Skip to content

Do not emit temporary decls for temporaries not used in a subkernel #741

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
inducer merged 3 commits into from
Jan 19, 2023
Merged

Do not emit temporary decls for temporaries not used in a subkernel #741

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2080ead
Fix some warning typos
inducer Jan 19, 2023
ebe7358
Do not emit temporary decls for temporaries not used in a subkernel
inducer Jan 19, 2023
f651135
Test codegen for global temporaries with different base storages
kaushikcfd Jan 17, 2023
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
8 changes: 3 additions & 5 deletions loopy/target/c/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -899,14 +899,12 @@ def get_temporary_decls(self, codegen_state, schedule_index):

ecm = self.get_expression_to_code_mapper(codegen_state)

for tv in sorted(
kernel.temporary_variables.values(),
key=lambda key_tv: key_tv.name):
for tv_name in sorted(sub_knl_temps):
tv = kernel.temporary_variables[tv_name]
if not tv.base_storage:
# global temp vars are mapped to arguments or global
# declarations, no need to declare locally.
if tv.address_space != AddressSpace.GLOBAL and (
tv.name in sub_knl_temps):
if tv.address_space != AddressSpace.GLOBAL:
decl = self.get_temporary_var_declarator(codegen_state, tv)

if tv.initializer is not None:
Expand Down
4 changes: 2 additions & 2 deletions loopy/transform/data.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1025,7 +1025,7 @@ def allocate_temporaries_for_base_storage(kernel: LoopKernel,
assert isinstance(tv.dtype, LoopyType)

if tv.address_space is auto:
raise LoopyError("Ahen allocating base storage for temporary "
raise LoopyError("When allocating base storage for temporary "
f"'{tv.name}', the address space of the temporary "
"was not yet determined (set to 'auto').")

Expand Down Expand Up @@ -1084,7 +1084,7 @@ def allocate_temporaries_for_base_storage(kernel: LoopKernel,
warn("Base storage allocation was performed implicitly during "
"preprocessing. This is deprecated and will stop working "
"in 2023. Call loopy.allocate_temporaries_for_base_storage "
"explicitly to aovid this warning.", DeprecationWarning)
"explicitly to avoid this warning.", DeprecationWarning)

return kernel.copy(temporary_variables=new_tvs)
else:
Expand Down
37 changes: 37 additions & 0 deletions test/test_loopy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3392,6 +3392,43 @@ def test_creation_kwargs():
ksdfjlasdf=None)


def test_global_temps_with_multiple_base_storages(ctx_factory):
# See https://github.com/inducer/loopy/issues/737

n = 10
ctx = ctx_factory()
cq = cl.CommandQueue(ctx)

prg = lp.make_kernel(
"{[r0, r1]: 0<=r0,r1< %s}" % n,
"""
tmp0 = sum(r0, r0**2)
... gbarrier
tmp1 = sum(r1, r1**3)
... gbarrier
out = tmp0 + tmp1
""",
[lp.TemporaryVariable("tmp0",
shape=lp.auto,
address_space=lp.AddressSpace.GLOBAL,
base_storage="base1"),
lp.TemporaryVariable("tmp1",
shape=lp.auto,
address_space=lp.AddressSpace.GLOBAL,
base_storage="base2"),
...],
seq_dependencies=True
)

prg = lp.infer_unknown_types(prg)
prg = lp.allocate_temporaries_for_base_storage(prg)
print(prg)

_, (out,) = prg(cq)

assert out == sum(i**2 for i in range(n)) + sum(i**3 for i in range(n))


if __name__ == "__main__":
if len(sys.argv) > 1:
exec(sys.argv[1])
Expand Down