Skip to content

Skip CSE for non-hoistable floats #57438

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

Merged
merged 9 commits into from
Aug 17, 2021
Merged

Skip CSE for non-hoistable floats #57438

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
00f9eab
Don't do CSE for simple non-hoistalbe expressions
EgorBo Aug 14, 2021
d297cd9
Limit to FP
EgorBo Aug 15, 2021
c6ff730
Clean up
EgorBo Aug 15, 2021
3e2725f
Address feedback
EgorBo Aug 15, 2021
9d2e5e5
Address feedback
EgorBo Aug 15, 2021
ea83988
formatting
EgorBo Aug 15, 2021
aa839b9
Tuning
EgorBo Aug 15, 2021
2a1f3b0
Better workaround
EgorBo Aug 16, 2021
205405a
Merge branch 'main' of https://github.com/dotnet/runtime into skip-cs…
EgorBo Aug 17, 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
11 changes: 9 additions & 2 deletions src/coreclr/jit/optcse.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2367,7 +2367,7 @@ class CSE_Heuristic

bool canEnregister = true;
unsigned slotCount = 1;
if (candidate->Expr()->TypeGet() == TYP_STRUCT)
if (candidate->Expr()->TypeIs(TYP_STRUCT))
{
// This is a non-enregisterable struct.
canEnregister = false;
Expand Down Expand Up @@ -2601,10 +2601,17 @@ class CSE_Heuristic
//
if (candidate->LiveAcrossCall())
{
if (candidate->Expr()->IsCnsFltOrDbl() && (CNT_CALLEE_SAVED_FLOAT == 0))
{
// We should do CSE for fp constants in case of LiveAcrossCall only when absolutely necessary
// on ABIs without callee-saved registers.
cse_use_cost += 2;
}
Copy link
Member

@jakobbotsch jakobbotsch Aug 16, 2021

Choose a reason for hiding this comment

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

I am ok with this more surgical fix for 6.0, but if we decide to go this route then you should open an issue about revisiting the float heuristics for .NET 7. It does not really make sense to adjust the use cost of a CSE based on the contents of the tree that is being CSE'd, I think.


// If we don't have a lot of variables to enregister or we have a floating point type
// then we will likely need to spill an additional caller save register.
//
if ((enregCount < (CNT_CALLEE_ENREG * 3 / 2)) || varTypeIsFloating(candidate->Expr()->TypeGet()))
if ((enregCount < (CNT_CALLEE_ENREG * 3 / 2)) || varTypeIsFloating(candidate->Expr()))
{
// Extra cost in case we have to spill/restore a caller saved register
extra_yes_cost = BB_UNITY_WEIGHT_UNSIGNED;
Expand Down
2 changes: 1 addition & 1 deletion src/coreclr/jit/optimizer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5955,7 +5955,7 @@ bool Compiler::optIsProfitableToHoistableTree(GenTree* tree, unsigned lnum)
int loopVarCount;
int varInOutCount;

if (varTypeIsFloating(tree->TypeGet()))
if (varTypeIsFloating(tree))
{
hoistedExprCount = pLoopDsc->lpHoistedFPExprCount;
loopVarCount = pLoopDsc->lpLoopVarFPCount;
Expand Down