smlal
Signed Multiply-Accumulate Long
SMLAL <Vd>.<Td>, <Vn>.<Ts>, <Vm>.<Ts>
Multiplies signed narrow vectors and adds to wide destination.
Details
Multiplies corresponding signed elements of two narrow SIMD vectors and adds the widened products to the existing contents of a wider destination register. This is an AArch64-only NEON instruction that performs sign-extended multiplication followed by accumulation on integer element types (8, 16, or 32 bits). Condition flags are not affected.
Pseudocode Operation
for i = 0 to (128 >> (size+1)) - 1 do
op1 ← SignExtend(Vn[i], element_width)
op2 ← SignExtend(Vm[i], element_width)
Vd[i] ← Vd[i] + (op1 * op2)
end for
Example
SMLAL v0.4s.Td, v1.4s.Ts, v2.4s.Ts
Encoding
Binary Layout
0
Q
0
01110
size
1
Rm
10
0
000
Rn
Rd
Operands
-
Vd
Dest/Acc -
Vn
First source SIMD/FP vector register -
Vm
Second source SIMD/FP vector register
Reference (Arm A64 ISA)
Instruction Forms
| Encoding | Instruction | ISA | Bit pattern | ||
|---|---|---|---|---|---|
| 0x0F002000 | SMLAL{2} <Vd>.<Ta>, <Vn>.<Tb>, <Vm>.<Ts>[<index>] | A64 | 0 | Q | 0 | 01111 | size | L | M | Rm | 0 | 0 | 10 | H | 0 | Rn | Rd | ||
| 0x0E208000 | SMLAL{2} <Vd>.<Ta>, <Vn>.<Tb>, <Vm>.<Tb> | A64 | 0 | Q | 0 | 01110 | size | 1 | Rm | 10 | 0 | 000 | Rn | Rd | ||
| 0xC1C01000 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>], <Zn>.H, <Zm>.H[<index>] | A64 | 110000011100 | Zm | i3h | Rv | 1 | i3l | Zn | 0 | 0 | off3 | ||
| 0xC1D01000 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>{, VGx2}], { <Zn1>.H-<Zn2>.H }, <Zm>.H[<index>] | A64 | 110000011101 | Zm | 0 | Rv | 1 | i3h | Zn | 0 | 0 | 0 | i3l | off2 | ||
| 0xC1D09000 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>{, VGx4}], { <Zn1>.H-<Zn4>.H }, <Zm>.H[<index>] | A64 | 110000011101 | Zm | 1 | Rv | 1 | i3h | Zn | 00 | 0 | 0 | i3l | off2 | ||
| 0xC1600C00 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>], <Zn>.H, <Zm>.H | A64 | 110000010110 | Zm | 0 | Rv | 011 | Zn | 0 | 0 | off3 | ||
| 0xC1600800 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>{, VGx2}], { <Zn1>.H-<Zn2>.H }, <Zm>.H | A64 | 110000010110 | Zm | 0 | Rv | 010 | Zn | 0 | 0 | 0 | off2 | ||
| 0xC1700800 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>{, VGx4}], { <Zn1>.H-<Zn4>.H }, <Zm>.H | A64 | 110000010111 | Zm | 0 | Rv | 010 | Zn | 0 | 0 | 0 | off2 | ||
| 0xC1E00800 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>{, VGx2}], { <Zn1>.H-<Zn2>.H }, { <Zm1>.H-<Zm2>.H } | A64 | 11000001111 | Zm | 00 | Rv | 010 | Zn | 0 | 0 | 0 | 0 | off2 | ||
| 0xC1E10800 | SMLAL ZA.S[<Wv>, <offs1>:<offs2>{, VGx4}], { <Zn1>.H-<Zn4>.H }, { <Zm1>.H-<Zm4>.H } | A64 | 11000001111 | Zm | 010 | Rv | 010 | Zn | 00 | 0 | 0 | 0 | off2 |
Description
Signed Multiply-Add Long (vector). This instruction multiplies corresponding signed integer values in the lower or upper half of the vectors of the two source SIMD&FP registers, and accumulates the results with the vector elements of the destination SIMD&FP register. The destination vector elements are twice as long as the elements that are multiplied.
The SMLAL instruction extracts each source vector from the lower half of each source register. The SMLAL2 instruction extracts each source vector from the upper half of each source register.
Depending on the settings in the CPACR_EL1, CPTR_EL2, and CPTR_EL3 registers, and the current Security state and Exception level, an attempt to execute the instruction might be trapped.
Operation
CheckFPAdvSIMDEnabled64();
bits(datasize) operand1 = Vpart[n, part, datasize];
bits(datasize) operand2 = Vpart[m, part, datasize];
bits(2*datasize) operand3 = V[d, 2*datasize];
bits(2*datasize) result;
integer element1;
integer element2;
bits(2*esize) product;
bits(2*esize) accum;
for e = 0 to elements-1
element1 = Int(Elem[operand1, e, esize], unsigned);
element2 = Int(Elem[operand2, e, esize], unsigned);
product = (element1*element2)<2*esize-1:0>;
if sub_op then
accum = Elem[operand3, e, 2*esize] - product;
else
accum = Elem[operand3, e, 2*esize] + product;
Elem[result, e, 2*esize] = accum;
V[d, 2*datasize] = result;