Add windows benchmark result (#243)

tests/ffi/README.md

@@ -54,29 +54,44 @@ bench_scale = 1000000
 Command: `cargo run run tests/ffi/benchmark/external_call`
 Command: `cargo run --profile=release run tests/ffi/benchmark/external_call`
 
-- Device1  
+- Device1-Linux-PVE  
   Lune release target: 0.205127 (sec)  
   Lune dev target: 1.556489 (sec)
+
+  > Commit: ddf0c4c
+
+- Device2-Windows-11
+  Lune release target: 0.1875 (sec)  
+  Lune dev target: ? SEGFUALT (sec)
+
   > Commit: ddf0c4c
 
 **LuaJit ffi**
 
 Command: `luajit tests/ffi/benchmark/external_call/luajit.lua`
 
-- Device1: 0.001682 (sec)
+- Device1-Linux-PVE: 0.001682 (sec)
   > LuaJIT 2.1.1727870382  
   > (flags = JIT ON SSE3 SSE4.1 BMI2 fold cse dce fwd dse narrow loop abc sink fuse)
 
 **Deno ffi**
 
 Command: `deno run --unstable-ffi --allow-ffi ./tests/ffi/benchmark/external_call/deno.ts`
 
-- Device1: 0.006384 (sec)
+- Device1-Linux-PVE: 0.006384 (sec)
   > Deno 1.46.3 (v8 = 12.9.202.5-rusty)
 
 **Sysinformation**
 
-- Device1
+- Device1-Linux-PVE
+
   > CPU: AMD Ryzen 5 7600 (12) @ 5.1  
   > MEM: 61898MiB 5600 MT/s  
   > KERNEL: 6.8.12-2-pve (Proxmox VE 8.2.7 x86_64)
+
+- Device2-Windows-11
+
+  > CPU: AMD Ryzen 5 7600 (4) @ 3.800GHz  
+  > MEM: 12250MiB 5600 MT/s  
+  > KERNEL: 10.0.22631 (Windows 11 x86_64)  
+  > HOST: QEMU Standard PC (Q35 + ICH9, 2009)

tests/ffi/utility/proc_clock/init.luau

@@ -1,4 +1,8 @@
+-- FIXME: in windows, we need another library to get process cpu time
+
 local ffi = require("@lune/ffi")
+local process = require("@lune/process")
+local is_windows = process.os == "windows"
 local c = ffi.c
 
 local proc_clock = {}
@@ -10,27 +14,35 @@ local lib = ffi.open(`{libdir}/lib.so`)
 
 -- sizeof_clock
 local sizeof_clock = c.fn({}, c.int):callable(lib:find("sizeof_clock"))
-function proc_clock.sizeof_clock()
+function proc_clock.sizeof_clock(): number
 	local result = ffi.box(c.int.size)
 	sizeof_clock(result)
 	return c.int:readData(result)
 end
-
 -- get_clock
-local clock_t = ffi["u" .. (proc_clock.sizeof_clock() * 8)]
+local clock_t = if is_windows then ffi.f32 else ffi["u" .. (proc_clock.sizeof_clock() * 8)]
 assert(clock_t, "clock_t is unknown type")
-local get_clock = c.fn({}, clock_t):callable(lib:find("get_clock"))
-proc_clock.get_clock = get_clock
+proc_clock.get_clock = (
+	if is_windows
+		then function(clock: ffi.BoxData | ffi.RefData)
+			ffi.f32:writeData(clock, os.clock())
+		end
+		else c.fn({}, clock_t):callable(lib:find("get_clock"))
+) :: (ffi.BoxData | ffi.RefData) -> ()
 
 -- get_offset
-local get_offset = c.fn({ clock_t, clock_t }, ffi.f64):callable(lib:find("get_offset"))
-function proc_clock.get_offset(before, after)
+local get_offset: (ffi.BoxData, ffi.RefData, ffi.RefData) -> () = if is_windows
+	then function(result: ffi.BoxData, before: ffi.RefData, after: ffi.RefData)
+		ffi.f64:writeData(result, (ffi.f32:readData(after) - ffi.f32:readData(before)))
+	end
+	else c.fn({ clock_t, clock_t }, ffi.f64):callable(lib:find("get_offset"))
+function proc_clock.get_offset(before: ffi.BoxData, after: ffi.BoxData): number
 	local result = ffi.box(ffi.f64.size)
 	get_offset(result, before:ref(), after:ref())
 	return ffi.f64:readData(result)
 end
 
-function proc_clock.new_box()
+function proc_clock.new_box(): (ffi.BoxData, ffi.BoxData)
 	return ffi.box(clock_t.size), ffi.box(clock_t.size)
 end