Skip to content
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

JIT: Fold (x + c1) + c2 in lowering #108779

Merged
merged 4 commits into from
Oct 11, 2024
+52 −12
Merged

JIT: Fold (x + c1) + c2 in lowering #108779

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f97de19
JIT: Fold `(x + c1) + c2` in lowering
jakobbotsch Oct 11, 2024
42d4386
Fold before other transform, clear reg-optionality/containment
jakobbotsch Oct 11, 2024
4783457
Handle TYP_BYREF, TYP_REF arithmetic
jakobbotsch Oct 11, 2024
8b68b2e
Fix
jakobbotsch Oct 11, 2024
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
64 changes: 52 additions & 12 deletions src/coreclr/jit/lower.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6998,20 +6998,60 @@ GenTree* Lowering::LowerAdd(GenTreeOp* node)
return next;
}

// Fold ADD(CNS1, CNS2). We mainly target a very specific pattern - byref ADD(frozen_handle, cns_offset)
// We could do this folding earlier, but that is not trivial as we'll have to introduce a way to restore
// the original object from a byref constant for optimizations.
if (comp->opts.OptimizationEnabled() && op1->IsCnsIntOrI() && op2->IsCnsIntOrI() && !node->gtOverflow() &&
(op1->IsIconHandle(GTF_ICON_OBJ_HDL) || op2->IsIconHandle(GTF_ICON_OBJ_HDL)) &&
!op1->AsIntCon()->ImmedValNeedsReloc(comp) && !op2->AsIntCon()->ImmedValNeedsReloc(comp))
if (comp->opts.OptimizationEnabled())
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could be relaxed to Tier0OptimizationEnabled

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let me relax this in a separate PR since it will also enable the existing opt in tier 0

{
assert(node->TypeIs(TYP_I_IMPL, TYP_BYREF));
// Fold (x + c1) + c2
while (op1->OperIs(GT_ADD) && op2->IsIntegralConst() && op1->gtGetOp2()->IsIntegralConst() &&
!node->gtOverflow() && !op1->gtOverflow() && !op2->AsIntConCommon()->ImmedValNeedsReloc(comp) &&
!op1->gtGetOp2()->AsIntConCommon()->ImmedValNeedsReloc(comp))
{
JITDUMP("Folding (x + c1) + c2. Before:\n");
DISPTREERANGE(BlockRange(), node);

// TODO-CQ: we should allow this for AOT too. For that we need to guarantee that the new constant
// will be lowered as the original handle with offset in a reloc.
BlockRange().Remove(op1);
BlockRange().Remove(op2);
node->BashToConst(op1->AsIntCon()->IconValue() + op2->AsIntCon()->IconValue(), node->TypeGet());
int64_t c1 = op1->gtGetOp2()->AsIntConCommon()->IntegralValue();
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You could probably save several lines of code to do gtFoldExprConst(gtNewOperNode(GT_ADD, op1->gtGetOp2(), op2)) instead

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah yeah, I'll keep that in mind for the future

int64_t c2 = op2->AsIntConCommon()->IntegralValue();

int64_t result;
if (genTypeSize(node) == sizeof(int64_t))
{
result = c1 + c2;
}
else
{
result = static_cast<int32_t>(c1) + static_cast<int32_t>(c2);
}

op2->AsIntConCommon()->SetIntegralValue(result);
node->gtOp1 = op1->gtGetOp1();

BlockRange().Remove(op1->gtGetOp2());
BlockRange().Remove(op1);

op1 = node->gtGetOp1();

// We will recompute containment/reg optionality below.
op1->ClearRegOptional();
op1->ClearContained();

JITDUMP("\nAfter:\n");
DISPTREERANGE(BlockRange(), node);
}

// Fold ADD(CNS1, CNS2). We mainly target a very specific pattern - byref ADD(frozen_handle, cns_offset)
// We could do this folding earlier, but that is not trivial as we'll have to introduce a way to restore
// the original object from a byref constant for optimizations.
if (op1->IsCnsIntOrI() && op2->IsCnsIntOrI() && !node->gtOverflow() &&
(op1->IsIconHandle(GTF_ICON_OBJ_HDL) || op2->IsIconHandle(GTF_ICON_OBJ_HDL)) &&
!op1->AsIntCon()->ImmedValNeedsReloc(comp) && !op2->AsIntCon()->ImmedValNeedsReloc(comp))
{
assert(node->TypeIs(TYP_I_IMPL, TYP_BYREF));

// TODO-CQ: we should allow this for AOT too. For that we need to guarantee that the new constant
// will be lowered as the original handle with offset in a reloc.
BlockRange().Remove(op1);
BlockRange().Remove(op2);
node->BashToConst(op1->AsIntCon()->IconValue() + op2->AsIntCon()->IconValue(), node->TypeGet());
}
}

#ifdef TARGET_XARCH
Expand Down
Loading