/* * mini-amd64.c: AMD64 backend for the Mono code generator * * Based on mini-x86.c. * * Authors: * Paolo Molaro (lupus@ximian.com) * Dietmar Maurer (dietmar@ximian.com) * Patrik Torstensson * Zoltan Varga (vargaz@gmail.com) * * (C) 2003 Ximian, Inc. */ #include "mini.h" #include #include #ifdef HAVE_UNISTD_H #include #endif #include #include #include #include #include #include #include #include "trace.h" #include "ir-emit.h" #include "mini-amd64.h" #include "cpu-amd64.h" #include "debugger-agent.h" /* * Can't define this in mini-amd64.h cause that would turn on the generic code in * method-to-ir.c. */ #define MONO_ARCH_IMT_REG AMD64_R11 static gint lmf_tls_offset = -1; static gint lmf_addr_tls_offset = -1; static gint appdomain_tls_offset = -1; #ifdef MONO_XEN_OPT static gboolean optimize_for_xen = TRUE; #else #define optimize_for_xen 0 #endif #define ALIGN_TO(val,align) ((((guint64)val) + ((align) - 1)) & ~((align) - 1)) #define IS_IMM32(val) ((((guint64)val) >> 32) == 0) #define IS_REX(inst) (((inst) >= 0x40) && ((inst) <= 0x4f)) #ifdef HOST_WIN32 /* Under windows, the calling convention is never stdcall */ #define CALLCONV_IS_STDCALL(call_conv) (FALSE) #else #define CALLCONV_IS_STDCALL(call_conv) ((call_conv) == MONO_CALL_STDCALL) #endif /* amd64_mov_reg_imm () */ #define BREAKPOINT_SIZE 8 /* This mutex protects architecture specific caches */ #define mono_mini_arch_lock() EnterCriticalSection (&mini_arch_mutex) #define mono_mini_arch_unlock() LeaveCriticalSection (&mini_arch_mutex) static CRITICAL_SECTION mini_arch_mutex; MonoBreakpointInfo mono_breakpoint_info [MONO_BREAKPOINT_ARRAY_SIZE]; /* * The code generated for sequence points reads from this location, which is * made read-only when single stepping is enabled. */ static gpointer ss_trigger_page; /* Enabled breakpoints read from this trigger page */ static gpointer bp_trigger_page; #ifdef HOST_WIN32 /* On Win64 always reserve first 32 bytes for first four arguments */ #define ARGS_OFFSET 48 #else #define ARGS_OFFSET 16 #endif #define GP_SCRATCH_REG AMD64_R11 /* * AMD64 register usage: * - callee saved registers are used for global register allocation * - %r11 is used for materializing 64 bit constants in opcodes * - the rest is used for local allocation */ /* * Floating point comparison results: * ZF PF CF * A > B 0 0 0 * A < B 0 0 1 * A = B 1 0 0 * A > B 0 0 0 * UNORDERED 1 1 1 */ const char* mono_arch_regname (int reg) { switch (reg) { case AMD64_RAX: return "%rax"; case AMD64_RBX: return "%rbx"; case AMD64_RCX: return "%rcx"; case AMD64_RDX: return "%rdx"; case AMD64_RSP: return "%rsp"; case AMD64_RBP: return "%rbp"; case AMD64_RDI: return "%rdi"; case AMD64_RSI: return "%rsi"; case AMD64_R8: return "%r8"; case AMD64_R9: return "%r9"; case AMD64_R10: return "%r10"; case AMD64_R11: return "%r11"; case AMD64_R12: return "%r12"; case AMD64_R13: return "%r13"; case AMD64_R14: return "%r14"; case AMD64_R15: return "%r15"; } return "unknown"; } static const char * packed_xmmregs [] = { "p:xmm0", "p:xmm1", "p:xmm2", "p:xmm3", "p:xmm4", "p:xmm5", "p:xmm6", "p:xmm7", "p:xmm8", "p:xmm9", "p:xmm10", "p:xmm11", "p:xmm12", "p:xmm13", "p:xmm14", "p:xmm15" }; static const char * single_xmmregs [] = { "s:xmm0", "s:xmm1", "s:xmm2", "s:xmm3", "s:xmm4", "s:xmm5", "s:xmm6", "s:xmm7", "s:xmm8", "s:xmm9", "s:xmm10", "s:xmm11", "s:xmm12", "s:xmm13", "s:xmm14", "s:xmm15" }; const char* mono_arch_fregname (int reg) { if (reg < AMD64_XMM_NREG) return single_xmmregs [reg]; else return "unknown"; } const char * mono_arch_xregname (int reg) { if (reg < AMD64_XMM_NREG) return packed_xmmregs [reg]; else return "unknown"; } G_GNUC_UNUSED static void break_count (void) { } G_GNUC_UNUSED static gboolean debug_count (void) { static int count = 0; count ++; if (!getenv ("COUNT")) return TRUE; if (count == atoi (getenv ("COUNT"))) { break_count (); } if (count > atoi (getenv ("COUNT"))) { return FALSE; } return TRUE; } static gboolean debug_omit_fp (void) { #if 0 return debug_count (); #else return TRUE; #endif } static inline gboolean amd64_is_near_call (guint8 *code) { /* Skip REX */ if ((code [0] >= 0x40) && (code [0] <= 0x4f)) code += 1; return code [0] == 0xe8; } static inline void amd64_patch (unsigned char* code, gpointer target) { guint8 rex = 0; /* Skip REX */ if ((code [0] >= 0x40) && (code [0] <= 0x4f)) { rex = code [0]; code += 1; } if ((code [0] & 0xf8) == 0xb8) { /* amd64_set_reg_template */ *(guint64*)(code + 1) = (guint64)target; } else if ((code [0] == 0x8b) && rex && x86_modrm_mod (code [1]) == 0 && x86_modrm_rm (code [1]) == 5) { /* mov 0(%rip), %dreg */ *(guint32*)(code + 2) = (guint32)(guint64)target - 7; } else if ((code [0] == 0xff) && (code [1] == 0x15)) { /* call *(%rip) */ *(guint32*)(code + 2) = ((guint32)(guint64)target) - 7; } else if ((code [0] == 0xe8)) { /* call */ gint64 disp = (guint8*)target - (guint8*)code; g_assert (amd64_is_imm32 (disp)); x86_patch (code, (unsigned char*)target); } else x86_patch (code, (unsigned char*)target); } void mono_amd64_patch (unsigned char* code, gpointer target) { amd64_patch (code, target); } typedef enum { ArgInIReg, ArgInFloatSSEReg, ArgInDoubleSSEReg, ArgOnStack, ArgValuetypeInReg, ArgValuetypeAddrInIReg, ArgNone /* only in pair_storage */ } ArgStorage; typedef struct { gint16 offset; gint8 reg; ArgStorage storage; /* Only if storage == ArgValuetypeInReg */ ArgStorage pair_storage [2]; gint8 pair_regs [2]; } ArgInfo; typedef struct { int nargs; guint32 stack_usage; guint32 reg_usage; guint32 freg_usage; gboolean need_stack_align; gboolean vtype_retaddr; ArgInfo ret; ArgInfo sig_cookie; ArgInfo args [1]; } CallInfo; #define DEBUG(a) if (cfg->verbose_level > 1) a #ifdef HOST_WIN32 #define PARAM_REGS 4 static AMD64_Reg_No param_regs [] = { AMD64_RCX, AMD64_RDX, AMD64_R8, AMD64_R9 }; static AMD64_Reg_No return_regs [] = { AMD64_RAX, AMD64_RDX }; #else #define PARAM_REGS 6 static AMD64_Reg_No param_regs [] = { AMD64_RDI, AMD64_RSI, AMD64_RDX, AMD64_RCX, AMD64_R8, AMD64_R9 }; static AMD64_Reg_No return_regs [] = { AMD64_RAX, AMD64_RDX }; #endif static void inline add_general (guint32 *gr, guint32 *stack_size, ArgInfo *ainfo) { ainfo->offset = *stack_size; if (*gr >= PARAM_REGS) { ainfo->storage = ArgOnStack; (*stack_size) += sizeof (gpointer); } else { ainfo->storage = ArgInIReg; ainfo->reg = param_regs [*gr]; (*gr) ++; } } #ifdef HOST_WIN32 #define FLOAT_PARAM_REGS 4 #else #define FLOAT_PARAM_REGS 8 #endif static void inline add_float (guint32 *gr, guint32 *stack_size, ArgInfo *ainfo, gboolean is_double) { ainfo->offset = *stack_size; if (*gr >= FLOAT_PARAM_REGS) { ainfo->storage = ArgOnStack; (*stack_size) += sizeof (gpointer); } else { /* A double register */ if (is_double) ainfo->storage = ArgInDoubleSSEReg; else ainfo->storage = ArgInFloatSSEReg; ainfo->reg = *gr; (*gr) += 1; } } typedef enum ArgumentClass { ARG_CLASS_NO_CLASS, ARG_CLASS_MEMORY, ARG_CLASS_INTEGER, ARG_CLASS_SSE } ArgumentClass; static ArgumentClass merge_argument_class_from_type (MonoType *type, ArgumentClass class1) { ArgumentClass class2 = ARG_CLASS_NO_CLASS; MonoType *ptype; ptype = mini_type_get_underlying_type (NULL, type); switch (ptype->type) { case MONO_TYPE_BOOLEAN: case MONO_TYPE_CHAR: case MONO_TYPE_I1: case MONO_TYPE_U1: case MONO_TYPE_I2: case MONO_TYPE_U2: case MONO_TYPE_I4: case MONO_TYPE_U4: case MONO_TYPE_I: case MONO_TYPE_U: case MONO_TYPE_STRING: case MONO_TYPE_OBJECT: case MONO_TYPE_CLASS: case MONO_TYPE_SZARRAY: case MONO_TYPE_PTR: case MONO_TYPE_FNPTR: case MONO_TYPE_ARRAY: case MONO_TYPE_I8: case MONO_TYPE_U8: class2 = ARG_CLASS_INTEGER; break; case MONO_TYPE_R4: case MONO_TYPE_R8: #ifdef HOST_WIN32 class2 = ARG_CLASS_INTEGER; #else class2 = ARG_CLASS_SSE; #endif break; case MONO_TYPE_TYPEDBYREF: g_assert_not_reached (); case MONO_TYPE_GENERICINST: if (!mono_type_generic_inst_is_valuetype (ptype)) { class2 = ARG_CLASS_INTEGER; break; } /* fall through */ case MONO_TYPE_VALUETYPE: { MonoMarshalType *info = mono_marshal_load_type_info (ptype->data.klass); int i; for (i = 0; i < info->num_fields; ++i) { class2 = class1; class2 = merge_argument_class_from_type (info->fields [i].field->type, class2); } break; } default: g_assert_not_reached (); } /* Merge */ if (class1 == class2) ; else if (class1 == ARG_CLASS_NO_CLASS) class1 = class2; else if ((class1 == ARG_CLASS_MEMORY) || (class2 == ARG_CLASS_MEMORY)) class1 = ARG_CLASS_MEMORY; else if ((class1 == ARG_CLASS_INTEGER) || (class2 == ARG_CLASS_INTEGER)) class1 = ARG_CLASS_INTEGER; else class1 = ARG_CLASS_SSE; return class1; } static void add_valuetype (MonoGenericSharingContext *gsctx, MonoMethodSignature *sig, ArgInfo *ainfo, MonoType *type, gboolean is_return, guint32 *gr, guint32 *fr, guint32 *stack_size) { guint32 size, quad, nquads, i; ArgumentClass args [2]; MonoMarshalType *info = NULL; MonoClass *klass; MonoGenericSharingContext tmp_gsctx; gboolean pass_on_stack = FALSE; /* * The gsctx currently contains no data, it is only used for checking whenever * open types are allowed, some callers like mono_arch_get_argument_info () * don't pass it to us, so work around that. */ if (!gsctx) gsctx = &tmp_gsctx; klass = mono_class_from_mono_type (type); size = mini_type_stack_size_full (gsctx, &klass->byval_arg, NULL, sig->pinvoke); #ifndef HOST_WIN32 if (!sig->pinvoke && !disable_vtypes_in_regs && ((is_return && (size == 8)) || (!is_return && (size <= 16)))) { /* We pass and return vtypes of size 8 in a register */ } else if (!sig->pinvoke || (size == 0) || (size > 16)) { pass_on_stack = TRUE; } #else if (!sig->pinvoke) { pass_on_stack = TRUE; } #endif if (pass_on_stack) { /* Allways pass in memory */ ainfo->offset = *stack_size; *stack_size += ALIGN_TO (size, 8); ainfo->storage = ArgOnStack; return; } /* FIXME: Handle structs smaller than 8 bytes */ //if ((size % 8) != 0) // NOT_IMPLEMENTED; if (size > 8) nquads = 2; else nquads = 1; if (!sig->pinvoke) { /* Always pass in 1 or 2 integer registers */ args [0] = ARG_CLASS_INTEGER; args [1] = ARG_CLASS_INTEGER; /* Only the simplest cases are supported */ if (is_return && nquads != 1) { args [0] = ARG_CLASS_MEMORY; args [1] = ARG_CLASS_MEMORY; } } else { /* * Implement the algorithm from section 3.2.3 of the X86_64 ABI. * The X87 and SSEUP stuff is left out since there are no such types in * the CLR. */ info = mono_marshal_load_type_info (klass); g_assert (info); #ifndef HOST_WIN32 if (info->native_size > 16) { ainfo->offset = *stack_size; *stack_size += ALIGN_TO (info->native_size, 8); ainfo->storage = ArgOnStack; return; } #else switch (info->native_size) { case 1: case 2: case 4: case 8: break; default: if (is_return) { ainfo->storage = ArgOnStack; ainfo->offset = *stack_size; *stack_size += ALIGN_TO (info->native_size, 8); } else { ainfo->storage = ArgValuetypeAddrInIReg; if (*gr < PARAM_REGS) { ainfo->pair_storage [0] = ArgInIReg; ainfo->pair_regs [0] = param_regs [*gr]; (*gr) ++; } else { ainfo->pair_storage [0] = ArgOnStack; ainfo->offset = *stack_size; *stack_size += 8; } } return; } #endif args [0] = ARG_CLASS_NO_CLASS; args [1] = ARG_CLASS_NO_CLASS; for (quad = 0; quad < nquads; ++quad) { int size; guint32 align; ArgumentClass class1; if (info->num_fields == 0) class1 = ARG_CLASS_MEMORY; else class1 = ARG_CLASS_NO_CLASS; for (i = 0; i < info->num_fields; ++i) { size = mono_marshal_type_size (info->fields [i].field->type, info->fields [i].mspec, &align, TRUE, klass->unicode); if ((info->fields [i].offset < 8) && (info->fields [i].offset + size) > 8) { /* Unaligned field */ NOT_IMPLEMENTED; } /* Skip fields in other quad */ if ((quad == 0) && (info->fields [i].offset >= 8)) continue; if ((quad == 1) && (info->fields [i].offset < 8)) continue; class1 = merge_argument_class_from_type (info->fields [i].field->type, class1); } g_assert (class1 != ARG_CLASS_NO_CLASS); args [quad] = class1; } } /* Post merger cleanup */ if ((args [0] == ARG_CLASS_MEMORY) || (args [1] == ARG_CLASS_MEMORY)) args [0] = args [1] = ARG_CLASS_MEMORY; /* Allocate registers */ { int orig_gr = *gr; int orig_fr = *fr; ainfo->storage = ArgValuetypeInReg; ainfo->pair_storage [0] = ainfo->pair_storage [1] = ArgNone; for (quad = 0; quad < nquads; ++quad) { switch (args [quad]) { case ARG_CLASS_INTEGER: if (*gr >= PARAM_REGS) args [quad] = ARG_CLASS_MEMORY; else { ainfo->pair_storage [quad] = ArgInIReg; if (is_return) ainfo->pair_regs [quad] = return_regs [*gr]; else ainfo->pair_regs [quad] = param_regs [*gr]; (*gr) ++; } break; case ARG_CLASS_SSE: if (*fr >= FLOAT_PARAM_REGS) args [quad] = ARG_CLASS_MEMORY; else { ainfo->pair_storage [quad] = ArgInDoubleSSEReg; ainfo->pair_regs [quad] = *fr; (*fr) ++; } break; case ARG_CLASS_MEMORY: break; default: g_assert_not_reached (); } } if ((args [0] == ARG_CLASS_MEMORY) || (args [1] == ARG_CLASS_MEMORY)) { /* Revert possible register assignments */ *gr = orig_gr; *fr = orig_fr; ainfo->offset = *stack_size; if (sig->pinvoke) *stack_size += ALIGN_TO (info->native_size, 8); else *stack_size += nquads * sizeof (gpointer); ainfo->storage = ArgOnStack; } } } /* * get_call_info: * * Obtain information about a call according to the calling convention. * For AMD64, see the "System V ABI, x86-64 Architecture Processor Supplement * Draft Version 0.23" document for more information. */ static CallInfo* get_call_info (MonoGenericSharingContext *gsctx, MonoMemPool *mp, MonoMethodSignature *sig, gboolean