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: Enable addressing modes for gc types #70749

Merged
merged 3 commits into from
Jun 15, 2022
Merged

JIT: Enable addressing modes for gc types #70749

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6ff0e0a
Enable LEA for gc types on arm64
EgorBo Jun 14, 2022
4debd6c
Fix arm32 build
EgorBo Jun 14, 2022
7dbaf15
Clean up
EgorBo Jun 15, 2022
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
10 changes: 10 additions & 0 deletions src/coreclr/jit/codegenarmarch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4485,6 +4485,16 @@ void CodeGen::genLeaInstruction(GenTreeAddrMode* lea)
}
else
{
#ifdef TARGET_ARM64
// Handle LEA with "contained" BFIZ
if (index->isContained() && index->OperIs(GT_BFIZ))
{
assert(scale == 0);
scale = (DWORD)index->gtGetOp2()->AsIntConCommon()->IconValue();
index = index->gtGetOp1()->gtGetOp1();
}
#endif

// Then compute target reg from [base + index*scale]
genScaledAdd(size, lea->GetRegNum(), memBase->GetRegNum(), index->GetRegNum(), scale);
}
Expand Down
2 changes: 1 addition & 1 deletion src/coreclr/jit/lower.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5443,7 +5443,7 @@ bool Lowering::TryCreateAddrMode(GenTree* addr, bool isContainable, GenTree* par

// Check if we can "contain" LEA(BFIZ) in order to extend 32bit index to 64bit as part of load/store.
if ((index != nullptr) && index->OperIs(GT_BFIZ) && index->gtGetOp1()->OperIs(GT_CAST) &&
index->gtGetOp2()->IsCnsIntOrI() && (varTypeIsIntegral(targetType) || varTypeIsFloating(targetType)))
index->gtGetOp2()->IsCnsIntOrI() && !varTypeIsStruct(targetType))
{
// BFIZ node is a binary op where op1 is GT_CAST and op2 is GT_CNS_INT
GenTreeCast* cast = index->gtGetOp1()->AsCast();
Expand Down
4 changes: 2 additions & 2 deletions src/coreclr/jit/morph.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4735,8 +4735,8 @@ GenTree* Compiler::fgMorphIndexAddr(GenTreeIndexAddr* indexAddr)
// we at least will be able to hoist/CSE "index + elemOffset" in some cases.
// See https://github.com/dotnet/runtime/pull/61293#issuecomment-964146497

// Use 2) form only for primitive types for now - it significantly reduced number of size regressions
if (!varTypeIsIntegral(elemTyp) && !varTypeIsFloating(elemTyp))
// Don't use 2) for structs to reduce number of size regressions
if (varTypeIsStruct(elemTyp))
{
groupArrayRefWithElemOffset = false;
}
Expand Down