News: 0001493761

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

Mesa 24.3 Finishes The Years Long Effort To Phase Out The Old GLSL IR Linker

([Mesa] 3 Hours Ago Removing The Old GLSL IR Linker)


Timothy Arceri with the Valve Linux graphics team has merged the code for Mesa GLSL to convert to NIR at compile-time and in turn dropping the old GLSL IR linker with this being a multi-year effort now wrapped up for Mesa 24.3.

This finishes up work for going straight to the modern NIR intermediate representation at compile-time and does away with the old GLSL IR linker code in favor of a fully-NIR-based linker.

Arceri explained in the [1]merge request merged this morning for Q4's Mesa 24.3 release:

"This is the final MR after years of work by many people to get us to this point. This series moves glsl_to_nir() to compile time and replaces the remaining bits of the GLSL IR linker with a fully nir based linker.

The result here is a cleaner code-base that current mesa devs can better work with, support into the future and better understand. It also sets us up for other possible improvements such as improving the glsl shader cache to finally support storing compiled shaders and not just linked shaders."

Kudos to all those involved in this lengthy effort to help better modernize the Mesa code and further focus around NIR.

This final merge request touched more than two thousand lines of code. This work and a lot of other graphics driver changes can be found in [2]Mesa 24.3 that should be seeing its stable release around November.



[1] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31137

[2] https://www.phoronix.com/search/Mesa+24.3



skeevy420

After this was written there appeared a remarkable posthumous memoir that
throws some doubt on Millikan's leading role in these experiments. Harvey
Fletcher (1884-1981), who was a graduate student at the University of Chicago,
at Millikan's suggestion worked on the measurement of electronic charge for
his doctoral thesis, and co-authored some of the early papers on this subject
with Millikan. Fletcher left a manuscript with a friend with instructions
that it be published after his death; the manuscript was published in
Physics Today, June 1982, page 43. In it, Fletcher claims that he was the
first to do the experiment with oil drops, was the first to measure charges on
single droplets, and may have been the first to suggest the use of oil.
According to Fletcher, he had expected to be co-authored with Millikan on
the crucial first article announcing the measurement of the electronic
charge, but was talked out of this by Millikan.
-- Steven Weinberg, "The Discovery of Subatomic Particles"

Robert Millikan is generally credited with making the first really
precise measurement of the charge on an electron and was awarded the
Nobel Prize in 1923.