From patchwork Fri Aug 6 05:24:22 2010 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Anthony Green X-Patchwork-Id: 61066 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) by ozlabs.org (Postfix) with SMTP id 836A11007E0 for ; Fri, 6 Aug 2010 15:24:42 +1000 (EST) Received: (qmail 7374 invoked by alias); 6 Aug 2010 05:24:39 -0000 Received: (qmail 7359 invoked by uid 22791); 6 Aug 2010 05:24:35 -0000 X-SWARE-Spam-Status: No, hits=-4.2 required=5.0 tests=AWL, BAYES_05, RCVD_IN_DNSWL_HI, SPF_HELO_PASS, TW_BF, TW_OV, TW_VZ, TW_ZB, TW_ZW, T_RP_MATCHES_RCVD X-Spam-Check-By: sourceware.org Received: from mx1.redhat.com (HELO mx1.redhat.com) (209.132.183.28) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Fri, 06 Aug 2010 05:24:26 +0000 Received: from int-mx03.intmail.prod.int.phx2.redhat.com (int-mx03.intmail.prod.int.phx2.redhat.com [10.5.11.16]) by mx1.redhat.com (8.13.8/8.13.8) with ESMTP id o765OPgC015130 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Fri, 6 Aug 2010 01:24:25 -0400 Received: from gmachine.redhat.com (vpn-10-81.rdu.redhat.com [10.11.10.81]) by int-mx03.intmail.prod.int.phx2.redhat.com (8.13.8/8.13.8) with ESMTP id o765ONC3023460; Fri, 6 Aug 2010 01:24:23 -0400 From: Anthony Green To: gcc-patches@gcc.gnu.org Subject: [patch, libffi] Build with msvcc User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.2 (gnu/linux) X-URL: http://www.redhat.com Date: Fri, 06 Aug 2010 01:24:22 -0400 Message-ID: MIME-Version: 1.0 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Delivered-To: mailing list gcc-patches@gcc.gnu.org I'm checking in the following libffi patch. This patch from Dan Witte adds build machinery for the MS compiler. I'm committing this in order to keep GCC's libffi in sync with upstream. Thanks, AG 2010-01-15 Anthony Green * README: Add notes on building with Microsoft Visual C++. 2010-01-15 Daniel Witte * msvcc.sh: New file. * src/x86/win32.S: Port assembly routines to MSVC and #ifdef. * src/x86/ffi.c: Tweak function declaration and remove excess parens. * include/ffi.h.in: Add __declspec(align(8)) to typedef struct ffi_closure. * src/x86/ffi.c: Merge ffi_call_SYSV and ffi_call_STDCALL into new function ffi_call_win32 on X86_WIN32. * src/x86/win32.S (ffi_call_SYSV): Rename to ffi_call_win32. (ffi_call_STDCALL): Remove. * src/prep_cif.c (ffi_prep_cif): Move stack space allocation code to ffi_prep_cif_machdep for x86. * src/x86/ffi.c (ffi_prep_cif_machdep): To here. Index: libffi/include/ffi.h.in =================================================================== --- libffi.orig/include/ffi.h.in +++ libffi/include/ffi.h.in @@ -251,6 +251,9 @@ size_t ffi_java_raw_size (ffi_cif *cif); #if FFI_CLOSURES +#ifdef _MSC_VER +__declspec(align(8)) +#endif typedef struct { char tramp[FFI_TRAMPOLINE_SIZE]; ffi_cif *cif; Index: libffi/src/prep_cif.c =================================================================== --- libffi.orig/src/prep_cif.c +++ libffi/src/prep_cif.c @@ -109,16 +109,13 @@ ffi_status ffi_prep_cif(ffi_cif *cif, ff /* Perform a sanity check on the return type */ FFI_ASSERT_VALID_TYPE(cif->rtype); - /* x86-64 and s390 stack space allocation is handled in prep_machdep. */ -#if !defined M68K && !defined __x86_64__ && !defined S390 && !defined PA + /* x86, x86-64 and s390 stack space