This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a change to branch unity
in repository https://gitbox.apache.org/repos/asf/tvm.git
omit befd95662c [Unity] Statement rewriter for DataflowBlock (#14043)
omit efb336a8cc [Unity] Relax dataflow pattern language (matching) (#14041)
omit 82bb352f76 [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
omit bae7fb5c6d [Unity] Disallow inline prim_func in relax IR (#14040)
omit 5e8ad8e90a [Unity][Pass] Block-level static memory planning (#14038)
omit 877fbf1867 [Unity] Initial PyTorch Frontend (#14037)
omit f866ae3edd [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
omit 1a8d1409c2 [Unity][Pass] Operator legalization (#14029)
omit 4d250c1890 [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
omit d80451bc3f [Unity][Pass] Wellformed Analysis (#14032)
omit cc3eb0c8e6 [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
omit c6c48deb94 [Unity][Pass] Normalize Pass (#14031)
omit 5dd5411d9b [Unity] Relay -> Relax translator (#14026)
omit 70c645d4ef [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
omit 47f8c8df58 [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
omit dd5cfaa0bc [Unity][VM] Supporting "compiled" exec mode. (#14015)
omit 13402fc028 [Unity][Pass] LambdaLift pass (#14012)
omit dc5612fa0a [Unity][Pass] Operator Fusion Passes (#14001)
omit 5bf538fa5b [Unity] NestedMsg Support utility (#13995)
omit 1991cecbbe [Unity] Relax op: manipulation (#13989)
omit e59b4e78fd [Unity] Relax op: search (#13992)
omit 15800607f5 [Unity] Relax op: linear algebra (#13988)
omit b3de34f1f2 [Unity] Relax op: creation (#13984)
omit 9a7a5fbe10 [Unity] Relax op: neural networks (#13993)
omit 7d8095e8be [Unity] Relax op: statistical (#13991)
omit 716f3aca7c [Unity] Relax op: arithmetic, comparison (#13983)
omit 08cbcdf11f [Unity] Relax op: image (#13994)
omit d4db3c092b [Unity] Relax op: set (#13990)
omit d9d79c0be9 [Unity] Relax op: datatype (#13986)
omit 342b670cde [Unity] Relax op: index (#13987)
omit 568ceebaf2 [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
omit 7c66a535eb [Unity] e2e Relax minimum build flow (#13961)
omit afe63b9a72 [Unity] Relax VM shape lowering pass (#13956)
omit 998069c519 [Unity] Relax VM codegen (#13954)
omit 81ebb3f938 [Unity] Relax TVMScript Printer (#13944)
omit 48a3423248 [Unity] Relax TVMScript Parser. (#13932)
omit abdd35aac2 [Unity] Relax BlockBuilder and ExprMutator (#13926)
omit d56a12aa49 [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
omit ad41f97f00 [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
omit 80a32cb0bc [Unity][IR] First-class StructInfo (#13907)
omit 13af1f26fa [Unity] Relax expressions and types (#13901)
omit f4a0323c22 [Unity] Relax VM (#13878)
add 6f232f91be [TIR] [Bugfix] Pass the correct block_sref_reuse to Replace
(#14023)
add ac57b013b2 [TIR] Update block flags and simplify predicate in
Reverse-Compute-Inline (#14030)
add 8613c79aa0 [TIR] Enable Host Func Attribute for PrimFunc (#14020)
add 054c11e101 [TOPI] Fix index dtype in topi strided_slice (#14022)
add 87bb8b1be8 [TIR] Introduce Pass InjectPTXLDG32 (#13973)
add 7249904622 [AutoScheduler][AutoTVM] Enable xgboost >= 1.7.x new
changes (#14036)
add adc73e35ca [Unity] Relax VM (#13878)
add 1528a8db87 [Unity] Relax expressions and types (#13901)
add d25251fb91 [Unity][IR] First-class StructInfo (#13907)
add 366c89d897 [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
add 4433f192bd [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
add e824efe4ac [Unity] Relax BlockBuilder and ExprMutator (#13926)
add 1f7aed4236 [Unity] Relax TVMScript Parser. (#13932)
add d15126636a [Unity] Relax TVMScript Printer (#13944)
add 48a4f2d4e1 [Unity] Relax VM codegen (#13954)
add 3b79924383 [Unity] Relax VM shape lowering pass (#13956)
add a9a21ed988 [Unity] e2e Relax minimum build flow (#13961)
