armsra.w

LBT intrinsic · group: ARM Shift

Synopsis

Instruction: armsra.w rj, rk, cond
CPU Flags: LBT

Description

ARM-style arithmetic shift right (32-bit): conditionally shift value in rj right arithmetically by amount in rk (masked to 8 bits). Update ARMFLAGS (C, N, Z). Only ARMFLAGS (LBT4) is updated; the GPR is not modified.

Operation

if (arm_cond_holds(ARMFLAGS, cond)) {
  uint32_t lhs = (uint32_t)a;
  uint32_t rhs = (uint32_t)b;
  uint32_t r;
  uint32_t c = rhs & 0xff;
  if (c == 0) {
    r = lhs;
  } else if (c < 32) {
    r = (uint32_t)((int32_t)lhs >> c);
    ARMFLAGS.C = (lhs >> (c - 1)) & 1;
  } else {
    ARMFLAGS.C = (lhs >> 31) & 1;
    r = ARMFLAGS.C ? UINT32_MAX : 0;
  }
  ARMFLAGS.N = (r >> 31) & 1;
  ARMFLAGS.Z = r == 0;
}

Tested on real machine.

Latency and Throughput

CPU	µarch	Latency	Throughput (IPC)
3C5000	LA464	2	0.5(1/2)
3C6000	LA664	2	0.5(1/2)

---

All intrinsics in alphabetical order