SMMLA

Signed 8-bit integer matrix multiply-accumulate to 32-bit integer

This instruction multiplies the 2×8 matrix of signed 8-bit integer values held in each 128-bit segment of the first source vector by the 8×2 matrix of signed 8-bit integer values in the corresponding segment of the second source vector. The resulting 2×2 widened 32-bit integer matrix product is then destructively added to the 32-bit integer matrix accumulator held in the corresponding segment of the addend and destination vector. This is equivalent to performing an 8-way dot product per destination element.

This instruction is unpredicated.

ID_AA64ZFR0_EL1.I8MM indicates whether this instruction is implemented.

This instruction is illegal when executed in Streaming SVE mode, unless FEAT_SME_FA64 is implemented and enabled.

SVE
(FEAT_SVE && FEAT_I8MM)

313029282726252423222120191817161514131211109876543210
01000101000Zm100110ZnZda
uns

Encoding

SMMLA <Zda>.S, <Zn>.B, <Zm>.B

Decode for this encoding

if !IsFeatureImplemented(FEAT_SVE) || !IsFeatureImplemented(FEAT_I8MM) then EndOfDecode(Decode_UNDEF); end; let n : integer = UInt(Zn); let m : integer = UInt(Zm); let da : integer = UInt(Zda); let op1_unsigned : boolean = FALSE; let op2_unsigned : boolean = FALSE;

Assembler Symbols

<Zda>

Is the name of the third source and destination scalable vector register, encoded in the "Zda" field.

<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.

Operation

CheckNonStreamingSVEEnabled(); let VL : integer{} = CurrentVL(); let segments : integer = VL DIV 128; let operand1 : bits(VL) = Z{}(n); let operand2 : bits(VL) = Z{}(m); let operand3 : bits(VL) = Z{}(da); var result : bits(VL) = Zeros{}; var op1, op2 : bits(128); var res, addend : bits(128); for s = 0 to segments-1 do op1 = operand1[s*:(128)]; op2 = operand2[s*:(128)]; addend = operand3[s*:(128)]; res = MatMulAdd(addend, op1, op2, op1_unsigned, op2_unsigned); result[s*:(128)] = res; end; Z{VL}(da) = result;

Operational information

This instruction is a data-independent-time instruction as described in About PSTATE.DIT.

This instruction might be immediately preceded in program order by a MOVPRFX instruction. The MOVPRFX must conform to all of the following requirements, otherwise the behavior of the MOVPRFX and this instruction is CONSTRAINED UNPREDICTABLE:


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.