Remove PromoteTensorLoads pass, convert ExtractOp in TensorToFlow. #6852
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
ScottTodd
added
the
compiler/dialects
benvanik
approved these changes
Aug 25, 2021
ScottTodd
force-pushed
the
promote-tensor-loads
branch
from
34d1273
to
434c94a
Compare
ScottTodd
added
the
(deprecated) buildkite:benchmark-android
Abbreviated Benchmark Summary@ commit 434c94aa23bc5215ca73902a9cc3b948cba8fdde (vs. base ce7b475992b38588fe73d8a16b4d2ab64284f8ea) Regressed Benchmarks 🚩
Improved Benchmarks 🎉
For more information: |
|
Benchmarks regressed a bit. May need to diff the IR for MobileNetV2 through the Vulkan target to see what happened... |
IR diff shows no change. Benchmark results appear to be in the noise on closer inspection.
|
Merged
copybara-service bot
pushed a commit
that referenced
this pull request
Aug 25, 2021
* 30463f1 Adding #util.composite attribute. (#6854) * 7a1c579 Implement function.py return type coercion (#6832) * d90f0fc Remove PromoteTensorLoads pass, convert ExtractOp in TensorToFlow. (#6852) * 7fa8c20 Update SwiftShader to 2021-08-25 (#6859) * a328761 Rename Bazel repo iree_vulkan_headers to vulkan_headers (#6862) * 7660e49 Merge pull request #6856 from google/benvanik-shared-target-backend * ebc5eb5 Removing the pipeline caching during executable translation. (#6855) * 2fb7a8d Reapply "Update TFLite concrete function conversion codes" (#6800) * ce7b475 NFC: Merge ConvertToFlow passes into dedicated before/after passes. (#6850) * 9fce2d6 Support f32 in the VM by default in the compiler. (#6744) * b6baea9 Support global ref ops and fix passing of refs on function boundaries in the C.. * 4b5d2ed Bump flatcc version (#6853) * c8a8f5d Add e2e tests for mhlo.bitcast_convert (#6846) COPYBARA_INTEGRATE_REVIEW=#6865 from hanhanW:main-to-google 30463f1 PiperOrigin-RevId: 392952973
iree-github-actions-bot
pushed a commit
that referenced
this pull request
Aug 25, 2021
* 30463f1 Adding #util.composite attribute. (#6854) * 7a1c579 Implement function.py return type coercion (#6832) * d90f0fc Remove PromoteTensorLoads pass, convert ExtractOp in TensorToFlow. (#6852) * 7fa8c20 Update SwiftShader to 2021-08-25 (#6859) * a328761 Rename Bazel repo iree_vulkan_headers to vulkan_headers (#6862) * 7660e49 Merge pull request #6856 from google/benvanik-shared-target-backend * ebc5eb5 Removing the pipeline caching during executable translation. (#6855) * 2fb7a8d Reapply "Update TFLite concrete function conversion codes" (#6800) * ce7b475 NFC: Merge ConvertToFlow passes into dedicated before/after passes. (#6850) * 9fce2d6 Support f32 in the VM by default in the compiler. (#6744) * b6baea9 Support global ref ops and fix passing of refs on function boundaries in the C.. * 4b5d2ed Bump flatcc version (#6853) * c8a8f5d Add e2e tests for mhlo.bitcast_convert (#6846) PiperOrigin-RevId: 392952973
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.
Fixes #6756 (the
tosa if.mlirfile compiles successfully using-iree-flow-enable-linalg-detensorizewith this change)The
PromoteTensorLoadspass was convertingi1loads toi8loads usingZeroExtendIOpandTruncateIOp. That was producing weird cycles during compilation when detensoring was applied, andflowops should be fine with i1 types. We still need to handlei1types when going to the HAL (since storage is incompatible) on the outside (external interface) and inside (codegen).