Skip to content
Snippets Groups Projects
Commit 7bd9c550 authored by Grégoire Kubler's avatar Grégoire Kubler Committed by Olivier BICHLER
Browse files

chore : conv forward 1/2D formatting

parent a2485823
No related branches found
No related tags found
1 merge request!142feat_operator_convtranspose
Hide whitespace changes
Inline Side-by-side
include/aidge/backend/cpu/operator/ConvImpl.hpp
+ 1
− 0
View file @ 7bd9c550
...@@ -20,6 +20,7 @@ ...@@ -20,6 +20,7 @@
#include "aidge/utils/Types.h" #include "aidge/utils/Types.h"
namespace Aidge { namespace Aidge {
// Operator implementation entry point for the backend // Operator implementation entry point for the backend
using Conv1D_Op = Conv_Op<1>; using Conv1D_Op = Conv_Op<1>;
using ConvImpl1D_cpu = OperatorImpl_cpu<Conv_Op<1>, using ConvImpl1D_cpu = OperatorImpl_cpu<Conv_Op<1>,
......
include/aidge/backend/cpu/operator/ConvImpl_kernels.hpp
+ 165
− 84
View file @ 7bd9c550
...@@ -37,16 +37,15 @@ using std::array; ...@@ -37,16 +37,15 @@ using std::array;
* @param output_ Output Tensor. * @param output_ Output Tensor.
*/ */
template <class I, class W, class B, class O> template <class I, class W, class B, class O>
void ConvImpl1D_cpu_forward_kernel(const std::array<DimSize_t, 1>& strideDims, void ConvImpl1D_cpu_forward_kernel(const array<DimSize_t, 1> &strideDim,
const std::array<DimSize_t, 1>& dilationDims, const array<DimSize_t, 1> &dilationDim,
const std::array<DimSize_t, 1>& kernelDims, const array<DimSize_t, 1> &kernelDim,
const std::array<DimSize_t, 3>& inputDims, const std::array<DimSize_t, 3> &inputDims,
DimSize_t outChannels, DimSize_t outChannels,
const void *input_, const void *input_,
const void *weights_, const void *weights_,
const void *biases_, const void *biases_,
void *output_) void *output_) {
{
// FIXME: missing convolution attributes as arguments // FIXME: missing convolution attributes as arguments
const I *input = static_cast<const I *>(input_); const I *input = static_cast<const I *>(input_);
const W *weights = static_cast<const W *>(weights_); const W *weights = static_cast<const W *>(weights_);
...@@ -54,34 +53,51 @@ void ConvImpl1D_cpu_forward_kernel(const std::array<DimSize_t, 1>& strideDims, ...@@ -54,34 +53,51 @@ void ConvImpl1D_cpu_forward_kernel(const std::array<DimSize_t, 1>& strideDims,
O *output = static_cast<O *>(output_); O *output = static_cast<O *>(output_);
// output H size // output H size
const std::size_t oxSize = const std::size_t oxSize = static_cast<std::size_t>(std::floor(
static_cast<std::size_t>(std::floor(static_cast<float>(inputDims[2] - dilationDims[0]*(kernelDims[0] - 1) - 1 + strideDims[0]) / static_cast<float>(inputDims[2] - dilationDim[0] * (kernelDim[0] - 1) -
static_cast<float>(strideDims[0]))); 1 + strideDim[0]) /
const DimSize_t dilated_kernel_x = dilationDims[0]*(kernelDims[0] - 1) + 1; static_cast<float>(strideDim[0])));
const DimSize_t dilated_kernel_x = dilationDim[0] * (kernelDim[0] - 1) + 1;
// TODO: kernel computation
// output (batch, outCh, Xout, Yout)
// input (batch, inCh, Xin, Yin)
// weight (outCh, inCh, kernelX, kernelY)
// does not take Dilation attribute into account
using signedsize = std::make_signed<std::size_t>::type; using signedsize = std::make_signed<std::size_t>::type;
for (std::size_t batch = 0; batch < inputDims[0]; ++batch) { for (std::size_t batch = 0; batch < inputDims[0]; ++batch) {
