_count_ = get_reduce_count_per_thread(thread_k_cluster_id);
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
mean_thread_buf(I) = type_convert<AccDataType>(0.0f);
var_thread_buf(I) = type_convert<AccDataType>(0.0f);
});
for(index_t reducedTiles = 0; reducedTiles < num_k_block_tile_iteration; ++reducedTiles)
{
threadwise_x_load.Run(x_grid_desc_m_k,
x_global_val_buf,
thread_buffer_desc_m_k,
make_tuple(I0, I0),
x_thread_buf);
threadwise_x_load.MoveSrcSliceWindow(x_grid_desc_m_k, thread_copy_fwd_step_m_k);
threadwise_welford.Run(x_thread_buf, mean_thread_buf, var_thread_buf);
}
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
if constexpr(I > 0)
block_sync_lds();
int count = threadwise_welford.cur_count_;
BlockwiseWelford::Run(mean_thread_buf(I), var_thread_buf(I), count);
});
// Step 2: do normalization and output y
threadwise_scale_load.Run(scale_grid_desc_m,
scale_global_val_buf,
thread_buffer_desc_m,
make_tuple(I0),
scale_thread_buf);
threadwise_bias_load.Run(bias_grid_desc_m,
bias_global_val_buf,
thread_buffer_desc_m,
make_tuple(I0),
bias_thread_buf);
auto thread_copy_tail_m_k = (num_k_block_tile_iteration - 1) * thread_copy_fwd_step_m_k;
threadwise_x_load.MoveSrcSliceWindow(x_grid_desc_m_k, thread_copy_bwd_step_m_k);
threadwise_y_store.MoveDstSliceWindow(y_grid_desc_m_k, thread_copy_tail_m_k);
for(index_t reducedTiles = 0; reducedTiles < num_k_block_tile_iteration; ++reducedTiles)
{
threadwise_x_load.Run(x_grid_desc_m_k,
x_global_val_buf,
thread_buffer_desc_m_k,
make_tuple(I0, I0),
x_thread_buf);
static_for<0, MThreadSliceSize, 1>{}([&](auto iM) {
AccDataType multiplier =
scale_thread_buf[Number<iM>{}] / sqrt(var_thread_buf[iM] + epsilon);
AccDataType fused_mean_bias =
bias_thread_buf[Number<iM>{}] - mean_thread_buf[iM] * multiplier;
static_for<0, KThreadSliceSize, 1>{}([&](auto iK) {
constexpr auto offset =
thread_buffer_desc_m_k.CalculateOffset(make_tuple(iM, iK));
// normalize
y_thread_buf(Number<offset>{}) =
x_thread_buf[Number<offset>{}] * multiplier + fused_mean_bias;
});
});
threadwise_y_store.Run(thread_buffer_desc_m_k,
make_tuple(I0, I0),
y_thread_buf,
y_grid_desc_m_k,
y_global_val_buf);
threadwise_x_load.MoveSrcSliceWindow(x_grid_desc_m_k, thread_copy_bwd_step_m_k);
threadwise_y_store.MoveDstSliceWindow(y_grid_desc_m_k, thread_copy_bwd_step_m_k);
}
// Step 3: update the moving average of mean and variance (optional)
if(updateMovingAverage && thread_k_cluster_id == 0)
{
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
running_mean_thread_buf;
StaticBuffer<AddressSpaceEnum::Vgpr, AccDataType, MThreadSliceSize, true>
running_var_thread_buf;
auto running_mean_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultRunningMean, mean_var_grid_desc_m.GetElementSpaceSize());
auto running_var_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultRunningVariance, mean_var_grid_desc_m.GetElementSpaceSize());
auto threadwise_mean_var_load =
ThreadwiseTensorSliceTransfer_v2<MeanVarDataType,
AccDataType,
MeanVarGridDesc_M,
