p95768021
/
InfiniTensor
forked from jiuyuan/InfiniTensor
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fix: conv2dreduce_kernel_ offset

This commit is contained in:
Liyan Zheng 2023-04-24 02:30:41 +08:00
parent 51cc042f56
commit c1275cddb6
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/kernels/cuda/conv2dreduce.cu
View File

@ -16,7 +16,7 @@ __global__ void conv2dreduce_kernel_(float *__restrict__ input,
int nid = blockIdx.x, fid = blockIdx.y;
int hid = threadIdx.x, wid = threadIdx.y;
const int fchunck = r * s, wchunk = f * fchunck, hchunk = w * wchunk,
nchunck = n * hchunk;
nchunck = h * hchunk;
float *nfinput = input + nid * nchunck + fid * fchunck;
if (nid < n && fid < f && hid < oh && wid < ow) {
float imm = 0.0;