Index: /trunk/include/iprt/asm-watcom-x86-16.h
===================================================================
--- /trunk/include/iprt/asm-watcom-x86-16.h	(revision 58708)
+++ /trunk/include/iprt/asm-watcom-x86-16.h	(revision 58708)
@@ -0,0 +1,537 @@
+/** @file
+ * IPRT - Assembly Functions, x86 16-bit Watcom C/C++ pragma aux.
+ */
+
+/*
+ * Copyright (C) 2006-2015 Oracle Corporation
+ *
+ * This file is part of VirtualBox Open Source Edition (OSE), as
+ * available from http://www.virtualbox.org. This file is free software;
+ * you can redistribute it and/or modify it under the terms of the GNU
+ * General Public License (GPL) as published by the Free Software
+ * Foundation, in version 2 as it comes in the "COPYING" file of the
+ * VirtualBox OSE distribution. VirtualBox OSE is distributed in the
+ * hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
+ *
+ * The contents of this file may alternatively be used under the terms
+ * of the Common Development and Distribution License Version 1.0
+ * (CDDL) only, as it comes in the "COPYING.CDDL" file of the
+ * VirtualBox OSE distribution, in which case the provisions of the
+ * CDDL are applicable instead of those of the GPL.
+ *
+ * You may elect to license modified versions of this file under the
+ * terms and conditions of either the GPL or the CDDL or both.
+ */
+
+#ifndef ___iprt_asm_h
+# error "Don't include this header directly."
+#endif
+#ifndef ___iprt_asm_watcom_x86_16_h
+#define ___iprt_asm_watcom_x86_16_h
+
+#if !RT_FAR_DATA
+# error "Only works with far data pointers!"
+#endif
+
+/*
+ * Turns out we cannot use 'ds' for segment stuff here because the compiler
+ * seems to insists on loading the DGROUP segment into 'ds' before calling
+ * stuff when using -ecc.  Using 'es' instead as this seems to work fine.
+ */
+
+#if 0 /* overkill version. */
+# pragma aux ASMCompilerBarrier = \
+    "nop" \
+    parm [] \
+    modify exact [ax bx cx dx es ds];
+#else
+# pragma aux ASMCompilerBarrier = \
+    "" \
+    parm [] \
+    modify exact [];
+#endif
+
+#pragma aux ASMNopPause = \
+    ".686p" \
+    ".xmm2" \
+    "pause" \
+    parm [] nomemory \
+    modify exact [] nomemory;
+
+#pragma aux ASMAtomicXchgU8 = \
+    "xchg es:[bx], al" \
+    parm [es bx] [al] \
+    value [al] \
+    modify exact [al];
+
+#pragma aux ASMAtomicXchgU16 = \
+    "xchg es:[bx], ax" \
+    parm [es bx] [ax] \
+    value [ax] \
+    modify exact [ax];
+
+#pragma aux ASMAtomicXchgU32 = \
+    "shl  ecx, 16" \
+    "mov  cx, ax" \
+    "xchg es:[bx], ecx" \
+    "mov  eax, ecx" \
+    "shr  ecx, 16" \
+    parm [es bx] [ax cx] \
+    value [ax cx] \
+    modify exact [ax cx];
+
+#pragma aux ASMAtomicXchgU64 = \
+    ".586" \
+    "shl eax, 16" \
+    "mov ax, bx" /* eax = high dword */ \
+    "shl ecx, 16" \
+    "mov cx, dx" /* ecx = low dword */ \
+    "mov ebx, ecx" /* ebx = low */ \
+    "mov ecx, eax" /* ecx = high */ \
+    "try_again:" \
+    "lock cmpxchg8b es:[si]" \
+    "jnz try_again" \
+    "xchg eax, edx" \
+    "mov  ebx, eax" \
+    "shr  eax, 16" \
+    "mov  ecx, edx" \
+    "shr  ecx, 16" \
+    parm [es si] [dx cx bx ax] \
+    value [dx cx bx ax] \
+    modify exact [dx cx bx ax];
+
+#pragma aux ASMAtomicCmpXchgU8 = \
+    ".486" \
+    "lock cmpxchg es:[bx], cl" \
+    "setz al" \
+    parm [es bx] [cl] [al] \
+    value [al] \
+    modify exact [al];
+
+#pragma aux ASMAtomicCmpXchgU16 = \
+    ".486" \
+    "lock cmpxchg es:[bx], cx" \
