News: 0001536469

  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)

Intel's AVX10.2 Patches Merged For GCC 15 To Drop 256-bit Rounding & AVX10.2-256 Options

([Intel] 3 Hours Ago Intel AVX10.2)


What a week. Last week Intel published a new AVX10 whitepaper where [1]they dropped the optional 512-bit support of AVX10.2 and confirmed future P and E cores will have AVX10.2-512 support unconditionally. A very welcome change by Intel albeit late in rushing to get patches out to change that behavior ahead of the GCC 15 stable compiler release as well as [2]working similar changes into the LLVM Clang compiler . As of today those GNU Compiler Collection patches have been merged to prepare for AVX10.2 always having 512-bit support available.

The set of patches by Intel compiler engineers to drop the AVX10.2 256-bit rounding support now that it's removed from the whitepaper as well as dropping the AVX10.2-256 / AVX10.2-512 options have been merged today to GCC Git ahead of the upcoming GCC 15.1 release. GCC 15.1 is to be the first stable release of the [3]GCC 15 compiler series as this annual feature release. GCC 15.1 should be out in the coming weeks, [4]is already in its final stage of development , and really was a last minute change for Intel in deciding to do away with AVX10.2-256 / optional 512-bit support considering all their previously merged patches for AVX10.2 had made those accommodations.

The clearest communication on the matter is [5]this patch dropping the avx10.2-256 and avx10.2-512 options where from Intel it's spelled clear as day that both P and E cores will support AVX10.2 512-bit vector widths:

"When AVX10.2 options are added into GCC 15, E-core is supposed to support up to 256 bit vector width, while P-core up to 512 bit vector width. Therefore, we added avx10.2-256 and avx10.2-512 options into compiler since there will be real platforms with 256 bit only support.

However, all the future platforms will now support 512 bit vector width, including P-core and E-core. It will result in no need for split the option for vector width. Therefore, we will remove them in this patch."

The code is merged in [6]GCC Git and thus GCC 15.1 is in good shape for the future Intel E and P cores supporting AVX10.2 in all of its 512-bit glory.

Other changes to find with the upcoming GCC 15.1 release include moving the default C language to C23, Intel Xeon 7 Diamond Rapids support, Fujitsu Monaka CPU support, initial Intel APX enablement, retiring of Xeon Phi support, new AMD Zen optimizations, much better Rust language support, the new COBOL front-end, and many other C and C++ language support improvements.



[1] https://www.phoronix.com/news/Intel-AVX10-Drops-256-Bit

[2] https://www.phoronix.com/news/LLVM-Clang-AVX10.2-Always-512

[3] https://www.phoronix.com/search/GCC+15

[4] https://www.phoronix.com/news/GCC-15.1-Coming-Soon

[5] https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=69d640d5f83eb9d49ab6dfde4453c68bfb587a27

[6] https://gcc.gnu.org/git/?p=gcc.git;a=shortlog



Anux

Research is what I'm doing when I don't know what I'm doing.
-- Wernher von Braun