for (std::size_t outCh = 0; outCh < outChannels; ++outCh) { for (std::size_t outCh = 0; outCh < outChannels; ++outCh) {
const std::size_t oIndex = (outCh + batch*outChannels) * oxSize; const std::size_t oIndex = (outCh + batch * outChannels) * oxSize;
// If bias = nullptr, set B(0) // If bias = nullptr, set B(0)
B biasVal = (biases != nullptr) ? biases[outCh] : B(0); B biasVal = (biases != nullptr) ? biases[outCh] : B(0);
std::fill(output + oIndex, output+(oIndex+oxSize), biasVal); std::fill(output + oIndex, output + (oIndex + oxSize), biasVal);
for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) { for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) {
const std::size_t iIndex = (inCh + batch*inputDims[1]) * inputDims[2]; const std::size_t iIndex =
const std::size_t wIndex = (inCh + outCh*inputDims[1]) * kernelDims[0]; (inCh + batch * inputDims[1]) * inputDims[2];
const std::size_t wIndex =
(inCh + outCh * inputDims[1]) * kernelDim[0];
for (std::size_t ox = 0; ox < oxSize; ++ox) { for (std::size_t ox = 0; ox < oxSize; ++ox) {
// const signedsize difx = static_cast<signedsize>(- ox * strideDims[0]); // const signedsize difx = static_cast<signedsize>(- ox *
// const std::size_t sxMin = static_cast<std::size_t>(std::max(difx, signedsize(0))); // strideDim[0s); const std::size_t sxMin =
// const std::size_t sxMax = (static_cast<signedsize>(inputDims[2]) + difx) < 0 ? 0 : ((inputDims[2] + difx) > kernelDims[0] ? kernelDims[0] : inputDims[2] + difx); // static_cast<std::size_t>(std::max(difx, signedsize(0)));
// const std::size_t sxMax =
// (static_cast<signedsize>(inputDims[2]) + difx) < 0 ? 0 :
// ((inputDims[2] + difx) > kernelDim[0s[0] ? kernelDim[0s
// : inputDims[2] + difx);
const std::size_t sxMin = 0; const std::size_t sxMin = 0;
const std::size_t sxMax = dilated_kernel_x; const std::size_t sxMax = dilated_kernel_x;
const std::size_t oIndexFull = oIndex + ox; const std::size_t oIndexFull = oIndex + ox;
const signedsize ix = static_cast<signedsize>(ox * strideDims[0]); const signedsize ix =
static_cast<signedsize>(ox * strideDim[0]);
for (std::size_t sx = sxMin; sx * dilationDim[0] < sxMax;
++sx) {
output[oIndexFull] +=
weights[wIndex + sx] *
input[iIndex + static_cast<std::size_t>(
ix + static_cast<signedsize>(
sx * dilationDim[0]))];
}
}
}
}
}
}
/** /**
* @brief perform 1D backpropagation for the data input * @brief perform 1D backpropagation for the data input
...@@ -119,9 +135,9 @@ void ConvImpl1D_cpu_forward_kernel(const std::array<DimSize_t, 1>& strideDims, ...@@ -119,9 +135,9 @@ void ConvImpl1D_cpu_forward_kernel(const std::array<DimSize_t, 1>& strideDims,
* @param[inout] iGrad gradients of the input to update * @param[inout] iGrad gradients of the input to update
*/ */
template <class I, class W, class O> template <class I, class W, class O>
void conv1DBackwardInput(const DimSize_t &stride, void conv1DBackwardInput(const array<DimSize_t, 1> &stride,
const DimSize_t &dilation, const array<DimSize_t, 1> &dilation,
const DimSize_t &kDim, const array<DimSize_t, 1> &kDim,
const array<DimSize_t, 2> &kStrides, const array<DimSize_t, 2> &kStrides,
const W *weights, const W *weights,
const array<DimSize_t, 3> &oDims, const array<DimSize_t, 3> &oDims,
...@@ -434,16 +450,15 @@ REGISTRAR(ConvImpl1D_cpu, ...@@ -434,16 +450,15 @@ REGISTRAR(ConvImpl1D_cpu,
* @param output_ Output Tensor. * @param output_ Output Tensor.
*/ */
template <class I, class W, class B, class O> template <class I, class W, class B, class O>
void ConvImpl2D_cpu_forward_kernel(const std::array<DimSize_t, 2>& strideDims, void ConvImpl2D_cpu_forward_kernel(const array<DimSize_t, 2> &strideDims,
const std::array<DimSize_t, 2>& dilationDims, const array<DimSize_t, 2> &dilationDims,
const std::array<DimSize_t, 2>& kernelDims, const array<DimSize_t, 2> &kernelDims,
const std::array<DimSize_t, 4> &inputDims, const array<DimSize_t, 4> &inputDims,
DimSize_t outChannels, DimSize_t outChannels,
const void *input_, const void *input_,
const void *weights_, const void *weights_,
const void *biases_, const void *biases_,
void *output_) void *output_) {
{
// FIXME: missing convolution attributes as arguments // FIXME: missing convolution attributes as arguments
const I *input = static_cast<const I *>(input_); const I *input = static_cast<const I *>(input_);
const W *weights = static_cast<const W *>(weights_); const W *weights = static_cast<const W *>(weights_);
...@@ -451,59 +466,102 @@ void ConvImpl2D_cpu_forward_kernel(const std::array<DimSize_t, 2>& strideDims, ...@@ -451,59 +466,102 @@ void ConvImpl2D_cpu_forward_kernel(const std::array<DimSize_t, 2>& strideDims,
O *output = static_cast<O *>(output_); O *output = static_cast<O *>(output_);
// output H size // output H size
const DimSize_t dilated_kernel_x = dilationDims[0]*(kernelDims[0] - 1) + 1; const DimSize_t dilated_kernel_x =
const std::size_t oxSize = dilationDims[0] * (kernelDims[0] - 1) + 1;
static_cast<std::size_t>(std::floor(static_cast<float>(inputDims[2] - dilated_kernel_x + strideDims[0]) / const std::size_t oxSize = static_cast<std::size_t>(std::floor(
static_cast<float>(strideDims[0]))); static_cast<float>(inputDims[2] - dilated_kernel_x + strideDims[0]) /
static_cast<float>(strideDims[0])));
// output W size // output W size
const DimSize_t dilated_kernel_y = dilationDims[1]*(kernelDims[1] - 1) + 1; const DimSize_t dilated_kernel_y =
const std::size_t oySize = dilationDims[1] * (kernelDims[1] - 1) + 1;
static_cast<std::size_t>(std::floor(static_cast<float>(inputDims[3] - dilated_kernel_y + strideDims[1]) / const std::size_t oySize = static_cast<std::size_t>(std::floor(
static_cast<float>(strideDims[1]))); static_cast<float>(inputDims[3] - dilated_kernel_y + strideDims[1]) /
static_cast<float>(strideDims[1])));
// TODO: kernel computation // TODO: kernel computation
// output (batch, outCh, Xout, Yout) // output (batch, outCh, Xout, Yout)
// input (batch, inCh, Xin, Yin) // input (batch, inCh, Xin, Yin)
// weight (outCh, inCh, kernelX, kernelY) // weight (outCh, inCh, kernelX, kernelY)
// does not take Dilation attribute into account // does not take Dilation attribute into account
const std::size_t outChannels_s = oxSize * oySize; const std::size_t outChannels_s = oxSize * oySize;
if (dilated_kernel_x == 3 && dilated_kernel_y == 3) { if (dilated_kernel_x == 3 && dilated_kernel_y == 3) {
for (std::size_t batch = 0; batch < inputDims[0]; ++batch) { for (std::size_t batch = 0; batch < inputDims[0]; ++batch) {
for (std::size_t outCh = 0; outCh < outChannels; ++outCh) { for (std::size_t outCh = 0; outCh < outChannels; ++outCh) {
// If bias = nullptr, set B(0) // If bias = nullptr, set B(0)