decltype(thread_buffer_desc_m),
ThreadBufferLengths_M,
Sequence<0>,
0,
MeanVarSrcDstVectorSize,
1,
true>(
mean_var_grid_desc_m,
make_multi_index(block_global_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize));
threadwise_mean_var_load.Run(mean_var_grid_desc_m,
running_mean_global_buf,
thread_buffer_desc_m,
make_tuple(I0),
running_mean_thread_buf);
threadwise_mean_var_load.Run(mean_var_grid_desc_m,
running_var_global_buf,
thread_buffer_desc_m,
make_tuple(I0),
running_var_thread_buf);
AccDataType oneMinusAverageFactor = type_convert<AccDataType>(1.0) - averageFactor;
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
running_mean_thread_buf(I) = running_mean_thread_buf[I] * oneMinusAverageFactor +
mean_thread_buf[I] * averageFactor;
running_var_thread_buf(I) = running_var_thread_buf[I] * oneMinusAverageFactor +
var_thread_buf[I] * averageFactor;
});
auto threadwise_mean_var_store =
ThreadwiseTensorSliceTransfer_v1r3<AccDataType,
MeanVarDataType,
decltype(thread_buffer_desc_m),
MeanVarGridDesc_M,
PassThroughOp,
ThreadBufferLengths_M,
Sequence<0>,
0,
MeanVarSrcDstVectorSize,
InMemoryDataOperationEnum::Set,
1,
true>(
mean_var_grid_desc_m,
make_multi_index(block_global_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize),
PassThroughOp{});
threadwise_mean_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
running_mean_thread_buf,
mean_var_grid_desc_m,
running_mean_global_buf);
threadwise_mean_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
running_var_thread_buf,
mean_var_grid_desc_m,
running_var_global_buf);
};
// Step 4: save mean and inv-variance (optional)
if(saveMeanInvVariance && thread_k_cluster_id == 0)
{
auto result_mean_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultSaveMean, mean_var_grid_desc_m.GetElementSpaceSize());
auto result_inv_var_global_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
resultSaveInvVariance, mean_var_grid_desc_m.GetElementSpaceSize());
static_for<0, MThreadSliceSize, 1>{}([&](auto I) {
var_thread_buf(I) =
type_convert<AccDataType>(1.0f) / sqrt(epsilon + var_thread_buf[I]);
});
auto threadwise_mean_inv_var_store =
ThreadwiseTensorSliceTransfer_v1r3<AccDataType,
MeanVarDataType,
decltype(thread_buffer_desc_m),
MeanVarGridDesc_M,
PassThroughOp,
ThreadBufferLengths_M,
Sequence<0>,
0,
MeanVarSrcDstVectorSize,
InMemoryDataOperationEnum::Set,
1,
true>(
mean_var_grid_desc_m,
make_multi_index(block_global_id * M_BlockTileSize +
thread_m_cluster_id * MThreadSliceSize),
PassThroughOp{});
threadwise_mean_inv_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
mean_thread_buf,
mean_var_grid_desc_m,
result_mean_global_buf);
threadwise_mean_inv_var_store.Run(thread_buffer_desc_m,
make_tuple(I0),
var_thread_buf,
mean_var_grid_desc_m,
result_inv_var_global_buf);
};
}
};
} // namespace ck
include/ck/tensor_operation/gpu/grid/gridwise_set_buffer_value.hpp
View file @ 7fa892e6
......@@ -3,6 +3,7 @@
#pragma once
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp"
namespace ck {
......