+    "setz al" \
+    parm [es bx] [cx] [ax] \
+    value [al] \
+    modify exact [ax];
+
+#pragma aux ASMAtomicCmpXchgU32 = \
+    ".486" \
+    "shl ecx, 16" \
+    "mov cx, dx" \
+    "shl eax, 16" \
+    "mov ax, di" \
+    "rol eax, 16" \
+    "lock cmpxchg es:[bx], ecx" \
+    "setz al" \
+    parm [es bx] [cx dx] [ax di] \
+    value [al] \
+    modify exact [ax cx];
+
+/* ASMAtomicCmpXchgU64: External assembly implementation, too few registers for parameters.  */
+/* ASMAtomicCmpXchgExU32: External assembly implementation, too few registers for parameters.  */
+/* ASMAtomicCmpXchgExU64: External assembly implementation, too few registers for parameters.  */
+
+#pragma aux ASMSerializeInstruction = \
+    ".586" \
+    "xor eax, eax" \
+    "cpuid" \
+    parm [] \
+    modify exact [ax bx cx dx];
+
+#pragma aux ASMAtomicReadU64 = \
+    ".586" \
+    "xor eax, eax" \
+    "xor edx, edx" \
+    "xor ebx, ebx" \
+    "xor ecx, ecx" \
+    "lock cmpxchg8b es:[si]" \
+    "xchg eax, edx" \
+    "mov  ebx, eax" \
+    "shr  eax, 16" \
+    "mov  ecx, edx" \
+    "shr  ecx, 16" \
+    parm [es si] [dx cx bx ax] \
+    value [dx cx bx ax] \
+    modify exact [dx cx bx ax];
+
+#pragma aux ASMAtomicUoReadU64 = \
+    ".586" \
+    "xor eax, eax" \
+    "xor edx, edx" \
+    "xor ebx, ebx" \
+    "xor ecx, ecx" \
+    "lock cmpxchg8b es:[si]" \
+    "xchg eax, edx" \
+    "mov  ebx, eax" \
+    "shr  eax, 16" \
+    "mov  ecx, edx" \
+    "shr  ecx, 16" \
+    parm [es si] [dx cx bx ax] \
+    value [dx cx bx ax] \
+    modify exact [dx cx bx ax];
+
+#pragma aux ASMAtomicAddU16 = \
+    ".486" \
+    "lock xadd es:[bx], ax" \
+    parm [es bx] [ax] \
+    value [ax] \
+    modify exact [ax];
+
+#pragma aux ASMAtomicAddU32 = \
+    ".486" \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock xadd es:[bx], edx" \
+    "mov ax, dx" \
+    "shr edx, 16" \
+    parm [es bx] [ax dx] \
+    value [ax dx] \
+    modify exact [ax dx];
+
+#pragma aux ASMAtomicIncU16 = \
+    ".486" \
+    "mov ax, 1" \
+    "lock xadd es:[bx], ax" \
+    "inc ax" \
+    parm [es bx] \
+    value [ax] \
+    modify exact [ax];
+
+#pragma aux ASMAtomicIncU32 = \
+    ".486" \
+    "mov edx, 1" \
+    "lock xadd es:[bx], edx" \
+    "inc edx" \
+    "mov ax, dx" \
+    "shr edx, 16" \
+    parm [es bx] \
+    value [ax dx] \
+    modify exact [ax dx];
+
+/* ASMAtomicIncU64: Should be done by C inline or in external file. */
+
+#pragma aux ASMAtomicDecU16 = \
+    ".486" \
+    "mov ax, 0ffffh" \
+    "lock xadd es:[bx], ax" \
+    "dec ax" \
+    parm [es bx] \
+    value [ax] \
+    modify exact [ax];
+
+#pragma aux ASMAtomicDecU32 = \
+    ".486" \
+    "mov edx, 0ffffffffh" \
+    "lock xadd es:[bx], edx" \
+    "dec edx" \
+    "mov ax, dx" \
+    "shr edx, 16" \
+    parm [es bx] \
+    value [ax dx] \
+    modify exact [ax dx];
+
+/* ASMAtomicDecU64: Should be done by C inline or in external file. */
+
+#pragma aux ASMAtomicOrU32 = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock or es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+/* ASMAtomicOrU64: Should be done by C inline or in external file. */
+
+#pragma aux ASMAtomicAndU32 = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock and es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+/* ASMAtomicAndU64: Should be done by C inline or in external file. */
+
