Commit 211dae82 authored by ltqin's avatar ltqin
Browse files

Merge branch 'develop' into miopen_downstream_all

parents 5890e300 d5297aba
#pragma once
#include "host_tensor.hpp"
#include "gemm_common.hpp"
template <typename AType, typename BType, typename CType>
void host_gemm(const Tensor<AType>& a,
const Tensor<BType>& b,
Tensor<CType>& c,
const GemmMatrixLayout layout)
{
if(layout == GemmMatrixLayout::MK_KN_MN)
{
auto f_mk_kn_mn = [&](auto m, auto n) {
const int K = a.mDesc.GetLengths()[1];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(m, k)) * static_cast<const double>(b(k, n));
}
c(m, n) = v;
};
make_ParallelTensorFunctor(f_mk_kn_mn, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else if(layout == GemmMatrixLayout::MK_NK_MN)
{
auto f_mk_nk_mn = [&](auto m, auto n) {
const int K = a.mDesc.GetLengths()[1];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(m, k)) * static_cast<const double>(b(n, k));
}
c(m, n) = v;
};
make_ParallelTensorFunctor(f_mk_nk_mn, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else if(layout == GemmMatrixLayout::KM_KN_MN)
{
auto f_km_kn_mn = [&](auto m, auto n) {
const int K = a.mDesc.GetLengths()[0];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(k, m)) * static_cast<const double>(b(k, n));
}
c(m, n) = v;
};
make_ParallelTensorFunctor(f_km_kn_mn, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else if(layout == GemmMatrixLayout::KM_NK_MN)
{
auto f_km_nk_mn = [&](auto m, auto n) {
const int K = a.mDesc.GetLengths()[0];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(k, m)) * static_cast<const double>(b(n, k));
}
c(m, n) = v;
};
make_ParallelTensorFunctor(f_km_nk_mn, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else if(layout == GemmMatrixLayout::MK_KN_NM)
{
auto f_mk_kn_nm = [&](auto n, auto m) {
const int K = a.mDesc.GetLengths()[1];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(m, k)) * static_cast<const double>(b(k, n));
}
c(n, m) = v;
};
make_ParallelTensorFunctor(f_mk_kn_nm, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else if(layout == GemmMatrixLayout::MK_NK_NM)
{
auto f_mk_nk_nm = [&](auto n, auto m) {
const int K = a.mDesc.GetLengths()[1];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(m, k)) * static_cast<const double>(b(n, k));
}
c(n, m) = v;
};
make_ParallelTensorFunctor(f_mk_nk_nm, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else if(layout == GemmMatrixLayout::KM_KN_NM)
{
auto f_km_kn_nm = [&](auto n, auto m) {
const int K = a.mDesc.GetLengths()[0];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(k, m)) * static_cast<const double>(b(k, n));
}
c(n, m) = v;
};
make_ParallelTensorFunctor(f_km_kn_nm, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else if(layout == GemmMatrixLayout::KM_NK_NM)
{
auto f_km_nk_nm = [&](auto n, auto m) {
const int K = a.mDesc.GetLengths()[0];
double v = 0;
for(int k = 0; k < K; ++k)
{
v += static_cast<const double>(a(k, m)) * static_cast<const double>(b(n, k));
}
c(n, m) = v;
};
make_ParallelTensorFunctor(f_km_nk_nm, c.mDesc.GetLengths()[0], c.mDesc.GetLengths()[1])(
std::thread::hardware_concurrency());
}
else
{
throw std::runtime_error("wrong! not supported layout");
}
}
...@@ -15,6 +15,17 @@ struct GeneratorTensor_1 ...@@ -15,6 +15,17 @@ struct GeneratorTensor_1
} }
}; };
struct GeneratorTensor_0
{
int value = 0;
template <typename... Is>
float operator()(Is...)
{
return value;
}
};
struct GeneratorTensor_2 struct GeneratorTensor_2
{ {
int min_value = 0; int min_value = 0;
......
...@@ -9,8 +9,8 @@ struct tunable_dyn_conv_fwd_v4r4_xdlops_nchw_kcyx_nkhw ...@@ -9,8 +9,8 @@ struct tunable_dyn_conv_fwd_v4r4_xdlops_nchw_kcyx_nkhw
int NPerBlock; int NPerBlock;
int KPerBlock; int KPerBlock;
int MPerWave; int MPerXDL;
int NPerWave; int NPerXDL;
int K1; int K1;
int MRepeat; int MRepeat;
...@@ -45,8 +45,8 @@ static tunable_dyn_conv_fwd_v4r4_xdlops_nchw_kcyx_nkhw ...@@ -45,8 +45,8 @@ static tunable_dyn_conv_fwd_v4r4_xdlops_nchw_kcyx_nkhw
128, // MPerBlock, 128, // MPerBlock,
128, // NPerBlock, 128, // NPerBlock,
4, // KPerBlock, 4, // KPerBlock,
32, // MPerWave, 32, // MPerXDL,
32, // NPerWave, 32, // NPerXDL,
4, // K1, 4, // K1,
2, // MRepeat, 2, // MRepeat,
2, // NRepeat, 2, // NRepeat,
......
WORKSPACE=$1
echo "workspace: " $WORKSPACE
docker run \
-it \
--rm \
--privileged \
--group-add sudo \
-w /root/workspace \
-v $WORKSPACE:/root/workspace \
rocm/tensorflow:rocm4.3.1-tf2.6-dev \
/bin/bash
#--network host \
...@@ -4,21 +4,12 @@ ...@@ -4,21 +4,12 @@
export ROCR_VISIBLE_DEVICE=0 export ROCR_VISIBLE_DEVICE=0
export GPU_DEVICE_ORDINAL=0 export GPU_DEVICE_ORDINAL=0
## Boost
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
## Compiling
#export OLC_DEBUG_HIP_VERBOSE=1
#export OLC_DEBUG_HIP_DUMP=1
#export OLC_DEBUG_SAVE_TEMP_DIR=1
make -j conv_fwd_driver_offline make -j conv_fwd_driver_offline
make -j conv_bwd_driver_offline #make -j conv_bwd_driver_offline
make -j conv_fwd_driver_online #make -j conv_wrw_driver_offline
#make -j gemm_driver_offline
#rm -rf /root/_hip_binary_kernels_/
#rm -rf /tmp/olCompile*
DRIVER="./host/driver_offline/conv_fwd_driver_offline"
LAYOUT=$1 LAYOUT=$1
ALGO=$2 ALGO=$2
VERIFY=$3 VERIFY=$3
...@@ -26,22 +17,121 @@ INIT=$4 ...@@ -26,22 +17,121 @@ INIT=$4
LOG=$5 LOG=$5
REPEAT=$6 REPEAT=$6
################################################ layout algo verify init log repeat N__ K___ C___ Y X Hi_ Wi__ Strides Dilations LeftPads RightPads #M01=$7
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 192 3 3 71 71 2 2 1 1 1 1 1 1 #N01=$8
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 192 3 3 71 71 2 2 1 1 1 1 1 1
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 1024 1 7 17 17 1 1 1 1 0 3 0 3 KBATCH=$7
./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 14 14 1 1 1 1 1 1 1 1
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 128 3 3 14 14 1 1 1 1 1 1 1 1 ######### layout algo verify init log repeat N__ K___ C___ Y X Hi_ Wi__ Strides Dilations LeftPads RightPads
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 7 7 1 1 1 1 1 1 1 1 #$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 192 3 3 71 71 2 2 1 1 1 1 1 1
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 192 3 3 71 71 2 2 1 1 1 1 1 1
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 1024 1 7 17 17 1 1 1 1 0 3 0 3
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 14 14 1 1 1 1 1 1 1 1
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 128 3 3 14 14 1 1 1 1 1 1 1 1
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 7 7 1 1 1 1 1 1 1 1
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 192 3 3 35 35 2 2 1 1 0 0 0 0
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 30 30 2 2 1 1 0 0 0 0
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 16 16 2 2 1 1 0 0 0 0
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 2048 1024 1 1 14 14 2 2 1 1 0 0 0 0
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 1024 1 1 14 14 1 1 1 1 0 0 0 0
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 2048 1 1 7 7 1 1 1 1 0 0 0 0
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 14 14 1 1 1 1 1 1 1 1
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 128 3 3 14 14 1 1 1 1 1 1 1 1
######### layout algo verify init log repeat M___ N___ K___
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 960 1024 1024 $M01 $N01
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 1920 2048 2048 $M01 $N01
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 3840 4096 4096 $M01 $N01
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 7680 8192 8192 $M01 $N01
# Resnet50
######### layout algo verify init log repeat N__ K___ C___ Y X Hi_ Wi__ Strides Dilations LeftPads RightPads
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 2048 1024 1 1 14 14 2 2 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 1024 1 1 14 14 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 1024 1 1 14 14 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 128 128 3 3 28 28 1 1 1 1 1 1 1 1
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 128 1 1 28 28 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 128 128 3 3 58 58 2 2 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 2048 1 1 7 7 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 1024 256 1 1 14 14 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 14 14 1 1 1 1 1 1 1 1
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 30 30 2 2 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 128 256 1 1 56 56 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 256 1 1 56 56 2 2 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 64 256 1 1 56 56 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 16 16 2 2 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 1024 512 1 1 28 28 2 2 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 128 512 1 1 28 28 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 512 1 1 28 28 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 2048 512 1 1 7 7 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 7 7 1 1 1 1 1 1 1 1
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 64 1 1 56 56 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 64 64 1 1 56 56 1 1 1 1 0 0 0 0
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 64 64 3 3 56 56 1 1 1 1 1 1 1 1
