purego: use syscall.Syscall6X for functions with 6 or less arguments (#3

)

runtime.go

@@ -10,6 +10,9 @@ import (
 	"unsafe"
 )
 
+//go:linkname syscall_syscall6X syscall.syscall6X
+func syscall_syscall6X(fn, a1, a2, a3, a4, a5, a6 uintptr) (r1, r2, err uintptr) // from runtime/sys_darwin_64.s
+
 //go:linkname runtime_libcCall runtime.libcCall
 //go:linkname runtime_entersyscall runtime.entersyscall
 //go:linkname runtime_exitsyscall runtime.exitsyscall

sys_darwin_amd64.s

@@ -51,7 +51,7 @@ TEXT ·syscall9X(SB), NOSPLIT, $0
 	MOVQ R11, X6 // a7
 	MOVQ R12, X7 // a8
 
-	// push the remainding paramters onto the stack
+	// push the remaining paramters onto the stack
 	MOVQ R11, 0(SP)  // push a7
 	MOVQ R12, 8(SP)  // push a8
 	MOVQ R13, 16(SP) // push a9

sys_darwin_arm64.s

@@ -37,6 +37,7 @@ TEXT ·syscall9X(SB), NOSPLIT, $0
 	MOVD 56(R0), R6 // a7
 	MOVD 64(R0), R7 // a8
 	MOVD 72(R0), R8 // a9
+	MOVD 8(R0), R0  // a1
 
 	// these may be float arguments
 	// so we put them also where C expects floats
@@ -46,12 +47,12 @@ TEXT ·syscall9X(SB), NOSPLIT, $0
 	FMOVD R3, F3 // a4
 	FMOVD R4, F4 // a5
 	FMOVD R5, F5 // a6
-	FMOVD R6, F6 // a4
-	FMOVD R7, F7 // a5
+	FMOVD R6, F6 // a7
+	FMOVD R7, F7 // a8
 
 	MOVD R8, (RSP) // push a9 onto stack
-	MOVD 8(R0), R0 // a1
-	BL   (R12)
+
+	BL (R12)
 
 	MOVD 8(RSP), R2     // pop structure pointer
 	ADD  $16, RSP

syscall.go

@@ -17,17 +17,23 @@ const maxArgs = 9
 // NOTE: SyscallN does not properly call functions that have both integer and float parameters.
 // See discussion comment https://github.com/ebiten/purego/pull/1#issuecomment-1128057607
 // for an explanation of why that is.
+//
+// On amd64, if there are more than 8 floats the 9th and so on will be placed incorrectly on the
+// stack.
 func SyscallN(fn uintptr, args ...uintptr) (r1, r2, err uintptr) {
 	if len(args) > maxArgs {
 		panic("too many arguments to SyscallN")
 	}
-	if len(args) < maxArgs {
-		// add padding so there is no out-of-bounds slicing
-		var tmp = make([]uintptr, maxArgs)
-		copy(tmp, args)
-		args = tmp
+	// add padding so there is no out-of-bounds slicing
+	var tmp [maxArgs]uintptr
+	copy(tmp[:], args)
+	if len(args) <= 6 {
+		// use the 6 argument version because
+		// gl.GenFramebuffersEXT would fail with the 9 version
+		// See https://github.com/hajimehoshi/ebiten/issues/2102#issuecomment-1134679352
+		return syscall_syscall6X(fn, tmp[0], tmp[1], tmp[2], tmp[3], tmp[4], tmp[5])
 	}
-	return syscall_syscall9X(fn, args[0], args[1], args[2], args[3], args[4], args[5], args[6], args[7], args[8])
+	return syscall_syscall9X(fn, tmp[0], tmp[1], tmp[2], tmp[3], tmp[4], tmp[5], tmp[6], tmp[7], tmp[8])
 }
 
 func callc(fn uintptr, args unsafe.Pointer) {