[oneDNN] Concat refactoring and disabling caching

[jczaja]

PR types

Others

PR changes

OPs

Describe

This PR disabled oneDNN primitives caching by PaddlePaddle in favour of oneDNN its own caching mechanism. Also refactoring to Acquire API was made for concat

[paddle-bot-old]

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[jakpiase]

I really like the initiative to change from factory to acquireAPI :)

[jakpiase]

Suggested change

	ConcatMKLDNNHandler(const paddle::framework::ExecutionContext& ctx,
	ConcatMKLDNNHandler(const framework::ExecutionContext& ctx,

[jczaja]

ok

[jakpiase]

Suggested change

	paddle::framework::ToMKLDNNDataType(inputs[0]->type());
	framework::ToMKLDNNDataType(inputs[0]->type());

[jczaja]

ok

[jakpiase]

Suggested change

	for (size_t i = 0; i < inputs.size(); i++) {
	for (size_t i = 0; i < inputs.size(); ++i) {

[jczaja]

ok

[jakpiase]

Suggested change

	paddle::framework::vectorize<int64_t>(inputs[i]->dims());
	framework::vectorize<int64_t>(inputs[i]->dims());

[jakpiase]

Suggested change

	auto dst_dims = paddle::framework::vectorize<int64_t>(output->dims());
	auto dst_dims = framework::vectorize<int64_t>(output->dims());

[jakpiase]

Suggested change

	const std::vector<mkldnn::memory::desc>& srcs_md) {
	const std::vector<memory::desc>& srcs_md) {

[jczaja]

ok

[jakpiase]

Suggested change

	const mkldnn::memory::desc& dst_md, const int concat_axis,
	const memory::desc& dst_md, const int concat_axis,

[jczaja]

ok

[jakpiase]

Suggested change

	const framework::Tensor& input, int i) {
	const Tensor& input, int i) {

[jczaja]

ok

[jakpiase]

Suggested change

	ctx.template device_context<paddle::platform::MKLDNNDeviceContext>();
	ctx.template device_context<platform::MKLDNNDeviceContext>();

[jakpiase]

You have forgotten about this one

[jczaja]

@jakpiase Please continue your review

[jakpiase]

LGTM, I really appreciate your effort to unify our operators

[lidanqing-vv]

LGTM

[lidanqing-vv]

LGTM