Safran/PyTorchEMD
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity

🎨 autoformatting cpp/cu files

Browse source
This commit is contained in:
Laurent FAINSIN 2023-04-24 11:53:37 +02:00
parent 2375a23ca5
commit 892b79f5c5
2 changed files with 368 additions and 333 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
cuda/emd.cpp
View file

@ -20,7 +20,8 @@ std::vector<at::Tensor> MatchCostBackward(
const at::Tensor xyz2,
const at::Tensor match);
PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
PYBIND11_MODULE(TORCH_EXTENSION_NAME, m)
{
m.def("approxmatch_forward", &ApproxMatchForward, "ApproxMatch forward (CUDA)");
m.def("matchcost_forward", &MatchCostForward, "MatchCost forward (CUDA)");
m.def("matchcost_backward", &MatchCostBackward, "MatchCost backward (CUDA)");

182
cuda/emd_kernel.cu
View file

@ -14,27 +14,37 @@
#define CHECK_CUDA(x) TORCH_CHECK(x.type().is_cuda(), #x " must be a CUDA tensor")
#define CHECK_CONTIGUOUS(x) TORCH_CHECK(x.is_contiguous(), #x " must be contiguous")
#define CHECK_INPUT(x) CHECK_CUDA(x); CHECK_CONTIGUOUS(x)
#define CHECK_INPUT(x) \
CHECK_CUDA(x); \
CHECK_CONTIGUOUS(x)
/********************************
* Forward kernel for approxmatch
*********************************/
template <typename scalar_t>
__global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1,const scalar_t * __restrict__ xyz2,scalar_t * __restrict__ match,scalar_t * temp){
scalar_t * remainL=temp+blockIdx.x*(n+m)*2, * remainR=temp+blockIdx.x*(n+m)*2+n,*ratioL=temp+blockIdx.x*(n+m)*2+n+m,*ratioR=temp+blockIdx.x*(n+m)*2+n+m+n;
__global__ void approxmatch(int b, int n, int m, const scalar_t *__restrict__ xyz1, const scalar_t *__restrict__ xyz2, scalar_t *__restrict__ match, scalar_t *temp)
{
scalar_t *remainL = temp + blockIdx.x * (n + m) * 2;
scalar_t *remainR = temp + blockIdx.x * (n + m) * 2 + n;
scalar_t *ratioL = temp + blockIdx.x * (n + m) * 2 + n + m;
scalar_t *ratioR = temp + blockIdx.x * (n + m) * 2 + n + m + n;
scalar_t multiL, multiR;
if (n>=m){
if (n >= m)
{
multiL = 1;
multiR = n / m;
}else{
}
else
{
multiL = m / n;
multiR = 1;
}
const int Block = 1024;
__shared__ scalar_t buf[Block * 4];
for (int i=blockIdx.x;i<b;i+=gridDim.x){
for (int i = blockIdx.x; i < b; i += gridDim.x)
{
for (int j = threadIdx.x; j < n * m; j += blockDim.x)
match[i * n * m + j] = 0;
for (int j = threadIdx.x; j < n; j += blockDim.x)
@ -42,23 +52,29 @@ __global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1
for (int j = threadIdx.x; j < m; j += blockDim.x)
remainR[j] = multiR;
__syncthreads();
for (int j=7;j>=-2;j--){
for (int j = 7; j >= -2; j--)
{
scalar_t level = -powf(4.0f, j);
if (j==-2){
if (j == -2)
{
level = 0;
}
for (int k0=0;k0<n;k0+=blockDim.x){
for (int k0 = 0; k0 < n; k0 += blockDim.x)
{
int k = k0 + threadIdx.x;
scalar_t x1 = 0, y1 = 0, z1 = 0;
if (k<n){
if (k < n)
{
x1 = xyz1[i * n * 3 + k * 3 + 0];
y1 = xyz1[i * n * 3 + k * 3 + 1];
z1 = xyz1[i * n * 3 + k * 3 + 2];
}
scalar_t suml = 1e-9f;
for (int l0=0;l0<m;l0+=Block){
