Lerking/OSACA
1
0
Fork 0
mirror of https://github.com/RRZE-HPC/OSACA.git synced 2026-01-06 19:20:07 +01:00
Code Issues Packages Projects Releases Wiki Activity

changed 'testcases' to 'benchmarks'

Browse Source
This commit is contained in:
Jan Laukemann
2017-10-12 18:55:45 +02:00
parent 1fac059ebd
commit e3c09a40a1
140 changed files with 17224 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

101
benchmarks/vmovsd-xmm_mem.S Normal file
View File

@@ -0,0 +1,101 @@
#define INSTR vmovsd
#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 xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
INSTR xmm0, [rip+PI]
cmp i, N
jl loop
done:
mov rsp, rbp
pop rbp
ret
.size latency, .-latency