Skip to content
GitLab

i#3044 AArch64 SVE codec: v8.6 BFloat16 and matrix multiply instructions

Merged Cam Mannett requested to merge i3044-bfloat16-matrix into master

This patch adds the appropriate macros, tests and codec entries to encode the following variants:

BFCVT   <Zd>.H, <Pg>/M, <Zn>.S
BFDOT   <Zda>.S, <Zn>.H, <Zm>.H
BFDOT   <Zda>.S, <Zn>.H, <Zm>.H[<index>]
BFMLALB <Zda>.S, <Zn>.H, <Zm>.H
BFMLALB <Zda>.S, <Zn>.H, <Zm>.H[<index>]
BFMLALT <Zda>.S, <Zn>.H, <Zm>.H
BFMLALT <Zda>.S, <Zn>.H, <Zm>.H[<index>]
BFMMLA  <Zda>.S, <Zn>.H, <Zm>.H
SMMLA   <Zda>.S, <Zn>.B, <Zm>.B
SUDOT   <Zda>.S, <Zn>.B, <Zm>.B[<index>]
UMMLA   <Zda>.S, <Zn>.B, <Zm>.B
USDOT   <Zda>.S, <Zn>.B, <Zm>.B
USDOT   <Zda>.S, <Zn>.B, <Zm>.B[<index>]
USMMLA  <Zda>.S, <Zn>.B, <Zm>.B

Issue #3044