[prev in list] [next in list] [prev in thread] [next in thread]
List: llvm-bugs
Subject: [llvm-bugs] [Bug 90680] Bug or insufficient utility in `transform.structured.split` op
From: LLVM Bugs via llvm-bugs <llvm-bugs () lists ! llvm ! org>
Date: 2024-04-30 22:46:41
Message-ID: 20240430224641.74467710d7142ea1 () email ! llvm ! org
[Download RAW message or body]
[Attachment #2 (text/html)]
<table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>Issue</th>
<td>
<a href=https://github.com/llvm/llvm-project/issues/90680>90680</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>
Bug or insufficient utility in `transform.structured.split` op
</td>
</tr>
<tr>
<th>Labels</th>
<td>
</td>
</tr>
<tr>
<th>Assignees</th>
<td>
</td>
</tr>
<tr>
<th>Reporter</th>
<td>
srcarroll
</td>
</tr>
</table>
<pre>
First of all I've identified an inconsistency in the way static and dynamic \
split points are handled. basically, no matter what in the static case, a single \
split point is provided and works no matter how many target payloads there are (see \
https://github.com/llvm/llvm-project/blob/main/mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp#L2208). \
Whereas with the dynamic case, the number of split points is determined by the number \
of payloads of the dynamic split point operand (see \
https://github.com/llvm/llvm-project/blob/main/mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp#L2208). \
but then the number of split points is compared to the number of target payloads \
(https://github.com/llvm/llvm-project/blob/main/mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp#L2200). \
I can't quite put my finger on it, but this seems wrong to me. For example, \
consider the edge case where there are no target payloads. In this case the static \
doesn't throw any definite failures while the dynamic case does. I think at the \
very least, there should be consistency here. but the question is, which one do we go \
with. I actually prefer how the static case works for this. because otherwise we \
would have to know that a payload is present to be able to use the op in the first \
place, which i don't think is very useful.
I was working on a patch for this to handle some case i needed. but then i realized \
that there was this discrepancy between the number of split point payloads and number \
of target payloads. I admit I'm a little confused about what the relationship \
between these two things should actually be. But I'm almost certain they \
shouldn't necessarily be the same size. </pre>
<img width="1px" height="1px" alt="" \
src="http://email.email.llvm.org/o/eJzMVUFv40YP_TXyhfgEWbGt5ODD7hcYCLBALwV6pjSUxGY0ox1 \
yonp_fTEj23FStEVvm0MEjDnk4-PjGxThwREdi_3XYv-8waijD0cJHYbgrd203pyPJw6i4HtAa-GlqJs3Ajbkl \
HsmA-iAXeedsCi57gzsQEeCBc8gisodoDNgzg4n7kBmywqzZ6cCGAhGdMaSKQFaFO7Q2nNR_x-chwlVKcAyol6 \
TXhJ2KJSCEITdYOk-K7DAHPwbmwzOwOLDq9zlG_0CE7ozKIaBFGY8W49GUoFAGVNRPwoRjKqzFA9fivpU1KeBd \
Yxt2fmpqE_Wvl0__5uD_506LepTa31b1KcJ2aWP5ZBiOJ09M9o16Bs7tENRn34N6KT3Yfplltv5_WHZzXNRP3y \
r6-qxqJ9K-C0hRIGFdcx8XFm9EpLOXJxaCmlgH7hmAUNKYWJHBtrzp9gbC77_kPmeWT9TSIz-rPS0URN29y80d \
H6aMZAB9Z8iPyuiqB9_oiar3GRRPRfVl_X_C3ToirpR-B5ZCeaoMJ2hZzekhhywJlWsvLCAEE0CS_BuSM1PVML \
JB6A_cJptFlBeZEMhE0NmoCwtWPJqvC-I85-5KgFe3Fol37jbVuNJVpQ6Br9A2j1DPbsEuUe2MZDAMrKlv4g6X \
065U2r3CpgnDG8UzmAJRS-qDwQy-mgNtAT3bpR-uikDvkcS5USMpIvLyN0I3qUysBAMPq9WroedxmRGMAfqL7b \
xyYIu1tL7kBsvoaUOoxD4hGjhFJGCEqwR3yhx_upyHlTAK3erY5GQ0xTREmBrc3C8EOnnqwH22Yxnix29N8Bg_ \
JXgRBLLSlAU6qMtPypmSf7hwyu7ISkkodBuvDWRyq6WDOKny_gZHJFJJn1bMYZAaPlHWqNxnUqgnDxnMSxdoBn \
TCFrShf5pK98XLrnL365jmaZiJtb8Ck2AYFnV5nn3UZLdtz7q-mCkYoEspmnLyPM9isTq4jNZg1x1c5t3SyXA1 \
3irYicvCh0FxXUI58uVlXJHHYlg4Hx1lQhOBMI_qNyY44N5enjCDR23zXa33zZ1td2Mxz3mP9ofqqqpd9Qb2j8 \
1_WHfHqr-0DQbPtZVvat2D1Vd7w67bdnsdoem2Vam2e5qwm2xq2hCtmXyn9KHYcMikY5P1eGx2lhsycr1YQ_Hb \
FJtHKTYVZZF5f2aslo6fo0D-ADsJPY9d5zEGJUta37Ri0OlV1cqRUPsNAYyZZ5icajAz5sY7PE_22UGnawv4_4zAAD__wZH_cM">
[Attachment #3 (text/plain)]
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic