rbdc 0.9.0 → 1.0.0

data/dyncall/dyncall/dyncall_call_arm32_arm_armhf.S

@@ -6,7 +6,7 @@
  Description: Call Kernel for ARM 32-bit ARM Architecture - Hard Float
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>,
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any
@@ -51,38 +51,36 @@ GLOBAL_C(dcCall_arm32_armhf)
 ENTRY_C(dcCall_arm32_armhf)
 
 	/* Prolog. This function never needs to spill inside its prolog, so just store the permanent registers. */
-	mov	r12 , r13	 /* Stack ptr (r13) -> temporary (r12). */
-	stmdb	r13!, {r4-r5, r11, r12, r14} /* Permanent registers and stack pointer (now in r12), etc... -> save area on stack (except counter). */
+	mov r12 , r13  /* Stack ptr (r13) -> temporary (r12). */
+	stmdb r13!, {r4-r5, r11, r12, r14} /* Permanent registers and stack pointer (now in r12), etc... -> save area on stack (except counter). */
 
-	mov	r11 , r12	 /* Set frame ptr. */
-	mov	r4  , r0 	 /* r4 = 'fptr' (1st argument is passed in r0). */
-	mov	r5  , r1	 /* r5 = 'args' (2nd argument is passed in r1). */
-	
-	/* Load 16 single-precision registers (= 8 double-precision registers). */
-	fldmiad	r3, {d0-d7}
+	mov r11, r12     /* Set frame ptr. */
+	mov r4,  r0      /* r4 = fptr */
+	add r5,  r1, #16 /* r5 = stack args (after intreg ones) */
 
-	sub	r2 , r2 , #16	/* skip spill area */
-	cmp     r2, #0
-	ble	armhf_call
+	/* Load 16 single-precision registers (= 8 double-precision registers). */
+	fldmiad r3, {d0-d7}
 
-	sub	r13, r13, r2
-	and	r13, r13, #-8	/* align 8-byte. */
+	/* prep stack parameter area (includes room for spill area, callee spills if needed) */
+	sub r13, r13, r2
+	and r13, r13, #-8  /* align 8-byte. */
 
-	mov	r3, #0		/* Init byte counter. */
-	add	r1 , r1 , #16
+	sub r2, r2, #16    /* loop counters for stack params to copy */
+	mov r3, #0
 armhf_pushArgs:
-	ldr	r0, [r1,  +r3]	/* Load word into r0. */
-	str	r0, [r13, +r3]	/* Push word onto stack. */
-	add	r3, r3, #4	/* Increment byte counter. */
-	cmp	r2, r3
-	bne	armhf_pushArgs
+	cmp r2, r3
+	ble armhf_call
+	ldr r0, [r5,  +r3] /* load word into r0 ... */
+	str r0, [r13, +r3] /* ... then push onto stack */
+	add r3, r3, #4
+	b   armhf_pushArgs
 
 armhf_call:
-	ldmia	r5, {r0-r3}	/* Load first 4 arguments for new call into r0-r3. */
-				/* 'blx %r4' workaround for ARMv4t: */
-	mov	r14, r15	/*   Branch return address(r15) -> link register (r14) -- r15 always points to address of current + 2 instructions (= Epilog code). */ 
-	bx	r4		/*   Call (ARM/THUMB), available for ARMv4t. */
+	ldmia r1, {r0-r3}  /* Load first 4 arguments for new call into r0-r3. */
+	                   /* 'blx %r4' workaround for ARMv4t: */
+	mov r14, r15       /* Branch return address(r15) -> link register (r14) -- r15 always points to address of current + 2 instructions (= Epilog code). */ 
+	bx  r4             /* Call (ARM/THUMB), available for ARMv4t. */
 
 	/* Epilog. */
-	ldmdb	r11, {r4-r5, r11, r13, r15}	/* Restore permanent registers (ignore temporary (r12), restore stack ptr and program counter).@@@db not needed since we rewrite r13? */
+	ldmdb r11, {r4-r5, r11, r13, r15} /* Restore permanent registers (ignore temporary (r12), restore stack ptr and program counter).@@@db not needed since we rewrite r13? */
 

data/dyncall/dyncall/dyncall_call_arm32_arm_armhf.h

@@ -6,7 +6,7 @@
  Description: 
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_arm32_thumb.h

@@ -6,7 +6,7 @@
  Description: 
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_arm32_thumb_armhf.S