include/ck/tensor_operation/gpu/thread/threadwise_welford.hpp
View file @ 7fa892e6
......@@ -75,4 +75,63 @@ struct ThreadwiseWelford
int max_count_;
};
template <typename T,
typename SrcMeanVarCountThreadDesc_M_K,
typename DstMeanVarThreadDesc_M,
bool GetActualVariance = false>
struct ThreadwiseWelfordMerge
{
static constexpr auto src_thread_desc_m_k = SrcMeanVarCountThreadDesc_M_K{};
static constexpr auto dst_thread_desc_m = DstMeanVarThreadDesc_M{};
static constexpr auto src_length_m = src_thread_desc_m_k.GetLength(Number<0>{});
static constexpr auto src_length_k = src_thread_desc_m_k.GetLength(Number<1>{});
static constexpr auto dst_length_m = dst_thread_desc_m.GetLength(Number<0>{});
static_assert(src_length_m == dst_length_m, "lengths of source and dst buffer must match!");
__device__ static void
Merge(T& mean_a, T& var_a, int32_t& count_a, T mean_b, T var_b, int32_t count_b)
{
int count = count_a + count_b;
T count_b_over_count = count == 0 ? type_convert<T>(0) : type_convert<T>(count_b) / count;
T delta = mean_b - mean_a;
mean_a += delta * count_b_over_count;
var_a += var_b + delta * delta * count_a * count_b_over_count;
count_a = count;
}
template <typename SrcMeanBufferType,
typename SrcVarBufferType,
typename SrcCountBufferType,
typename DstMeanBufferType,
typename DstVarBufferType,
typename DstCountBufferType>
__device__ static void Run(const SrcMeanBufferType& src_mean_buf,
const SrcVarBufferType& src_var_buf,
const SrcCountBufferType& src_count_buf,
DstMeanBufferType& dst_mean_buf,
DstVarBufferType& dst_var_buf,
DstCountBufferType& dst_count_buf)
{
static_for<0, src_length_m, 1>{}([&](auto iM) {
static_for<0, src_length_k, 1>{}([&](auto iK) {
constexpr auto src_offset = src_thread_desc_m_k.CalculateOffset(make_tuple(iM, iK));
Merge(dst_mean_buf(iM),
dst_var_buf(iM),
dst_count_buf(iM),
src_mean_buf[Number<src_offset>{}],
src_var_buf[Number<src_offset>{}],
src_count_buf[Number<src_offset>{}]);
});
if constexpr(GetActualVariance)
{
dst_var_buf(iM) = dst_var_buf[iM] / dst_count_buf[iM];
};
});
};
};
} // namespace ck
library/include/ck/library/reference_tensor_operation/cpu/reference_batchnorm_forward_nhwc_c.hpp
View file @ 7fa892e6
......@@ -9,46 +9,61 @@
#include <algorithm>
#include <thread>
#include "ck/utility/math_v2.hpp"
#include "ck/utility/ignore.hpp"
#include "ck/tensor_operation/gpu/device/device_batchnorm_forward.hpp"
namespace ck {
namespace tensor_operation {
namespace host {
template <typename InOutDataType, typename AccDataType>
struct ReferenceBatchNormFwd_Input_N_H_W_C_Output_C : public device::DeviceBatchNormFwd<4, 3>
template <typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType,
typename YElementwiseOp>
struct ReferenceBatchNormFwd_Input_N_H_W_C_Output_C
: public device::DeviceBatchNormFwd<4, 3, YElementwiseOp>
{
struct Argument : public device::BaseArgument
{
Argument(const std::array<index_t, 4> xyLengths,
const std::array<index_t, 4> xStrides,
const std::array<index_t, 4> yStrides,
const std::array<int, 3> reduceDims,
const std::array<index_t, 1> bnScaleBiasMeanVarLengths,
const std::array<index_t, 1> bnScaleBiasMeanVarStrides,
const InOutDataType* p_x,
const AccDataType* bnScale,
const AccDataType* bnBias,
InOutDataType* p_y,
double exponentialAverageFactor,
AccDataType* resultRunningMean,
AccDataType* resultRunningVariance,
const std::array<index_t, 1> bnScaleStrides,
const std::array<index_t, 1> bnBiasStrides,
const std::array<index_t, 1> bnMeanVarStrides,
const XDataType* p_x,
const ScaleDataType* bnScale,
