News for hackers. Stuff that matters.
Re:ARM Cortex-A72 fetch and branch processing by cpgxiii on Monday December 14, 2020 @07:45PM (#25425577)
The reality is that the embarrassingly parallel nature of graphics has been long understood, as well as the advantages of having graphics be handled by a different, dedicated processor (see Intel's i860 in the early 1990s).

Unless in some alternate universe frequency scaling and memory bandwidth were never an issue, you're always going to come back to the reality that a parallel processor optimized for minimally-branchy math will be more efficient at graphics tasks. The early days of fixed-function hardware are, if anything, an aberration.

Re:ARM Cortex-A72 fetch and branch processing by Veedrac on Monday December 14, 2020 @08:11PM (#25425795)
And yet most non-games graphics workloads in industry were CPU rendered until 2010ish, I think. The draw of defaults is extremely strong.
Re:ARM Cortex-A72 fetch and branch processing by cpgxiii on Monday December 14, 2020 @09:25PM (#25426296)
No. Back to the very first graphical workstations, people have people putting dedicated processors in computers for graphics (ex Apollo Computer using the TMS320 DSP for graphics math). The entire reason for legacy decisions in OpenGL is from non-gaming users, namely CAD vendors, who have been doing this for a long time.
Re:ARM Cortex-A72 fetch and branch processing by Veedrac on Monday December 14, 2020 @09:43PM (#25426389)
The context is very different if you're talking about the early 1980s. I recognize (and have said before in this thread) that my arguments about lock-in don't apply to then.