for (int l0 = 0; l0 < m; l0 += Block)
{
int lend = min(m, l0 + Block) - l0;
for (int l=threadIdx.x;l<lend;l+=blockDim.x){
for (int l = threadIdx.x; l < lend; l += blockDim.x)
{
scalar_t x2 = xyz2[i * m * 3 + l0 * 3 + l * 3 + 0];
scalar_t y2 = xyz2[i * m * 3 + l0 * 3 + l * 3 + 1];
scalar_t z2 = xyz2[i * m * 3 + l0 * 3 + l * 3 + 2];
@ -68,7 +84,8 @@ __global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1
buf[l * 4 + 3] = remainR[l0 + l];
}
__syncthreads();
for (int l=0;l<lend;l++){
for (int l = 0; l < lend; l++)
{
scalar_t x2 = buf[l * 4 + 0];
scalar_t y2 = buf[l * 4 + 1];
scalar_t z2 = buf[l * 4 + 2];
@ -82,25 +99,30 @@ __global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1
ratioL[k] = remainL[k] / suml;
}
__syncthreads();
for (int l0=0;l0<m;l0+=blockDim.x){
for (int l0 = 0; l0 < m; l0 += blockDim.x)
{
int l = l0 + threadIdx.x;
scalar_t x2 = 0, y2 = 0, z2 = 0;
if (l<m){
if (l < m)
{
x2 = xyz2[i * m * 3 + l * 3 + 0];
y2 = xyz2[i * m * 3 + l * 3 + 1];
z2 = xyz2[i * m * 3 + l * 3 + 2];
}
scalar_t sumr = 0;
for (int k0=0;k0<n;k0+=Block){
for (int k0 = 0; k0 < n; k0 += Block)
{
int kend = min(n, k0 + Block) - k0;
for (int k=threadIdx.x;k<kend;k+=blockDim.x){
for (int k = threadIdx.x; k < kend; k += blockDim.x)
{
buf[k * 4 + 0] = xyz1[i * n * 3 + k0 * 3 + k * 3 + 0];
buf[k * 4 + 1] = xyz1[i * n * 3 + k0 * 3 + k * 3 + 1];
buf[k * 4 + 2] = xyz1[i * n * 3 + k0 * 3 + k * 3 + 2];
buf[k * 4 + 3] = ratioL[k0 + k];
}
__syncthreads();
for (int k=0;k<kend;k++){
for (int k = 0; k < kend; k++)
{
scalar_t x1 = buf[k * 4 + 0];
scalar_t y1 = buf[k * 4 + 1];
scalar_t z1 = buf[k * 4 + 2];
@ -109,7 +131,8 @@ __global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1
}
__syncthreads();
}
if (l<m){
if (l < m)
{
sumr *= remainR[l];
scalar_t consumption = fminf(remainR[l] / (sumr + 1e-9f), 1.0f);
ratioR[l] = consumption * remainR[l];
@ -117,18 +140,22 @@ __global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1
}
}
__syncthreads();
for (int k0=0;k0<n;k0+=blockDim.x){
for (int k0 = 0; k0 < n; k0 += blockDim.x)
{
int k = k0 + threadIdx.x;
scalar_t x1 = 0, y1 = 0, z1 = 0;
if (k<n){
if (k < n)
{
x1 = xyz1[i * n * 3 + k * 3 + 0];
y1 = xyz1[i * n * 3 + k * 3 + 1];
z1 = xyz1[i * n * 3 + k * 3 + 2];
}
scalar_t suml = 0;
for (int l0=0;l0<m;l0+=Block){
for (int l0 = 0; l0 < m; l0 += Block)
{
int lend = min(m, l0 + Block) - l0;
for (int l=threadIdx.x;l<lend;l+=blockDim.x){
for (int l = threadIdx.x; l < lend; l += blockDim.x)
{
buf[l * 4 + 0] = xyz2[i * m * 3 + l0 * 3 + l * 3 + 0];
buf[l * 4 + 1] = xyz2[i * m * 3 + l0 * 3 + l * 3 + 1];
buf[l * 4 + 2] = xyz2[i * m * 3 + l0 * 3 + l * 3 + 2];
@ -136,8 +163,10 @@ __global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1
}
__syncthreads();
scalar_t rl = ratioL[k];
if (k<n){
for (int l=0;l<lend;l++){
if (k < n)
{
for (int l = 0; l < lend; l++)
{
scalar_t x2 = buf[l * 4 + 0];
scalar_t y2 = buf[l * 4 + 1];
scalar_t z2 = buf[l * 4 + 2];
@ -156,10 +185,6 @@ __global__ void approxmatch(int b,int n,int m,const scalar_t * __restrict__ xyz1
}
}
//void approxmatchLauncher(int b,int n,int m,const scalar_t * xyz1,const scalar_t * xyz2,scalar_t * match,scalar_t * temp){
// approxmatch<<<32,512>>>(b,n,m,xyz1,xyz2,match,temp);
//}
/* ApproxMatch forward interface
Input:
xyz1: (B, N1, 3) # dataset_points
@ -169,7 +194,8 @@ Output:
*/
at::Tensor ApproxMatchForward(
const at::Tensor xyz1,
const at::Tensor xyz2){
const at::Tensor xyz2)
{
const auto b = xyz1.size(0);
const auto n = xyz1.size(1);
const auto m = xyz2.size(1);
@ -183,41 +209,46 @@ at::Tensor ApproxMatchForward(
auto match = at::zeros({b, m, n}, xyz1.type());
auto temp = at::zeros({b, (n + m) * 2}, xyz1.type());
AT_DISPATCH_FLOATING_TYPES(xyz1.scalar_type(), "ApproxMatchForward", ([&] {
approxmatch<scalar_t><<<32,512>>>(b, n, m, xyz1.data<scalar_t>(), xyz2.data<scalar_t>(), match.data<scalar_t>(), temp.data<scalar_t>());
}));
AT_DISPATCH_FLOATING_TYPES(xyz1.scalar_type(), "ApproxMatchForward", ([&]
{ approxmatch<scalar_t><<<32, 512>>>(b, n, m, xyz1.data<scalar_t>(), xyz2.data<scalar_t>(), match.data<scalar_t>(), temp.data<scalar_t>()); }));
C10_CUDA_CHECK(cudaGetLastError());
return match;
}
/********************************
* Forward kernel for matchcost
*********************************/
template <typename scalar_t>
__global__ void matchcost(int b,int n,int m,const scalar_t * __restrict__ xyz1,const scalar_t * __restrict__ xyz2,const scalar_t * __restrict__ match,scalar_t * __restrict__ out){
__global__ void matchcost(int b, int n, int m, const scalar_t *__restrict__ xyz1, const scalar_t *__restrict__ xyz2, const scalar_t *__restrict__ match, scalar_t *__restrict__ out)
{
__shared__ scalar_t allsum[512];
const int Block = 1024;
__shared__ scalar_t buf[Block * 3];
for (int i=blockIdx.x;i<b;i+=gridDim.x){
for (int i = blockIdx.x; i < b; i += gridDim.x)
{
scalar_t subsum = 0;
for (int k0=0;k0<n;k0+=blockDim.x){
for (int k0 = 0; k0 < n; k0 += blockDim.x)
{
int k = k0 + threadIdx.x;
scalar_t x1 = 0, y1 = 0, z1 = 0;
if (k<n){
if (k < n)
{
x1 = xyz1[i * n * 3 + k * 3 + 0];
y1 = xyz1[i * n * 3 + k * 3 + 1];
z1 = xyz1[i * n * 3 + k * 3 + 2];
}
for (int l0=0;l0<m;l0+=Block){
for (int l0 = 0; l0 < m; l0 += Block)
{
int lend = min(m, l0 + Block) - l0;
for (int l = threadIdx.x; l < lend * 3; l += blockDim.x)
buf[l] = xyz2[i * m * 3 + l0 * 3 + l];
__syncthreads();
if (k<n){
for (int l=0;l<lend;l++){
if (k < n)
{
for (int l = 0; l < lend; l++)
{
scalar_t x2 = buf[l * 3 + 0];
scalar_t y2 = buf[l * 3 + 1];
scalar_t z2 = buf[l * 3 + 2];
@ -229,9 +260,11 @@ __global__ void matchcost(int b,int n,int m,const scalar_t * __restrict__ xyz1,c
}
}
allsum[threadIdx.x] = subsum;
for (int j=1;j<blockDim.x;j<<=1){
for (int j = 1; j < blockDim.x; j <<= 1)
{
__syncthreads();
if ((threadIdx.x&j)==0 && threadIdx.x+j<blockDim.x){
if ((threadIdx.x & j) == 0 && threadIdx.x + j < blockDim.x)
{
allsum[threadIdx.x] += allsum[threadIdx.x + j];
}
}
@ -241,10 +274,6 @@ __global__ void matchcost(int b,int n,int m,const scalar_t * __restrict__ xyz1,c
}
}
//void matchcostLauncher(int b,int n,int m,const scalar_t * xyz1,const scalar_t * xyz2,const scalar_t * match,scalar_t * out){
// matchcost<<<32,512>>>(b,n,m,xyz1,xyz2,match,out);
//}
/* MatchCost forward interface
Input:
xyz1: (B, N1, 3) # dataset_points
@ -256,7 +285,8 @@ Output:
at::Tensor MatchCostForward(
const at::Tensor xyz1,
const at::Tensor xyz2,
const at::Tensor match){
const at::Tensor match)
{
const auto b = xyz1.size(0);
const auto n = xyz1.size(1);
const auto m = xyz2.size(1);
@ -269,31 +299,33 @@ at::Tensor MatchCostForward(
auto cost = at::zeros({b}, xyz1.type());
AT_DISPATCH_FLOATING_TYPES(xyz1.scalar_type(), "MatchCostForward", ([&] {
matchcost<scalar_t><<<32,512>>>(b, n, m, xyz1.data<scalar_t>(), xyz2.data<scalar_t>(), match.data<scalar_t>(), cost.data<scalar_t>());
}));
AT_DISPATCH_FLOATING_TYPES(xyz1.scalar_type(), "MatchCostForward", ([&]
{ matchcost<scalar_t><<<32, 512>>>(b, n, m, xyz1.data<scalar_t>(), xyz2.data<scalar_t>(), match.data<scalar_t>(), cost.data<scalar_t>()); }));
C10_CUDA_CHECK(cudaGetLastError());
return cost;
}
/********************************
* matchcostgrad2 kernel
*********************************/
template <typename scalar_t>
__global__ void matchcostgrad2(int b,int n,int m,const scalar_t * __restrict__ grad_cost,const scalar_t * __restrict__ xyz1,const scalar_t * __restrict__ xyz2,const scalar_t * __restrict__ match,scalar_t * __restrict__ grad2){
__global__ void matchcostgrad2(int b, int n, int m, const scalar_t *__restrict__ grad_cost, const scalar_t *__restrict__ xyz1, const scalar_t *__restrict__ xyz2, const scalar_t *__restrict__ match, scalar_t *__restrict__ grad2)
{
__shared__ scalar_t sum_grad[256 * 3];
for (int i=blockIdx.x;i<b;i+=gridDim.x){
for (int i = blockIdx.x; i < b; i += gridDim.x)
{
int kbeg = m * blockIdx.y / gridDim.y;
int kend = m * (blockIdx.y + 1) / gridDim.y;
for (int k=kbeg;k<kend;k++){
for (int k = kbeg; k < kend; k++)
{
scalar_t x2 = xyz2[(i * m + k) * 3 + 0];
scalar_t y2 = xyz2[(i * m + k) * 3 + 1];
scalar_t z2 = xyz2[(i * m + k) * 3 + 2];
scalar_t subsumx = 0, subsumy = 0, subsumz = 0;
for (int j=threadIdx.x;j<n;j+=blockDim.x){
for (int j = threadIdx.x; j < n; j += blockDim.x)
{
scalar_t x1 = x2 - xyz1[(i * n + j) * 3 + 0];
scalar_t y1 = y2 - xyz1[(i * n + j) * 3 + 1];
scalar_t z1 = z2 - xyz1[(i * n + j) * 3 + 2];
@ -305,17 +337,20 @@ __global__ void matchcostgrad2(int b,int n,int m,const scalar_t * __restrict__ g
sum_grad[threadIdx.x * 3 + 0] = subsumx;
sum_grad[threadIdx.x * 3 + 1] = subsumy;
sum_grad[threadIdx.x * 3 + 2] = subsumz;
for (int j=1;j<blockDim.x;j<<=1){
for (int j = 1; j < blockDim.x; j <<= 1)
{
__syncthreads();
int j1 = threadIdx.x;
int j2 = threadIdx.x + j;
if ((j1&j)==0 && j2<blockDim.x){
if ((j1 & j) == 0 && j2 < blockDim.x)
{
sum_grad[j1 * 3 + 0] += sum_grad[j2 * 3 + 0];
sum_grad[j1 * 3 + 1] += sum_grad[j2 * 3 + 1];
sum_grad[j1 * 3 + 2] += sum_grad[j2 * 3 + 2];
}
}
if (threadIdx.x==0){
if (threadIdx.x == 0)
{
