Last active
January 21, 2018 15:40
-
-
Save Terminus-IMRC/a4f187149f1d51208c0eb2b6f3978d26 to your computer and use it in GitHub Desktop.
QPU benchmarks descriptions
rep NNOPS {
nop
}
alu.pe
mov host_int, 1
nop
Aborted.
mov host_int, 1
alu.pe
nop
nop
nop
rep N {
sem15 inc
}
rep {
sem15 inc
sem15 dec
}
rep {
mutex_acquire
mutex_release
}
rep {
uniforms_address <- addr
}
http://imrc.noip.me/blog/vc4/QI8/
Reading qpu_number.
n | Thread 0 | Thread 1 |
---|---|---|
1 | nop | nop |
2 | sem0 inc | nop |
3 | nop | sem0 dec |
4 | nop | nop |
5 | nop | sem1 inc |
6 | sem1 dec | nop |
7 | nop | nop |
n | Thread 0 | Thread 1 |
---|---|---|
1 | nop | nop |
2 | sem0 dec | nop |
3 | nop | sem0 inc |
4 | nop | nop |
5 | nop | sem1 dec |
6 | sem1 inc | nop |
7 | nop | nop |
n | Thread 0 | Thread 1 |
---|---|---|
1 | nop | nop |
2 | nop | sem0 dec |
3 | sem0 inc | nop |
4 | nop | nop |
rep {
sem0 inc
}
rep {
sem0 dec
}
n | Thread 0 | Thread 1 |
---|---|---|
1 | sem0 inc | sem0 dec |
2 | nop | X |
3 | nop (bra) | nop |
4 | nop | nop (bra) |
5 | sem1 dec | sem0 inc |
6 | X | nop |
7 | nop | nop |
X: Expected stall caused by semaphore inst.
n | Thread 0 | Thread 1 |
---|---|---|
1 | sem0 dec | nop |
2 | nop | sem0 inc |
3 | nop | nop |
4 | nop | nop |
5 | sem1 inc | sem1 dec |
6 | nop (bra) | nop (bra) |
7 | nop | nop |
8 | nop | nop |
9 | nop | nop |
n | Thread 0 | Thread 1 |
---|---|---|
1 | sem0 dec | nop |
2 | X | sem0 inc |
3 | X | nop |
4 | nop | nop |
5 | nop | sem1 dec |
6 | sem1 inc | X |
7 | nop | X |
8 | nop (bra) | nop (bra) |
9 | nop | nop |
10 | nop | nop |
11 | nop | nop |
n | Thread 0 | Thread 1 | Thread 2 |
---|---|---|---|
1 | sem0 inc | nop | nop |
2 | nop | sem0 dec | nop |
3 | nop | sem1 inc | nop |
4 | nop | nop | sem1 dec |
5 | nop | nop | sem2 inc |
6 | sem2 dec | nop | nop |
7 | nop (bra) | nop (bra) | nop (bra) |
8 | nop | nop | nop |
9 | nop | nop | nop |
10 | nop | nop | nop |
n | Thread 0 | Thread 1 | Thread 2 | Thread 3 | Thread 4 |
---|---|---|---|---|---|
1 | sem0 inc | nop | nop | nop | nop |
2 | nop | sem0 dec | nop | nop | nop |
3 | nop | sem1 inc | nop | nop | nop |
4 | nop | nop | sem1 dec | nop | nop |
5 | nop | nop | sem2 inc | nop | nop |
6 | nop | nop | nop | sem2 dec | nop |
7 | nop | nop | nop | sem3 inc | nop |
8 | nop | nop | nop | nop | sem3 dec |
9 | nop | nop | nop | nop | sem4 inc |
10 | sem4 dec | nop | nop | nop | nop |
11 | nop (bra) | nop (bra) | nop (bra) | nop (bra) | nop (bra) |
12 | nop | nop | nop | nop | nop |
13 | nop | nop | nop | nop | nop |
14 | nop | nop | nop | nop | nop |
