http://www.lastrayofhope.com Home of Kaluriel Hargrove Wed, 08 Feb 2012 23:27:18 +0000 hourly 1 http://wordpress.org/?v=3.3.1 http://www.lastrayofhope.com/ios/comment-page-1/#comment-1769 Rob Napier Wed, 08 Feb 2012 23:27:18 +0000 http://www.lastrayofhope.com/?page_id=3933#comment-1769 I'm very interested in your experiences converting iOS code to vector code. I've been doing a lot of experimentation, and while some vector routines are clearly faster than their by-hand counterpart (vDSP_vramp() and vDSP_vsq() for instance), I have yet to come up with a piece of test data where vDSP_mmul() is faster than hand-multiplying the matrix (even when using a for() loop to do it). Is this a surprising result in your experience? I'm seeing the same behavior on Mac and iOS. I’m very interested in your experiences converting iOS code to vector code. I’ve been doing a lot of experimentation, and while some vector routines are clearly faster than their by-hand counterpart (vDSP_vramp() and vDSP_vsq() for instance), I have yet to come up with a piece of test data where vDSP_mmul() is faster than hand-multiplying the matrix (even when using a for() loop to do it). Is this a surprising result in your experience? I’m seeing the same behavior on Mac and iOS.

]]>