const BiasDataType* bnBias,
double epsilon,
AccDataType* resultSaveMean,
AccDataType* resultSaveInvVariance)
const YElementwiseOp y_elementwise_op,
YDataType* p_y,
MeanVarDataType* resultSaveMean,
MeanVarDataType* resultSaveInvVariance,
double averageFactor,
MeanVarDataType* resultRunningMean,
MeanVarDataType* resultRunningVariance)
: p_x_(p_x),
bnScale_(bnScale),
bnBias_(bnBias),
y_elementwise_op_(y_elementwise_op),
p_y_(p_y),
resultRunningMean_(resultRunningMean),
resultRunningVariance_(resultRunningVariance),
resultSaveMean_(resultSaveMean),
resultSaveInvVariance_(resultSaveInvVariance),
exponentialAverageFactor_(exponentialAverageFactor),
epsilon_(epsilon)
resultRunningMean_(resultRunningMean),
resultRunningVariance_(resultRunningVariance)
{
(void)xStrides;
(void)yStrides;
(void)bnScaleBiasMeanVarStrides;
ignore = xStrides;
ignore = yStrides;
ignore = bnScaleStrides;
ignore = bnBiasStrides;
ignore = bnMeanVarStrides;
ignore = reduceDims;
if(xyLengths.size() != 4 || bnScaleBiasMeanVarLengths.size() != 1 ||
bnScaleBiasMeanVarLengths[0] != xyLengths[3])
......@@ -59,26 +74,30 @@ struct ReferenceBatchNormFwd_Input_N_H_W_C_Output_C : public device::DeviceBatch
w = xyLengths[2];
c = xyLengths[3];
epsilon_ = type_convert<AccDataType>(epsilon);
averageFactor_ = type_convert<AccDataType>(averageFactor);
resultSave = (resultSaveMean != nullptr && resultSaveInvVariance != nullptr);
resultRunning = (resultRunningMean != nullptr && resultRunningVariance != nullptr);
}
const InOutDataType* p_x_;
const AccDataType* bnScale_;
const AccDataType* bnBias_;
InOutDataType* p_y_;
const XDataType* p_x_;
const ScaleDataType* bnScale_;
const BiasDataType* bnBias_;
const YElementwiseOp y_elementwise_op_;
YDataType* p_y_;
AccDataType* resultRunningMean_;
AccDataType* resultRunningVariance_;
AccDataType* resultSaveMean_;
AccDataType* resultSaveInvVariance_;
MeanVarDataType* resultSaveMean_;
MeanVarDataType* resultSaveInvVariance_;
MeanVarDataType* resultRunningMean_;
MeanVarDataType* resultRunningVariance_;
bool resultSave, resultRunning;
index_t n, h, w, c;
double exponentialAverageFactor_;
double epsilon_;
AccDataType averageFactor_;
AccDataType epsilon_;
};
struct Invoker : public device::BaseInvoker
......@@ -86,14 +105,12 @@ struct ReferenceBatchNormFwd_Input_N_H_W_C_Output_C : public device::DeviceBatch
float Run(const Argument& arg)
{
auto thread_reduce_func = [&](auto iC) {
AccDataType reduceSize = type_convert<AccDataType>(arg.n) *
type_convert<AccDataType>(arg.h) *
type_convert<AccDataType>(arg.w);
index_t offset_C = iC;
AccDataType mean = type_convert<AccDataType>(0.0f);
AccDataType meansquare = type_convert<AccDataType>(0.0f);
AccDataType variance = type_convert<AccDataType>(0.0f);
int32_t curr_count = 0;
// compute mean, meanquare, variance, invVariance
// compute mean, variance using welford method
for(index_t iN = 0; iN < arg.n; iN++)
{
index_t offset_N = iN * arg.h * arg.w * arg.c;
......@@ -106,40 +123,46 @@ struct ReferenceBatchNormFwd_Input_N_H_W_C_Output_C : public device::DeviceBatch
auto offset = offset_N + offset_H + offset_W + offset_C;
curr_count++;
AccDataType x = type_convert<AccDataType>(arg.p_x_[offset]);
mean += x;
meansquare += x * x;
AccDataType delta = x - mean;
mean += delta / curr_count;
AccDataType delta2 = x - mean;
variance += delta * delta2;
};
}
};
mean = mean / reduceSize;
meansquare = meansquare / reduceSize;
// actual variance
variance = variance / curr_count;
AccDataType variance = meansquare - mean * mean;
AccDataType invVariance =
type_convert<AccDataType>(1.0f) /
