Skip to content

Fix ARM64 unsigned div by const perf regression #57372

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
pentp wants to merge 3 commits into from
Closed

Fix ARM64 unsigned div by const perf regression #57372

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
976bf2a
Fix ARM64 unsigned div by const perf regression
pentp Aug 13, 2021
6dba123
Try fix formatting
pentp Aug 13, 2021
9a9d09d
Try to make formatting happy
pentp Aug 14, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Try to make formatting happy
  • Loading branch information
@pentp
pentp committed Aug 14, 2021
commit 9a9d09d32812cf4010e20e6949b7c617b13ba73d
11 changes: 5 additions & 6 deletions src/coreclr/jit/lower.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5261,7 +5261,7 @@ bool Lowering::LowerUnsignedDivOrMod(GenTreeOp* divMod)
#ifdef TARGET_ARM64
&& !simpleMul
#endif
)
)
{
adjustedDividend = comp->gtNewCastNode(TYP_I_IMPL, adjustedDividend, true, TYP_U_IMPL);
BlockRange().InsertBefore(divMod, adjustedDividend);
Expand All @@ -5286,16 +5286,15 @@ bool Lowering::LowerUnsignedDivOrMod(GenTreeOp* divMod)
}
else
{
// Insert a new GT_MULHI node before the existing GT_UDIV/GT_UMOD node.
// The existing node will later be transformed into a GT_RSZ/GT_SUB that
// computes the final result. This way don't need to find and change the use
// of the existing node.

#ifdef TARGET_ARM64
genTreeOps mulOp = simpleMul ? GT_MULWIDE : GT_MULHI; // 64-bit MUL is more expensive than UMULL on ARM64
#else
genTreeOps mulOp = simpleMul ? GT_MUL : GT_MULHI; // 64-bit IMUL is less expensive than MUL eax:edx on x64
#endif
// Insert a new GT_MULHI node before the existing GT_UDIV/GT_UMOD node.
// The existing node will later be transformed into a GT_RSZ/GT_SUB that
// computes the final result. This way don't need to find and change the use
// of the existing node.
GenTree* mulhi = comp->gtNewOperNode(mulOp, TYP_I_IMPL, adjustedDividend, divisor);
mulhi->gtFlags |= GTF_UNSIGNED;
BlockRange().InsertBefore(divMod, mulhi);
Expand Down