Multi-vector polynomial multiply long and accumulate
This instruction performs a polynomial multiplication over [0, 1] of the even-numbered elements of the first source vector by the corresponding elements of the second source vector, and then performs a bitwise exclusive-OR between the result and the overlapping double-width elements of the first destination vector. The same operation is performed with odd-numbered elements of the source vectors, writing to the second destination vector. This instruction is unpredicated.
This instruction is legal when executed in Streaming SVE mode if both FEAT_SSVE_AES and FEAT_SVE_AES2 are implemented.
| 31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
| 0 | 1 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 1 | Zm | 1 | 1 | 1 | 1 | 1 | 1 | Zn | Zda | 0 | |||||||||||
| size | |||||||||||||||||||||||||||||||
if !IsFeatureImplemented(FEAT_SVE_AES2) then EndOfDecode(Decode_UNDEF); end; let esize : integer{} = 128; let n : integer = UInt(Zn); let m : integer = UInt(Zm); let da : integer = UInt(Zda::'0');
| <Zda1> |
Is the name of the first scalable vector register of the destination multi-vector group, encoded as "Zda" times 2. |
| <Zda2> |
Is the name of the second scalable vector register of the destination multi-vector group, encoded as "Zda" times 2 plus 1. |
| <Zn> |
Is the name of the first source scalable vector register, encoded in the "Zn" field. |
| <Zm> |
Is the name of the second source scalable vector register, encoded in the "Zm" field. |
if IsFeatureImplemented(FEAT_SSVE_AES) then CheckSVEEnabled(); else CheckNonStreamingSVEEnabled(); end; let VL : integer{} = CurrentVL(); let elements : integer = VL DIV esize; let operand1 : bits(VL) = Z{}(n); let operand2 : bits(VL) = Z{}(m); var result_lo : bits(VL) = Z{}(da + 0); var result_hi : bits(VL) = Z{}(da + 1); for e = 0 to elements-1 do let element1_lo : bits(esize DIV 2) = operand1[(2*e + 0)*:(esize DIV 2)]; let element2_lo : bits(esize DIV 2) = operand2[(2*e + 0)*:(esize DIV 2)]; let element1_hi : bits(esize DIV 2) = operand1[(2*e + 1)*:(esize DIV 2)]; let element2_hi : bits(esize DIV 2) = operand2[(2*e + 1)*:(esize DIV 2)]; let product_lo : bits(esize) = PolynomialMult{esize DIV 2, esize DIV 2}(element1_lo, element2_lo); let product_hi : bits(esize) = PolynomialMult{esize DIV 2, esize DIV 2}(element1_hi, element2_hi); result_lo[e*:esize] = result_lo[e*:esize] XOR product_lo; result_hi[e*:esize] = result_hi[e*:esize] XOR product_hi; end; Z{VL}(da + 0) = result_lo; Z{VL}(da + 1) = result_hi;
2026-03_rel 2026-03-26 20:48:11
Copyright © 2010-2026 Arm Limited or its affiliates. All rights reserved. This document is Non-Confidential.