Skip to content

Add support for ellipsis (...) indexing in Helion #437

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

Closed
wants to merge 1 commit into from
Closed

Add support for ellipsis (...) indexing in Helion #437

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
63 changes: 59 additions & 4 deletions helion/_compiler/indexing_strategy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -227,6 +227,9 @@ def valid_block_size(
for i, k in enumerate(subscript):
if k is None:
continue
if k is Ellipsis:
# Ellipsis is not supported in tensor descriptor mode
return False
Comment on lines +230 to +232
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this hard to support? Maybe we we pre-expand the indexing string the existing code could work unmodified.

size, stride = size_stride.popleft()
if isinstance(k, slice):
# Slices with steps are not supported in tensor descriptor mode
Expand Down Expand Up @@ -447,6 +450,14 @@ def codegen_store(
)


def _calculate_ellipsis_dims(
index: list[object], current_index: int, total_dims: int
) -> int:
"""Calculate how many dimensions an ellipsis should expand to."""
remaining_indices = len(index) - current_index - 1
return total_dims - current_index - remaining_indices
Comment on lines +453 to +458
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this is right when combined with None (which will increment i in the caller but not consume an input dimension.



class SubscriptIndexing(NamedTuple):
index_expr: ast.AST
mask_expr: ast.AST
Expand All @@ -465,9 +476,18 @@ def compute_shape(
input_size = collections.deque(tensor.size())
output_size = []
env = CompileEnvironment.current()
for k in index:
for i, k in enumerate(index):
if k is None:
output_size.append(1)
elif k is Ellipsis:
ellipsis_dims = _calculate_ellipsis_dims(index, i, len(tensor.size()))
for _ in range(ellipsis_dims):
size = input_size.popleft()
if size != 1:
rdim = env.allocate_reduction_dimension(size)
output_size.append(rdim.var)
else:
output_size.append(1)
elif isinstance(k, int):
input_size.popleft()
elif isinstance(k, torch.SymInt):
Expand Down Expand Up @@ -517,6 +537,21 @@ def create(
for n, k in enumerate(index):
if k is None:
output_idx += 1
elif k is Ellipsis:
ellipsis_dims = _calculate_ellipsis_dims(index, n, fake_value.ndim)
for _ in range(ellipsis_dims):
expand = tile_strategy.expand_str(output_size, output_idx)
size = fake_value.size(len(index_values))
if size != 1:
rdim = env.allocate_reduction_dimension(size)
block_idx = rdim.block_id
index_var = state.codegen.index_var(block_idx)
index_values.append(f"({index_var}){expand}")
if mask := state.codegen.mask_var(block_idx):
mask_values.setdefault(f"({mask}){expand}")
else:
index_values.append(f"tl.zeros([1], {dtype}){expand}")
output_idx += 1
elif isinstance(k, int):
index_values.append(repr(k))
elif isinstance(k, torch.SymInt):
Expand Down Expand Up @@ -729,8 +764,16 @@ def is_supported(
# TODO(jansel): support block_ptr with extra_mask
return False
input_sizes = collections.deque(fake_tensor.size())
for k in index:
input_size = 1 if k is None else input_sizes.popleft()
for n, k in enumerate(index):
if k is None:
input_size = 1
elif k is Ellipsis:
ellipsis_dims = _calculate_ellipsis_dims(index, n, fake_tensor.ndim)
for _ in range(ellipsis_dims):
input_sizes.popleft()
continue
else:
input_size = input_sizes.popleft()
if isinstance(k, torch.SymInt):
symbol = k._sympy_()
origin = None
Expand Down Expand Up @@ -780,9 +823,21 @@ def create(
fake_value,
reshaped_size=SubscriptIndexing.compute_shape(fake_value, index),
)
for k in index:
for n, k in enumerate(index):
if k is None:
pass # handled by reshaped_size
elif k is Ellipsis:
ellipsis_dims = _calculate_ellipsis_dims(index, n, fake_value.ndim)
env = CompileEnvironment.current()
for _ in range(ellipsis_dims):
size = fake_value.size(len(res.offsets))
if size != 1:
rdim = env.allocate_reduction_dimension(size)
res.offsets.append(state.codegen.offset_var(rdim.block_id))
res.block_shape.append(rdim.var)
else:
res.offsets.append("0")
res.block_shape.append(1)
elif isinstance(k, int):
res.offsets.append(repr(k))
res.block_shape.append(1)
Expand Down
20 changes: 20 additions & 0 deletions helion/_compiler/type_propagation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -433,6 +433,26 @@ def _device_indexing_size(self, key: TypeInfo) -> list[int | torch.SymInt]:
inputs_consumed += 1
elif k.value is None:
output_sizes.append(1)
elif k.value is Ellipsis:
# Count indices after ellipsis (excluding None)
remaining_keys = sum(
1
for key in keys[keys.index(k) + 1 :]
if not (isinstance(key, LiteralType) and key.value is None)
)
ellipsis_dims = (
self.fake_value.ndim - inputs_consumed - remaining_keys
)
for _ in range(ellipsis_dims):
size = self.fake_value.size(inputs_consumed)
inputs_consumed += 1
if self.origin.is_device():
output_sizes.append(size)
elif size != 1:
rdim = env.allocate_reduction_dimension(size)
output_sizes.append(rdim.var)
else:
output_sizes.append(1)
else:
raise exc.InvalidIndexingType(k)
elif isinstance(k, SymIntType):
Expand Down
3 changes: 0 additions & 3 deletions test/test_indexing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -759,9 +759,6 @@ def kernel(
torch.testing.assert_close(src_result, expected_src)
torch.testing.assert_close(dst_result, expected_dst)

@skipIfNormalMode(
"RankMismatch: Cannot assign a tensor of rank 2 to a buffer of rank 3"
)
def test_ellipsis_indexing(self):
"""Test both setter from scalar and getter for [..., i]"""

Expand Down
Loading