std::sqrt(type_convert<AccDataType>(arg.epsilon_) + variance);
type_convert<AccDataType>(1.0f) / ck::math::sqrt(arg.epsilon_ + variance);
// save the mean/invVariance if required
if(arg.resultSave)
{
arg.resultSaveMean_[iC] = mean;
arg.resultSaveInvVariance_[iC] = invVariance;
arg.resultSaveMean_[iC] = type_convert<MeanVarDataType>(mean);
arg.resultSaveInvVariance_[iC] = type_convert<MeanVarDataType>(invVariance);
};
// update the moving average if required
if(arg.resultRunning)
{
arg.resultRunningMean_[iC] =
arg.resultRunningMean_[iC] *
type_convert<AccDataType>(1.0 - arg.exponentialAverageFactor_) +
mean * arg.exponentialAverageFactor_;
arg.resultRunningVariance_[iC] =
arg.resultRunningVariance_[iC] *
type_convert<AccDataType>(1.0 - arg.exponentialAverageFactor_) +
variance * arg.exponentialAverageFactor_;
AccDataType oneMinusAverageFactor =
type_convert<AccDataType>(1.0) - arg.averageFactor_;
arg.resultRunningMean_[iC] = type_convert<MeanVarDataType>(
type_convert<AccDataType>(arg.resultRunningMean_[iC]) *
oneMinusAverageFactor +
mean * arg.averageFactor_);
arg.resultRunningVariance_[iC] = type_convert<MeanVarDataType>(
arg.resultRunningVariance_[iC] * oneMinusAverageFactor +
variance * arg.averageFactor_);
};
// Normalization
......@@ -160,7 +183,7 @@ struct ReferenceBatchNormFwd_Input_N_H_W_C_Output_C : public device::DeviceBatch
AccDataType norm_x =
arg.bnScale_[iC] * (x - mean) * invVariance + arg.bnBias_[iC];
arg.p_y_[offset] = type_convert<InOutDataType>(norm_x);
arg.p_y_[offset] = type_convert<YDataType>(norm_x);
};
}
};
......@@ -207,34 +230,42 @@ struct ReferenceBatchNormFwd_Input_N_H_W_C_Output_C : public device::DeviceBatch
MakeArgumentPointer(const std::array<index_t, 4> xyLengths,
const std::array<index_t, 4> xStrides,
const std::array<index_t, 4> yStrides,
const std::array<int, 3> reduceDims,
const std::array<index_t, 1> bnScaleBiasMeanVarLengths,
const std::array<index_t, 1> bnScaleBiasMeanVarStrides,
const std::array<index_t, 1> bnScaleStrides,
const std::array<index_t, 1> bnBiasStrides,
const std::array<index_t, 1> bnMeanVarStrides,
const void* p_x,
const void* bnScale,
const void* bnBias,
void* p_y,
double exponentialAverageFactor,
void* resultRunningMean,
void* resultRunningVariance,
double epsilon,
const YElementwiseOp y_elementwise_op,
void* p_y,
void* resultSaveMean,
void* resultSaveInvVariance) override
void* resultSaveInvVariance,
double averageFactor,
void* resultRunningMean,
void* resultRunningVariance) override
{
return std::make_unique<Argument>(xyLengths,
xStrides,
yStrides,
reduceDims,
bnScaleBiasMeanVarLengths,
bnScaleBiasMeanVarStrides,
static_cast<const InOutDataType*>(p_x),
static_cast<const AccDataType*>(bnScale),
static_cast<const AccDataType*>(bnBias),
static_cast<InOutDataType*>(p_y),
exponentialAverageFactor,
static_cast<AccDataType*>(resultRunningMean),
static_cast<AccDataType*>(resultRunningVariance),
bnScaleStrides,
bnBiasStrides,
bnMeanVarStrides,
static_cast<const XDataType*>(p_x),
static_cast<const ScaleDataType*>(bnScale),
static_cast<const BiasDataType*>(bnBias),
epsilon,
static_cast<AccDataType*>(resultSaveMean),
static_cast<AccDataType*>(resultSaveInvVariance));
y_elementwise_op,
static_cast<YDataType*>(p_y),
static_cast<MeanVarDataType*>(resultSaveMean),
static_cast<MeanVarDataType*>(resultSaveInvVariance),
averageFactor,
static_cast<MeanVarDataType*>(resultRunningMean),
static_cast<MeanVarDataType*>(resultRunningVariance));
};
std::unique_ptr<device::BaseInvoker> MakeInvokerPointer() override
......