@@ -6,7 +6,7 @@
  Description: Call Kernel for ARM 32-bit ARM Architecture - Hard Float in Thumb code
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_arm32_thumb_gas.s

@@ -36,17 +36,13 @@
 dcCall_arm32_thumb:
 
 	/* Prolog. This function never needs to spill inside its prolog, so just store the permanent registers. */
-	push	{%r4-%r7, %r14}	   	/* Frame ptr, permanent registers, link register -> save area on stack. */
+	/* Code below is not using high registers, so not storing them in prolog, which is more involved with thumb, anyways. */
+	push	{%r4-%r7, %r14}		/* Frame ptr, permanent registers, link register -> save area on stack. */
 	mov		%r7, %r13	/* Set frame ptr. */
+	sub		%r13, #4	/* Realign stack to 8 bytes (b/c we stored 5 regs = 20b). */
 
 	/* Call. */
 	mov		%r4, %r0	/* Move 'fptr' to r4 (1st argument is passed in r0). */
-
-	/* Disable 'thumb' address forcing... */
-
-	/* mov		%r0, #1	*/	/* Assure that LSB is set to 1 (THUMB call). - Not Required and not useful for interworking calls */
-	/* orr		%r4, %r0 */
-	
 	mov		%r5, %r1	/* Move 'args' to r5 (2nd argument is passed in r1). */
 	mov		%r6, %r2	/* Move 'size' to r6 (3rd argument is passed in r2). */
 

data/dyncall/dyncall/dyncall_call_arm64.S

@@ -6,8 +6,8 @@
  Description: Call Kernel for ARM 64-bit Architecture (aka ARM64, AArch64)
  License:
 
-   Copyright (c) 2015 Daniel Adler <dadler@uni-goettingen.de>, 
-                      Tassilo Philipp <tphilipp@potion-studios.com>
+   Copyright (c) 2015-2018 Daniel Adler <dadler@uni-goettingen.de>, 
+                           Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any
    purpose with or without fee is hereby granted, provided that the above

data/dyncall/dyncall/dyncall_call_arm64.h

@@ -6,8 +6,8 @@
  Description: ARM 64-bit
  License:
 
-   Copyright (c) 2015 Daniel Adler <dadler@uni-goettingen.de>, 
-                      Tassilo Philipp <tphilipp@potion-studios.com>
+   Copyright (c) 2015-2018 Daniel Adler <dadler@uni-goettingen.de>, 
+                           Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any
    purpose with or without fee is hereby granted, provided that the above

data/dyncall/dyncall/dyncall_call_mips.h

@@ -6,7 +6,7 @@
  Description: mips call-kernel C interfaces.
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_mips_eabi.h

@@ -6,7 +6,7 @@
  Description: mips "eabi" abi call kernel C interface.
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_mips_eabi_gas.s

@@ -3,7 +3,7 @@
  Package: dyncall
  Library: dyncall
  File: dyncall/dyncall_call_mips_eabi_gas.s
- Description: 
+ Description: mips "eabi" abi call kernel implementation in GNU Assembler
  License:
 
    Copyright (c) 2007-2011 Daniel Adler <dadler@uni-goettingen.de>, 
@@ -37,7 +37,7 @@
 
 dcCall_mips_eabi:
 	/* $4 target function */
- 	/* $5 register data */
+	/* $5 register data */
 	/* $6 stack size */
 	/* $7 stack data */
 	addiu	$sp,$sp,-16
@@ -46,7 +46,7 @@ dcCall_mips_eabi:
 	sw	$fp,0($sp)
 
 	move	$fp,$sp
-	
+
 	move	$2, $0
 	add	$2, 8
 	neg	$2
@@ -57,9 +57,9 @@ dcCall_mips_eabi:
 	move	$12,$4		/* target function */
 	move	$13,$5		/* register data   */
 	move    $16,$6		/* stack size      */
-	
+
 	sub	$sp,$sp,$16	/* allocate stack frame */
-	
+
 	/* copy stack data */
 
 .next:
@@ -73,8 +73,8 @@ dcCall_mips_eabi:
 	addiu	$sp,$sp, 4
 	j	.next
 	nop
-	
-.skip:	
+
+.skip:
 
 	sub	$sp,$sp,$16
 
@@ -89,16 +89,16 @@ dcCall_mips_eabi:
 	lw	$10,24($13)
 	lw	$11,28($13)
 
-	/* load single-precise floating pointer parameter registers */
+	/* load single-precision floating pointer parameter registers */
 
