(no commit message)
authorlkcl <lkcl@web>
Fri, 29 Jul 2022 05:40:06 +0000 (06:40 +0100)
committerIkiWiki <ikiwiki.info>
Fri, 29 Jul 2022 05:40:06 +0000 (06:40 +0100)
openpower/sv/comparison_table.mdwn

index 2781a581c5a43877ca8bd87b89bfd2f7cfee5331..9b48012decd5a21af86ac58aa93d02323457cd2e 100644 (file)
@@ -47,7 +47,7 @@
 * (29): SVE: 4140 intrinsics, SVE2 1900 intrinsics
 * (30): Count includes SSE, SSE2, AVX, AVX2 and all AVX512 variants
 * (31): [RVV intrinsics listing](https://raw.githubusercontent.com/riscv-non-isa/rvv-intrinsic-doc/master/intrinsic_funcs.md) page is 25,000 lines long.
-* (32): Unknown. estimated to be of the order of length of RVV due to also being a Cray-style Scalable ISA.
+* (32): Unknown. estimated to be of the order of length of RVV due to also being a Cray-style Scalable ISA, NEC maintains an [LLVM hard fork](https://github.com/sx-aurora-dev)
 * (33): [Scalable Matrix Optional Extension](https://community.arm.com/arm-community-blogs/b/architectures-and-processors-blog/posts/scalable-matrix-extension-armv9-a-architecture)
   the key is an outer-product instruction [SMOPA](https://developer.arm.com/documentation/ddi0602/2022-06/SME-Instructions/SMOPA--Signed-integer-sum-of-outer-products-and-accumulate-?lang=en) which is very hard to tell at a glance if it is power-2 or non-power-2
 * (34:) [Advanced matrix Extensions](https://en.wikipedia.org/wiki/Advanced_Matrix_Extensions) supports BF16 and INT8 only. Separate regfile, power-of-two "tiles". Not general-purpose at all.