Skip to content

[Clean][Refactor] Phaseout Legacy Pass ParallelLoopTransformer #1672

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
LeiWang1999 merged 16 commits into from
Jan 15, 2026
Merged

[Clean][Refactor] Phaseout Legacy Pass ParallelLoopTransformer #1672

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
ed31be7
[Feat] Add tilelang autodd for delta debugging
KEKE046 Jan 8, 2026
6736dae
fix typos
KEKE046 Jan 8, 2026
41428d8
fix lint error
KEKE046 Jan 8, 2026
636c3b2
fix typos
KEKE046 Jan 8, 2026
0427dfb
fix lint error
KEKE046 Jan 8, 2026
285b3cf
fix bugs
KEKE046 Jan 8, 2026
889c37f
Apply suggestions from code review
KEKE046 Jan 8, 2026
03a3c76
fix codeview comments
KEKE046 Jan 8, 2026
8970b34
Merge branch 'autodd' of https://github.com/KEKE046/tilelang into autodd
KEKE046 Jan 8, 2026
5d2aa63
Merge branch 'main' of https://github.com/tile-ai/tilelang into autodd
kurisu6912 Jan 9, 2026
a096abb
[Refactor] Move AutoDD detection to env module and update import logic
LeiWang1999 Jan 13, 2026
5c4ebfa
[Documentation] Add AutoDD section to debug_tools_for_tilelang.md
LeiWang1999 Jan 13, 2026
076257b
[Refactor] Update flash attention implementation in TileLang examples
LeiWang1999 Jan 14, 2026
6581b5f
[Refactor] Clean up code formatting and improve readability in TileLa…
LeiWang1999 Jan 14, 2026
fd59247
Merge branch 'main' of https://github.com/tile-ai/tilelang into autodd
LeiWang1999 Jan 14, 2026
8c7d301
[Refactor] Remove unused import in fill.cc
LeiWang1999 Jan 15, 2026
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
376 changes: 184 additions & 192 deletions examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_paged.py
View file
Open in desktop

Large diffs are not rendered by default.

337 changes: 155 additions & 182 deletions examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_varlen_indice.py
View file
Open in desktop

Large diffs are not rendered by default.

334 changes: 173 additions & 161 deletions examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_varlen_mask.py
View file
Open in desktop

Large diffs are not rendered by default.

11 changes: 4 additions & 7 deletions src/op/atomic_add.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,6 @@
#include "../target/utils.h"
#include "../transform/atomicadd_vectorize.h"
#include "../transform/common/loop_fusion_utils.h"
#include "../transform/common/loop_parallel_transform_utils.h"
#include "../transform/loop_partition.h"
#include "builtin.h"

Expand Down Expand Up @@ -658,8 +657,6 @@ Stmt AtomicAddNode::Lower(const LowerArgs &T, arith::Analyzer *analyzer) const {
}
auto simt_loop = MakeSIMTLoop(analyzer);
auto fused_loop = Downcast<For>(ParallelLoopFuser::Fuse(simt_loop));
auto transformed_loop =
Downcast<For>(ParallelLoopTransformer::Substitute(fused_loop));

auto GetArchInt = [&](const Target &tgt) -> int {
int arch_int = 0;
Expand Down Expand Up @@ -785,12 +782,12 @@ Stmt AtomicAddNode::Lower(const LowerArgs &T, arith::Analyzer *analyzer) const {
return {loop_layout, pred};
};

auto ret = AtomicAddInferLayout(transformed_loop,
{T.target, T.thread_bounds, T.layout_map,
analyzer, false, T.buffer_remap});
auto ret =
AtomicAddInferLayout(fused_loop, {T.target, T.thread_bounds, T.layout_map,
analyzer, false, T.buffer_remap});
Fragment loop_layout = ret.loop_layout;
auto thread_loop =
PartitionLoop(transformed_loop, T.thread_var, analyzer, loop_layout);
PartitionLoop(fused_loop, T.thread_var, analyzer, loop_layout);
auto vectorized_thread_loop =
VectorizeAtomicAdd(thread_loop, GetArchInt(target));
return vectorized_thread_loop;
Expand Down
8 changes: 2 additions & 6 deletions src/op/copy.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,6 @@
#include "../layout/tcgen05_layout.h"
#include "../target/utils.h"
#include "../transform/common/loop_fusion_utils.h"
#include "../transform/common/loop_parallel_transform_utils.h"
#include "../transform/loop_partition.h"
#include "../transform/loop_vectorize.h"
#include "utils.h"
Expand Down Expand Up @@ -716,19 +715,16 @@ Stmt CopyNode::LowerNormalCopy(const LowerArgs &T,
auto simt_loop = MakeSIMTLoop(analyzer);
auto fused_loop = Downcast<For>(ParallelLoopFuser::Fuse(simt_loop));

auto transformed_loop =
Downcast<For>(ParallelLoopTransformer::Substitute(fused_loop));

For vectorized_thread_loop;
auto par_op = ParallelOp(transformed_loop);
auto par_op = ParallelOp(fused_loop);

if (is_cpu_target || IsLocalBuffer(src) || IsLocalBuffer(dst)) {
if (IsLocalBuffer(src) && !IsLocalBuffer(dst)) {
LOG(WARNING) << "Copy from local buffer `" << src->name << "` to "
<< dst.scope() << " buffer `" << dst->name
<< "` may cause conflicted write.";
}
vectorized_thread_loop = VectorizeLoop(transformed_loop);
vectorized_thread_loop = VectorizeLoop(fused_loop);
return vectorized_thread_loop;
} else {
std::vector<InferLevel> levels = {InferLevel::kCommon, InferLevel::kStrict,
Expand Down
1 change: 0 additions & 1 deletion src/op/fill.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,6 @@
#include "../layout/tcgen05_layout.h"
#include "../target/utils.h"
#include "../transform/common/loop_fusion_utils.h"
#include "../transform/common/loop_parallel_transform_utils.h"
#include "../transform/loop_partition.h"
#include "../transform/loop_vectorize.h"
#include "builtin.h"
Expand Down
170 changes: 0 additions & 170 deletions src/transform/common/loop_parallel_transform_utils.h
View file
Open in desktop

This file was deleted.

2 changes: 0 additions & 2 deletions src/transform/layout_inference.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,6 @@
#include "arith/ir_mutator_with_analyzer.h"
#include "arith/ir_visitor_with_analyzer.h"
#include "common/loop_fusion_utils.h"
#include "common/loop_parallel_transform_utils.h"
#include "common/union_find.h"
#include "layout_reducer.h"
#include "parallel_loop_layout_validator.h"
Expand Down Expand Up @@ -1253,7 +1252,6 @@ class LayoutInferencer : public IRMutatorWithAnalyzer {
tvm::transform::Pass LayoutInference() {
using namespace tir::transform;
auto pass_func = [=](PrimFunc f, const IRModule &m, const PassContext &ctx) {
f.CopyOnWrite()->body = ParallelLoopTransformer::Substitute(f->body);
ThreadBindingCollector collector;
collector(f->body);
bool has_thread_binding = !collector.thread_binding_.empty();
Expand Down
Loading