The motivation is twofold:
1) Allow plugging in a different training-time evaluator, e.g.
TFLite-based, etc.
2) Allow using TensorSpec for AOT, too, to support evolution: we start
by extracting a superset of the features currently supported by a
model. For the tensors the model does not support, we just return a
valid, but useless, buffer. This makes using a 'smaller' model (less
supported tensors) transparent to the compiler. The key is to
dimension the buffer appropriately, and we already have TensorSpec
modeling that info.
The only coupling was due to the reliance of a TF internal API for
getting the element size, but for the types we are interested in,
`sizeof` is sufficient.
A subsequent change will yank out TensorSpec in its own module.
Differential Revision: https://reviews.llvm.org/D124045
Analysis Opportunities:
//===---------------------------------------------------------------------===//
In test/Transforms/LoopStrengthReduce/quadradic-exit-value.ll, the
ScalarEvolution expression for %r is this:
{1,+,3,+,2}<loop>
Outside the loop, this could be evaluated simply as (%n * %n), however
ScalarEvolution currently evaluates it as
(-2 + (2 * (trunc i65 (((zext i64 (-2 + %n) to i65) * (zext i64 (-1 + %n) to i65)) /u 2) to i64)) + (3 * %n))
In addition to being much more complicated, it involves i65 arithmetic,
which is very inefficient when expanded into code.
//===---------------------------------------------------------------------===//
In formatValue in test/CodeGen/X86/lsr-delayed-fold.ll,
ScalarEvolution is forming this expression:
((trunc i64 (-1 * %arg5) to i32) + (trunc i64 %arg5 to i32) + (-1 * (trunc i64 undef to i32)))
This could be folded to
(-1 * (trunc i64 undef to i32))
//===---------------------------------------------------------------------===//
e4794ff5c6