On Tue, Oct 11, 2011 at 08:56:21PM +0900, Carsten Haitzler wrote:
> On Mon, 3 Oct 2011 10:45:01 -0700 Jim Kukunas 
> <[email protected]>
> said:
> 
> found a rendering bug in your sse3 code. _op_blend_c_dpan_sse3 <- or more
> specifically that its #defined to equate to _op_blend_c_dp_sse3(). rendering
> bug can be seen in this app:
> 
> http://www.enlightenment.org/~raster/ello-0.1.0.0.tar.gz
> 
> notice the text has overflowed pixel values (green garbage where it should
> be orange) if sse3 (or you re-enable the function i just disabled in svn in
> op_blend_color_sse3.c). same with the orange rectangle - it goes green if that
> sse3 routine is on.

Interesting. When I run your app (ello, ello_tut01), I see orange, not green,
when using the SSE3 implementation and _op_blend_c_dpan_sse3 enabled. 

I'd like to be able to reproduce the issue to test any potential fixes, so could
you provide some platform information?

Thanks.

> 
> so for now that code path is disabled, probably dropping speed somewhat.
> 
> > On Sun, Oct 02, 2011 at 08:32:50PM +0900, Carsten Haitzler wrote:
> > > On Sat, 1 Oct 2011 10:12:11 -0700 Jim Kukunas
> > > <[email protected]> said:
> > > 
> > > ok. big problems with sse3 on 32bit. we have to have it disabled. why? you
> > > did it with intrinsics, and intrinsics fail without -msse3, BUT... -msse3
> > > builds code OPTIMIZED for sse3 - ie produces sse3 instructions even for
> > > regular c code. this means people compile evas and then have an x86 cpu
> > > incapable of sse3.. and presto. that binary doesnt work. that pretty much
> > > breaks backwards compatibility for x86 - packagers will have our throats
> > > for this.
> > > 
> > > so this is all bad. the runtime sse3 tests are pointless and moot as long 
> > > as
> > > we compile with -msse3. 
> > > 
> > > so we need sse3 asm that doesnt rely on -msse3 - ie like the mmx/sse was
> > > done. via macros that add real inlined assembly. :)
> > 
> > Darn. I should have caught that.
> > 
> > It appears that you fixed this issue in 63762/63775. Is there anything
> > else you need me to do?
> > 
> > Sorry for the inconvenience.
> > 
> > > 
> > > -- 
> > > ------------- Codito, ergo sum - "I code, therefore I am" --------------
> > > The Rasterman (Carsten Haitzler)    [email protected]
> > 
> > -- 
> > Jim Kukunas
> > Intel Open Source Technology Center
> > 
> > ------------------------------------------------------------------------------
> > All the data continuously generated in your IT infrastructure contains a
> > definitive record of customers, application performance, security
> > threats, fraudulent activity and more. Splunk takes this data and makes
> > sense of it. Business sense. IT sense. Common sense.
> > http://p.sf.net/sfu/splunk-d2dcopy1
> > _______________________________________________
> > enlightenment-devel mailing list
> > [email protected]
> > https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
> > 
> 
> 
> -- 
> ------------- Codito, ergo sum - "I code, therefore I am" --------------
> The Rasterman (Carsten Haitzler)    [email protected]

-- 
Jim Kukunas
Intel Open Source Technology Center

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2d-oct
_______________________________________________
enlightenment-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/enlightenment-devel

Reply via email to