allocation is handled in prep_machdep. */ +#if !defined M68K && !defined __i386__ && !defined __x86_64__ && !defined S390 && !defined PA /* Make space for the return structure pointer */ if (cif->rtype->type == FFI_TYPE_STRUCT #ifdef SPARC && (cif->abi != FFI_V9 || cif->rtype->size > 32) #endif -#ifdef X86_DARWIN - && (cif->rtype->size > 8) -#endif ) bytes = STACK_ARG_SIZE(sizeof(void*)); #endif @@ -134,7 +131,7 @@ ffi_status ffi_prep_cif(ffi_cif *cif, ff check after the initialization. */ FFI_ASSERT_VALID_TYPE(*ptr); -#if !defined __x86_64__ && !defined S390 && !defined PA +#if !defined __i386__ && !defined __x86_64__ && !defined S390 && !defined PA #ifdef SPARC if (((*ptr)->type == FFI_TYPE_STRUCT && ((*ptr)->size > 16 || cif->abi != FFI_V9)) Index: libffi/src/x86/ffi.c =================================================================== --- libffi.orig/src/x86/ffi.c +++ libffi/src/x86/ffi.c @@ -148,13 +148,13 @@ void ffi_prep_args(char *stack, extended /* Perform machine dependent cif processing */ ffi_status ffi_prep_cif_machdep(ffi_cif *cif) { + unsigned int i; + ffi_type **ptr; + /* Set the return type flag */ switch (cif->rtype->type) { case FFI_TYPE_VOID: -#ifdef X86 - case FFI_TYPE_STRUCT: -#endif #if defined(X86) || defined (X86_WIN32) || defined(X86_FREEBSD) || defined(X86_DARWIN) || defined(X86_WIN64) case FFI_TYPE_UINT8: case FFI_TYPE_UINT16: @@ -165,7 +165,6 @@ ffi_status ffi_prep_cif_machdep(ffi_cif case FFI_TYPE_UINT32: case FFI_TYPE_SINT32: #endif - case FFI_TYPE_SINT64: case FFI_TYPE_FLOAT: case FFI_TYPE_DOUBLE: @@ -184,8 +183,8 @@ ffi_status ffi_prep_cif_machdep(ffi_cif cif->flags = FFI_TYPE_SINT64; break; -#ifndef X86 case FFI_TYPE_STRUCT: +#ifndef X86 if (cif->rtype->size == 1) { cif->flags = FFI_TYPE_SMALL_STRUCT_1B; /* same as char size */ @@ -207,15 +206,13 @@ ffi_status ffi_prep_cif_machdep(ffi_cif cif->flags = FFI_TYPE_SINT64; /* same as int64 type */ } else +#endif { cif->flags = FFI_TYPE_STRUCT; -#ifdef X86_WIN64 // allocate space for return value pointer cif->bytes += ALIGN(sizeof(void*), FFI_SIZEOF_ARG); -#endif } break; -#endif default: #ifdef X86_WIN64 @@ -229,41 +226,36 @@ ffi_status ffi_prep_cif_machdep(ffi_cif break; } -#ifdef X86_DARWIN - cif->bytes = (cif->bytes + 15) & ~0xF; -#endif + for (ptr = cif->arg_types, i = cif->nargs; i > 0; i--, ptr++) + { + if (((*ptr)->alignment - 1) & cif->bytes) + cif->bytes = ALIGN(cif->bytes, (*ptr)->alignment); + cif->bytes += ALIGN((*ptr)->size, FFI_SIZEOF_ARG); + } #ifdef X86_WIN64 - { - unsigned int i; - ffi_type **ptr; - - for (ptr = cif->arg_types, i = cif->nargs; i > 0; i--, ptr++) - { - if (((*ptr)->alignment - 1) & cif->bytes) - cif->bytes = ALIGN(cif->bytes, (*ptr)->alignment); - cif->bytes += ALIGN((*ptr)->size, FFI_SIZEOF_ARG); - } - } // ensure space for storing four registers cif->bytes += 4 * sizeof(ffi_arg); #endif +#ifdef X86_DARWIN + cif->bytes = (cif->bytes + 15) & ~0xF; +#endif + return FFI_OK; } -extern void ffi_call_SYSV(void (*)(char *, extended_cif *), extended_cif *, - unsigned, unsigned, unsigned *, void (*fn)(void)); - -#ifdef X86_WIN32 -extern void ffi_call_STDCALL(void (*)(char *, extended_cif *), extended_cif *, - unsigned, unsigned, unsigned *, void (*fn)(void)); - -#endif /* X86_WIN32 */ #ifdef X86_WIN64 extern int ffi_call_win64(void (*)(char *, extended_cif *), extended_cif *, unsigned, unsigned, unsigned *, void (*fn)(void)); +#elif defined(X86_WIN32) +extern void +ffi_call_win32(void (*)(char *, extended_cif *), extended_cif *, + unsigned, unsigned, unsigned *, void (*fn)(void)); +#else +extern void ffi_call_SYSV(void (*)(char *, extended_cif *), extended_cif *, + unsigned, unsigned, unsigned *, void (*fn)(void)); #endif void ffi_call(ffi_cif *cif, void (*fn)(void), void *rvalue, void **avalue) @@ -321,18 +313,18 @@ void ffi_call(ffi_cif *cif, void (*fn)(v cif->flags, ecif.rvalue, fn); } break; +#elif defined(X86_WIN32) + case FFI_SYSV: + case FFI_STDCALL: + ffi_call_win32(ffi_prep_args, &ecif, cif->bytes, cif->flags, + ecif.rvalue, fn); + break; #else case FFI_SYSV: ffi_call_SYSV(ffi_prep_args, &ecif, cif->bytes, cif->flags, ecif.rvalue, fn); break; -#ifdef X86_WIN32 - case FFI_STDCALL: - ffi_call_STDCALL(ffi_prep_args, &ecif, cif->bytes, cif->flags, - ecif.rvalue, fn); - break; -#endif /* X86_WIN32 */ -#endif /* X86_WIN64 */ +#endif default: FFI_ASSERT(0); break; @@ -342,6 +334,8 @@ void ffi_call(ffi_cif *cif, void (*fn)(v /** private members **/ +/* The following __attribute__((regparm(1))) decorations will have no effect + on MSVC - standard cdecl convention applies. */ static void ffi_prep_incoming_args_SYSV (char *stack, void **ret, void** args, ffi_cif* cif); void FFI_HIDDEN ffi_closure_SYSV (ffi_closure *) @@ -390,11 +384,8 @@ ffi_closure_win64_inner (ffi_closure *cl } #else -unsigned int FFI_HIDDEN -ffi_closure_SYSV_inner (closure, respp, args) - ffi_closure *closure; - void **respp; - void *args; +unsigned int FFI_HIDDEN __attribute__ ((regparm(1))) +ffi_closure_SYSV_inner (ffi_closure *closure, void **respp, void *args) { /* our various things... */ ffi_cif *cif; @@ -505,7 +496,7 @@ ffi_prep_incoming_args_SYSV(char *stack, /* How to make a trampoline. Derived from gcc/config/i386/i386.c. */ #define FFI_INIT_TRAMPOLINE(TRAMP,FUN,CTX) \ -({ unsigned char *__tramp = (unsigned char*)(TRAMP); \ +{ unsigned char *__tramp = (unsigned char*)(TRAMP); \ unsigned int __fun = (unsigned int)(FUN); \ unsigned int __ctx = (unsigned int)(CTX); \ unsigned int __dis = __fun - (__ctx + 10); \ @@ -513,10 +504,10 @@ ffi_prep_incoming_args_SYSV(char *stack, *(unsigned int*) &__tramp[1] = __ctx; /* movl __ctx, %eax */ \ *(unsigned char *) &__tramp[5] = 0xe9; \ *(unsigned int*) &__tramp[6] = __dis; /* jmp __fun */ \ - }) + } #define FFI_INIT_TRAMPOLINE_STDCALL(TRAMP,FUN,CTX,SIZE) \ -({ unsigned char *__tramp = (unsigned char*)(TRAMP); \ +{ unsigned char *__tramp = (unsigned char*)(TRAMP); \ unsigned int __fun = (unsigned int)(FUN); \ unsigned int __ctx = (unsigned int)(CTX); \ unsigned int __dis = __fun - (__ctx + 10); \ @@ -527,7 +518,7 @@ ffi_prep_incoming_args_SYSV(char *stack, *(unsigned int*) &__tramp[6] = __dis; /* call __fun */ \ *(unsigned char *) &__tramp[10] = 0xc2; \ *(unsigned short*) &__tramp[11] = __size; /* ret __size */ \ - }) + } /* the cif must already be prep'ed */ @@ -627,16 +618,6 @@ ffi_prep_args_raw(char *stack, extended_ * libffi-1.20, this is not the case.) */ -extern void -ffi_call_SYSV(void (*)(char *, extended_cif *), extended_cif *, unsigned, - unsigned, unsigned *, void (*fn)(void)); - -#ifdef X86_WIN32 -extern void -ffi_call_STDCALL(void (*)(char *, extended_cif *), extended_cif *, unsigned, - unsigned, unsigned *, void (*fn)(void)); -#endif /* X86_WIN32 */ - void ffi_raw_call(ffi_cif *cif, void (*fn)(void), void *rvalue, ffi_raw *fake_avalue) { @@ -660,16 +641,18 @@ ffi_raw_call(ffi_cif *cif, void (*fn)(vo switch (cif->abi) { +#ifdef X86_WIN32 + case FFI_SYSV: + case FFI_STDCALL: + ffi_call_win32(ffi_prep_args_raw, &ecif, cif->bytes, cif->flags, + ecif.rvalue, fn); + break; +#else case FFI_SYSV: ffi_call_SYSV(ffi_prep_args_raw, &ecif, cif->bytes, cif->flags, ecif.rvalue, fn); break; -#ifdef X86_WIN32 - case FFI_STDCALL: - ffi_call_STDCALL(ffi_prep_args_raw, &ecif, cif->bytes, cif->flags, - ecif.rvalue, fn); - break; -#endif /* X86_WIN32 */ +#endif default: FFI_ASSERT(0); break; Index: libffi/src/x86/win32.S =================================================================== --- libffi.orig/src/x86/win32.S +++ libffi/src/x86/win32.S @@ -2,6 +2,7 @@ win32.S - Copyright (c) 1996, 1998, 2001, 2002, 2009 Red Hat, Inc. Copyright (c) 2001 John Beniton Copyright (c) 2002 Ranjit Mathew + Copyright (c) 2009 Daniel Witte X86 Foreign Function Interface @@ -31,14 +32,371 @@ #define LIBFFI_ASM #include #include - + +#ifdef _MSC_VER + +.386 +.MODEL FLAT, C + +EXTRN ffi_closure_SYSV_inner:NEAR + +_TEXT SEGMENT + +ffi_call_win32 PROC NEAR, + ffi_prep_args : NEAR PTR DWORD, + ecif : NEAR PTR DWORD, + cif_bytes : DWORD, + cif_flags : DWORD, + rvalue : NEAR PTR DWORD, + fn : NEAR PTR DWORD + + ;; Make room for all of the new args. + mov ecx, cif_bytes + sub esp, ecx + + mov eax, esp + + ;; Place all of the ffi_prep_args in position + push ecif + push eax + call ffi_prep_args + + ;; Return stack to previous state and call the function + add esp, 8 + + call fn + + ;; cdecl: we restore esp in the epilogue, so there's no need to + ;; remove the space we pushed for the args. + ;; stdcall: the callee has already cleaned the stack. + + ;; Load ecx with the return type code + mov ecx, cif_flags + + ;; If the return value pointer is NULL, assume no return value. + cmp rvalue, 0 + jne ca_jumptable + + ;; Even if there is no space for the return value, we are + ;; obliged to handle floating-point values. + cmp ecx, FFI_TYPE_FLOAT + jne ca_epilogue + fstp st(0) + + jmp ca_epilogue + +ca_jumptable: + jmp [ca_jumpdata + 4 * ecx] +ca_jumpdata: + ;; Do not insert anything here between label and jump table. + dd offset ca_epilogue ;; FFI_TYPE_VOID + dd offset ca_retint ;; FFI_TYPE_INT + dd offset ca_retfloat ;; FFI_TYPE_FLOAT + dd offset ca_retdouble ;; FFI_TYPE_DOUBLE + dd offset ca_retlongdouble ;; FFI_TYPE_LONGDOUBLE + dd offset ca_retint8 ;; FFI_TYPE_UINT8 + dd offset ca_retint8 ;; FFI_TYPE_SINT8 + dd offset ca_retint16 ;; FFI_TYPE_UINT16 + dd offset ca_retint16 ;; FFI_TYPE_SINT16 + dd offset ca_retint ;; FFI_TYPE_UINT32 + dd offset ca_retint ;; FFI_TYPE_SINT32 + dd offset ca_retint64 ;; FFI_TYPE_UINT64 + dd offset ca_retint64 ;; FFI_TYPE_SINT64 + dd offset ca_epilogue ;; FFI_TYPE_STRUCT + dd offset ca_retint ;; FFI_TYPE_POINTER + dd offset ca_retint8 ;; FFI_TYPE_SMALL_STRUCT_1B + dd offset ca_retint16 ;; FFI_TYPE_SMALL_STRUCT_2B + dd offset ca_retint ;; FFI_TYPE_SMALL_STRUCT_4B + +ca_retint8: + ;; Load %ecx with the pointer to storage for the return value + mov ecx, rvalue + mov [ecx + 0], al + jmp ca_epilogue + +ca_retint16: + ;; Load %ecx with the pointer to storage for the return value + mov ecx, rvalue + mov [ecx + 0], ax + jmp ca_epilogue + +ca_retint: + ;; Load %ecx with the pointer to storage for the return value + mov ecx, rvalue + mov [ecx + 0], eax + jmp ca_epilogue + +ca_retint64: + ;; Load %ecx with the pointer to storage for the return value + mov ecx, rvalue + mov [ecx + 0], eax + mov [ecx + 4], edx + jmp ca_epilogue + +ca_retfloat: + ;; Load %ecx with the pointer to storage for the return value + mov ecx, rvalue + fstp DWORD PTR [ecx] + jmp ca_epilogue + +ca_retdouble: + ;; Load %ecx with the pointer to storage for the return value + mov ecx, rvalue + fstp QWORD PTR [ecx] + jmp ca_epilogue + +ca_retlongdouble: + ;; Load %ecx with the pointer to storage for the return value + mov ecx, rvalue + fstp TBYTE PTR [ecx] + jmp ca_epilogue + +ca_epilogue: + ;; Epilogue code is autogenerated. + ret +ffi_call_win32 ENDP + +ffi_closure_SYSV PROC NEAR FORCEFRAME + ;; the ffi_closure ctx is passed in eax by the trampoline. + + sub esp, 40 + lea edx, [ebp - 24] + mov [ebp - 12], edx ;; resp + lea edx, [ebp + 8] + mov [esp + 8], edx ;; args + lea edx, [ebp - 12] + mov [esp + 4], edx ;; &resp + mov [esp], eax ;; closure + call ffi_closure_SYSV_inner + mov ecx, [ebp - 12] + +cs_jumptable: + jmp [cs_jumpdata + 4 * eax] +cs_jumpdata: + ;; Do not insert anything here between the label and jump table. + dd offset cs_epilogue ;; FFI_TYPE_VOID + dd offset cs_retint ;; FFI_TYPE_INT + dd offset cs_retfloat ;; FFI_TYPE_FLOAT + dd offset cs_retdouble ;; FFI_TYPE_DOUBLE + dd offset cs_retlongdouble ;; FFI_TYPE_LONGDOUBLE + dd offset cs_retint8 ;; FFI_TYPE_UINT8 + dd offset cs_retint8 ;; FFI_TYPE_SINT8 + dd offset cs_retint16 ;; FFI_TYPE_UINT16 + dd offset cs_retint16 ;; FFI_TYPE_SINT16 + dd offset cs_retint ;; FFI_TYPE_UINT32 + dd offset cs_retint ;; FFI_TYPE_SINT32 + dd offset cs_retint64 ;; FFI_TYPE_UINT64 + dd offset cs_retint64 ;; FFI_TYPE_SINT64 + dd offset cs_retstruct ;; FFI_TYPE_STRUCT + dd offset cs_retint ;; FFI_TYPE_POINTER + dd offset cs_retint8 ;; FFI_TYPE_SMALL_STRUCT_1B + dd offset cs_retint16 ;; FFI_TYPE_SMALL_STRUCT_2B + dd offset cs_retint ;; FFI_TYPE_SMALL_STRUCT_4B + +cs_retint8: + mov al, [ecx] + jmp cs_epilogue + +cs_retint16: + mov ax, [ecx] + jmp cs_epilogue + +cs_retint: + mov eax, [ecx] + jmp cs_epilogue + +cs_retint64: + mov eax, [ecx + 0] + mov edx, [ecx + 4] + jmp cs_epilogue + +cs_retfloat: + fld DWORD PTR [ecx] + jmp cs_epilogue + +cs_retdouble: + fld QWORD PTR [ecx] + jmp cs_epilogue + +cs_retlongdouble: + fld TBYTE PTR [ecx] + jmp cs_epilogue + +cs_retstruct: + ;; Caller expects us to pop struct return value pointer hidden arg. + ;; Epilogue code is autogenerated. + ret 4 + +cs_epilogue: + ;; Epilogue code is autogenerated. + ret +ffi_closure_SYSV ENDP + +#if !FFI_NO_RAW_API + +#define RAW_CLOSURE_CIF_OFFSET ((FFI_TRAMPOLINE_SIZE + 3) AND NOT 3) +#define RAW_CLOSURE_FUN_OFFSET (RAW_CLOSURE_CIF_OFFSET + 4) +#define RAW_CLOSURE_USER_DATA_OFFSET (RAW_CLOSURE_FUN_OFFSET + 4) +#define CIF_FLAGS_OFFSET 20 + +ffi_closure_raw_SYSV PROC NEAR USES esi + ;; the ffi_closure ctx is passed in eax by the trampoline. + + sub esp, 40 + mov esi, [eax + RAW_CLOSURE_CIF_OFFSET] ;; closure->cif + mov edx, [eax + RAW_CLOSURE_USER_DATA_OFFSET] ;; closure->user_data + mov [esp + 12], edx ;; user_data + lea edx, [ebp + 8] + mov [esp + 8], edx ;; raw_args + lea edx, [ebp - 24] + mov [esp + 4], edx ;; &res + mov [esp], esi ;; cif + call DWORD PTR [eax + RAW_CLOSURE_FUN_OFFSET] ;; closure->fun + mov eax, [esi + CIF_FLAGS_OFFSET] ;; cif->flags + lea ecx, [ebp - 24] + +cr_jumptable: + jmp [cr_jumpdata + 4 * eax] +cr_jumpdata: + ;; Do not insert anything here between the label and jump table. + dd offset cr_epilogue ;; FFI_TYPE_VOID + dd offset cr_retint ;; FFI_TYPE_INT + dd offset cr_retfloat ;; FFI_TYPE_FLOAT + dd offset cr_retdouble ;; FFI_TYPE_DOUBLE + dd offset cr_retlongdouble ;; FFI_TYPE_LONGDOUBLE + dd offset cr_retint8 ;; FFI_TYPE_UINT8 + dd offset cr_retint8 ;; FFI_TYPE_SINT8 + dd offset cr_retint16 ;; FFI_TYPE_UINT16 + dd offset cr_retint16 ;; FFI_TYPE_SINT16 + dd offset cr_retint ;; FFI_TYPE_UINT32 + dd offset cr_retint ;; FFI_TYPE_SINT32 + dd offset cr_retint64 ;; FFI_TYPE_UINT64 + dd offset cr_retint64 ;; FFI_TYPE_SINT64 + dd offset cr_epilogue ;; FFI_TYPE_STRUCT + dd offset cr_retint ;; FFI_TYPE_POINTER + dd offset cr_retint8 ;; FFI_TYPE_SMALL_STRUCT_1B + dd offset cr_retint16 ;; FFI_TYPE_SMALL_STRUCT_2B + dd offset cr_retint ;; FFI_TYPE_SMALL_STRUCT_4B + +cr_retint8: + mov al, [ecx] + jmp cr_epilogue + +cr_retint16: + mov ax, [ecx] + jmp cr_epilogue + +cr_retint: + mov eax, [ecx] + jmp cr_epilogue + +cr_retint64: + mov eax, [ecx + 0] + mov edx, [ecx + 4] + jmp cr_epilogue + +cr_retfloat: + fld DWORD PTR [ecx] + jmp cr_epilogue + +cr_retdouble: + fld QWORD PTR [ecx] + jmp cr_epilogue + +cr_retlongdouble: + fld TBYTE PTR [ecx] + jmp cr_epilogue + +cr_epilogue: + ;; Epilogue code is autogenerated. + ret +ffi_closure_raw_SYSV ENDP + +#endif /* !FFI_NO_RAW_API */ + +ffi_closure_STDCALL PROC NEAR FORCEFRAME + ;; the ffi_closure ctx is passed in eax by the trampoline. + + sub esp, 40 + lea edx, [ebp - 24] + mov [ebp - 12], edx ;; resp + lea edx, [ebp + 12] ;; account for stub return address on stack + mov [esp + 8], edx ;; args + lea edx, [ebp - 12] + mov [esp + 4], edx ;; &resp + mov [esp], eax ;; closure + call ffi_closure_SYSV_inner + mov ecx, [ebp - 12] + +cd_jumptable: + jmp [cd_jumpdata + 4 * eax] +cd_jumpdata: + ;; Do not insert anything here between the label and jump table. + dd offset cd_epilogue ;; FFI_TYPE_VOID + dd offset cd_retint ;; FFI_TYPE_INT + dd offset cd_retfloat ;; FFI_TYPE_FLOAT + dd offset cd_retdouble ;; FFI_TYPE_DOUBLE + dd offset cd_retlongdouble ;; FFI_TYPE_LONGDOUBLE + dd offset cd_retint8 ;; FFI_TYPE_UINT8 + dd offset cd_retint8 ;; FFI_TYPE_SINT8 + dd offset cd_retint16 ;; FFI_TYPE_UINT16 + dd offset cd_retint16 ;; FFI_TYPE_SINT16 + dd offset cd_retint ;; FFI_TYPE_UINT32 + dd offset cd_retint ;; FFI_TYPE_SINT32 + dd offset cd_retint64 ;; FFI_TYPE_UINT64 + dd offset cd_retint64 ;; FFI_TYPE_SINT64 + dd offset cd_epilogue ;; FFI_TYPE_STRUCT + dd offset cd_retint ;; FFI_TYPE_POINTER + dd offset cd_retint8 ;; FFI_TYPE_SMALL_STRUCT_1B + dd offset cd_retint16 ;; FFI_TYPE_SMALL_STRUCT_2B + dd offset cd_retint ;; FFI_TYPE_SMALL_STRUCT_4B + +cd_retint8: + mov al, [ecx] + jmp cd_epilogue + +cd_retint16: + mov ax, [ecx] + jmp cd_epilogue + +cd_retint: + mov eax, [ecx] + jmp cd_epilogue + +cd_retint64: + mov eax, [ecx + 0] + mov