x86 Ecosystem Advisory Group Posts ACE v2 & AVX10.3 Specs, GCC Patches Ready

While there aren't any processors released yet with AI Compute Extensions (ACE) or AVX10.2 quite yet, the x86 Ecosystem Advisory Group led by AMD and Intel have now posted their ACE v2 and AVX10.3 specifications. The ACE v2 and AVX10.3 architecture specifications have been posted for continuing to iterate on AVX and ACE for processors even further out on the roadmap. ACE v2 builds upon ACE with reduced-precision API with new 4-bit formats and expanded 8-bit format capabilities. ACE v2 also supports half-precision floating point (FP16) as well as single precision (FP32) and double precision (FP64) modes. AVX10.3 complements ACE v2 with more improvements for general purpose vectorization. New AVX10.3 instructions cover FP8 to BF16 up-convert, new BF16 to FP8 conversions, and FP16/BF16 to INT4 conversion. All of the ACE v2 and AVX10.3 changes are outlined on the x86ecosystem.org spec. With the new spec drop, GNU Compiler Collection (GCC) patches are already ready for review. Thanks to Intel compiler engineers there is this patch series for the new ACE instructions and another one for the new AVX10.3 instructions. Great seeing the punctual open-source compiler support continue ahead of CPUs with ACEv2 and AVX10.3 likely coming in the 2028~2029 timeframe.