MOVA (tile to vector, four registers)

Move four ZA tile slices to four vector registers

The instruction operates on four consecutive horizontal or vertical slices within a named ZA tile of the specified element size.

The consecutive slice numbers within the tile are selected starting from the sum of the slice index register and immediate offset, modulo the number of such elements in a vector. The immediate offset is a multiple of 4 in the range 0 to the number of elements in a 128-bit vector segment minus 4.

This instruction is unpredicated.

This instruction is used by the alias MOV (tile to vector, four registers).

It has encodings from 4 classes: 8-bit , 16-bit , 32-bit and 64-bit

8-bit
(FEAT_SME2)

313029282726252423222120191817161514131211109876543210
1100000000000110VRs001000off2Zd00
size<1>size<0>

MOVA { <Zd1>.B-<Zd4>.B }, ZA0<HV>.B[<Ws>, <offs1>:<offs4>]

if !IsFeatureImplemented(FEAT_SME2) then UNDEFINED; constant integer s = UInt('011':Rs); constant integer nreg = 4; constant integer esize = 8; constant integer d = UInt(Zd:'00'); constant integer n = 0; constant integer offset = UInt(off2:'00'); constant boolean vertical = V == '1';

16-bit
(FEAT_SME2)

313029282726252423222120191817161514131211109876543210
1100000001000110VRs001000ZAno1Zd00
size<1>size<0>

MOVA { <Zd1>.H-<Zd4>.H }, <ZAn><HV>.H[<Ws>, <offs1>:<offs4>]

if !IsFeatureImplemented(FEAT_SME2) then UNDEFINED; constant integer s = UInt('011':Rs); constant integer nreg = 4; constant integer esize = 16; constant integer d = UInt(Zd:'00'); constant integer n = UInt(ZAn); constant integer offset = UInt(o1:'00'); constant boolean vertical = V == '1';

32-bit
(FEAT_SME2)

313029282726252423222120191817161514131211109876543210
1100000010000110VRs001000ZAnZd00
size<1>size<0>

MOVA { <Zd1>.S-<Zd4>.S }, <ZAn><HV>.S[<Ws>, <offs1>:<offs4>]

if !IsFeatureImplemented(FEAT_SME2) then UNDEFINED; constant integer s = UInt('011':Rs); constant integer nreg = 4; constant integer esize = 32; constant integer d = UInt(Zd:'00'); constant integer n = UInt(ZAn); constant integer offset = 0; constant boolean vertical = V == '1';

64-bit
(FEAT_SME2)

313029282726252423222120191817161514131211109876543210
1100000011000110VRs00100ZAnZd00
size<1>size<0>

MOVA { <Zd1>.D-<Zd4>.D }, <ZAn><HV>.D[<Ws>, <offs1>:<offs4>]

if !IsFeatureImplemented(FEAT_SME2) then UNDEFINED; if MaxImplementedSVL() < 256 then UNDEFINED; constant integer s = UInt('011':Rs); constant integer nreg = 4; constant integer esize = 64; constant integer d = UInt(Zd:'00'); constant integer n = UInt(ZAn); constant integer offset = 0; constant boolean vertical = V == '1';

Assembler Symbols

<Zd1>

Is the name of the first scalable vector register of the destination multi-vector group, encoded as "Zd" times 4.

<Zd4>

Is the name of the fourth scalable vector register of the destination multi-vector group, encoded as "Zd" times 4 plus 3.

<ZAn>

For the 16-bit variant: is the name of the ZA tile ZA0-ZA1 to be accessed, encoded in the "ZAn" field.

For the 32-bit variant: is the name of the ZA tile ZA0-ZA3 to be accessed, encoded in the "ZAn" field.

For the 64-bit variant: is the name of the ZA tile ZA0-ZA7 to be accessed, encoded in the "ZAn" field.

<HV>

Is the horizontal or vertical slice indicator, encoded in V:

V <HV>
0 H
1 V
<Ws>

Is the 32-bit name of the slice index register W12-W15, encoded in the "Rs" field.

<offs1>

For the 8-bit variant: is the first slice index offset, encoded as "off2" field times 4.

For the 16-bit variant: is the first slice index offset, encoded as "o1" field times 4.

For the 32-bit and 64-bit variant: is the first slice index offset, with implicit value 0.

<offs4>

For the 8-bit variant: is the fourth slice index offset, encoded as "off2" field times 4 plus 3.

For the 16-bit variant: is the fourth slice index offset, encoded as "o1" field times 4 plus 3.

For the 32-bit and 64-bit variant: is the fourth slice index offset, with implicit value 3.

Operation

CheckStreamingSVEAndZAEnabled(); constant integer VL = CurrentVL; if nreg == 4 && esize == 64 && VL < 256 then UNDEFINED; constant integer slices = VL DIV esize; constant bits(32) index = X[s, 32]; constant integer slice = ((UInt(index) - (UInt(index) MOD nreg)) + offset) MOD slices; for r = 0 to nreg-1 constant bits(VL) result = ZAslice[n, esize, vertical, slice + r, VL]; Z[d + r, VL] = result;

Operational information

If PSTATE.DIT is 1:


Internal version only: aarchmrs v2024-03_relA, pseudocode v2024-03_rel, sve v2024-03_rel ; Build timestamp: 2024-03-26T09:45

Copyright © 2010-2024 Arm Limited or its affiliates. All rights reserved. This document is Non-Confidential.