edx, [ecx + 4] + jmp cd_epilogue + +cd_retfloat: + fld DWORD PTR [ecx] + jmp cd_epilogue + +cd_retdouble: + fld QWORD PTR [ecx] + jmp cd_epilogue + +cd_retlongdouble: + fld TBYTE PTR [ecx] + jmp cd_epilogue + +cd_epilogue: + ;; Epilogue code is autogenerated. + ret +ffi_closure_STDCALL ENDP + +_TEXT ENDS +END + +#else + .text # This assumes we are using gas. .balign 16 - .globl _ffi_call_SYSV - .def _ffi_call_SYSV; .scl 2; .type 32; .endef -_ffi_call_SYSV: + .globl _ffi_call_win32 + .def _ffi_call_win32; .scl 2; .type 32; .endef +_ffi_call_win32: .LFB1: pushl %ebp .LCFI0: @@ -61,8 +419,10 @@ _ffi_call_SYSV: # FIXME: Align the stack to a 128-bit boundary to avoid # potential performance hits. - call *28(%ebp) + call *28(%ebp) + # stdcall functions pop arguments off the stack themselves + # Load %ecx with the return type code movl 20(%ebp),%ecx @@ -181,164 +541,11 @@ _ffi_call_SYSV: movl %ebp,%esp popl %ebp ret -.ffi_call_SYSV_end: +.ffi_call_win32_end: .LFE1: # This assumes we are using gas. .balign 16 - .globl _ffi_call_STDCALL - .def _ffi_call_STDCALL; .scl 2; .type 32; .endef -_ffi_call_STDCALL: -.LFB2: - pushl %ebp -.LCFI2: - movl %esp,%ebp -.LCFI3: - # Make room for all of the new args. - movl 16(%ebp),%ecx - subl %ecx,%esp - - movl %esp,%eax - - # Place all of the ffi_prep_args in position - pushl 12(%ebp) - pushl %eax - call *8(%ebp) - - # Return stack to previous state and call the function - addl $8,%esp - - # FIXME: Align the stack to a 128-bit boundary to avoid - # potential performance hits. - - call *28(%ebp) - - # stdcall functions pop arguments off the stack themselves - - # Load %ecx with the return type code - movl 20(%ebp),%ecx - - # If the return value pointer is NULL, assume no return value. - cmpl $0,24(%ebp) - jne 0f - - # Even if there is no space for the return value, we are - # obliged to handle floating-point values. - cmpl $FFI_TYPE_FLOAT,%ecx - jne .Lsc_noretval - fstp %st(0) - - jmp .Lsc_epilogue - -0: - call 1f - # Do not insert anything here between the call and the jump table. -.Lsc_store_table: - .long .Lsc_noretval /* FFI_TYPE_VOID */ - .long .Lsc_retint /* FFI_TYPE_INT */ - .long .Lsc_retfloat /* FFI_TYPE_FLOAT */ - .long .Lsc_retdouble /* FFI_TYPE_DOUBLE */ - .long .Lsc_retlongdouble /* FFI_TYPE_LONGDOUBLE */ - .long .Lsc_retuint8 /* FFI_TYPE_UINT8 */ - .long .Lsc_retsint8 /* FFI_TYPE_SINT8 */ - .long .Lsc_retuint16 /* FFI_TYPE_UINT16 */ - .long .Lsc_retsint16 /* FFI_TYPE_SINT16 */ - .long .Lsc_retint /* FFI_TYPE_UINT32 */ - .long .Lsc_retint /* FFI_TYPE_SINT32 */ - .long .Lsc_retint64 /* FFI_TYPE_UINT64 */ - .long .Lsc_retint64 /* FFI_TYPE_SINT64 */ - .long .Lsc_retstruct /* FFI_TYPE_STRUCT */ - .long .Lsc_retint /* FFI_TYPE_POINTER */ - .long .Lsc_retstruct1b /* FFI_TYPE_SMALL_STRUCT_1B */ - .long .Lsc_retstruct2b /* FFI_TYPE_SMALL_STRUCT_2B */ - .long .Lsc_retstruct4b /* FFI_TYPE_SMALL_STRUCT_4B */ - -1: - add %ecx, %ecx - add %ecx, %ecx - add (%esp),%ecx - add $4, %esp - jmp *(%ecx) - - /* Sign/zero extend as appropriate. */ -.Lsc_retsint8: - movsbl %al, %eax - jmp .Lsc_retint - -.Lsc_retsint16: - movswl %ax, %eax - jmp .Lsc_retint - -.Lsc_retuint8: - movzbl %al, %eax - jmp .Lsc_retint - -.Lsc_retuint16: - movzwl %ax, %eax - jmp .Lsc_retint - -.Lsc_retint: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - movl %eax,0(%ecx) - jmp .Lsc_epilogue - -.Lsc_retfloat: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - fstps (%ecx) - jmp .Lsc_epilogue - -.Lsc_retdouble: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - fstpl (%ecx) - jmp .Lsc_epilogue - -.Lsc_retlongdouble: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - fstpt (%ecx) - jmp .Lsc_epilogue - -.Lsc_retint64: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - movl %eax,0(%ecx) - movl %edx,4(%ecx) - jmp .Lsc_epilogue - -.Lsc_retstruct1b: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - movb %al,0(%ecx) - jmp .Lsc_epilogue - -.Lsc_retstruct2b: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - movw %ax,0(%ecx) - jmp .Lsc_epilogue - -.Lsc_retstruct4b: - # Load %ecx with the pointer to storage for the return value - movl 24(%ebp),%ecx - movl %eax,0(%ecx) - jmp .Lsc_epilogue - -.Lsc_retstruct: - # Nothing to do! - -.Lsc_noretval: -.Lsc_epilogue: - movl %ebp,%esp - popl %ebp - ret -.ffi_call_STDCALL_end: -.LFE2: - - # This assumes we are using gas. - .balign 16 .globl _ffi_closure_SYSV .def _ffi_closure_SYSV; .scl 2; .type 32; .endef _ffi_closure_SYSV: @@ -742,38 +949,6 @@ _ffi_closure_STDCALL: .LEFDE1: -.LSFDE2: - .long .LEFDE2-.LASFDE2 /* FDE Length */ -.LASFDE2: - .long .LASFDE2-.Lframe1 /* FDE CIE offset */ -#if defined __PIC__ && defined HAVE_AS_X86_PCREL - .long .LFB2-. /* FDE initial location */ -#else - .long .LFB2 -#endif - .long .LFE2-.LFB2 /* FDE address range */ -#ifdef __PIC__ - .byte 0x0 /* .uleb128 0x0; Augmentation size */ -#endif - /* DW_CFA_xxx CFI instructions go here. */ - - .byte 0x4 /* DW_CFA_advance_loc4 */ - .long .LCFI2-.LFB2 - .byte 0xe /* DW_CFA_def_cfa_offset CFA = r4 + 8 = 8(%esp) */ - .byte 0x8 /* .uleb128 0x8 */ - .byte 0x85 /* DW_CFA_offset, column 0x5 %ebp at CFA + 2 * -4 */ - .byte 0x2 /* .uleb128 0x2 */ - - .byte 0x4 /* DW_CFA_advance_loc4 */ - .long .LCFI3-.LCFI2 - .byte 0xd /* DW_CFA_def_cfa_register CFA = r5 = %ebp */ - .byte 0x5 /* .uleb128 0x5 */ - - /* End of DW_CFA_xxx CFI instructions. */ - .align 4 -.LEFDE2: - - .LSFDE3: .long .LEFDE3-.LASFDE3 /* FDE Length */ .LASFDE3: @@ -875,3 +1050,6 @@ _ffi_closure_STDCALL: /* End of DW_CFA_xxx CFI instructions. */ .align 4 .LEFDE5: + +#endif /* !_MSC_VER */ + Index: libffi/README =================================================================== --- libffi.orig/README +++ libffi/README @@ -109,6 +109,14 @@ will add some extra code which will supp are using Purify with libffi. Only use this switch when using Purify, as it will slow down the library. +It's also possible to build libffi on Windows platforms with +Microsoft's Visual C++ compiler. In this case, use the msvcc.sh +wrapper script during configuration like so: + +path/to/configure --enable-shared --enable-static \ + CC=path/to/msvcc.sh LD=link \ + CPP=\"cl -nologo -EP\" + Configure has many other options. Use "configure --help" to see them all. Once configure has finished, type "make". Note that you must be using @@ -128,6 +136,8 @@ See the ChangeLog files for details. 3.0.10 ???-??-?? Fix the N64 build on mips-sgi-irix6.5. Testsuite fixes for Tru64 Unix. + Enable builds with Microsoft's compiler. + Enable x86 builds with Sun's compiler. 3.0.9 Dec-31-09 Add AVR32 and win64 ports. Add ARM softfp support.