ld4
Load Multiple 4-Element Structures
LD4 { <Vt1>.<T>, <Vt2>.<T>, <Vt3>.<T>, <Vt4>.<T> }, [<Xn|SP>]
Loads four-element structures (e.g., RGBA) into four registers.
Details
Loads four consecutive 4-element SIMD structures from memory, de-interleaving them into four registers (Vt1, Vt2, Vt3, Vt4). The element type T and Q field determine the 64-bit (Q=0) or 128-bit (Q=1) load size per register; the base address in Xn is post-incremented by the total bytes loaded. This AArch64 NEON instruction does not modify condition flags.
Pseudocode Operation
elements_per_struct ← 4; element_size ← GetElementSize(T); struct_bytes ← 4 * elements_per_struct * element_size; mem_addr ← Xn; (Vt1, Vt2, Vt3, Vt4) ← DeinterleaveLoad(mem_addr, struct_bytes, Q); Xn ← Xn + struct_bytes;
Example
LD4 [x1]
Encoding
Binary Layout
0
Q
0011000
1
000000
0000
size
Rn
Rt
Operands
-
Vt1
R -
Vt2
G -
Vt3
B -
Vt4
A
Reference (Arm A64 ISA)
Instruction Forms
| Encoding | Instruction | ISA | Bit pattern | ||
|---|---|---|---|---|---|
| 0x0C400000 | LD4 { <Vt>.<T>, <Vt2>.<T>, <Vt3>.<T>, <Vt4>.<T> }, [<Xn|SP>] | A64 | 0 | Q | 0011000 | 1 | 000000 | 0000 | size | Rn | Rt | ||
| 0x0CDF0000 | LD4 { <Vt>.<T>, <Vt2>.<T>, <Vt3>.<T>, <Vt4>.<T> }, [<Xn|SP>], <imm> | A64 | 0 | Q | 0011001 | 1 | 0 | 11111 | 0000 | size | Rn | Rt | ||
| 0x0CC00000 | LD4 { <Vt>.<T>, <Vt2>.<T>, <Vt3>.<T>, <Vt4>.<T> }, [<Xn|SP>], <Xm> | A64 | 0 | Q | 0011001 | 1 | 0 | Rm | 0000 | size | Rn | Rt | ||
| 0x0D602000 | LD4 { <Vt>.B, <Vt2>.B, <Vt3>.B, <Vt4>.B }[<index>], [<Xn|SP>] | A64 | 0 | Q | 0011010 | 1 | 1 | 0000 | 0 | 001 | S | size | Rn | Rt | ||
| 0x0D606000 | LD4 { <Vt>.H, <Vt2>.H, <Vt3>.H, <Vt4>.H }[<index>], [<Xn|SP>] | A64 | 0 | Q | 0011010 | 1 | 1 | 0000 | 0 | 011 | S | size | Rn | Rt | ||
| 0x0D60A000 | LD4 { <Vt>.S, <Vt2>.S, <Vt3>.S, <Vt4>.S }[<index>], [<Xn|SP>] | A64 | 0 | Q | 0011010 | 1 | 1 | 0000 | 0 | 101 | S | 00 | Rn | Rt | ||
| 0x0D60A400 | LD4 { <Vt>.D, <Vt2>.D, <Vt3>.D, <Vt4>.D }[<index>], [<Xn|SP>] | A64 | 0 | Q | 0011010 | 1 | 1 | 0000 | 0 | 101 | 0 | 01 | Rn | Rt | ||
| 0x0DFF2000 | LD4 { <Vt>.B, <Vt2>.B, <Vt3>.B, <Vt4>.B }[<index>], [<Xn|SP>], #4 | A64 | 0 | Q | 0011011 | 1 | 1 | 11111 | 001 | S | size | Rn | Rt | ||
| 0x0DE02000 | LD4 { <Vt>.B, <Vt2>.B, <Vt3>.B, <Vt4>.B }[<index>], [<Xn|SP>], <Xm> | A64 | 0 | Q | 0011011 | 1 | 1 | Rm | 001 | S | size | Rn | Rt | ||
| 0x0DFF6000 | LD4 { <Vt>.H, <Vt2>.H, <Vt3>.H, <Vt4>.H }[<index>], [<Xn|SP>], #8 | A64 | 0 | Q | 0011011 | 1 | 1 | 11111 | 011 | S | size | Rn | Rt | ||
| 0x0DE06000 | LD4 { <Vt>.H, <Vt2>.H, <Vt3>.H, <Vt4>.H }[<index>], [<Xn|SP>], <Xm> | A64 | 0 | Q | 0011011 | 1 | 1 | Rm | 011 | S | size | Rn | Rt | ||
| 0x0DFFA000 | LD4 { <Vt>.S, <Vt2>.S, <Vt3>.S, <Vt4>.S }[<index>], [<Xn|SP>], #16 | A64 | 0 | Q | 0011011 | 1 | 1 | 11111 | 101 | S | 00 | Rn | Rt | ||
| 0x0DE0A000 | LD4 { <Vt>.S, <Vt2>.S, <Vt3>.S, <Vt4>.S }[<index>], [<Xn|SP>], <Xm> | A64 | 0 | Q | 0011011 | 1 | 1 | Rm | 101 | S | 00 | Rn | Rt | ||
| 0x0DFFA400 | LD4 { <Vt>.D, <Vt2>.D, <Vt3>.D, <Vt4>.D }[<index>], [<Xn|SP>], #32 | A64 | 0 | Q | 0011011 | 1 | 1 | 11111 | 101 | 0 | 01 | Rn | Rt |
Description
Load multiple 4-element structures to four registers. This instruction loads multiple 4-element structures from memory and writes the result to the four SIMD&FP registers, with de-interleaving.
For an example of de-interleaving, see LD3 (multiple structures).
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(64) address;
bits(64) eaddr;
bits(64) offs;
bits(datasize) rval;
integer tt;
constant integer ebytes = esize DIV 8;
AccessDescriptor accdesc = CreateAccDescASIMD(memop, nontemporal, tagchecked);
if n == 31 then
CheckSPAlignment();
address = SP[];
else
address = X[n, 64];
offs = Zeros(64);
for r = 0 to rpt-1
for e = 0 to elements-1
tt = (t + r) MOD 32;
for s = 0 to selem-1
rval = V[tt, datasize];
eaddr = GenerateAddress(address, offs, accdesc);
if memop == MemOp_LOAD then
Elem[rval, e, esize] = Mem[eaddr, ebytes, accdesc];
V[tt, datasize] = rval;
else // memop == MemOp_STORE
Mem[eaddr, ebytes, accdesc] = Elem[rval, e, esize];
offs = offs + ebytes;
tt = (tt + 1) MOD 32;
if wback then
if m != 31 then
offs = X[m, 64];
address = GenerateAddress(address, offs, accdesc);
if n == 31 then
SP[] = address;
else
X[n, 64] = address;