[JIT] ARM64 - Fix Assertion failed 'node->OperIs(GT_DIV, GT_UDIV, GT_MOD)' during 'Lowering nodeinfo' #85664
Conversation
… parameter for the next node to lower
dotnet-issue-labeler
bot
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch Issue DetailsResolves #85630 DescriptionThere was a case where the node was lowered from Best way to solve this is return a
|
|
@dotnet/jit-contrib @kunalspathak this is ready - no diffs Current CI failure is unrelated. |
| { | ||
| assert((node->OperGet() == GT_DIV) || (node->OperGet() == GT_MOD)); | ||
| assert(nextNode != nullptr); |
There was a problem hiding this comment.
Should this be initialized *nextNode = nullptr which should be the case if this method returns false?
There was a problem hiding this comment.
We have a similar TryLowerAndNegativeOne with an out parameter nextNode and we do not set that to nullptr or when we return false.
| { | ||
| return newNode; | ||
| return nextNode; |
There was a problem hiding this comment.
assert(nextNode != nullptr)?
There was a problem hiding this comment.
The issue is that nextCode can be a nullptr even if the transformation was successful.
But, we could add an assert assert(nextNode == nullptr); if the transformation wasn't successful.
ghost
added
needs-author-action
|
@kunalspathak this is ready again |
|
Surprisingly the simplified test is still not simple, but don't want to waste your time trying to trim it further. |
|
Merging. CI failure is unrelated. |
Resolves #85630
Description
There was a case where the node was lowered from
LowerModPow2butnode->gtNextwould returnnullptrfromLowerConstIntDivOrModwhich indicates the node was not transformed when it actually was.Best way to solve this is return a
boolindicating if the transformation was successful or not and also provide an out parameter for the next node to lower. So, this renamesLowerConstIntDivOrModtoTryLowerConstIntDivOrModwith an out parameter for the next node to lower.