Open Bug 496503 Opened 16 years ago Updated 2 years ago

SIMD optimization of conversion to ARGB

Categories

(Core :: Graphics: ImageLib, defect)

x86
Windows NT
defect

Tracking

()

People

(Reporter: m_kato, Assigned: m_kato)

Details

Attachments

(1 file)

Attached patch patch v0Splinter Review
We can optimize ARGB conversion using SIMD (SSE2 or SSSE3). At result of simple unit test, this code will be improved about 50% on Core 2 processor.
Makoto, has this conversion been a bottleneck for you?
No, this is just a idea patch. This code can optimize code by SIMD.
Attachment #381710 - Flags: review?(joe)
Comment on attachment 381710 [details] [diff] [review] patch v0 It'll be easier for Jeff to review this than me.
Attachment #381710 - Flags: review?(joe) → review?(jmuizelaar)
Comment on attachment 381710 [details] [diff] [review] patch v0 In libjpegturbo we can decode directly to RGBX I think we should do this before worrying about optimizing this transformation.
Attachment #381710 - Flags: review?(jmuizelaar)
(In reply to Jeff Muizelaar [:jrmuizel] from comment #4) > Comment on attachment 381710 [details] [diff] [review] [diff] [details] [review] > patch v0 > > In libjpegturbo we can decode directly to RGBX I think we should do this > before worrying about optimizing this transformation. OK. about libjpeg-turbo, I will handle bug 584652. To use it, we need BGRA support on qcms and etc.
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: