*: add support for windowing of tensors #372
Conversation
rohany
force-pushed
the
windowing-different-notation
branch
8 times, most recently
from
744f88f
to
4fd7744
Compare
rohany
force-pushed
the
windowing-different-notation
branch
5 times, most recently
from
cf3369d
to
f364047
Compare
| @@ -220,14 +223,24 @@ class Access : public IndexExpr { | |||
| Access() = default; | |||
| Access(const Access&) = default; | |||
| Access(const AccessNode*); | |||
| Access(const TensorVar& tensorVar, const std::vector<IndexVar>& indices={}); | |||
| Access(const TensorVar& tensorVar, const std::vector<IndexVar>& indices={}, const std::map<int, AccessWindow>& windows={}); | |||
There was a problem hiding this comment.
Should probably break this line (and other longer lines) up into multiple lines. (Lines should be restricted to 80 characters at most where possible.)
There was a problem hiding this comment.
Done. clang-format isn't set up on the repository, so I'm missing these small formatting things everywhere.
include/taco/tensor.h
Outdated
| auto resultReversed = this->_access(args...); | ||
| std::vector<std::shared_ptr<IndexVarInterface>> result; | ||
| result.reserve(resultReversed.size()); | ||
| for (auto it = resultReversed.rbegin(); it != resultReversed.rend(); it++) { |
There was a problem hiding this comment.
You can also use util::reverse (defined in taco/util/collections.h) for this.
src/codegen/codegen_c.cpp
Outdated
| @@ -516,7 +516,7 @@ void CodeGen_C::visit(const Allocate* op) { | |||
| stream << ", "; | |||
| } | |||
| else { | |||
| stream << "malloc("; | |||
| stream << "calloc(1, "; | |||
There was a problem hiding this comment.
If you need calloc instead of malloc, you should make that optional since it can impose performance overhead and often it's not necessary to actually zero-initialize the array.
There was a problem hiding this comment.
Done, made it configurable based on the context.
test/tests-windowing.cpp
Outdated
| // These dimensions are chosen so that one is above the constant in `mode_format_dense.cpp:54` | ||
| // where the known stride is generated vs using the dimension. | ||
| // TODO (rohany): Change that constant to be in a header file and import it here. | ||
| for (auto& dim : {6, 20}) { |
There was a problem hiding this comment.
Should probably implement this as parameterized tests using TEST_P, so that each combination of dimensions/formats can be tested individually. (Same with the other test cases below.)
There was a problem hiding this comment.
Done. I couldn't figure out how to get good names for the parameters to show up in the gtest output though.
This commit adds support for windowing of tensors in the existing index notation DSL. For example: ``` A(i, j) = B(i(1, 4), j) * C(i, j(5, 10)) ``` causes `B` to be windowed along its first mode, and `C` to be windowed along its second mode. In this commit any mix of windowed and non-windowed modes are supported, along with windowing the same tensor in different ways in the same expression. The windowing expressions correspond to the `:` operator to slice dimensions in `numpy`. Currently, only windowing by integers is supported. Windowing is achieved by tying windowing information to particular `Iterator` objects, as these are created for each `Tensor`-`IndexVar` pair. When iterating over an `Iterator` that may be windowed, extra steps are taken to either generate an index into the windowed space, or to recover an index from a point in the windowed space.
rohany
force-pushed
the
windowing-different-notation
branch
from
f364047
to
2d4a7d3
Compare
|
Looks good! I'll merge it into master. |
This commit adds support for windowing of tensors in the existing index
notation DSL. For example:
causes
Bto be windowed along its first mode, andCto be windowedalong its second mode. In this commit any mix of windowed and
non-windowed modes are supported, along with windowing the same tensor
in different ways in the same expression. The windowing expressions
correspond to the
:operator to slice dimensions innumpy.Currently, only windowing by integers is supported.
Windowing is achieved by tying windowing information to particular
Iteratorobjects, as these are created for eachTensor-IndexVarpair. When iterating over an
Iteratorthat may be windowed, extrasteps are taken to either generate an index into the windowed space, or
to recover an index from a point in the windowed space.