[TIR] Add utility for anchor block extraction #13194
Merged
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
Collaborator
|
Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.
Generated by tvm-bot |
masahi
force-pushed
the
tir-anchor-block
branch
from
ff9689f to
0ba3873
Compare
masahi
commented
Oct 25, 2022
| throw; | ||
| } | ||
|
|
||
| const PrimFuncNode* FindEntryFunc(const IRModule& mod, GlobalVar* result_g_var) { |
Member
Author
There was a problem hiding this comment.
Moved to src/tir/analysis/stmt_finding.cc
masahi
commented
Oct 25, 2022
| * \param result_g_var The result GlobalVar of the entry function. | ||
| * \return The entry function. | ||
| */ | ||
| const PrimFuncNode* FindEntryFunc(const IRModule& mod, GlobalVar* result_g_var); |
Member
Author
There was a problem hiding this comment.
Moved to include/tir/analysis.h
xinetzone
pushed a commit
to daobook/tvm
that referenced
this pull request
Nov 10, 2022
I'm working on enabling "anchor block" tuning for MS. This is a utility to extract anchor blocks. I define the "anchor block" to be the block (1) with an init statement and (2) having the biggest flops count. The latter condition is only used when there are multiple blocks with an init statement. For example, if the input module is conv2d + fused spatial blocks, conv2d is the anchor block. A module created from winograd convolution has multiple blocks with an init statement (input transform, batched GEMM, and output transform). We use the second condition, the flops count, to determine that the batched GEMM block is the anchor block.
xinetzone
pushed a commit
to daobook/tvm
that referenced
this pull request
Nov 25, 2022
I'm working on enabling "anchor block" tuning for MS. This is a utility to extract anchor blocks. I define the "anchor block" to be the block (1) with an init statement and (2) having the biggest flops count. The latter condition is only used when there are multiple blocks with an init statement. For example, if the input module is conv2d + fused spatial blocks, conv2d is the anchor block. A module created from winograd convolution has multiple blocks with an init statement (input transform, batched GEMM, and output transform). We use the second condition, the flops count, to determine that the batched GEMM block is the anchor block.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
I'm working on enabling "anchor block" tuning for MS. This is a utility to extract anchor blocks.
I define the "anchor block" to be the block (1) with an init statement and (2) having the biggest flops count. The latter condition is only used when there are multiple blocks with an init statement.
For example, if the input module is conv2d + fused spatial blocks, conv2d is the anchor block. A module created from winograd convolution has multiple blocks with an init statement (input transform, batched GEMM, and output transform). We use the second condition, the flops count, to determine that the batched GEMM block is the anchor block.
cc @junrushao @Hzfengsy