add bfa3b9c6d1 [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
add d3870a7229 [Unity] Relax op: index (#13987)
add baa08eec21 [Unity] Relax op: datatype (#13986)
add e561dd93db [Unity] Relax op: set (#13990)
add af46f0df85 [Unity] Relax op: image (#13994)
add 5e71c8060e [Unity] Relax op: arithmetic, comparison (#13983)
add e12a97e23c [Unity] Relax op: statistical (#13991)
add 25b1fae67d [Unity] Relax op: neural networks (#13993)
add e980513ad7 [Unity] Relax op: creation (#13984)
add 096f8464e9 [Unity] Relax op: linear algebra (#13988)
add 07dd06c3bf [Unity] Relax op: search (#13992)
add 9c4e7e3944 [Unity] Relax op: manipulation (#13989)
add 04d68873ed [Unity] NestedMsg Support utility (#13995)
add dca96539f8 [Unity][Pass] Operator Fusion Passes (#14001)
add 87a911f1c3 [Unity][Pass] LambdaLift pass (#14012)
add 617f21e260 [Unity][VM] Supporting "compiled" exec mode. (#14015)
add a4dc336bd3 [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
add dcb6ae9786 [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
add 53b3dc82a0 [Unity] Relay -> Relax translator (#14026)
add 16e0c5f064 [Unity][Pass] Normalize Pass (#14031)
add 001a414484 [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
add 407e7f43b6 [Unity][Pass] Wellformed Analysis (#14032)
add 8ca086b227 [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
add d789a293f5 [Unity][Pass] Operator legalization (#14029)
add e06d67213e [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
add cd55803ae1 [Unity] Initial PyTorch Frontend (#14037)
add e0775c8c95 [Unity][Pass] Block-level static memory planning (#14038)
add 50bcd1b0af [Unity] Disallow inline prim_func in relax IR (#14040)
add b42f8a44cd [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
add 413329dfa2 [Unity] Relax dataflow pattern language (matching) (#14041)
add 7e4d71cc1e [Unity] Statement rewriter for DataflowBlock (#14043)
This update added new revisions after undoing existing revisions.
That is to say, some revisions that were in the old version of the
branch are not in the new version. This situation occurs
when a user --force pushes a change and generates a repository
containing something like this:
* -- * -- B -- O -- O -- O (befd95662c)
\
N -- N -- N refs/heads/unity (7e4d71cc1e)
You should already have received notification emails for all of the O
revisions, and so the following emails describe only the N revisions
from the common base, B.
Any revisions marked "omit" are not gone; other references still
refer to them. Any revisions marked "discard" are gone forever.
No new revisions were added by this update.
Summary of changes:
docs/install/from_source.rst | 4 +-
include/tvm/tir/builtin.h | 12 ++
include/tvm/tir/function.h | 7 +
include/tvm/tir/transform.h | 6 +
include/tvm/topi/transform.h | 16 +-
python/gen_requirements.py | 2 +-
python/tvm/auto_scheduler/cost_model/xgb_model.py | 202 ++++++++++++---------
python/tvm/autotvm/tuner/xgboost_cost_model.py | 201 +++++++++++---------
src/driver/driver_api.cc | 12 ++
src/target/source/codegen_cuda.cc | 31 ++++
src/tir/op/builtin.cc | 2 +
src/tir/schedule/primitive/compute_inline.cc | 10 +-
.../schedule/primitive/layout_transformation.cc | 49 +++--
src/tir/transforms/inject_ptx_ldg32.cc | 130 +++++++++++++
src/tir/transforms/primfunc_utils.cc | 4 +
...ir_ptx_cp_async.py => test_inject_ptx_ldg32.py} | 56 +++---
tests/python/unittest/test_tir_host_func.py | 79 ++++++++
.../unittest/test_tir_schedule_compute_inline.py | 74 ++++++++
.../unittest/test_tir_schedule_transform_layout.py | 51 ++++++
19 files changed, 731 insertions(+), 217 deletions(-)
create mode 100644 src/tir/transforms/inject_ptx_ldg32.cc
copy tests/python/unittest/{test_tir_ptx_cp_async.py =>
test_inject_ptx_ldg32.py} (55%)
create mode 100644 tests/python/unittest/test_tir_host_func.py
