Skip to content

[AOT] Correctly calculate workspace for vector types #17077

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jun 11, 2024
Merged

[AOT] Correctly calculate workspace for vector types #17077

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ad9a613
[AOT] Correctly calculate workspace for vector types
lhutton1 Jun 10, 2024
d0a9031
Use tvm.testing.main()
lhutton1 Jun 10, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion src/tir/usmp/utils.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -181,7 +181,11 @@ Map<String, PoolAllocation> GetIOPoolAllocations(
}

static Integer CalculateExtentsSize(const DataType& dtype, const Array<PrimExpr>& extents) {
size_t element_size_bytes = dtype.bytes();
if (dtype.is_scalable_vector()) {
// We cannot statically calculate workspace for scalable types
return Integer();
}
size_t element_size_bytes = dtype.bytes() * dtype.lanes();
size_t num_elements = 1;
for (const auto& ext : extents) {
if (ext->IsInstance<IntImmNode>()) {
Expand Down
20 changes: 18 additions & 2 deletions tests/python/tir-analysis/test_tir_analysis_calculate_workspace.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,6 +91,18 @@ def primfunc_local_allocates(placeholder_162: T.handle, placeholder_163: T.handl
# fmt: on


@T.prim_func
def prim_func_decl_vector_type(a: T.handle, b: T.handle):
T.func_attr({"tir.noalias": True})
A = T.match_buffer(a, (4,), "float32x4")
B = T.match_buffer(b, (4,), "float32x4")
C = T.decl_buffer((4,), "float32x4")
for i in range(3):
with T.block("block"):
vi = T.axis.remap("S", [i])
B[vi] = A[vi] + C[vi]


@pytest.mark.parametrize("alignment,size,consts", [(1, 663552, 0), (10, 663560, 0)])
def test_global_allocates(alignment, size, consts):
primfunc = primfunc_global_allocates
Expand All @@ -105,6 +117,10 @@ def test_local_allocates(alignment, size, consts):
assert tvm.tir.analysis.calculate_workspace_bytes(primfunc, alignment) == size


def test_vector_type():
primfunc = prim_func_decl_vector_type
assert tvm.tir.analysis.calculate_workspace_bytes(primfunc, 1) == 64


if __name__ == "__main__":
test_global_allocates()
test_local_allocates()
tvm.testing.main()