Some improvement to the caller

Source/DynCall.cpp

@@ -90,4 +90,12 @@ int main(void) {
 	ctx.Push(123.456f);
 	ctx.Push(999);
 	printf("6th is: %d.\n", ctx.Call<int>());
+	//
+	nioCall<void>(&printf, "0x%08x  %s\n", 12947765, "TEST");
+	//
+	ctx.Init(printf);
+	ctx.Push("0x%08x  %s\n");
+	ctx.Push(12947765);
+	ctx.Push("TEST");
+	ctx.Call<void>();
 }

Source/DynCall.h

@@ -8,6 +8,12 @@ extern "C" void* _DynCall(uint64_t);
 // 22222222 - Arg2
 // ........ - And so on
 //
+
+template <typename TRet = void, typename... TArgs, typename TAddr>
+constexpr inline TRet nioCall(TAddr address, TArgs... args) {
+	return reinterpret_cast<TRet(*)(TArgs...)>(address)(args ...);
+}
+
 #define MAX_ARGS 32
 class DynCallContext {
 	// Data fields
@@ -23,7 +29,7 @@ class DynCallContext {
 		nArgs = 0;
 		Address = (uint64_t)address;
 		for(int i = 0; i < MAX_ARGS; i++)
-			Args[i] = 0xCCCCCCCCCCCCCCCC;
+			Args[i] = 0x0;
 	}
 	// Push a value onto the context
 	template<typename T>

Source/caller.asm

@@ -6,24 +6,26 @@ _DynCall PROC
 
 	push rsi							; Save some nonvolatile registers
 	push rdi							;
-	push rbp 							;
 	push r12							;
 
+	sub	rsp, 10h						; Give the stack some slack
+	mov	r12, rsp						; and save it
+
 	mov rax, [rcx]						; Gets the number of arguments onto RAX
 
 	add rcx, 8							; Gets the address of the called function 
 	mov	r10, [rcx]						; onto R10
 
-	add rcx, 8							; Points RCX to the first argument
-
-	mov r12, rsp						; Saves the RSP onto R12
-
-	test eax, eax						; Do we have any args?
+	test eax, eax						; Do we have any arguments?
 	jz noargs							;
 
-	sub rsp, 108h						; Open enough space in the stack for 32 args
+	shl eax,3							; Open enough space in the stack for the arguments
+	sub rsp, rax						;
+	shr eax,3							;
 
-	mov rsi, rcx						; Copies all the parameters to the stack
+	add rcx, 8							; Points RCX to the first argument
+
+	mov rsi, rcx						; Copies all the arguments to the stack
 	mov rcx, rax						;
 	mov rdi, rsp						;
 	rep movsq							;
@@ -47,20 +49,19 @@ _DynCall PROC
 	movq xmm3, r9						;
 
 noargs:
-	mov rax, return						; Pushes the return address onto the stack
-	push rax
-	jmp r10								; Executes the called function
+    call r10							; Executes the called function
 										;   Which will probably crash ...
 										;      ... or maybe not!  :)
+
 return:
-	mov rsp, r12						; Gets the rsp back to where it was before
+	mov rsp, r12						; Restore the stack and...
+	add	rsp, 10h						; get the stack slack back
 
 	pop r12								; Restore nonvolatile registers
-	pop rbp								;
 	pop rdi								;
 	pop rsi								;
 
-	ret									; Back to the caller
+	ret									; Back to the original caller
 
 _DynCall ENDP
 

premake5.lua

@@ -1,10 +1,10 @@
 workspace "DynCall"
-	location "Build"
+	location "VStudio"
 	configurations { "Debug", "Release" }
 	platforms { "X64" }
 	startproject "DynCall"
 	pic "On"
-	systemversion "10.0.15063.0"
+	systemversion "10.0.17763.0"
 	characterset "ASCII"
 
 project "DynCall"
@@ -14,7 +14,7 @@ project "DynCall"
 	targetdir "Binary"
 	includedirs { }
 
-	files { "**.hpp", "**.h", "**.cpp", "**.c", "**.asm", "**.inc", "**.def"}
+	files { "Source/**"}
 	removefiles { }
 
 	vpaths {
@@ -34,5 +34,5 @@ project "DynCall"
 -- Cleanup
 if _ACTION == "clean" then
 	os.rmdir("Binary");
-	os.rmdir("Build");
+	os.rmdir("VStudio");
 end

setup.cmd

@@ -1 +1 @@
-premake5 vs2017
+premake5 vs2019