CODEWITHSUNDEEP
assemblyneonarm64
Hooli
Hooli

Reputation: 731

What kind of assembly instruction is this ld1 {v0.16b}, %[in]?

This instruction is an ARMv8 neon instruction but I'm not able to understand the elements of it. Can anyone elaborate on this?

Upvotes: 4

Views: 4085

Answers (3)

InfinitelyManic
InfinitelyManic

Reputation: 822

The instruction is an ARMv8 Load-Store Multiple Structure; which is an Advanced SIMD instruction.

LD1 - loads multiple 1-element structures

v0 - vector register 0 in this case, could be v1, v2,...

.16b - indicates you are loading 16 (1 byte) elements from memory. This can be 8 or 4H(alve words) or 8H... See the manual for more.

%[n] - I don't know what (%) is for; however, if you're using the instruction directly (i.e., not as an inline, etc) then [] - contains your vector address/pointer or the register holding array address.

In my partial code sample below, you will see a sample array in the .data section. A pointer in x10. The load instruction of 16 single byte elements, then proof of the load in a gdb print statement.

I compiled with gcc; therefore, other compilers may require a different syntax. #armv8 #simd

.data
array: .byte 7,3,1,6,7,1,7,6,5,3,1,3,3,0,6,2,4,9,1,9,2,2,5,1,1,9,6,7,4,4,2,6,5,7,4,7,4,2,3,5,5,3,4,9,1,9,4,9,3,4
...
ldr x10,=array
...
// Load multiple 1-element structures (to one register)
ld1 {v0.16b}, [x10] // load 16 byte values into vector 0 
...
(gdb) p $v0.b.s
$4 = {7, 3, 1, 6, 7, 1, 7, 6, 5, 3, 1, 3, 3, 0, 6, 2}

Upvotes: 2

G B
G B

Reputation: 3024

This is not called NEON anymore, the SIMD instructions are part of the armv8 standard set.

  • ld1 is the instruction: load single from memory into vector register
  • v0.16b is the register name and type: first SIMD register, 16 bytes
  • %[in] is the address in memory (probably a symbolic name for inline assembly)

So, you load a single register (16 bytes) from a memory location specified by the %[in] parameter.

Upvotes: 2

unwind
unwind

Reputation: 399703

Looking at the ARM NEON programming quick reference, we learn:

  • The general form of a NEON instruction is {<prefix>}<op>{<suffix>} Vd.<T>, Vn.<T>, Vm.<T>.
  • v0 is a 128-bit NEON vector register
  • The .16b matches the <T> part, which means "type" (16B means 16 bytes, i.e. the full 128 bits are being used).

The in part probably comes from surrounding C code.

Upvotes: 0

Related Questions