search

NervanaSystems / ngraph

nGraph has moved to OpenVINO
https://www.ngraph.ai/
Apache License 2.0
1.35k stars 222 forks source link

Unboxing tensor ops #91

Closed diyessi closed 6 years ago

diyessi commented 7 years ago

Would it make sense to unbox element-wise ops at some point? Dot as reductions? Convolution as dots.

Currently, lower level primitives (kernels) are highly-optimized tensor operations. We would need to generate code that was competitive with these optimized kernels.

aprocter commented 7 years ago

Currently, lower level primitives (kernels) are highly-optimized tensor operations. We would need to generate code that was competitive with these optimized kernels.

Yup, there's the rub; maximum parsimony here would require transformers to do fat pattern-matches like BinaryReduce<+,0>(ElementWiseBinary<*>(arg1,arg2)) instead of Dot1D(arg1,arg2) (or whatever). This may prove a bit brittle in practice.

aprocter commented 7 years ago

Will think more about the case for it, but to reiterate I don't think this is a blocker for merge of #87.

diyessi commented 7 years ago

To make matters worse, we're in a semi-JIT mode. We will need to see what the compile/compute ratio actually is. If we only end up compiling once, we can spend more time at it than if we need to compile every batch.