[async] More SFG optimizations #1877
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Codecov Report
@@ Coverage Diff @@
## master #1877 +/- ##
==========================================
+ Coverage 43.25% 44.21% +0.95%
==========================================
Files 44 44
Lines 6320 6111 -209
Branches 1092 1092
==========================================
- Hits 2734 2702 -32
+ Misses 3416 3240 -176
+ Partials 170 169 -1
Continue to review full report at Codecov.
|
yuanming-hu
force-pushed
the
asyncmgpcg
branch
from
04cddb4 to
fb77de7
Compare
yuanming-hu
commented
Sep 18, 2020
k-ye
approved these changes
Sep 19, 2020
| continue; | ||
|
|
||
| auto list_node = *node->input_edges[list_state].begin(); | ||
| tasks[std::make_pair(node->rec.ir_handle, list_node)].push_back(node); |
There was a problem hiding this comment.
Just to confirm, if there's such a launch sequence:
@ti.kernel
def foo():
for i in x:
y[i] = ...
@ti.kernel
def bar():
for i in range(...):
if some_cond(i):
x[i] = ... # potentially changes x's mask/list state
foo()
bar()
foo() # <-- cannot demote the activation of |y| nowDoes this not demote activation as expected?
There was a problem hiding this comment.
I believe the things will go as expected (although not tested). The second foo() uses a different list of x, and activation demotion will only happen when the same list states are used :-)
(Btw we should set up a testing system so that cases like this can be tested...)
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Related issue = #742
List of changes:
misc/async_mgpcg.py, which is a fully asynchronous MGPCG. Note that the old MGPCG has Python-scope operations such asbeta = sum[None], which force the system to synchronizeStateFlowGraph::optimize_listgenis re-written. Now theClearListStmtgets removed together with thelistgentaskStateFlowGraph::optimize_dead_storeno longer removesClearListStmtAsyncEngine::synchronize()no longer dooptimize_listgenafterfusion, sinceListGenwith fused serials ofClearListis not yet supported (TODO: support...)TaskLaunchRecordnow has a uniqueid. This will be used when generating thedotfiles. Having a consistentidsimplified debugging.StateFlowGraph::Node::launch_idis removed and we display theTaskLaunchRecordid onlyProgramnow has asnodesmapping that maps SNode id to SNode pointerStateFlowGraph::extractnow takes abool sort. Note that sorting an intermediate graph (which doesn't have order independency) can lead to wrong resultsStateFlowGraph::demote_activationConstExprPropagationfordemote_activationStateFlowGraph::dump_dotcan now visual a state flow chain of a single state only. (TODO: make a standard API? it's harder coded for now. What should the API look like?)[Click here for the format server]