#!/bin/bash
# o3 against plain code: each kernel in tests/programs/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 $T' EXIT
CC=${CC:-$(sh src/tools/findcc.sh)} || exit 1
make -s build/core/genome.bin || exit 1
$CC -O2 -w -DJTYPE=int -DZYINTERP src/core/core.c src/core/io_linux.c -Isrc/core -Ibuild/core/map --embed-dir=build/core -o $T/zy
vms=("$T/zy"); names=("interp")
# the host's own jit, where there is one (x86-64 linux: blocks are translated, o3 regions run on the region interpreter)
if [ "$(uname -s)" = Linux ] && [ "$(uname -m)" = x86_64 ]; then
	$CC -O2 -w -DJTYPE=int src/core/core.c src/core/io_linux.c -Isrc/core -Ibuild/core/map --embed-dir=build/core -o $T/zyn && { vms+=("$T/zyn"); names+=("x86-64 jit"); }
fi
if command -v llvm-config >/dev/null; then
	$CC -O2 -w -DJTYPE=int -DZYLLVM $(llvm-config --cflags) src/core/core.c src/core/io_linux.c -Isrc/core -Ibuild/core/map --embed-dir=build/core -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 src/core/core.c src/core/io_linux.c -Isrc/core -Ibuild/core/map --embed-dir=build/core -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/programs/bench.zc; printf '`%s%s' $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 $b; }
# grows the run count until the difference is at least a second, so the noise
# of starting up is small next to it
per(){
	local one n=4 many
	one=$(best "$1" $2 1)
	while :; do
		many=$(best "$1" $2 $n)
		if [ "$(echo "$many - $one > 1" | bc)" = 1 ] || [ $n -ge 4096 ]; then break; fi
		n=$((n * 4))
	done
	echo "scale=6; ($many - $one) / ($n - 1)" | bc
}
printf "%-14s %-8s %10s %10s %8s\n" backend kernel "plain ms" "o3 ms" speedup
for i in "${!vms[@]}"; do
	for k in "collatz a b" "sieve c d" "digits e f" "fib g h"; do
		set -- $k
		p=$(per "${vms[$i]}" $2); o=$(per "${vms[$i]}" $3)
		printf "%-14s %-8s %10.2f %10.2f %7.1fx\n" "${names[$i]}" $1 $(echo "$p * 1000" | bc) $(echo "$o * 1000" | bc) $(echo "scale=2; $p / $o" | bc)
	done
done