B biasVal = (biases != nullptr) ? biases[outCh] : B(0); B biasVal = (biases != nullptr) ? biases[outCh] : B(0);
std::fill(output, output+outChannels_s, biasVal); std::fill(output, output + outChannels_s, biasVal);
for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) { for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) {
std::size_t iIndex = (inCh + batch*inputDims[1]) * inputDims[2] * inputDims[3]; std::size_t iIndex = (inCh + batch * inputDims[1]) *
const std::size_t wIndex = (inCh + outCh*inputDims[1]) * 9; inputDims[2] * inputDims[3];
if (strideDims[0] == 1 && strideDims[1]==1) { const std::size_t wIndex =
for (std::size_t ox = 0, oIndex = 0; ox < oxSize; ++ox, oIndex+=oySize, iIndex-=inputDims[3]) { (inCh + outCh * inputDims[1]) * 9;
if (strideDims[0] == 1 && strideDims[1] == 1) {
for (std::size_t ox = 0, oIndex = 0; ox < oxSize;
++ox, oIndex += oySize, iIndex -= inputDims[3]) {
for (std::size_t oy = 0; oy < oySize; ++oy) { for (std::size_t oy = 0; oy < oySize; ++oy) {
output[oIndex + oy] += weights[wIndex+0]*input[iIndex+oy]+weights[wIndex+1]*input[iIndex+oy+1]+weights[wIndex+2]*input[iIndex+oy+2]; output[oIndex + oy] +=
weights[wIndex + 0] * input[iIndex + oy] +
weights[wIndex + 1] *
input[iIndex + oy + 1] +
weights[wIndex + 2] *
input[iIndex + oy + 2];
} }
iIndex+=inputDims[3]; iIndex += inputDims[3];
for (std::size_t oy = 0; oy < oySize; ++oy) { for (std::size_t oy = 0; oy < oySize; ++oy) {
output[oIndex + oy] += weights[wIndex+3]*input[iIndex+oy]+weights[wIndex+4]*input[iIndex+oy+1]+weights[wIndex+5]*input[iIndex+oy+2]; output[oIndex + oy] +=
weights[wIndex + 3] * input[iIndex + oy] +
weights[wIndex + 4] *
input[iIndex + oy + 1] +
weights[wIndex + 5] *
input[iIndex + oy + 2];
} }
iIndex+=inputDims[3]; iIndex += inputDims[3];
for (std::size_t oy = 0; oy < oySize; ++oy) { for (std::size_t oy = 0; oy < oySize; ++oy) {
output[oIndex + oy] += weights[wIndex+6]*input[iIndex+oy]+weights[wIndex+7]*input[iIndex+oy+1]+weights[wIndex+8]*input[iIndex+oy+2]; output[oIndex + oy] +=
weights[wIndex + 6] * input[iIndex + oy] +
weights[wIndex + 7] *
input[iIndex + oy + 1] +
weights[wIndex + 8] *
input[iIndex + oy + 2];
} }
} }
} else { } else {
for (std::size_t ox = 0, oIndex = 0; ox < oxSize; ++ox, oIndex+=oySize, iIndex+=(strideDims[0]-2)*inputDims[3]) { for (std::size_t ox = 0, oIndex = 0; ox < oxSize; ++ox,
oIndex += oySize,
iIndex += (strideDims[0] -
2) * inputDims[3]) {
for (std::size_t oy = 0; oy < oySize; ++oy) { for (std::size_t oy = 0; oy < oySize; ++oy) {
output[oIndex + oy] += weights[wIndex+0]*input[iIndex+oy*strideDims[1]]+weights[wIndex+1]*input[iIndex+oy*strideDims[1]+1]+weights[wIndex+2]*input[iIndex+oy*strideDims[1]+2]; output[oIndex + oy] +=
weights[wIndex + 0] *
input[iIndex + oy * strideDims[1]] +
weights[wIndex + 1] *
input[iIndex + oy * strideDims[1] +
1] +
weights[wIndex + 2] *
input[iIndex + oy * strideDims[1] + 2];
} }
iIndex+=inputDims[3]; iIndex += inputDims[3];
for (std::size_t oy = 0; oy < oySize; ++oy) { for (std::size_t oy = 0; oy < oySize; ++oy) {
output[oIndex + oy] += weights[wIndex+3]*input[iIndex+oy*strideDims[1]]+weights[wIndex+4]*input[iIndex+oy*strideDims[1]+1]+weights[wIndex+5]*input[iIndex+oy*strideDims[1]+2]; output[oIndex + oy] +=