library/include/ck/library/reference_tensor_operation/cpu/reference_batchnorm_infer_nhwc_c.hpp
View file @ 7fa892e6
......@@ -14,7 +14,12 @@ namespace ck {
namespace tensor_operation {
namespace host {
template <typename InOutDataType, typename AccDataType>
template <typename XDataType,
typename YDataType,
typename AccDataType,
typename ScaleDataType,
typename BiasDataType,
typename MeanVarDataType>
struct ReferenceBatchNormInfer_Input_N_H_W_C_Output_C : public device::DeviceBatchNormInfer<4, 3>
{
struct Argument : public device::BaseArgument
......@@ -23,14 +28,16 @@ struct ReferenceBatchNormInfer_Input_N_H_W_C_Output_C : public device::DeviceBat
const std::array<index_t, 4> xStrides,
const std::array<index_t, 4> yStrides,
const std::array<index_t, 1> bnScaleBiasMeanVarLengths,
const std::array<index_t, 1> bnScaleBiasMeanVarStrides,
const InOutDataType* p_x,
const AccDataType* bnScale,
const AccDataType* bnBias,
const std::array<index_t, 1> bnScaleStrides,
const std::array<index_t, 1> bnBiasStrides,
const std::array<index_t, 1> bnMeanVarStrides,
const XDataType* p_x,
const ScaleDataType* bnScale,
const BiasDataType* bnBias,
double epsilon,
const AccDataType* estimatedMean,
const AccDataType* estimatedVariance,
InOutDataType* p_y)
const MeanVarDataType* estimatedMean,
const MeanVarDataType* estimatedVariance,
YDataType* p_y)
: p_x_(p_x),
bnScale_(bnScale),
bnBias_(bnBias),
......@@ -39,32 +46,34 @@ struct ReferenceBatchNormInfer_Input_N_H_W_C_Output_C : public device::DeviceBat
estimatedVariance_(estimatedVariance),
p_y_(p_y)
{
(void)xStrides;
(void)yStrides;
(void)bnScaleBiasMeanVarStrides;
ignore = xStrides;
ignore = yStrides;
ignore = bnScaleStrides;
ignore = bnBiasStrides;
ignore = bnMeanVarStrides;
if(xyLengths.size() != 4 || bnScaleBiasMeanVarLengths.size() != 1 ||
bnScaleBiasMeanVarLengths[0] != xyLengths[3])
throw std::runtime_error("Invalid tensor dimensions!");
n = xyLengths[0];
h = xyLengths[1];
w = xyLengths[2];
c = xyLengths[3];
n_ = xyLengths[0];
h_ = xyLengths[1];
w_ = xyLengths[2];
c_ = xyLengths[3];
}
const InOutDataType* p_x_;
const AccDataType* bnScale_;
const AccDataType* bnBias_;
const XDataType* p_x_;
const ScaleDataType* bnScale_;
const BiasDataType* bnBias_;
double epsilon_;
const AccDataType* estimatedMean_;
const AccDataType* estimatedVariance_;
const MeanVarDataType* estimatedMean_;
const MeanVarDataType* estimatedVariance_;
InOutDataType* p_y_;
YDataType* p_y_;
index_t n, h, w, c;
index_t n_, h_, w_, c_;
};
struct Invoker : public device::BaseInvoker
......@@ -81,15 +90,15 @@ struct ReferenceBatchNormInfer_Input_N_H_W_C_Output_C : public device::DeviceBat
std::sqrt(type_convert<AccDataType>(arg.epsilon_) + variance);
// Normalization
for(index_t iN = 0; iN < arg.n; iN++)