# 256x128x32 c64
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 2048 1024 1 1 14 14 2 2 1 1 0 0 0 0 7
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 1024 1 1 14 14 1 1 1 1 0 0 0 0 56
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 1024 1 1 14 14 1 1 1 1 0 0 0 0 56
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 128 3 3 28 28 1 1 1 1 1 1 1 1 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 128 1 1 28 28 1 1 1 1 0 0 0 0 224
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 128 3 3 58 58 2 2 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 2048 1 1 7 7 1 1 1 1 0 0 0 0 14
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 1024 256 1 1 14 14 1 1 1 1 0 0 0 0 56
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 256 3 3 14 14 1 1 1 1 1 1 1 1 28
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 256 3 3 30 30 2 2 1 1 0 0 0 0 28
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 256 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 256 1 1 56 56 2 2 1 1 0 0 0 0 224
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 256 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 512 3 3 16 16 2 2 1 1 0 0 0 0 7
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 1024 512 1 1 28 28 2 2 1 1 0 0 0 0 56
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 512 1 1 28 28 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 512 1 1 28 28 1 1 1 1 0 0 0 0 224
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 2048 512 1 1 7 7 1 1 1 1 0 0 0 0 14
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 512 3 3 7 7 1 1 1 1 1 1 1 1 7
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 64 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 64 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 64 3 3 56 56 1 1 1 1 1 1 1 1 $KBATCH
# 128x128x32 c64
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 2048 1024 1 1 14 14 2 2 1 1 0 0 0 0 7
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 1024 1 1 14 14 1 1 1 1 0 0 0 0 56
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 1024 1 1 14 14 1 1 1 1 0 0 0 0 28
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 128 3 3 28 28 1 1 1 1 1 1 1 1 112
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 128 1 1 28 28 1 1 1 1 0 0 0 0 224
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 128 3 3 58 58 2 2 1 1 0 0 0 0 112
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 2048 1 1 7 7 1 1 1 1 0 0 0 0 14
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 1024 256 1 1 14 14 1 1 1 1 0 0 0 0 56
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 256 3 3 14 14 1 1 1 1 1 1 1 1 28
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 256 3 3 30 30 2 2 1 1 0 0 0 0 28
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 256 1 1 56 56 1 1 1 1 0 0 0 0 448
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 256 1 1 56 56 2 2 1 1 0 0 0 0 224
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 256 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 512 3 3 16 16 2 2 1 1 0 0 0 0 7
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 1024 512 1 1 28 28 2 2 1 1 0 0 0 0 28
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 512 1 1 28 28 1 1 1 1 0 0 0 0 224
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 512 1 1 28 28 1 1 1 1 0 0 0 0 112
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 2048 512 1 1 7 7 1 1 1 1 0 0 0 0 14
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 512 3 3 7 7 1 1 1 1 1 1 1 1 7
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 64 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 64 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 64 3 3 56 56 1 1 1 1 1 1 1 1 $KBATCH
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 192 3 3 35 35 2 2 1 1 0 0 0 0
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 30 30 2 2 1 1 0 0 0 0
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 16 16 2 2 1 1 0 0 0 0
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 2048 1024 1 1 14 14 2 2 1 1 0 0 0 0 # 128x64x32 c64
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 1024 1 1 14 14 1 1 1 1 0 0 0 0 #$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 64 1 1 56 56 1 1 1 1 0 0 0 0 112
#./host/driver_offline/conv_fwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 2048 1 1 7 7 1 1 1 1 0 0 0 0
#./host/driver_offline/conv_bwd_driver_offline $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 14 14 1 1 1 1 1 1 1 1 # 64x128x32 c64
$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 256 1 1 56 56 1 1 1 1 0 0 0 0 $KBATCH
#./host/driver_online/conv_fwd_driver_online $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 192 3 3 71 71 2 2 1 1 1 1 1 1 # 64x64x32 c32
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 256 1 1 56 56 1 1 1 1 0 0 0 0 112
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 64 1 1 56 56 1 1 1 1 0 0 0 0 112
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 64 1 1 56 56 1 1 1 1 0 0 0 0 448
#$DRIVER $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 64 64 3 3 56 56 1 1 1 1 1 1 1 1 448
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