Comments for Skitten http://skitten.org/blog Blog for stuff that isn't food, drink or not having a car Tue, 26 Jan 2010 10:39:30 +0000 http://wordpress.org/?v=2.9.2 hourly 1 Comment on Projection mapping Arthur’s Seat by Val hunzinger http://skitten.org/blog/2009/06/21/projection-mapping-arthurs-seat/comment-page-1/#comment-1683 Val hunzinger Tue, 26 Jan 2010 10:39:30 +0000 http://skitten.org/blog/?p=9#comment-1683 Hey Ciaran, This model looks amazing, I have a cheeky query - would it be possible to get a copy of the 3d mesh? I am working on a theoretical project near St.Anthonys chapel as part of my academic portfolio (I am an architect alumni of Edinburgh University). Look forward to hearing from you. Yours faithfully, Val Hey Ciaran,
This model looks amazing, I have a cheeky query – would it be possible to get a copy of the 3d mesh? I am working on a theoretical project near St.Anthonys chapel as part of my academic portfolio (I am an architect alumni of Edinburgh University).
Look forward to hearing from you.
Yours faithfully,
Val

]]> Comment on Comparing 160-bit hash values with SSE by onitake http://skitten.org/blog/2009/09/13/comparing-with-sse/comment-page-1/#comment-1429 onitake Mon, 21 Dec 2009 22:16:20 +0000 http://skitten.org/blog/?p=37#comment-1429 thank you for the post. i was optimizing a small vector math library i wrote and had some trouble with the result of cmpeqps. what is it good for if i have to transfer a whole xmm register into memory to evaluate the result? movmskps was exactly what i was looking for. in my case, the work wasn't futile: my hand-optimized dot product runs about 10% faster than the mulss/addss version generated by gcc 4.2's tree-ssa optimizer, while my cross product routine reduces benchmark runtime by 20%. not as much as i expected, but i'm sure there's still room for improvement. i suspect there are still some unneccessary memory accesses. they hurt performance much more than the occasional serialization... thank you for the post.
i was optimizing a small vector math library i wrote and had some trouble with the result of cmpeqps. what is it good for if i have to transfer a whole xmm register into memory to evaluate the result?
movmskps was exactly what i was looking for.

in my case, the work wasn’t futile: my hand-optimized dot product runs about 10% faster than the mulss/addss version generated by gcc 4.2’s tree-ssa optimizer, while my cross product routine reduces benchmark runtime by 20%. not as much as i expected, but i’m sure there’s still room for improvement.
i suspect there are still some unneccessary memory accesses. they hurt performance much more than the occasional serialization…

]]> Comment on Comparing 160-bit hash values with SSE by Joey http://skitten.org/blog/2009/09/13/comparing-with-sse/comment-page-1/#comment-1347 Joey Tue, 08 Dec 2009 02:44:53 +0000 http://skitten.org/blog/?p=37#comment-1347 This is funny. Then sad. I shall continue to avoid assembler. This is funny. Then sad. I shall continue to avoid assembler.

]]> Comment on Siggraph 2009 google calendars by Iliyan http://skitten.org/blog/2009/07/06/siggraph-2009-google-calendars/comment-page-1/#comment-55 Iliyan Sat, 25 Jul 2009 10:35:27 +0000 http://skitten.org/blog/?p=28#comment-55 Thank you soooo much :) Thank you soooo much :)

]]>