This is a collection of Intel®’ IA32® Software Developer's Manuals (URL of the day) and AMD' AMD64 Architecture Programmer's Manual together with the related specifications, application notes, white papers, and change logs. The collection aims to keep all available revisions. It was originally created by Michal Necasek, see OS/2 Museum.

If you have a public document, related to the IA32® specifications and missing from the collection, please mail it to me. The content of this URL and all sub-ULRs is available for convenient bulk download by rsync x86docs password "" (empty).

SMLALBB, SMLALBT, SMLALTB, SMLALTT -- AArch32

SMLALBB, SMLALBT, SMLALTB, SMLALTT

Signed Multiply Accumulate Long (halfwords) multiplies two signed 16-bit values to produce a 32-bit value, and accumulates this with a 64-bit value. The multiply acts on two signed 16-bit quantities, taken from either the bottom or the top half of their respective source registers. The other halves of these source registers are ignored. The 32-bit product is sign-extended and accumulated with a 64-bit accumulate value.

Overflow is possible during this instruction, but only as a result of the 64-bit addition. This overflow is not detected if it occurs. Instead, the result wraps around modulo 264.

It has encodings from the following instruction sets: A32 ( A1 ) and T32 ( T1 ) .

A1

313029282726252423222120191817161514131211109876543210
!= 111100010100RdHiRdLoRm1MN0Rn
condopc

Encoding for the SMLALBB variant

Applies when (M == 0 && N == 0)

SMLALBB{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Encoding for the SMLALBT variant

Applies when (M == 1 && N == 0)

SMLALBT{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Encoding for the SMLALTB variant

Applies when (M == 0 && N == 1)

SMLALTB{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Encoding for the SMLALTT variant

Applies when (M == 1 && N == 1)

SMLALTT{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Decode for all variants of this encoding

let dLo : integer = UInt(RdLo); let dHi : integer = UInt(RdHi); let n : integer = UInt(Rn); let m : integer = UInt(Rm); let n_index : integer = UInt(N); let m_index : integer = UInt(M); if dLo == 15 || dHi == 15 || n == 15 || m == 15 then UnpredictableProcedure(); end; if dHi == dLo then UnpredictableProcedure(); end;

CONSTRAINED UNPREDICTABLE behavior

If dHi == dLo, then one of the following behaviors must occur:

T1

15141312111098765432101514131211109876543210
111110111100RnRdLoRdHi10NMRm
op1

Encoding for the SMLALBB variant

Applies when (N == 0 && M == 0)

SMLALBB{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Encoding for the SMLALBT variant

Applies when (N == 0 && M == 1)

SMLALBT{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Encoding for the SMLALTB variant

Applies when (N == 1 && M == 0)

SMLALTB{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Encoding for the SMLALTT variant

Applies when (N == 1 && M == 1)

SMLALTT{<c>}{<q>} <RdLo>, <RdHi>, <Rn>, <Rm>

Decode for all variants of this encoding

let dLo : integer = UInt(RdLo); let dHi : integer = UInt(RdHi); let n : integer = UInt(Rn); let m : integer = UInt(Rm); let n_index : integer = UInt(N); let m_index : integer = UInt(M); if dLo == 15 || dHi == 15 || n == 15 || m == 15 then UnpredictableProcedure(); end; // Armv8-A removes UNPREDICTABLE for R13 if dHi == dLo then UnpredictableProcedure(); end;

CONSTRAINED UNPREDICTABLE behavior

If dHi == dLo, then one of the following behaviors must occur:

For more information about the CONSTRAINED UNPREDICTABLE behavior of this instruction, see Architectural Constraints on UNPREDICTABLE behaviors.

Assembler Symbols

<c>

See Standard assembler syntax fields.

<q>

See Standard assembler syntax fields.

<RdLo>

Is the general-purpose source register holding the lower 32 bits of the addend, and the destination register for the lower 32 bits of the result, encoded in the "RdLo" field.

<RdHi>

Is the general-purpose source register holding the upper 32 bits of the addend, and the destination register for the upper 32 bits of the result, encoded in the "RdHi" field.

<Rn>

For the "A1 SMLALBB", "A1 SMLALBT", "A1 SMLALTB", and "A1 SMLALTT" variants: is the first general-purpose source register holding the multiplicand in the bottom or top half (selected by <x>), encoded in the "Rn" field.

For the "T1 SMLALBB", "T1 SMLALBT", "T1 SMLALTB", and "T1 SMLALTT" variants: is the first general-purpose source register holding the multiplicand in the bottom or top half (selected by <x>), encoded in the "Rn" field.

<Rm>

For the "A1 SMLALBB", "A1 SMLALBT", "A1 SMLALTB", and "A1 SMLALTT" variants: is the second general-purpose source register holding the multiplier in the bottom or top half (selected by <y>), encoded in the "Rm" field.

For the "T1 SMLALBB", "T1 SMLALBT", "T1 SMLALTB", and "T1 SMLALTT" variants: is the second general-purpose source register holding the multiplier in the bottom or top half (selected by <x>), encoded in the "Rm" field.

Operation

if ConditionPassed() then EncodingSpecificOperations(); let operand1 : bits(16) = R(n)[n_index*:16]; let operand2 : bits(16) = R(m)[m_index*:16]; let result : integer = SInt(operand1) * SInt(operand2) + SInt(R(dHi)::R(dLo)); R(dHi) = result[63:32]; R(dLo) = result[31:0]; end;

Operational information

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


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.