Integrate LLVM at llvm/llvm-project@f7b2c2e4 #18143
Merged
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
Signed-off-by: hanhanW <[email protected]>
Abbreviated Benchmark Summary@ commit 473cbe7ec997040d4419af91d44835c757b65e67 (vs. base 4716f685cd74b09d85ec8b3ee515d6dcb72b8e2c) Data-Tiling Comparison TableClick to show
No improved or regressed benchmarks 🏖️ No improved or regressed compilation metrics 🏖️ For more information: |
Signed-off-by: hanhanW <[email protected]>
hanhanW
commented
Aug 7, 2024
Comment on lines
+88
to
+92
| // CHECK: ^bb2 | ||
| ^bb2(%bb2_0: !stream.resource<*>, %bb2_1: !stream.resource<*>): | ||
| // CHECK-NOT: stream.async.transfer | ||
| %external_transfer = stream.async.transfer %bb2_1 : !stream.resource<*>{%size} -> !stream.resource<external>{%size} | ||
| // CHECK: util.return %[[BB2_ARG_FILL0]], %[[BB2_ARG_SELECT]] : !stream.resource<transient>, !stream.resource<external> | ||
| // CHECK: util.return %[[FILL0]], %[[SELECT]] : !stream.resource<transient>, !stream.resource<external> |
There was a problem hiding this comment.
I confirmed that llvm/llvm-project@441b672 breaks the test. The new output looks okay to me..
Input IR:
util.func private @propagateBlocks(%cond: i1, %size: index) -> (!stream.resource<*>, !stream.resource<external>) {
%c0 = arith.constant 0 : index
%c128 = arith.constant 128 : index
%c123_i32 = arith.constant 123 : i32
%c456_i32 = arith.constant 456 : i32
%splat0 = stream.async.splat %c123_i32 : i32 -> !stream.resource<*>{%size}
%splat1 = stream.async.splat %c456_i32 : i32 -> !stream.resource<*>{%size}
cf.br ^bb1(%splat0, %splat1 : !stream.resource<*>, !stream.resource<*>)
^bb1(%bb1_0: !stream.resource<*>, %bb1_1: !stream.resource<*>):
%clone0 = stream.async.clone %bb1_0 : !stream.resource<*>{%size} -> !stream.resource<*>{%size}
%fill0 = stream.async.fill %c123_i32, %clone0[%c0 to %c128 for %c128] : i32 -> !stream.resource<*>{%size}
%clone1 = stream.async.clone %bb1_1 : !stream.resource<*>{%size} -> !stream.resource<*>{%size}
%fill1 = stream.async.fill %c456_i32, %clone1[%c0 to %c128 for %c128] : i32 -> !stream.resource<*>{%size}
%bb1_1_new = arith.select %cond, %splat1, %fill1 : !stream.resource<*>
cf.cond_br %cond, ^bb1(%fill0, %bb1_1_new : !stream.resource<*>, !stream.resource<*>),
^bb2(%fill0, %bb1_1_new : !stream.resource<*>, !stream.resource<*>)
^bb2(%bb2_0: !stream.resource<*>, %bb2_1: !stream.resource<*>):
%external_transfer = stream.async.transfer %bb2_1 : !stream.resource<*>{%size} -> !stream.resource<external>{%size}
util.return %bb2_0, %external_transfer : !stream.resource<*>, !stream.resource<external>
}The original output:
util.func private @propagateBlocks(%arg0: i1, %arg1: index) -> (!stream.resource<transient>, !stream.resource<external>) {
%c0 = arith.constant 0 : index
%c128 = arith.constant 128 : index
%c123_i32 = arith.constant 123 : i32
%c456_i32 = arith.constant 456 : i32
%0 = stream.async.splat %c123_i32 : i32 -> !stream.resource<transient>{%arg1}
%1 = stream.async.splat %c456_i32 : i32 -> !stream.resource<external>{%arg1}
cf.br ^bb1(%0, %1 : !stream.resource<transient>, !stream.resource<external>)
^bb1(%2: !stream.resource<transient>, %3: !stream.resource<external>): // 2 preds: ^bb0, ^bb1
%4 = stream.async.clone %2 : !stream.resource<transient>{%arg1} -> !stream.resource<transient>{%arg1}
%5 = stream.async.fill %c123_i32, %4[%c0 to %c128 for %c128] : i32 -> %4 as !stream.resource<transient>{%arg1}
%6 = stream.async.clone %3 : !stream.resource<external>{%arg1} -> !stream.resource<external>{%arg1}
%7 = stream.async.fill %c456_i32, %6[%c0 to %c128 for %c128] : i32 -> %6 as !stream.resource<external>{%arg1}
%8 = arith.select %arg0, %1, %7 : !stream.resource<external>
cf.cond_br %arg0, ^bb1(%5, %8 : !stream.resource<transient>, !stream.resource<external>), ^bb2(%5, %8 : !stream.resource<transient>, !stream.resource<external>)
^bb2(%9: !stream.resource<transient>, %10: !stream.resource<external>): // pred: ^bb1
util.return %9, %10 : !stream.resource<transient>, !stream.resource<external>
}New output:
util.func private @propagateBlocks(%arg0: i1, %arg1: index) -> (!stream.resource<transient>, !stream.resource<external>) {
%c0 = arith.constant 0 : index
%c128 = arith.constant 128 : index
%c123_i32 = arith.constant 123 : i32
%c456_i32 = arith.constant 456 : i32
%0 = stream.async.splat %c123_i32 : i32 -> !stream.resource<transient>{%arg1}
%1 = stream.async.splat %c456_i32 : i32 -> !stream.resource<external>{%arg1}
cf.br ^bb1(%0, %1 : !stream.resource<transient>, !stream.resource<external>)
^bb1(%2: !stream.resource<transient>, %3: !stream.resource<external>): // 2 preds: ^bb0, ^bb1
%4 = stream.async.clone %2 : !stream.resource<transient>{%arg1} -> !stream.resource<transient>{%arg1}
%5 = stream.async.fill %c123_i32, %4[%c0 to %c128 for %c128] : i32 -> %4 as !stream.resource<transient>{%arg1}
%6 = stream.async.clone %3 : !stream.resource<external>{%arg1} -> !stream.resource<external>{%arg1}
%7 = stream.async.fill %c456_i32, %6[%c0 to %c128 for %c128] : i32 -> %6 as !stream.resource<external>{%arg1}
%8 = arith.select %arg0, %1, %7 : !stream.resource<external>
cf.cond_br %arg0, ^bb1(%5, %8 : !stream.resource<transient>, !stream.resource<external>), ^bb2
^bb2: // pred: ^bb1
util.return %5, %8 : !stream.resource<transient>, !stream.resource<external>
}I think we can drop the bb2 arguments in this case. @benvanik does the fix make sense?
|
SGTM! |
benvanik
approved these changes
Aug 7, 2024
|
W7900 job failed because of #18140. The integrate PR itself is good, I'm landing it. |
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.
No description provided.