+#pragma aux ASMAtomicUoOrU32 = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "or es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+/* ASMAtomicUoOrU64: Should be done by C inline or in external file. */
+
+#pragma aux ASMAtomicUoAndU32 = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "and es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+/* ASMAtomicUoAndU64: Should be done by C inline or in external file. */
+
+#pragma aux ASMAtomicUoIncU32 = \
+    ".486" \
+    "mov edx, 1" \
+    "xadd es:[bx], edx" \
+    "inc edx" \
+    "mov ax, dx" \
+    "shr edx, 16" \
+    parm [es bx] \
+    value [ax dx] \
+    modify exact [ax dx];
+
+#pragma aux ASMAtomicUoDecU32 = \
+    ".486" \
+    "mov edx, 0ffffffffh" \
+    "xadd es:[bx], edx" \
+    "dec edx" \
+    "mov ax, dx" \
+    "shr edx, 16" \
+    parm [es bx] \
+    value [ax dx] \
+    modify exact [ax dx];
+
+#pragma aux ASMMemZeroPage = \
+    "mov ecx, 1024" \
+    "xor eax, eax" \
+    "rep stosd"  \
+    parm [es di] \
+    modify exact [ax cx di];
+
+#pragma aux ASMMemZero32 = \
+    "and ecx, 0ffffh" /* probably not necessary, lazy bird should check... */ \
+    "shr ecx, 2" \
+    "xor eax, eax" \
+    "rep stosd"  \
+    parm [es di] [cx] \
+    modify exact [ax cx di];
+
+#pragma aux ASMMemZero32 = \
+    "and ecx, 0ffffh" /* probably not necessary, lazy bird should check... */ \
+    "shr ecx, 2" \
+    "shr eax, 16" \
+    "mov ax, dx" \
+    "rol eax, 16" \
+    "rep stosd"  \
+    parm [es di] [cx] [ax dx]\
+    modify exact [ax cx di];
+
+#pragma aux ASMProbeReadByte = \
+    "mov al, es:[bx]" \
+    parm [es bx] \
+    value [al] \
+    modify exact [al];
+
+#pragma aux ASMBitSet = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "bts es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+#pragma aux ASMAtomicBitSet = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock bts es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+#pragma aux ASMBitClear = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "btr es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+#pragma aux ASMAtomicBitClear = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock btr es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+#pragma aux ASMBitToggle = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "btc es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+
+#pragma aux ASMAtomicBitToggle = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock btc es:[bx], edx" \
+    parm [es bx] [ax dx] \
+    modify exact [dx];
+///
+///
+#pragma aux ASMBitTestAndSet = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "bts es:[bx], edx" \
+    "setc al" \
+    parm [es bx] [ax dx] \
+    value [al] \
+    modify exact [ax dx];
+
+#pragma aux ASMAtomicBitTestAndSet = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock bts es:[bx], edx" \
+    "setc al" \
+    parm [es bx] [ax dx] \
+    value [al] \
+    modify exact [ax dx];
+
+#pragma aux ASMBitTestAndClear = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "btr es:[bx], edx" \
+    "setc al" \
+    parm [es bx] [ax dx] \
+    value [al] \
+    modify exact [ax dx];
+
+#pragma aux ASMAtomicBitTestAndClear = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock btr es:[bx], edx" \
+    "setc al" \
+    parm [es bx] [ax dx] \
+    value [al] \
+    modify exact [ax dx];
+
+#pragma aux ASMBitTestAndToggle = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "btc es:[bx], edx" \
+    "setc al" \
+    parm [es bx] [ax dx] \
