Closed
Bug 742034
Opened 12 years ago
Closed 12 years ago
IonMonkey: Compile JSOP_THIS hard case
Categories
(Core :: JavaScript Engine, defect)
Core
JavaScript Engine
Tracking
()
RESOLVED
WORKSFORME
People
(Reporter: h4writer, Unassigned)
References
(Blocks 1 open bug)
Details
(Whiteboard: [ion:t])
According to AWFY we are still slower on SS 3d-cube. I reduced it to the following code: function CreateP(X, Y, Z) { this.V = [X, Y, Z, 1]; } function Init(CubeSize) { for (var i = 0; i < 320000; i++) CreateP(0, 0, 0); } Init() IM: 768ms JM+TI: 177ms V8: 177ms Haven't look into the cause yet.
Reporter | ||
Comment 1•12 years ago
|
||
The slowdown is because JSOP_THIS isn't supported in globalFrame yet. We bailout, so the function CreateP never gets compiled.
Reporter | ||
Updated•12 years ago
|
Summary: IonMonkey: slower on reduced SS 3d-cube benchmark → IonMonkey: Compile JSOP_THIS in a global frame
Reporter | ||
Comment 2•12 years ago
|
||
Or I wasn't awake when I reported this, or there were some changes. Because now it bails in JSOP_THIS the hard case. (What is more understandable. I was already puzzled how this could hit global frame) Even more reduced is the following testcase. Where the reason of performance loss is clearly the JSOP_THIS and not the array init. function CreateP() { this.V = 1 } function Init(CubeSize) { for (var i = 0; i < 320000; i++) CreateP(); } Init() JM+TI: 13ms V8: 47ms IM: 99ms
Summary: IonMonkey: Compile JSOP_THIS in a global frame → IonMonkey: Compile JSOP_THIS hard case
Updated•12 years ago
|
Whiteboard: [ion:t]
Reporter | ||
Comment 3•12 years ago
|
||
JM and IM now have the same performance
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → WORKSFORME
You need to log in
before you can comment on or make changes to this bug.
Description
•