-	lwc1	$f12, 32($13)
-	lwc1	$f13, 36($13)
-	lwc1	$f14, 40($13)
-	lwc1	$f15, 44($13)
-	lwc1	$f16, 48($13)
-	lwc1	$f17, 52($13)
-	lwc1	$f18, 56($13)
-	lwc1	$f19, 60($13)	
+	l.s	$f12, 32($13)
+	l.s	$f13, 36($13)
+	l.s	$f14, 40($13)
+	l.s	$f15, 44($13)
+	l.s	$f16, 48($13)
+	l.s	$f17, 52($13)
+	l.s	$f18, 56($13)
+	l.s	$f19, 60($13)
 
 	jal	$12
 	nop

data/dyncall/dyncall/dyncall_call_mips_gas.S

@@ -3,10 +3,10 @@
  Package: dyncall
  Library: dyncall
  File: dyncall/dyncall_call_mips_gas.S
- Description: auto-select (via gnu c preprocessor) mips abi call kernel
+ Description: auto-select (via C preprocessor) mips abi call kernel
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>,
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>,
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_mips_n32.h

@@ -3,10 +3,10 @@
  Package: dyncall
  Library: dyncall
  File: dyncall/dyncall_call_mips_n32.h
- Description: mips "n32" ABI call-kernel C interface.
+ Description: mips64 "n32" ABI call-kernel C interface.
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any
@@ -52,12 +52,13 @@ extern "C" {
 
 struct DCRegData_mips_n32
 {
-  DClonglong   mIntData[8];
+  DClonglong                       mIntData[8];
   union { DCfloat f; DCdouble d; } mFloatData[8];
-  DClonglong   mUseDouble; /* bitmask: lower 8 bits specifies to use float or double from union array. */
+  DClonglong                       mUseDouble; /* bitmask: lower 8 bits specify to use float or double from union array. */
 };
 
 void dcCall_mips_n32(DCpointer target, struct DCRegData_mips_n32* regdata, DCsize stksize, DCpointer stkdata);
+/* @@@ this is the same as n64, combine code */
 
 #ifdef __cplusplus
 }

data/dyncall/dyncall/dyncall_call_mips_n32_gas.s

@@ -3,10 +3,10 @@
  Package: dyncall
  Library: dyncall
  File: dyncall/dyncall_call_mips_n32_gas.s
- Description: mips "n32" abi call kernel implementation in GNU Assembler
+ Description: mips64 "n32" abi call kernel implementation in GNU Assembler
  License:
 
-   Copyright (c) 2007-2011 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2016 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any
@@ -177,8 +177,8 @@ dcCall_mips_n32:
 	/* jump-and-link to register $25 */
 
 	jal	$31, $25
-
-	/* no nop according to gcc assembly output */	
+	nop /* branch-delay slot - no nop according to gcc assembly */
+	    /* output, but let's play safe */	
 
 	/* Stack-frame epilog */	
 	move	$sp,$fp 

data/dyncall/dyncall/dyncall_call_mips_n64.h

@@ -6,7 +6,7 @@
  Description: mips "n64" ABI call-kernel C interface.
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any
@@ -52,9 +52,9 @@ extern "C" {
 
 struct DCRegData_mips_n64
 {
-  DClonglong   mIntData[8];
+  DClonglong                       mIntData[8];
   union { DCfloat f; DCdouble d; } mFloatData[8];
-  DClonglong   mUseDouble; /* bitmask: lower 8 bits specifies to use float or double from union array. */
+  DClonglong                       mUseDouble; /* bitmask: lower 8 bits specify to use float or double from union array. */
 };
 
 void dcCall_mips_n64(DCpointer target, struct DCRegData_mips_n64* regdata, DCsize stksize, DCpointer stkdata);

data/dyncall/dyncall/dyncall_call_mips_n64_gas.s

@@ -177,8 +177,8 @@ dcCall_mips_n64:
 	/* jump-and-link to register $25 */
 
 	jal	$31, $25
-
-	/* no nop according to gcc assembly output */	
+	nop /* branch-delay slot - no nop according to gcc assembly */
+	    /* output, but let's play safe */	
 
 	/* Stack-frame epilog */	
 	move	$sp,$fp 

data/dyncall/dyncall/dyncall_call_mips_o32.h

@@ -6,7 +6,7 @@
  Description: mips "o32" abi call kernel C interface.
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any
@@ -43,8 +43,9 @@ extern "C" {
   32-bit floating pointer registers.
   Float arguments map as following:
   
-    float argument 0 is at floats[1] and
-    float argument 1 is at floats[3] of DCRegData_mips_o32 union.
+    float argument 0 is at u[0][0] for little, u[0][1] for big endian and
+    float argument 1 is at u[1][0] for little, u[1][1] for big endian of
+	DCRegData_mips_o32 union.
 