n | Thread 0 | Thread 1 | Thread 2 | Thread 3 | Thread 4 |
---|---|---|---|---|---|
1 | sem0 inc | nop | nop | nop | nop |
2 | nop | sem0 dec | nop | nop | nop |
3 | nop | sem1 inc | nop | nop | nop |
4 | nop | nop | sem1 dec | nop | nop |
5 | nop | nop | sem2 inc | nop | nop |
6 | nop | nop | nop | sem2 dec | nop |
7 | nop | nop | nop | sem3 inc | nop |
8 | nop | nop | nop | nop | sem3 dec |
9 | nop | nop | nop | nop | sem4 inc |
10 | sem4 dec | nop | nop | nop | nop |
11 | sem5 inc | sem5 inc | sem5 inc | sem5 inc | sem5 inc |
12 | sem5 dec | sem5 dec | sem5 dec | sem5 dec | sem5 dec |
13 | nop (bra) | nop (bra) | nop (bra) | nop (bra) | nop (bra) |
14 | nop | nop | nop | nop | nop |
15 | nop | nop | nop | nop | nop |
16 | nop | nop | nop | nop | nop |
n | Thread 0 | Thread 1 |
---|---|---|
1 | nop | nop |
2 | sem0 dec | nop |
3 | nop | sem0 inc |
4 | sem2 inc | sem2 inc |
5 | nop | nop |
6 | sem2 dec | sem1 dec |
7 | sem1 inc | nop |
8 | nop | sem2 dec |
sem
- QT8
rep {
tmu0 with stride=4
delays (implemented with delays of alus)
sig tmu0
}
- QT9
rep {
tmu0 with stride=4
delays (implemented with branch)
sig tmu0
}
- QT10
rep {
tmu0 with +4*element_number with stride=4*16
delays (implemented with branch)
sig tmu0
}
-
QT11 -- The same as QT10. Aborted.
-
QT12
tmu0 <- addr + ELEM_STEP * element_number
delays
sig tmu0
- QT13
tmu0 with +ftoi(unif*itof(element_number))
delays
sig tmu0
- QT14 -- extends QT10.
rep {
tmu0 with stride=128
delays
sig tmu0
}
- QT15 -- extends QT10.
rep {
tmu0 with stride=64
tmu0 with stride=64
delays
sig tmu0
sig tmu0
}
- QT16 -- extends QT15
rep {
tmu0 with stride=64
tmu1 with stride=64
delays
sig tmu0
sig tmu1
}
- QT17 -- extends QT15
rep {
tmu0 with stride=64
tmu0 with stride=64
delays
sig tmu0
20 nops
sig tmu0
}
- QT18 -- extends QT17
addr += e * 4
rep {
tmu0 = addr
tmu0 = addr + 64
tmu0 = addr + 128
delays
sig tmu0
20 nops
sig tmu0
20 nops
sig tmu0
addr += 172
}
- QT19 -- extends QT18
addr += e * 4
rep {
tmu0 = addr
tmu0 = addr
tmu0 = addr
delays
sig tmu0
20 nops
sig tmu0
20 nops
sig tmu0
addr += 64
- QT20 -- extends QT19
addr += e * 4
rep {
tmu0 = addr
tmu0 = addr
tmu0 = addr
tmu0 = addr
delays
sig tmu0
20 nops
sig tmu0
20 nops
sig tmu0
20 nops
sig tmu0
addr += 64
- QT21 -- extends QT15
rep {
tmu0 with stride=64
tmu0 with stride=64
tmu0 with stride=64
delays
sig tmu0
sig tmu0
sig tmu0
}
- QT22 -- extends QT15
rep {
tmu0 with stride=64
tmu0 with stride=64
tmu0 with stride=64
tmu0 with stride=64
delays
sig tmu0
sig tmu0
sig tmu0
sig tmu0
}
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Todo: Checklist