MCPcopy Create free account
hub / github.com/MegEngine/MegEngine / dispatch_float_noncontig_batch

Method dispatch_float_noncontig_batch

dnn/src/x86/local/opr_impl.cpp:16–43  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

14}
15
16LocalImpl::float_noncontig_batch_kern LocalImpl::dispatch_float_noncontig_batch(
17 const TensorLayout& src, const TensorLayout&, const TensorLayout&) {
18 megdnn_assert(
19 src.stride[0] > 0 &&
20 static_cast<size_t>(src.stride[0]) >= src.total_nr_elems() / src.shape[0]);
21
22 if (param().mode == Mode::CROSS_CORRELATION) {
23 if (is_supported(SIMDType::FMA)) {
24 return local_xcorr_FMA;
25 } else if (is_supported(SIMDType::AVX)) {
26 return local_xcorr_AVX;
27 } else if (is_supported(SIMDType::SSE)) {
28 return local_xcorr_SSE;
29 } else {
30 megdnn_throw("no fma/avx/sse detected");
31 }
32 } else {
33 if (is_supported(SIMDType::FMA)) {
34 return local_conv_FMA;
35 } else if (is_supported(SIMDType::AVX)) {
36 return local_conv_AVX;
37 } else if (is_supported(SIMDType::SSE)) {
38 return local_conv_SSE;
39 } else {
40 megdnn_throw("no fma/avx/sse detected");
41 }
42 }
43}
44
45void LocalImpl::exec(
46 _megdnn_tensor_in src, _megdnn_tensor_in filter, _megdnn_tensor_out dst,

Callers 1

execMethod · 0.45

Calls 2

paramFunction · 0.50
total_nr_elemsMethod · 0.45

Tested by

no test coverage detected