Closed
Bug 533802
Opened 16 years ago
Closed 15 years ago
x64 version of asm_load64 doesn't attempt to use FpRegs for 64-bit loads
Categories
(Core Graveyard :: Nanojit, defect)
Tracking
(Not tracked)
RESOLVED
INVALID
Future
People
(Reporter: stejohns, Unassigned)
Details
It calls regalloc_load to get a register, which implicitly limits to GpRegs. We could trivially pass in a mask to allow FpRegs as well, allowing greater register utilization.
Quoth Edwin: "the risk is that if you load something that is later used as a pointer, then you have to move it to a GPR first. since you'll see the pointer-use first (bottom up), it wont be a problem when the uses are close together. for spill/restore situations we may need a hint flag that a GPR is preferred."
So this probably needs some profiling at before such a patch can land.
Updated•16 years ago
|
Component: JIT Compiler (NanoJIT) → Nanojit
Product: Tamarin → Core
QA Contact: nanojit → nanojit
Target Milestone: --- → Future
Comment 1•15 years ago
|
||
now that LIR distinguishes quad from double, this is obsolete.
Status: NEW → RESOLVED
Closed: 15 years ago
Resolution: --- → INVALID
| Assignee | ||
Updated•12 years ago
|
Product: Core → Core Graveyard
You need to log in
before you can comment on or make changes to this bug.
Description
•