skip copies of reshaped buffers (#12430)

* skip copies of reshaped buffers

* always run NOOP

* comment

* comment
This commit is contained in:
George Hotz
2025-10-03 13:05:47 +08:00
committed by GitHub
parent 0047bcc535
commit 4c63f7e786
+6 -5
View File
@@ -46,6 +46,9 @@ earliest_rewrites = PatternMatcher([
# just removing it works...
(UPat((Ops.DETACH, Ops.CONTIGUOUS_BACKWARD, Ops.FUSE), name="x"), lambda x: x.src[0]),
# remove CONTIGUOUS if the BUFFER is already contiguous
(UPat(Ops.BUFFER).f(Ops.RESHAPE, name="r").f(Ops.CONTIGUOUS, name="c"), lambda r,c: r.replace(tag=c.tag)),
# split_reduceop
(UPat(Ops.REDUCE_AXIS, name="reduce", src=(UPat.var("x"),)), split_reduceop),
@@ -67,9 +70,6 @@ earliest_rewrites = PatternMatcher([
(UPat(Ops.COPY, src=(UPat(GroupOp.Movement, name="r"), UPat(name="d")), name="c"),
lambda c,r,d: c.replace(src=(r.contiguous(), d)) if r.size != r.base.size else None),
# make inputs to mstack contiguous
(UPat(Ops.MSTACK, name="ms"), lambda ms: ms.replace(src=tuple(s if s.op in ALWAYS_CONTIGUOUS else s.contiguous() for s in ms.src))),
# assign only to buffer, otherwise make it a CONTIGUOUS
(UPat(Ops.ASSIGN, src=(UPat(GroupOp.All-{Ops.BUFFER}, name="target"), UPat(name="x")), name="assign"),
lambda x,target,assign: x.f(Ops.CONTIGUOUS, tag=assign.tag) if ((t:=target.base).op is not Ops.BUFFER and \
@@ -378,7 +378,8 @@ pm_rangeify = pm_mops+PatternMatcher([
# *****************
# 3.5 cleanups
ALWAYS_RUN_OPS = {Ops.CONTIGUOUS, Ops.COPY, Ops.ASSIGN}
# Ops.NOOP happens when we have a COPY to the device the Tensor is already on. We treat it like COPY here for MSTACK.
ALWAYS_RUN_OPS = {Ops.CONTIGUOUS, Ops.COPY, Ops.ASSIGN, Ops.NOOP}
# you don't know in the first pass if axes are going to die, this happens if there's an EXPAND to the left
def cleanup_dead_axes(b:UOp):
@@ -736,7 +737,7 @@ def get_rangeify_map(sink:UOp) -> dict[UOp, UOp]:
# rebuild the sink with all the BUFFERIZEs with tags, this is what's ending up in the tensor graph
# MSTACK stacks multiple BUFFERIZEs in one tagged tensor
# if it's not tagged by here, it's out
tsink = UOp.sink(*[x for x in tsink.parents if x.base.op in {Ops.BUFFERIZE, Ops.MSTACK, Ops.CONST} and x.tag is not None])
tsink = UOp.sink(*[x for x in tsink.parents if x.base.op in {Ops.BUFFERIZE, Ops.MSTACK, Ops.CONST, Ops.BUFFER} and x.tag is not None])
if getenv("VIZ"): graph_rewrite(tsink, PatternMatcher([]), name="View Tagged Rangeify")