mirrors/tinygrad
2
0
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-06-24 02:14:17 +00:00
Code Issues Projects Releases Packages Wiki Activity

do not emit loop end code for global+local loops in assembly kernel (#1100)

Browse source
This commit is contained in:
David Hou 2023-07-02 18:33:57 -07:00 • committed by GitHub
commit 363fbfc2e4
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
tinygrad/codegen/assembly.py
View file

@ -125,7 +125,7 @@ class AssemblyCodegen(Linearizer):
ins.append(AssemblyInstruction(UOps.CONST, newreg(var, dtype=dtypes.int32, scalar=True), [], 0))
ins.append(AssemblyInstruction(UOps.LABEL, None, [], "$loop_"+var.expr))
elif uop == UOps.ENDLOOP:
if args[1] not in ["global", "local"]:
if args[1] not in ["global", "local", "global+local"]:
for var in reversed(args[0]):
if not isinstance(var, NumNode): # TODO: why is this coming through?
ins.append(AssemblyInstruction(UOps.ALU, tor[var], [tor[var], 1], BinaryOps.ADD))