Files
OSACA/benchmarks/pushq-imd.S
2017-10-12 18:55:45 +02:00

101 lines
1.6 KiB
ArmAsm

#define INSTR pushq
#define NINST 64
#define N edi
#define i r8d
.intel_syntax noprefix
.globl ninst
.data
ninst:
.long NINST
.align 32
PI:
.long 0xf01b866e, 0x400921f9, 0xf01b866e, 0x400921f9, 0xf01b866e, 0x400921f9, 0xf01b866e, 0x400921f9, 0xf01b866e, 0x400921f9, 0xf01b866e, 0x400921f9, 0xf01b866e, 0x400921f9, 0xf01b866e, 0x400921f9
.text
.globl latency
.type latency, @function
.align 32
latency:
push rbp
mov rbp, rsp
xor i, i
test N, N
jle done
# create DP 1.0
vpcmpeqw xmm0, xmm0, xmm0 # all ones
vpsllq xmm0, xmm0, 54 # logical left shift: 11111110..0 (54=64-(10-1))
vpsrlq xmm0, xmm0, 2 # logical right shift: 1 bit for sign; leading mantissa bit is zero
loop:
inc i
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
INSTR 1
cmp i, N
jl loop
done:
mov rsp, rbp
pop rbp
ret
.size latency, .-latency