+    value [al] \
+    modify exact [ax dx];
+
+#pragma aux ASMAtomicBitTestAndToggle = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "lock btc es:[bx], edx" \
+    "setc al" \
+    parm [es bx] [ax dx] \
+    value [al] \
+    modify exact [ax dx];
+
+/** @todo this is way to much inline assembly, better off in an external function. */
+#pragma aux ASMBitFirstClear = \
+    "mov bx, di" /* save start of bitmap for later */ \
+    "shl ecx, 16" \
+    "mov cx, ax" /* ecx = cBits */ \
+    "add ecx, 31" \
+    "shr ecx, 5" /* cDWord = RT_ALIGN_32(cBits, 32) / 32; */  \
+    "mov eax, 0ffffffffh" \
+    "mov edx, eax" /* default return value */ \
+    "repe scasd" \
+    "je done" \
+    "sub di, 4" /* rewind di */ \
+    "xor eax, es:[di]" /* load inverted bits */ \
+    "sub di, bx" /* calc byte offset */ \
+    "movzx edi, di" \
+    "shl edi, 3" /* convert byte to bit offset */ \
+    "bsf edx, eax" \
+    "add edx, edi" \
+    "done:" \
+    "mov eax, edx" \
+    "shr edx, 16" \
+    parm [es di] [ax cx] \
+    value [ax dx] \
+    modify exact [ax bx cx dx di];
+
+/* ASMBitNextClear: Too much work, do when needed. */
+
+/** @todo this is way to much inline assembly, better off in an external function. */
+#pragma aux ASMBitFirstSet = \
+    "mov bx, di" /* save start of bitmap for later */ \
+    "shl ecx, 16" \
+    "mov cx, ax" /* ecx = cBits */ \
+    "add ecx, 31" \
+    "shr ecx, 5" /* cDWord = RT_ALIGN_32(cBits, 32) / 32; */  \
+    "xor eax, eax" \
+    "mov edx, 0ffffffffh" /* default return value */ \
+    "repe scasd" \
+    "je done" \
+    "sub di, 4" /* rewind di */ \
+    "mov eax, es:[di]" /* reload previous dword */ \
+    "sub di, bx" /* calc byte offset */ \
+    "movzx edi, di" \
+    "shl edi, 3" /* convert byte to bit offset */ \
+    "bsf edx, eax" /* find first set bit in dword */ \
+    "add edx, edi" /* calc final bit number */ \
+    "done:" \
+    "mov eax, edx" \
+    "shr edx, 16" \
+    parm [es di] [ax cx] \
+    value [ax dx] \
+    modify exact [ax bx cx dx di];
+
+/* ASMBitNextSet: Too much work, do when needed. */
+
+#pragma aux ASMBitFirstSetU32 = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "bsf eax, edx" \
+    "jz  not_found" \
+    "inc ax" \
+    "jmp done" \
+    "not_found:" \
+    "xor ax, ax" \
+    "done:" \
+    parm [ax dx] nomemory \
+    value [ax] \
+    modify exact [ax dx] nomemory;
+
+#pragma aux ASMBitLastSetU32 = \
+    "shl edx, 16" \
+    "mov dx, ax" \
+    "bsr eax, edx" \
+    "jz  not_found" \
+    "inc ax" \
+    "jmp done" \
+    "not_found:" \
+    "xor ax, ax" \
+    "done:" \
+    parm [ax dx] nomemory \
+    value [ax] \
+    modify exact [ax dx] nomemory;
+
+#pragma aux ASMByteSwapU16 = \
+    "ror ax, 8" \
+    parm [ax] nomemory \
+    value [ax] \
+    modify exact [ax] nomemory;
+
+#pragma aux ASMByteSwapU32 = \
+    "xchg ax, dx" \
+    parm [ax dx] nomemory \
+    value [ax dx] \
+    modify exact [ax dx] nomemory;
+
+#pragma aux ASMRotateLeftU32 = \
+    "shl    edx, 16" \
+    "mov    dx, ax" \
+    "rol    edx, cl" \
+    "mov    eax, edx" \
+    "shr    edx, 16" \
+    parm [ax dx] [cx] nomemory \
+    value [ax dx] \
+    modify exact [ax dx] nomemory;
+
+#pragma aux ASMRotateRightU32 = \
+    "shl    edx, 16" \
+    "mov    dx, ax" \
+    "ror    edx, cl" \
+    "mov    eax, edx" \
+    "shr    edx, 16" \
+    parm [ax dx] [cx] nomemory \
+    value [ax dx] \
+    modify exact [ax dx] nomemory;
+
+#endif