weights[wIndex + 3] *
input[iIndex + oy * strideDims[1]] +
weights[wIndex + 4] *
input[iIndex + oy * strideDims[1] +
1] +
weights[wIndex + 5] *
input[iIndex + oy * strideDims[1] + 2];
} }
iIndex+=inputDims[3]; iIndex += inputDims[3];
for (std::size_t oy = 0; oy < oySize; ++oy) { for (std::size_t oy = 0; oy < oySize; ++oy) {
output[oIndex + oy] += weights[wIndex+6]*input[iIndex+oy*strideDims[1]]+weights[wIndex+7]*input[iIndex+oy*strideDims[1]+1]+weights[wIndex+8]*input[iIndex+oy*strideDims[1]+2]; output[oIndex + oy] +=
weights[wIndex + 6] *
input[iIndex + oy * strideDims[1]] +
weights[wIndex + 7] *
input[iIndex + oy * strideDims[1] +
1] +
weights[wIndex + 8] *
input[iIndex + oy * strideDims[1] + 2];
} }
} }
} }
...@@ -516,18 +574,26 @@ void ConvImpl2D_cpu_forward_kernel(const std::array<DimSize_t, 2>& strideDims, ...@@ -516,18 +574,26 @@ void ConvImpl2D_cpu_forward_kernel(const std::array<DimSize_t, 2>& strideDims,
for (std::size_t outCh = 0; outCh < outChannels; ++outCh) { for (std::size_t outCh = 0; outCh < outChannels; ++outCh) {
// If bias = nullptr, set B(0) // If bias = nullptr, set B(0)
B biasVal = (biases != nullptr) ? biases[outCh] : B(0); B biasVal = (biases != nullptr) ? biases[outCh] : B(0);
std::fill(output, output+outChannels_s, biasVal); std::fill(output, output + outChannels_s, biasVal);
for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) { for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) {
std::size_t iIndex = (inCh + batch*inputDims[1]) * inputDims[2] * inputDims[3]; std::size_t iIndex = (inCh + batch * inputDims[1]) *
const std::size_t wIndex = (inCh + outCh*inputDims[1]); inputDims[2] * inputDims[3];
const std::size_t wIndex = (inCh + outCh * inputDims[1]);
if (strideDims[0] == 1 && strideDims[1] == 1) { if (strideDims[0] == 1 && strideDims[1] == 1) {
for (std::size_t oIndex = 0; oIndex < oxSize*oySize; ++oIndex, ++iIndex) { for (std::size_t oIndex = 0; oIndex < oxSize * oySize;
++oIndex, ++iIndex) {
output[oIndex] += weights[wIndex] * input[iIndex]; output[oIndex] += weights[wIndex] * input[iIndex];
} }
} else { } else {
for (std::size_t ox = 0, oIndex = 0; ox < oxSize; ++ox, oIndex+=oySize, iIndex+=inputDims[3]*strideDims[0]) { for (std::size_t ox = 0, oIndex = 0; ox < oxSize;
for (std::size_t oy = 0, iy = 0; oy < oySize; ++oy, iy+=strideDims[1]) { ++ox,
output[oIndex + oy] += weights[wIndex+0]*input[iIndex+iy]; oIndex += oySize,
iIndex +=
inputDims[3] * strideDims[0]) {
for (std::size_t oy = 0, iy = 0; oy < oySize;
++oy, iy += strideDims[1]) {
output[oIndex + oy] +=
weights[wIndex + 0] * input[iIndex + iy];
} }
} }
} }
...@@ -540,21 +606,36 @@ void ConvImpl2D_cpu_forward_kernel(const std::array<DimSize_t, 2>& strideDims, ...@@ -540,21 +606,36 @@ void ConvImpl2D_cpu_forward_kernel(const std::array<DimSize_t, 2>& strideDims,
for (std::size_t outCh = 0; outCh < outChannels; ++outCh) { for (std::size_t outCh = 0; outCh < outChannels; ++outCh) {
// If bias = nullptr, set B(0) // If bias = nullptr, set B(0)
B biasVal = (biases != nullptr) ? biases[outCh] : B(0); B biasVal = (biases != nullptr) ? biases[outCh] : B(0);
std::fill(output, output+outChannels_s, biasVal); std::fill(output, output + outChannels_s, biasVal);
