T0b1/regalloc2
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Avoid stack-to-stack moves by allocating an extra spillslot and re-using the scratch reg instead.

Browse Source
This commit is contained in:
Chris Fallin
2021-06-10 22:36:02 -07:00
parent 09b2dd4e73
commit 1bd1248cb5
3 changed files with 123 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/lib.rs
View File

@@ -797,16 +797,9 @@ pub trait Function {
/// 128-bit vector value will require two slots. The regalloc will always
/// align on this size.
///
/// This passes the associated virtual register to the client as well,
/// because the way in which we spill a real register may depend on the
/// value that we are using it for. E.g., if a machine has V128 registers
/// but we also use them for F32 and F64 values, we may use a different
/// store-slot size and smaller-operand store/load instructions for an F64
/// than for a true V128.
///
/// (This trait method's design and doc text derives from
/// regalloc.rs' trait of the same name.)
fn spillslot_size(&self, regclass: RegClass, for_vreg: VReg) -> usize;
fn spillslot_size(&self, regclass: RegClass) -> usize;
/// When providing a spillslot number for a multi-slot spillslot,
/// do we provide the first or the last? This is usually related