for(index_t iN = 0; iN < arg.n_; iN++)
{
index_t offset_N = iN * arg.h * arg.w * arg.c;
for(index_t iH = 0; iH < arg.h; iH++)
index_t offset_N = iN * arg.h_ * arg.w_ * arg.c_;
for(index_t iH = 0; iH < arg.h_; iH++)
{
index_t offset_H = iH * arg.w * arg.c;
for(index_t iW = 0; iW < arg.w; iW++)
index_t offset_H = iH * arg.w_ * arg.c_;
for(index_t iW = 0; iW < arg.w_; iW++)
{
index_t offset_W = iW * arg.c;
index_t offset_W = iW * arg.c_;
auto offset = offset_N + offset_H + offset_W + offset_C;
......@@ -98,21 +107,21 @@ struct ReferenceBatchNormInfer_Input_N_H_W_C_Output_C : public device::DeviceBat
AccDataType norm_x =
arg.bnScale_[iC] * (x - mean) * invVariance + arg.bnBias_[iC];
arg.p_y_[offset] = type_convert<InOutDataType>(norm_x);
arg.p_y_[offset] = type_convert<YDataType>(norm_x);
};
}
};
};
std::size_t num_thread = std::thread::hardware_concurrency();
std::size_t work_per_thread = (arg.c + num_thread - 1) / num_thread;
std::size_t work_per_thread = (arg.c_ + num_thread - 1) / num_thread;
std::vector<joinable_thread> threads(num_thread);
for(std::size_t it = 0; it < num_thread; ++it)
{
std::size_t ic_begin = it * work_per_thread;
std::size_t ic_end = std::min(static_cast<int>((it + 1) * work_per_thread), arg.c);
std::size_t ic_end = std::min(static_cast<int>((it + 1) * work_per_thread), arg.c_);
auto f = [=] {
for(std::size_t ic = ic_begin; ic < ic_end; ++ic)
......@@ -146,7 +155,9 @@ struct ReferenceBatchNormInfer_Input_N_H_W_C_Output_C : public device::DeviceBat
const std::array<index_t, 4> xStrides,
const std::array<index_t, 4> yStrides,
const std::array<index_t, 1> bnScaleBiasMeanVarLengths,
const std::array<index_t, 1> bnScaleBiasMeanVarStrides,
const std::array<index_t, 1> bnScaleStrides,
const std::array<index_t, 1> bnBiasStrides,
const std::array<index_t, 1> bnMeanVarStrides,
const void* p_x,
const void* bnScale,
const void* bnBias,
......@@ -159,14 +170,16 @@ struct ReferenceBatchNormInfer_Input_N_H_W_C_Output_C : public device::DeviceBat
xStrides,
yStrides,
bnScaleBiasMeanVarLengths,
bnScaleBiasMeanVarStrides,
static_cast<const InOutDataType*>(p_x),
static_cast<const AccDataType*>(bnScale),
static_cast<const AccDataType*>(bnBias),
bnScaleStrides,
bnBiasStrides,
bnMeanVarStrides,
static_cast<const XDataType*>(p_x),
static_cast<const ScaleDataType*>(bnScale),
static_cast<const BiasDataType*>(bnBias),
epsilon,
static_cast<const AccDataType*>(estimatedMean),
static_cast<const AccDataType*>(estimatedVariance),
static_cast<InOutDataType*>(p_y));
static_cast<const MeanVarDataType*>(estimatedMean),
static_cast<const MeanVarDataType*>(estimatedVariance),
static_cast<YDataType*>(p_y));
};
std::unique_ptr<device::BaseInvoker> MakeInvokerPointer() override
......
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