Array<T> createSubArray(const Array<T> &parent, const vector<af_seq> &index,
bool copy) {
parent.eval();
dim4 dDims = parent.getDataDims();
dim4 dStrides = calcStrides(dDims);
dim4 parent_strides = parent.strides();
if (dStrides != parent_strides) {
const Array<T> parentCopy = copyArray(parent);
return createSubArray(parentCopy, index, copy);
}
dim4 pDims = parent.dims();
dim4 dims = toDims(index, pDims);
dim4 strides = toStride(index, dDims);
// Find total offsets after indexing
dim4 offsets = toOffset(index, pDims);
dim_t offset = parent.getOffset();
for (int i = 0; i < 4; i++) offset += offsets[i] * parent_strides[i];
Array<T> out = Array<T>(parent, dims, offset, strides);
if (!copy) return out;
if (strides[0] != 1 || strides[1] < 0 || strides[2] < 0 || strides[3] < 0) {
out = copyArray(out);
}
return out;
}
Array<T> createSubArray(const Array<T>& parent,
const std::vector<af_seq> &index,
bool copy)
{
parent.eval();
dim4 dDims = parent.getDataDims();
dim4 pDims = parent.dims();
dim4 dims = toDims (index, pDims);
dim4 offset = toOffset(index, dDims);
dim4 stride = toStride (index, dDims);
Array<T> out = Array<T>(parent, dims, offset, stride);
if (!copy) return out;
if (stride[0] != 1 ||
stride[1] < 0 ||
stride[2] < 0 ||
stride[3] < 0) {
out = copyArray(out);
}
return out;
}
Array<T> index(const Array<T>& in, const af_index_t idxrs[])
{
kernel::IndexKernelParam_t p;
std::vector<af_seq> seqs(4, af_span);
// create seq vector to retrieve output
// dimensions, offsets & offsets
for (dim_t x=0; x<4; ++x) {
if (idxrs[x].isSeq) {
seqs[x] = idxrs[x].idx.seq;
}
}
// retrieve dimensions, strides and offsets
dim4 iDims = in.dims();
dim4 dDims = in.getDataDims();
dim4 oDims = toDims (seqs, iDims);
dim4 iOffs = toOffset(seqs, dDims);
dim4 iStrds= toStride(seqs, dDims);
for (dim_t i=0; i<4; ++i) {
p.isSeq[i] = idxrs[i].isSeq;
p.offs[i] = iOffs[i];
p.strds[i] = iStrds[i];
}
Buffer* bPtrs[4];
std::vector< Array<uint> > idxArrs(4, createEmptyArray<uint>(dim4()));
// look through indexs to read af_array indexs
for (dim_t x=0; x<4; ++x) {
// set index pointers were applicable
if (!p.isSeq[x]) {
idxArrs[x] = castArray<uint>(idxrs[x].idx.arr);
bPtrs[x] = idxArrs[x].get();
// set output array ith dimension value
oDims[x] = idxArrs[x].elements();
}
else {
// alloc an 1-element buffer to avoid OpenCL from failing
bPtrs[x] = bufferAlloc(sizeof(uint));
}
}
Array<T> out = createEmptyArray<T>(oDims);
if(oDims.elements() == 0) { return out; }
kernel::index<T>(out, in, p, bPtrs);
for (dim_t x=0; x<4; ++x) {
if (p.isSeq[x]) bufferFree(bPtrs[x]);
}
return out;
}