for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) { for (std::size_t inCh = 0; inCh < inputDims[1]; ++inCh) {
std::size_t iIndex_channel = (inCh + batch*inputDims[1]) * inputDims[2] * inputDims[3]; std::size_t iIndex_channel =
const std::size_t wIndex = (inCh + outCh*inputDims[1]) * kernelDims[0] * kernelDims[1]; (inCh + batch * inputDims[1]) * inputDims[2] *
inputDims[3];
const std::size_t wIndex = (inCh + outCh * inputDims[1]) *
kernelDims[0] * kernelDims[1];
// loop over each ouput line // loop over each ouput line
for (std::size_t ox = 0, oIndex = 0; ox < oxSize; ++ox, oIndex+=oySize, iIndex_channel+=inputDims[3]*strideDims[0]) { for (std::size_t ox = 0, oIndex = 0; ox < oxSize;
++ox,
oIndex += oySize,
iIndex_channel +=
inputDims[3] * strideDims[0]) {
// loop over associated input line // loop over associated input line
for (std::size_t ky = 0, ix = 0; ky < kernelDims[0]; ++ky, ix += inputDims[3]*dilationDims[0]) { for (std::size_t ky = 0, ix = 0; ky < kernelDims[0];
++ky, ix += inputDims[3] * dilationDims[0]) {
// loop over the entire line // loop over the entire line
for (std::size_t oy = 0, iy = 0; oy < oySize; ++oy, iy+=strideDims[1]) { for (std::size_t oy = 0, iy = 0; oy < oySize;
const std::size_t iIndex = iIndex_channel + ix + iy; ++oy, iy += strideDims[1]) {
// loop over elements assosicated with one output const std::size_t iIndex =
for (std::size_t kx = 0; kx < kernelDims[0]; ++kx) { iIndex_channel + ix + iy;
output[oIndex + oy] += weights[wIndex+kernelDims[0]*ky+kx]*input[iIndex+kx*dilationDims[1]]; // loop over elements assosicated with one
// output
for (std::size_t kx = 0; kx < kernelDims[0];
++kx) {
output[oIndex + oy] +=
weights[wIndex + kernelDims[0] * ky +
kx] *
input[iIndex + kx * dilationDims[1]];
} }
} }
} }
......
src/operator/ConvImpl.cpp
+ 11
− 10
View file @ 7bd9c550
...@@ -40,16 +40,17 @@ void Aidge::ConvImpl1D_cpu::forward() { ...@@ -40,16 +40,17 @@ void Aidge::ConvImpl1D_cpu::forward() {
const auto& input2 = (op_.getInput(2)) ? op_.getInput(2)->refCastFrom(input2Fallback, *op_.getOutput(0)) : Tensor(); const auto& input2 = (op_.getInput(2)) ? op_.getInput(2)->refCastFrom(input2Fallback, *op_.getOutput(0)) : Tensor();
// Call kernel // Call kernel
impl.forward(op_.strideDims(), impl.forward(
op_.dilationDims(), op_.strideDims(),
op_.kernelDims(), op_.dilationDims(),
op_.getInput(0)->template dims<3>(), // input dimensions op_.kernelDims(),
dynamic_cast<const Conv_Op<1>&>(mOp).outChannels(), // outChannels op_.getInput(0)->template dims<3>(), // input dimensions
input0.getImpl()->rawPtr(), // input dynamic_cast<const Conv_Op<1> &>(mOp).outChannels(), // outChannels
input1.getImpl()->rawPtr(), // weight input0.getImpl()->rawPtr(), // input
op_.getInput(2) ? input2.getImpl()->rawPtr() : nullptr, // bias input1.getImpl()->rawPtr(), // weight
getCPUPtr(mOp.getRawOutput(0)) // output op_.getInput(2) ? input2.getImpl()->rawPtr() : nullptr, // bias
); getCPUPtr(mOp.getRawOutput(0)) // output
);
} }
template <> void ConvImpl1D_cpu::backward() { template <> void ConvImpl1D_cpu::backward() {
......
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

Copyright © Eclipse Foundation, Inc. All Rights Reserved.     Privacy Policy | Terms of Use | Copyright Agent