blob: c5e5a83d93d1e29c9285b1975f22e01ffc0aa17b (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
|
#!/bin/bash
# o3 against plain code: each kernel in tests/bench.zc on each backend, with
# the time to start up and compile (the 'z' run) taken off. needs bc.
cd "$(dirname "$0")"; T=$(mktemp -d); trap 'rm -rf $T' EXIT
gcc -O2 -w -DJTYPE=int core.c io_linux.c -o $T/zy
vms=("$T/zy"); names=("interp")
if command -v llvm-config >/dev/null; then
gcc -O2 -w -DJTYPE=int -DZYLLVM $(llvm-config --cflags) core.c io_linux.c -o $T/zyl $(llvm-config --ldflags --libs) && { vms+=("$T/zyl"); names+=("llvm"); }
fi
if command -v aarch64-linux-gnu-gcc >/dev/null && command -v qemu-aarch64 >/dev/null; then
aarch64-linux-gnu-gcc -O2 -w -DJTYPE=int -static core.c io_linux.c -o $T/zyj && { vms+=("qemu-aarch64 $T/zyj"); names+=("arm64 (qemu)"); }
fi
# seconds per run of kernel $2 on vm $1: the slope between $3 runs and 1 run,
# best of three each, so starting up and compiling cancel out
t(){ local s e; s=$(date +%s.%N); (cat tests/bench.zc; printf '`'$2$3) | $1 >/dev/null; e=$(date +%s.%N); echo "$e - $s" | bc; }
best(){ local b=999 x i; for i in 1 2 3; do x=$(t "$1" $2 $3); b=$(echo "if ($x < $b) $x else $b" | bc); done; echo $b; }
per(){ echo "scale=5; ($(best "$1" $2 $3) - $(best "$1" $2 1)) / ($3 - 1)" | bc; }
printf "%-14s %-8s %10s %10s %8s\n" backend kernel "plain ms" "o3 ms" speedup
for i in "${!vms[@]}"; do
n=41; [ "${names[$i]}" = interp ] && n=5
for k in "collatz a b" "sieve c d" "digits e f" "fib g h"; do
set -- $k
p=$(per "${vms[$i]}" $2 5); o=$(per "${vms[$i]}" $3 $n)
printf "%-14s %-8s %10.2f %10.2f %7.1fx\n" "${names[$i]}" $1 $(echo "$p * 1000" | bc) $(echo "$o * 1000" | bc) $(echo "scale=2; $p / $o" | bc)
done
done
|
