Intel Granite Rapids D Support Merged Into GCC 14

As a follow-up to the news last week when I pointed out that Intel has begun working on Granite Rapids D compiler support for the GNU Compiler Collection, that code has now been merged for next year's GCC 14 release.

Last year Intel contributed Granite Rapids support for GCC 13. Now though the newest target is Granite Rapids D as the eventual successor to Ice Lake D CPUs. Making last week's Granite Rapids D compiler patches all the more notable is that it now confirms the new AMX-COMPLEX instructions will only be found in Granite Rapids D and not Granite Rapids itself. The GCC compiler had originally positioned the AMX-COMPLEX support as premiering with Granite Rapids while now the target is being adjusted to point to the GNR-D introduction.

As of this morning the Granite Rapids D support landed in GCC Git that will be found in the GCC 14.1 stable release due out in early 2024. The compiler support enables the following extensions for Granite Rapids D:

"MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID, AVX512VPOPCNTDQ, PCONFIG, WBNOINVD, CLWB, MOVDIRI, MOVDIR64B, AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, PTWRITE, WAITPKG, SERIALIZE, TSXLDTRK, UINTR, AMX-BF16, AMX-TILE, AMX-INT8, AVX-VNNI, AVX512FP16, AVX512BF16, AMX-FP16, PREFETCHI and AMX-COMPLEX instruction set support."

GCC 13 in addition to the Granite Rapids (non-D) support also already has Meteor Lake and Sierra Forest support, Emerald Rapids, and Grand Ridge. Intel continues to be very punctual in enabling new CPU support within the upstream GCC compiler well ahead of launch as well as for the LLVM/Clang compiler too.