 */
 

data/dyncall/dyncall/dyncall_call_mips_o32_gas.s

@@ -42,7 +42,6 @@ dcCall_mips_o32:
 	.mask	0xc0000000,-4
 	.fmask	0x00000000,0
 	.set	noreorder
-	.set	nomacro
 	
 	addiu	$sp,$sp,-8
 	sw	$31,4($sp)	/* save link register */
@@ -77,8 +76,8 @@ dcCall_mips_o32:
 
 	l.d     $f12, 0($5)
 	l.d     $f14, 8($5)
-	/* prepare call */
 
+	/* prepare call */
 
 	move	$12, $7		/* $12  stack data */
 	move	$25, $4		/* $25  target function */
@@ -102,7 +101,6 @@ dcCall_mips_o32:
 	j	$31		/* return */
 	nop
 
-	.set	macro
 	.set	reorder
 	.end	dcCall_mips_o32 
 	.ident 	"handwritten"

data/dyncall/dyncall/dyncall_call_ppc32.S

@@ -6,7 +6,7 @@
  Description: Call Kernel for PowerPC 32-bit Architecture
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_ppc32.h

@@ -6,7 +6,7 @@
  Description: 
  License:
 
-   Copyright (c) 2007-2015 Daniel Adler <dadler@uni-goettingen.de>, 
+   Copyright (c) 2007-2018 Daniel Adler <dadler@uni-goettingen.de>, 
                            Tassilo Philipp <tphilipp@potion-studios.com>
 
    Permission to use, copy, modify, and distribute this software for any

data/dyncall/dyncall/dyncall_call_sparc.h

@@ -6,7 +6,7 @@
  Description: 
  License:
 
-   Copyright (c) 2011-2015 Daniel Adler <dadler@uni-goettingen.de>
+   Copyright (c) 2011-2018 Daniel Adler <dadler@uni-goettingen.de>
 
    Permission to use, copy, modify, and distribute this software for any
    purpose with or without fee is hereby granted, provided that the above

data/dyncall/dyncall/{dyncall_call_sparc.S → dyncall_call_sparc.s}

@@ -27,6 +27,8 @@
 
 /* --------------------------------------------------------------------------- 
 
+@@@ this should all go in manual
+
 call kernel for sparc 32-bit
 ----------------------------
 tested on linux/debian [gcc54.fsffrance.org - thanx to the farm!] 
@@ -45,7 +47,7 @@ in sparc, this is simply a leaf-function layer using %o3.
 
 old C Interface:
   void dcCall_sparc (DCpointer target, DCsize size, DCpointer data);
-		     %i0             , %i1        , %i2
+                     %i0             , %i1        , %i2
 
 
 Input:
@@ -95,11 +97,11 @@ Stack Layout 32-Bit Model:
 	 0: 16 registers save area
 
 Stack Layout 64-Bit Model:
-        XX: should be 16 byte aligned (min stack frame size is 172).
-       168: on stack argument 6
-       136: input argument 0 spill
-       128: struct/union poiner return value
-	 0: 16 registers save area
+	 XX: should be 16 byte aligned (min stack frame size is 172).
+	168: on stack argument 6
+	136: input argument 0 spill
+	128: struct/union poiner return value
+	  0: 16 registers save area
 
 
 
@@ -121,26 +123,20 @@ r24-r31 -> ins
 
 */
 
-#if defined __arch64__
-#define REGSIZE 8
-#error invalid arch
-#else
-#define REGSIZE 4
-#endif
-
-#define ALIGN   16
+REGSIZE        =  4
+ALIGN          = 16
 CALLVM_size    = 12
 CALLVM_dataoff = 16
+
+.text
 .global dcCall_sparc
 dcCall_sparc:
 