grad2[(i * m + k) * 3 + 0] = sum_grad[0] * grad_cost[i];
grad2[(i * m + k) * 3 + 1] = sum_grad[1] * grad_cost[i];
grad2[(i * m + k) * 3 + 2] = sum_grad[2] * grad_cost[i];
@ -330,14 +365,18 @@ __global__ void matchcostgrad2(int b,int n,int m,const scalar_t * __restrict__ g
*********************************/
template <typename scalar_t>
__global__ void matchcostgrad1(int b,int n,int m,const scalar_t * __restrict__ grad_cost,const scalar_t * __restrict__ xyz1,const scalar_t * __restrict__ xyz2,const scalar_t * __restrict__ match,scalar_t * __restrict__ grad1){
for (int i=blockIdx.x;i<b;i+=gridDim.x){
for (int l=threadIdx.x;l<n;l+=blockDim.x){
__global__ void matchcostgrad1(int b, int n, int m, const scalar_t *__restrict__ grad_cost, const scalar_t *__restrict__ xyz1, const scalar_t *__restrict__ xyz2, const scalar_t *__restrict__ match, scalar_t *__restrict__ grad1)
{
for (int i = blockIdx.x; i < b; i += gridDim.x)
{
for (int l = threadIdx.x; l < n; l += blockDim.x)
{
scalar_t x1 = xyz1[i * n * 3 + l * 3 + 0];
scalar_t y1 = xyz1[i * n * 3 + l * 3 + 1];
scalar_t z1 = xyz1[i * n * 3 + l * 3 + 2];
scalar_t dx = 0, dy = 0, dz = 0;
for (int k=0;k<m;k++){
for (int k = 0; k < m; k++)
{
scalar_t x2 = xyz2[i * m * 3 + k * 3 + 0];
scalar_t y2 = xyz2[i * m * 3 + k * 3 + 1];
scalar_t z2 = xyz2[i * m * 3 + k * 3 + 2];
@ -353,12 +392,6 @@ __global__ void matchcostgrad1(int b,int n,int m,const scalar_t * __restrict__ g
}
}
//void matchcostgradLauncher(int b,int n,int m,const scalar_t * xyz1,const scalar_t * xyz2,const scalar_t * match,scalar_t * grad1,scalar_t * grad2){
// matchcostgrad1<<<32,512>>>(b,n,m,xyz1,xyz2,match,grad1);
// matchcostgrad2<<<dim3(32,32),256>>>(b,n,m,xyz1,xyz2,match,grad2);
//}
/* MatchCost backward interface
Input:
grad_cost: (B) # gradients on cost
@ -373,7 +406,8 @@ std::vector<at::Tensor> MatchCostBackward(
const at::Tensor grad_cost,
const at::Tensor xyz1,
const at::Tensor xyz2,
const at::Tensor match){
const at::Tensor match)
{
const auto b = xyz1.size(0);
const auto n = xyz1.size(1);
const auto m = xyz2.size(1);
@ -387,10 +421,10 @@ std::vector<at::Tensor> MatchCostBackward(
auto grad1 = at::zeros({b, n, 3}, xyz1.type());
auto grad2 = at::zeros({b, m, 3}, xyz1.type());
AT_DISPATCH_FLOATING_TYPES(xyz1.scalar_type(), "MatchCostBackward", ([&] {
AT_DISPATCH_FLOATING_TYPES(xyz1.scalar_type(), "MatchCostBackward", ([&]
{
matchcostgrad1<scalar_t><<<32,512>>>(b, n, m, grad_cost.data<scalar_t>(), xyz1.data<scalar_t>(), xyz2.data<scalar_t>(), match.data<scalar_t>(), grad1.data<scalar_t>());
matchcostgrad2<scalar_t><<<dim3(32,32),256>>>(b, n, m, grad_cost.data<scalar_t>(), xyz1.data<scalar_t>(), xyz2.data<scalar_t>(), match.data<scalar_t>(), grad2.data<scalar_t>());
}));
matchcostgrad2<scalar_t><<<dim3(32,32),256>>>(b, n, m, grad_cost.data<scalar_t>(), xyz1.data<scalar_t>(), xyz2.data<scalar_t>(), match.data<scalar_t>(), grad2.data<scalar_t>()); }));
C10_CUDA_CHECK(cudaGetLastError());
return std::vector<at::Tensor>({grad1, grad2});