Change contiguity into std::vector<c10::optional<bool>>
#2569
Conversation
This file contains hidden or 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
zasdfgbnm
commented
Mar 10, 2023
Comment on lines
-114
to
127
| " Send/Receive Val {T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ]} from cluster 0 to cluster 2\n" | ||
| " AggregateVal representing Val T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ] on cluster 2\n" | ||
| " AggregateExpr representing Cluster 2.Inputs={T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ], }. Outputs={T5_l[ rS12{i2}, iS13{i3} ], }.\n" | ||
| " AggregateVal representing Val T5_l[ rS12{i2}, iS13{i3} ] on cluster 2\n" | ||
| " Send/Receive Val {T5_l[ rS12{i2}, iS13{i3} ]} from cluster 2 to cluster 3\n" | ||
| " AggregateVal representing Val T5_l[ rS12{i2}, iS13{i3} ] on cluster 3\n" | ||
| " Send/Receive Val {T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ]} from cluster 0 to cluster 1\n" | ||
| " AggregateVal representing Val T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ] on cluster 1\n" | ||
| " AggregateExpr representing Cluster 1.Inputs={T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ], }. Outputs={T3_l[ rS8{i2}, iS9{i3} ], }.\n" | ||
| " AggregateVal representing Val T3_l[ rS8{i2}, iS9{i3} ] on cluster 1\n" | ||
| " Send/Receive Val {T3_l[ rS8{i2}, iS9{i3} ]} from cluster 1 to cluster 3\n" | ||
| " AggregateVal representing Val T3_l[ rS8{i2}, iS9{i3} ] on cluster 3\n" | ||
| " AggregateExpr representing Cluster 3.Inputs={T5_l[ rS12{i2}, iS13{i3} ], T3_l[ rS8{i2}, iS9{i3} ], }. Outputs={T6_g[ iS14{i3} ], }.\n" | ||
| " Send/Receive Val {T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ]} from cluster 0 to cluster 2\n" | ||
| " AggregateVal representing Val T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ] on cluster 2\n" | ||
| " AggregateExpr representing Cluster 2.Inputs={T1_l[ iS3{i0}, iS4{i2}, iS5{i3} ], }. Outputs={T5_l[ rS12{i2}, iS13{i3} ], }.\n" | ||
| " AggregateVal representing Val T5_l[ rS12{i2}, iS13{i3} ] on cluster 2\n" | ||
| " Send/Receive Val {T5_l[ rS12{i2}, iS13{i3} ]} from cluster 2 to cluster 3\n" | ||
| " AggregateVal representing Val T5_l[ rS12{i2}, iS13{i3} ] on cluster 3\n" | ||
| " AggregateExpr representing Cluster 3.Inputs={T3_l[ rS8{i2}, iS9{i3} ], T5_l[ rS12{i2}, iS13{i3} ], }. Outputs={T6_g[ iS14{i3} ], }.\n" | ||
| "}\n" |
There was a problem hiding this comment.
I don't know why this changed, and I don't know if this change is OK, it seems it is just some order change.
There was a problem hiding this comment.
@samnordmann will look into it. I think it's fine to change it for now.
@zasdfgbnm, please update the test if necessary.
|
I am marking this PR as draft because I need to wait #2561 and update the python frontend. But I ran C++ and torchscript tests, and they pass. So this test should be ready for review. I run tests with git revert --no-commit 3b85308a8e303d0df43c2d3cac1edba87dde2e49 |
zasdfgbnm
commented
Mar 10, 2023
|
Marking as ready-for-review, python frontend updated. |
naoyam
reviewed
Mar 10, 2023
naoyam
reviewed
Mar 10, 2023
naoyam
reviewed
Mar 10, 2023
naoyam
reviewed
Mar 10, 2023
naoyam
reviewed
Mar 10, 2023
naoyam
approved these changes
Mar 10, 2023
jjsjann123
reviewed
Mar 10, 2023
| @@ -425,12 +416,14 @@ class VectorizeValidator : public OptInDispatch { | |||
|
|
|||
| // Contiguity is based on rfactor domain. | |||
| IterDomain* last_root_dim = nullptr; | |||
| size_t last_root_dim_pos; | |||
There was a problem hiding this comment.
uninitialized last_root_dim_pos. If we have a tensor that's full size-1 (broadcasted), then we are *tv->domain()->contiguity().at(last_root_dim_pos), that would be a segfault/UB
There was a problem hiding this comment.
I think this will be avoided by the if (last_root_dim == nullptr) before the *tv->domain()->contiguity().at(last_root_dim_pos)
zasdfgbnm
commented
Mar 10, 2023
| @@ -6,6 +6,7 @@ | |||
| #include <torch/csrc/jit/ir/ir.h> | |||
| #include <type.h> | |||
| #include <array> | |||
| #include <optional> | |||
There was a problem hiding this comment.
@naoyam I fixed a compilation error:
/home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/executor_kernel_arg.h:282:18: error: ‘optional’ in namespace ‘std’ does not name a template type
282 | const std::optional<KernelIndexMode>& index_mode = std::nullopt);
zasdfgbnm
commented
Mar 10, 2023
Comment on lines
+81
to
+82
| executor_ptr->compileFusion( | ||
| fusion_from_cluster.get(), args, launch_params, {}); |
There was a problem hiding this comment.
This is another compilation error fix:
/home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/multidevice/multidevice_runtime.cpp:81:30: error: no matching function for call to ‘nvfuser::FusionExecutor::compileFusion(std::unique_ptr<nvfuser::Fusion>::pointer, nvfuser::KernelArgumentHolder&, nvfuser::LaunchParams&)’
81 | executor_ptr->compileFusion(fusion_from_cluster.get(), args, launch_params);
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from /home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/kernel_cache.h:4,
from /home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/fusion_segmenter.h:5,
from /home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/multidevice/multidevice_runtime.cpp:2:
/home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/executor.h:62:8: note: candidate: ‘void nvfuser::FusionExecutor::compileFusion(nvfuser::Fusion*, const nvfuser::KernelArgumentHolder&, const nvfuser::LaunchParams&, nvfuser::CompileParams)’
62 | void compileFusion(
| ^~~~~~~~~~~~~
/home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/executor.h:62:8: note: candidate expects 4 arguments, 3 provided
/home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/executor.h:71:8: note: candidate: ‘void nvfuser::FusionExecutor::compileFusion(nvfuser::Fusion*, const c10::ArrayRef<c10::IValue>&, const nvfuser::LaunchParams&, nvfuser::CompileParams)’
71 | void compileFusion(
| ^~~~~~~~~~~~~
/home/gaoxiang/nvfuser7/third_party/nvfuser/csrc/executor.h:73:40: note: no known conversion for argument 2 from ‘nvfuser::KernelArgumentHolder’ to ‘const c10::ArrayRef<c10::IValue>&’
73 | const at::ArrayRef<c10::IValue>& inputs = {},There was a problem hiding this comment.
Thanks. I guess I didn't see it as I had USE_DISTRIBUTED=0.
|
The build fix comes in just about time~~ Thanks for that |
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.
So that it has the same size as rfactor domain