 /* Basic Prolog: supports up to 6 arguments. */
 
-	/* new C interface */
 	/* o0-1: callvm,target */
-	
-	or   %o0, %g0, %o3	       /* %o3: callvm */
-	or   %o1, %g0, %o0	       /* %o0: target */
+	or   %o0, %g0, %o3             /* %o3: callvm */
+	or   %o1, %g0, %o0             /* %o0: target */
 	ld  [%o3+CALLVM_size], %o1     /* %o1: size */
 	add  %o3, CALLVM_dataoff, %o2  /* %o2: data */
 	/*o0-2:target,size,data*/
@@ -148,46 +144,45 @@ dcCall_sparc:
 	/*leaf functions: may use the first six output registers.*/
 	/*o3-5:free to use */
 
-	/* Compute a matiching stack size (approximate): o3 = align(92+o1,16) */
-
+	/* Compute a matching stack size (approximate): o3 = align(92+o1,16) */
 	add     %o1, (16+1+6)*REGSIZE+ALIGN-1, %o3
 	and     %o3,   -ALIGN, %o3
 	neg     %o3
-	
+
 	/* Prolog. */
-	save	%sp, %o3, %sp	/* min stack size (16+1+6)*sizeof(ptr)=92 paddded to 8-byte alignment => min frame size of 96 bytes. */
-	
-	/* Load output registers. */
+	save    %sp, %o3, %sp    /* min stack size (16+1+6)*sizeof(ptr)=92 paddded to 8-byte alignment => min frame size of 96 bytes. */
 
-	ld	[%i2           ],%o0
-	ld	[%i2+REGSIZE*1 ],%o1
-	ld	[%i2+REGSIZE*2 ],%o2
-	ld	[%i2+REGSIZE*3 ],%o3
-	ld	[%i2+REGSIZE*4 ],%o4
-	ld	[%i2+REGSIZE*5 ],%o5
+	/* Load output registers. */
+	ld      [%i2           ],%o0
+	ld      [%i2+REGSIZE*1 ],%o1
+	ld      [%i2+REGSIZE*2 ],%o2
+	ld      [%i2+REGSIZE*3 ],%o3
+	ld      [%i2+REGSIZE*4 ],%o4
+	ld      [%i2+REGSIZE*5 ],%o5
 
 	/* Copy on stack? */
-	sub	%i1,  REGSIZE*6, %i1   		/* i1 = decrement copy size by 6 regs (=6 regs x 4 bytes = 24 bytes total). */
-	cmp %i1, 0
-    ble .do_call
+	sub     %i1, REGSIZE*6, %i1   /* i1 = decrement copy size by 6 regs (=6 regs x 4 bytes = 24 bytes total). */
+	cmp     %i1, 0
+	ble     .do_call
 	nop
 
 	/* Copy loop: */
-	add     %i2,  REGSIZE*6, %i2	/* i2 = address of 7th word of args buffer. */
-	or      %g0, %g0, %l0			/* l0 = offset initialized to 0. */
-	add     %sp,  (16+1+6)*REGSIZE, %l2	/* l2 = argument area on stack space (7th word). (64+4+6*4 = byte offset 92). */
+	add     %i2, REGSIZE*6, %i2         /* i2 = address of 7th word of args buffer. */
+	or      %g0, %g0, %l0               /* l0 = offset initialized to 0. */
+	add     %sp, (16+1+6)*REGSIZE, %l2  /* l2 = argument area on stack space (7th word). (64+4+6*4 = byte offset 92). */
 .next:
-	ld      [%i2+%l0],%l1			/* Read from arg buffer(%i2) to %l1. */
-	st      %l1, [%l2+%l0]			/* Write %l1 to stack space(%l2). */
-	add     %l0, REGSIZE, %l0		/* Increment offset. */
-	sub     %i1, REGSIZE, %i1		/* Decrement copy size. */
+	ld      [%i2+%l0],%l1      /* Read from arg buffer(%i2) to %l1. */
+	st      %l1, [%l2+%l0]     /* Write %l1 to stack space(%l2). */
+	add     %l0, REGSIZE, %l0  /* Increment offset. */
+	sub     %i1, REGSIZE, %i1  /* Decrement copy size. */
 	cmp     %i1, 0
 	bgt     .next
 	nop
 .do_call:
-	call    %i0						/* Call target. */
+	call    %i0   /* Call target. */
 	nop
-    	or     %o0, %g0, %i0
-    	or     %o1, %g0, %i1
-	jmpl	%i7 + 8, %g0
+	or      %o0, %g0, %i0
+	or      %o1, %g0, %i1
+	jmpl    %i7 + 8, %g0       /* optimized restore;retl;nop */
 	restore
+