Closed Bug 704087 Opened 13 years ago Closed 13 years ago

Optimize LirHelper::atomToNative

Categories

(Tamarin Graveyard :: Baseline JIT (CodegenLIR), defect, P3)

Product:
Tamarin Graveyard
Component:
Baseline JIT (CodegenLIR)
Type:
defect

Tracking

(Not tracked)

Status:
VERIFIED FIXED
Milestone:
Q2 12 - Cyril

People

(Reporter: lhansen, Assigned: edwsmith)

References

Details

There are two TODO comments in this method that point to optimizations that would seem to be useful.  Needs investigation and implementation.
Edwin, can you spend a couple minutes on this and see if (a) the float4 case is correct (note the else branch) and if so (b) whether the float case can be rewritten in a similar way?
Assignee: nobody → edwsmith
I only see one TODO, maybe the code morphed since this bug was written.

LirHelper::atomToNative() is intended to be called when you have an atom that is known to be the native type boxed.  So, its inbounds to do const folding (embed native literal), and the only runtime branching required is for Number, where the atom has two possible reps.

so a) float4 case looks right, and b) yes.  calling singlePrecisionFloat() is overkill.
tr-float

changeset:   7066:eaae2888ad94
tag:         tip
user:        Edwin Smith <edwsmith@adobe.com>
date:        Mon Dec 12 13:55:13 2011 -0500
summary:     Bug 704087 - Optimize LirHelper::atomToNative
Status: NEW → RESOLVED
Closed: 13 years ago
Resolution: --- → FIXED
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.