is_pinvoke) { guint32 i, gr, fr; MonoType *ret_type; int n = sig->hasthis + sig->param_count; guint32 stack_size = 0; CallInfo *cinfo; if (mp) cinfo = mono_mempool_alloc0 (mp, sizeof (CallInfo) + (sizeof (ArgInfo) * n)); else cinfo = g_malloc0 (sizeof (CallInfo) + (sizeof (ArgInfo) * n)); cinfo->nargs = n; gr = 0; fr = 0; /* return value */ { ret_type = mini_type_get_underlying_type (gsctx, sig->ret); switch (ret_type->type) { case MONO_TYPE_BOOLEAN: case MONO_TYPE_I1: case MONO_TYPE_U1: case MONO_TYPE_I2: case MONO_TYPE_U2: case MONO_TYPE_CHAR: case MONO_TYPE_I4: case MONO_TYPE_U4: case MONO_TYPE_I: case MONO_TYPE_U: case MONO_TYPE_PTR: case MONO_TYPE_FNPTR: case MONO_TYPE_CLASS: case MONO_TYPE_OBJECT: case MONO_TYPE_SZARRAY: case MONO_TYPE_ARRAY: case MONO_TYPE_STRING: cinfo->ret.storage = ArgInIReg; cinfo->ret.reg = AMD64_RAX; break; case MONO_TYPE_U8: case MONO_TYPE_I8: cinfo->ret.storage = ArgInIReg; cinfo->ret.reg = AMD64_RAX; break; case MONO_TYPE_R4: cinfo->ret.storage = ArgInFloatSSEReg; cinfo->ret.reg = AMD64_XMM0; break; case MONO_TYPE_R8: cinfo->ret.storage = ArgInDoubleSSEReg; cinfo->ret.reg = AMD64_XMM0; break; case MONO_TYPE_GENERICINST: if (!mono_type_generic_inst_is_valuetype (ret_type)) { cinfo->ret.storage = ArgInIReg; cinfo->ret.reg = AMD64_RAX; break; } /* fall through */ case MONO_TYPE_VALUETYPE: { guint32 tmp_gr = 0, tmp_fr = 0, tmp_stacksize = 0; add_valuetype (gsctx, sig, &cinfo->ret, sig->ret, TRUE, &tmp_gr, &tmp_fr, &tmp_stacksize); if (cinfo->ret.storage == ArgOnStack) { cinfo->vtype_retaddr = TRUE; /* The caller passes the address where the value is stored */ add_general (&gr, &stack_size, &cinfo->ret); } break; } case MONO_TYPE_TYPEDBYREF: /* Same as a valuetype with size 24 */ add_general (&gr, &stack_size, &cinfo->ret); ; break; case MONO_TYPE_VOID: break; default: g_error ("Can't handle as return value 0x%x", sig->ret->type); } } /* this */ if (sig->hasthis) add_general (&gr, &stack_size, cinfo->args + 0); if (!sig->pinvoke && (sig->call_convention == MONO_CALL_VARARG) && (n == 0)) { gr = PARAM_REGS; fr = FLOAT_PARAM_REGS; /* Emit the signature cookie just before the implicit arguments */ add_general (&gr, &stack_size, &cinfo->sig_cookie); } for (i = 0; i < sig->param_count; ++i) { ArgInfo *ainfo = &cinfo->args [sig->hasthis + i]; MonoType *ptype; #ifdef HOST_WIN32 /* The float param registers and other param registers must be the same index on Windows x64.*/ if (gr > fr) fr = gr; else if (fr > gr) gr = fr; #endif if (!sig->pinvoke && (sig->call_convention == MONO_CALL_VARARG) && (i == sig->sentinelpos)) { /* We allways pass the sig cookie on the stack for simplicity */ /* * Prevent implicit arguments + the sig cookie from being passed * in registers. */ gr = PARAM_REGS; fr = FLOAT_PARAM_REGS; /* Emit the signature cookie just before the implicit arguments */ add_general (&gr, &stack_size, &cinfo->sig_cookie); } if (sig->params [i]->byref) { add_general (&gr, &stack_size, ainfo); continue; } ptype = mini_type_get_underlying_type (gsctx, sig->params [i]); switch (ptype->type) { case MONO_TYPE_BOOLEAN: case MONO_TYPE_I1: case MONO_TYPE_U1: add_general (&gr, &stack_size, ainfo); break; case MONO_TYPE_I2: case MONO_TYPE_U2: case MONO_TYPE_CHAR: add_general (&gr, &stack_size, ainfo); break; case MONO_TYPE_I4: case MONO_TYPE_U4: add_general (&gr, &stack_size, ainfo); break; case MONO_TYPE_I: case MONO_TYPE_U: case MONO_TYPE_PTR: case MONO_TYPE_FNPTR: case MONO_TYPE_CLASS: case MONO_TYPE_OBJECT: case MONO_TYPE_STRING: case MONO_TYPE_SZARRAY: case MONO_TYPE_ARRAY: add_general (&gr, &stack_size, ainfo); break; case MONO_TYPE_GENERICINST: if (!mono_type_generic_inst_is_valuetype (ptype)) { add_general (&gr, &stack_size, ainfo); break; } /* fall through */ case MONO_TYPE_VALUETYPE: add_valuetype (gsctx, sig, ainfo, sig->params [i], FALSE, &gr, &fr, &stack_size); break; case MONO_TYPE_TYPEDBYREF: #ifdef HOST_WIN32 add_valuetype (gsctx, sig, ainfo, sig->params [i], FALSE, &gr, &fr, &stack_size); #else stack_size += sizeof (MonoTypedRef); ainfo->storage = ArgOnStack; #endif break; case MONO_TYPE_U8: case MONO_TYPE_I8: add_general (&gr, &stack_size, ainfo); break; case MONO_TYPE_R4: add_float (&fr, &stack_size, ainfo, FALSE); break; case MONO_TYPE_R8: add_float (&fr, &stack_size, ainfo, TRUE); break; default: g_assert_not_reached (); } } if (!sig->pinvoke && (sig->call_convention == MONO_CALL_VARARG) && (n > 0) && (sig->sentinelpos == sig->param_count)) { gr = PARAM_REGS; fr = FLOAT_PARAM_REGS; /* Emit the signature cookie just before the implicit arguments */ add_general (&gr, &stack_size, &cinfo->sig_cookie); } #ifdef HOST_WIN32 // There always is 32 bytes reserved on the stack when calling on Winx64 stack_size += 0x20; #endif if (stack_size & 0x8) { /* The AMD64 ABI requires each stack frame to be 16 byte aligned */ cinfo->need_stack_align = TRUE; stack_size += 8; } cinfo->stack_usage = stack_size; cinfo->reg_usage = gr; cinfo->freg_usage = fr; return cinfo; } /* * mono_arch_get_argument_info: * @csig: a method signature * @param_count: the number of parameters to consider * @arg_info: an array to store the result infos * * Gathers information on parameters such as size, alignment and * padding. arg_info should be large enought to hold param_count + 1 entries. * * Returns the size of the argument area on the stack. */ int mono_arch_get_argument_info (MonoMethodSignature *csig, int param_count, MonoJitArgumentInfo *arg_info) { int k; CallInfo *cinfo = get_call_info (NULL, NULL, csig, FALSE); guint32 args_size = cinfo->stack_usage; /* The arguments are saved to a stack area in mono_arch_instrument_prolog */ if (csig->hasthis) { arg_info [0].offset = 0; } for (k = 0; k < param_count; k++) { arg_info [k + 1].offset = ((k + csig->hasthis) * 8); /* FIXME: */ arg_info [k + 1].size = 0; } g_free (cinfo); return args_size; } static int cpuid (int id, int* p_eax, int* p_ebx, int* p_ecx, int* p_edx) { #ifndef _MSC_VER __asm__ __volatile__ ("cpuid" : "=a" (*p_eax), "=b" (*p_ebx), "=c" (*p_ecx), "=d" (*p_edx) : "a" (id)); #else int info[4]; __cpuid(info, id); *p_eax = info[0]; *p_ebx = info[1]; *p_ecx = info[2]; *p_edx = info[3]; #endif return 1; } /* * Initialize the cpu to execute managed code. */ void mono_arch_cpu_init (void) { #ifndef _MSC_VER guint16 fpcw; /* spec compliance requires running with double precision */ __asm__ __volatile__ ("fnstcw %0\n": "=m" (fpcw)); fpcw &= ~X86_FPCW_PRECC_MASK; fpcw |= X86_FPCW_PREC_DOUBLE; __asm__ __volatile__ ("fldcw %0\n": : "m" (fpcw)); __asm__ __volatile__ ("fnstcw %0\n": "=m" (fpcw)); #else /* TODO: This is crashing on Win64 right now. * _control87 (_PC_53, MCW_PC); */ #endif } /* * Initialize architecture specific code. */ void mono_arch_init (void) { InitializeCriticalSection (&mini_arch_mutex); ss_trigger_page = mono_valloc (NULL, mono_pagesize (), MONO_MMAP_READ|MONO_MMAP_32BIT); bp_trigger_page = mono_valloc (NULL, mono_pagesize (), MONO_MMAP_READ|MONO_MMAP_32BIT); mono_mprotect (bp_trigger_page, mono_pagesize (), 0); } /* * Cleanup architecture specific code. */ void mono_arch_cleanup (void) { DeleteCriticalSection (&mini_arch_mutex); } /* * This function returns the optimizations supported on this cpu. */ guint32 mono_arch_cpu_optimizazions (guint32 *exclude_mask) { int eax, ebx, ecx, edx; guint32 opts = 0; /* FIXME: AMD64 */ *exclude_mask = 0; /* Feature Flags function, flags returned in EDX. */ if (cpuid (1, &eax, &ebx, &ecx, &edx)) { if (edx & (1 << 15)) { opts |= MONO_OPT_CMOV; if (edx & 1) opts |= MONO_OPT_FCMOV; else *exclude_mask |= MONO_OPT_FCMOV; } else *exclude_mask |= MONO_OPT_CMOV; } return opts; } /* * This function test for all SSE functions supported. * * Returns a bitmask corresponding to all supported versions. * */ guint32 mono_arch_cpu_enumerate_simd_versions (void) { int eax, ebx, ecx, edx; guint32 sse_opts = 0; if (cpuid (1, &eax, &ebx, &ecx, &edx)) { if (edx & (1 << 25)) sse_opts |= 1 << SIMD_VERSION_SSE1; if (edx & (1 << 26)) sse_opts |= 1 << SIMD_VERSION_SSE2; if (ecx & (1 << 0)) sse_opts |= 1 << SIMD_VERSION_SSE3; if (ecx & (1 << 9)) sse_opts |= 1 << SIMD_VERSION_SSSE3; if (ecx & (1 << 19)) sse_opts |= 1 << SIMD_VERSION_SSE41; if (ecx & (1 << 20)) sse_opts |= 1 << SIMD_VERSION_SSE42; } /* Yes, all this needs to be done to check for sse4a. See: "Amd: CPUID Specification" */ if (cpuid (0x80000000, &eax, &ebx, &ecx, &edx)) { /* eax greater or equal than 0x80000001, ebx = 'htuA', ecx = DMAc', edx = 'itne'*/ if ((((unsigned int) eax) >= 0x80000001) && (ebx == 0x68747541) && (ecx == 0x444D4163) && (edx == 0x69746E65)) { cpuid (0x80000001, &eax, &ebx, &ecx, &edx); if (ecx & (1 << 6)) sse_opts |= 1 << SIMD_VERSION_SSE4a; } } return sse_opts; } GList * mono_arch_get_allocatable_int_vars (MonoCompile *cfg) { GList *vars = NULL; int i; for (i = 0; i < cfg->num_varinfo; i++) { MonoInst *ins = cfg->varinfo [i]; MonoMethodVar *vmv = MONO_VARINFO (cfg, i); /* unused vars */ if (vmv->range.first_use.abs_pos >= vmv->range.last_use.abs_pos) continue; if ((ins->flags & (MONO_INST_IS_DEAD|MONO_INST_VOLATILE|MONO_INST_INDIRECT)) || (ins->opcode != OP_LOCAL && ins->opcode != OP_ARG)) continue; if (mono_is_regsize_var (ins->inst_vtype)) { g_assert (MONO_VARINFO (cfg, i)->reg == -1); g_assert (i == vmv->idx); vars = g_list_prepend (vars, vmv); } } vars = mono_varlist_sort (cfg, vars, 0); return vars; } /** * mono_arch_compute_omit_fp: * * Determine whenever the frame pointer can be eliminated. */ static void mono_arch_compute_omit_fp (MonoCompile *cfg) { MonoMethodSignature *sig; MonoMethodHeader *header; int i, locals_size; CallInfo *cinfo; if (cfg->arch.omit_fp_computed) return; header = mono_method_get_header (cfg->method); sig = mono_method_signature (cfg->method); if (!cfg->arch.cinfo) cfg->arch.cinfo = get_call_info (cfg->generic_sharing_context, cfg->mempool, sig, FALSE); cinfo = cfg->arch.cinfo; /* * FIXME: Remove some of the restrictions. */ cfg->arch.omit_fp = TRUE; cfg->arch.omit_fp_computed = TRUE; if (cfg->disable_omit_fp) cfg->arch.omit_fp = FALSE; if (!debug_omit_fp ()) cfg->arch.omit_fp = FALSE; /* if (cfg->method->save_lmf) cfg->arch.omit_fp = FALSE; */ if (cfg->flags & MONO_CFG_HAS_ALLOCA) cfg->arch.omit_fp = FALSE; if (header->num_clauses) cfg->arch.omit_fp = FALSE; if (cfg->param_area) cfg->arch.omit_fp = FALSE; if (!sig->pinvoke && (sig->call_convention == MONO_CALL_VARARG)) cfg->arch.omit_fp = FALSE; if ((mono_jit_trace_calls != NULL && mono_trace_eval (cfg->method)) || (cfg->prof_options & MONO_PROFILE_ENTER_LEAVE)) cfg->arch.omit_fp = FALSE; for (i = 0; i < sig->param_count + sig->hasthis; ++i) { ArgInfo *ainfo = &cinfo->args [i]; if (ainfo->storage == ArgOnStack) { /* * The stack offset can only be determined when the frame * size is known. */ cfg->arch.omit_fp = FALSE; } } locals_size = 0; for (i = cfg->locals_start; i < cfg->num_varinfo; i++) { MonoInst *ins = cfg->varinfo [i]; int ialign; locals_size += mono_type_size (ins->inst_vtype, &ialign); } } GList * mono_arch_get_global_int_regs (MonoCompile *cfg) { GList *regs = NULL; mono_arch_compute_omit_fp (cfg); if (cfg->globalra) { if (cfg->arch.omit_fp) regs = g_list_prepend (regs, (gpointer)AMD64_RBP); regs = g_list_prepend (regs, (gpointer)AMD64_RBX); regs = g_list_prepend (regs, (gpointer)AMD64_R12); regs = g_list_prepend (regs, (gpointer)AMD64_R13); regs = g_list_prepend (regs, (gpointer)AMD64_R14); regs = g_list_prepend (regs, (gpointer)AMD64_R15); regs = g_list_prepend (regs, (gpointer)AMD64_R10); regs = g_list_prepend (regs, (gpointer)AMD64_R9); regs = g_list_prepend (regs, (gpointer)AMD64_R8); regs = g_list_prepend (regs, (gpointer)AMD64_RDI); regs = g_list_prepend (regs, (gpointer)AMD64_RSI); regs = g_list_prepend (regs, (gpointer)AMD64_RDX); regs = g_list_prepend (regs, (gpointer)AMD64_RCX); regs = g_list_prepend (regs, (gpointer)AMD64_RAX); } else { if (cfg->arch.omit_fp) regs = g_list_prepend (regs, (gpointer)AMD64_RBP); /* We use the callee saved registers for global allocation */ regs = g_list_prepend (regs, (gpointer)AMD64_RBX); regs = g_list_prepend (regs, (gpointer)AMD64_R12); regs = g_list_prepend (regs, (gpointer)AMD64_R13); regs = g_list_prepend (regs, (gpointer)AMD64_R14); regs = g_list_prepend (regs, (gpointer)AMD64_R15); #ifdef HOST_WIN32 regs = g_list_prepend (regs, (gpointer)AMD64_RDI); regs = g_list_prepend (regs, (gpointer)AMD64_RSI); #endif } return regs; } GList* mono_arch_get_global_fp_regs (MonoCompile *cfg) { GList *regs = NULL; int i; /* All XMM registers */ for (i = 0; i < 16; ++i) regs = g_list_prepend (regs, GINT_TO_POINTER (i)); return regs; } GList* mono_arch_get_iregs_clobbered_by_call (MonoCallInst *call) { static GList *r = NULL; if (r == NULL) { GList *regs = NULL; regs = g_list_prepend (regs, (gpointer)AMD64_RBP); regs = g_list_prepend (regs, (gpointer)AMD64_RBX); regs = g_list_prepend (regs, (gpointer)AMD64_R12); regs = g_list_prepend (regs, (gpointer)AMD64_R13); regs = g_list_prepend (regs, (gpointer)AMD64_R14); regs = g_list_prepend (regs, (gpointer)AMD64_R15); regs = g_list_prepend (regs, (gpointer)AMD64_R10); regs = g_list_prepend (regs, (gpointer)AMD64_R9); regs = g_list_prepend (regs, (gpointer)AMD64_R8); regs = g_list_prepend (regs, (gpointer)AMD64_RDI); regs = g_list_prepend (regs, (gpointer)AMD64_RSI); regs = g_list_prepend (regs, (gpointer)AMD64_RDX); regs = g_list_prepend (regs, (gpointer)AMD64_RCX); regs = g_list_prepend (regs, (gpointer)AMD64_RAX); InterlockedCompareExchangePointer ((gpointer*)&r, regs, NULL); } return r; } GList* mono_arch_get_fregs_clobbered_by_call (MonoCallInst *call) { int i; static GList *r = NULL; if (r == NULL) { GList *regs = NULL; for (i = 0; i < AMD64_XMM_NREG; ++i) regs = g_list_prepend (regs, GINT_TO_POINTER (MONO_MAX_IREGS + i)); InterlockedCompareExchangePointer ((gpointer*)&r, regs, NULL); } return r; } /* * mono_arch_regalloc_cost: * * Return the cost, in number of memory references, of the action of * allocating the variable VMV into a register during global register * allocation. */ guint32 mono_arch_regalloc_cost (MonoCompile *cfg, MonoMethodVar *vmv) { MonoInst *ins = cfg->varinfo [vmv->idx]; if (cfg->method->save_lmf) /* The register is already saved */ /* substract 1 for the invisible store in the prolog */ return (ins->opcode == OP_ARG) ? 0 : 1; else /* push+pop */ return (ins->opcode == OP_ARG) ? 1 : 2; } /* * mono_arch_fill_argument_info: * * Populate cfg->args, cfg->ret and cfg->vret_addr with information about the arguments * of the method. */ void mono_arch_fill_argument_info (MonoCompile *cfg) { MonoMethodSignature *sig; MonoMethodHeader *header; MonoInst *ins; int i; CallInfo *cinfo; header = mono_method_get_header (cfg->method); sig = mono_method_signature (cfg->method); cinfo = cfg->arch.cinfo; /* * Contrary to mono_arch_allocate_vars (), the information should describe * where the arguments are at the beginning of the method, not where they can be * accessed during the execution of the method. The later makes no sense for the * global register allocator, since a variable can be in more than one location. */ if (sig->ret->type != MONO_TYPE_VOID) { switch (cinfo->ret.storage) { case ArgInIReg: case ArgInFloatSSEReg: case ArgInDoubleSSEReg: if ((MONO_TYPE_ISSTRUCT (sig->ret) && !mono_class_from_mono_type (sig->ret)->enumtype) || (sig->ret->type == MONO_TYPE_TYPEDBYREF)) { cfg->vret_addr->opcode = OP_REGVAR; cfg->vret_addr->inst_c0 = cinfo->ret.reg; } else { cfg->ret->opcode = OP_REGVAR; cfg->ret->inst_c0 = cinfo->ret.reg; } break; case ArgValuetypeInReg: cfg->ret->opcode = OP_REGOFFSET; cfg->ret->inst_basereg = -1; cfg->ret->inst_offset = -1; break; default: g_assert_not_reached (); } } for (i = 0; i < sig->param_count + sig->hasthis; ++i) { ArgInfo *ainfo = &cinfo->args [i]; MonoType *arg_type; ins = cfg->args [i]; if (sig->hasthis && (i == 0)) arg_type = &mono_defaults.object_class->byval_arg; else arg_type = sig->params [i - sig->hasthis]; switch (ainfo->storage) { case ArgInIReg: case ArgInFloatSSEReg: case ArgInDoubleSSEReg: ins->opcode = OP_REGVAR; ins->inst_c0 = ainfo->reg; break; case ArgOnStack: ins->opcode = OP_REGOFFSET; ins->inst_basereg = -1; ins->inst_offset = -1; break; case ArgValuetypeInReg: /* Dummy */ ins->opcode = OP_NOP; break; default: g_assert_not_reached (); } } } void mono_arch_allocate_vars (MonoCompile *cfg) { MonoMethodSignature *sig; MonoMethodHeader *header; MonoInst *ins; int i, offset; guint32 locals_stack_size, locals_stack_align; gint32 *offsets; CallInfo *cinfo; header = mono_method_get_header (cfg->method); sig = mono_method_signature (cfg->method); cinfo = cfg->arch.cinfo; mono_arch_compute_omit_fp (cfg); /* * We use the ABI calling conventions for managed code as well. * Exception: valuetypes are only sometimes passed or returned in registers. */ /* * The stack looks like this: * * * * * * -> grows dynamically * */ if (cfg->arch.omit_fp) { cfg->flags |= MONO_CFG_HAS_SPILLUP; cfg->frame_reg = AMD64_RSP; offset = 0; } else { /* Locals are allocated backwards from %fp */ cfg->frame_reg = AMD64_RBP; offset = 0; } if (cfg->method->save_lmf) { /* Reserve stack space for saving LMF */ if (cfg->arch.omit_fp) { cfg->arch.lmf_offset = offset; offset += sizeof (MonoLMF); } else { offset += sizeof (MonoLMF); cfg->arch.lmf_offset = -offset; } } else { if (cfg->arch.omit_fp) cfg->arch.reg_save_area_offset = offset; /* Reserve space for caller saved registers */ for (i = 0; i < AMD64_NREG; ++i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { offset += sizeof (gpointer); } } if (sig->ret->type != MONO_TYPE_VOID) { switch (cinfo->ret.storage) { case ArgInIReg: case ArgInFloatSSEReg: case ArgInDoubleSSEReg: if ((MONO_TYPE_ISSTRUCT (sig->ret) && !mono_class_from_mono_type (sig->ret)->enumtype) || (sig->ret->type == MONO_TYPE_TYPEDBYREF)) { if (cfg->globalra) { cfg->vret_addr->opcode = OP_REGVAR; cfg->vret_addr->inst_c0 = cinfo->ret.reg; } else { /* The register is volatile */ cfg->vret_addr->opcode = OP_REGOFFSET; cfg->vret_addr->inst_basereg = cfg->frame_reg; if (cfg->arch.omit_fp) { cfg->vret_addr->inst_offset = offset; offset += 8; } else { offset += 8; cfg->vret_addr->inst_offset = -offset; } if (G_UNLIKELY (cfg->verbose_level > 1)) { printf ("vret_addr ="); mono_print_ins (cfg->vret_addr); } } } else { cfg->ret->opcode = OP_REGVAR; cfg->ret->inst_c0 = cinfo->ret.reg; } break; case ArgValuetypeInReg: /* Allocate a local to hold the result, the epilog will copy it to the correct place */ cfg->ret->opcode = OP_REGOFFSET; cfg->ret->inst_basereg = cfg->frame_reg; if (cfg->arch.omit_fp) { cfg->ret->inst_offset = offset; offset += 16; } else { offset += 16; cfg->ret->inst_offset = - offset; } break; default: g_assert_not_reached (); } if (!cfg->globalra) cfg->ret->dreg = cfg->ret->inst_c0; } /* Allocate locals */ if (!cfg->globalra) { offsets = mono_allocate_stack_slots_full (cfg, cfg->arch.omit_fp ? FALSE: TRUE, &locals_stack_size, &locals_stack_align); if (locals_stack_size > MONO_ARCH_MAX_FRAME_SIZE) { char *mname = mono_method_full_name (cfg->method, TRUE); cfg->exception_type = MONO_EXCEPTION_INVALID_PROGRAM; cfg->exception_message = g_strdup_printf ("Method %s stack is too big.", mname); g_free (mname); return; } if (locals_stack_align) { offset += (locals_stack_align - 1); offset &= ~(locals_stack_align - 1); } if (cfg->arch.omit_fp) { cfg->locals_min_stack_offset = offset; cfg->locals_max_stack_offset = offset + locals_stack_size; } else { cfg->locals_min_stack_offset = - (offset + locals_stack_size); cfg->locals_max_stack_offset = - offset; } for (i = cfg->locals_start; i < cfg->num_varinfo; i++) { if (offsets [i] != -1) { MonoInst *ins = cfg->varinfo [i]; ins->opcode = OP_REGOFFSET; ins->inst_basereg = cfg->frame_reg; if (cfg->arch.omit_fp) ins->inst_offset = (offset + offsets [i]); else ins->inst_offset = - (offset + offsets [i]); //printf ("allocated local %d to ", i); mono_print_tree_nl (ins); } } offset += locals_stack_size; } if (!sig->pinvoke && (sig->call_convention == MONO_CALL_VARARG)) { g_assert (!cfg->arch.omit_fp); g_assert (cinfo->sig_cookie.storage == ArgOnStack); cfg->sig_cookie = cinfo->sig_cookie.offset + ARGS_OFFSET; } for (i = 0; i < sig->param_count + sig->hasthis; ++i) { ins = cfg->args [i]; if (ins->opcode != OP_REGVAR) { ArgInfo *ainfo = &cinfo->args [i]; gboolean inreg = TRUE; MonoType *arg_type; if (sig->hasthis && (i == 0)) arg_type = &mono_defaults.object_class->byval_arg; else arg_type = sig->params [i - sig->hasthis]; if (cfg->globalra) { /* The new allocator needs info about the original locations of the arguments */ switch (ainfo->storage) { case ArgInIReg: case ArgInFloatSSEReg: case ArgInDoubleSSEReg: ins->opcode = OP_REGVAR; ins->inst_c0 = ainfo->reg; break; case ArgOnStack: g_assert (!cfg->arch.omit_fp); ins->opcode = OP_REGOFFSET; ins->inst_basereg = cfg->frame_reg; ins->inst_offset = ainfo->offset + ARGS_OFFSET; break; case ArgValuetypeInReg: ins->opcode = OP_REGOFFSET; ins->inst_basereg = cfg->frame_reg; /* These arguments are saved to the stack in the prolog */ offset = ALIGN_TO (offset, sizeof (gpointer)); if (cfg->arch.omit_fp) { ins->inst_offset = offset; offset += (ainfo->storage == ArgValuetypeInReg) ? 2 * sizeof (gpointer) : sizeof (gpointer); } else { offset += (ainfo->storage == ArgValuetypeInReg) ? 2 * sizeof (gpointer) : sizeof (gpointer); ins->inst_offset = - offset; } break; default: g_assert_not_reached (); } continue; } /* FIXME: Allocate volatile arguments to registers */ if (ins->flags & (MONO_INST_VOLATILE|MONO_INST_INDIRECT)) inreg = FALSE; /* * Under AMD64, all registers used to pass arguments to functions * are volatile across calls. * FIXME: Optimize this. */ if ((ainfo->storage == ArgInIReg) || (ainfo->storage == ArgInFloatSSEReg) || (ainfo->storage == ArgInDoubleSSEReg) || (ainfo->storage == ArgValuetypeInReg)) inreg = FALSE; ins->opcode = OP_REGOFFSET; switch (ainfo->storage) { case ArgInIReg: case ArgInFloatSSEReg: case ArgInDoubleSSEReg: if (inreg) { ins->opcode = OP_REGVAR; ins->dreg = ainfo->reg; } break; case ArgOnStack: g_assert (!cfg->arch.omit_fp); ins->opcode = OP_REGOFFSET; ins->inst_basereg = cfg->frame_reg; ins->inst_offset = ainfo->offset + ARGS_OFFSET; break; case ArgValuetypeInReg: break; case ArgValuetypeAddrInIReg: { MonoInst *indir; g_assert (!cfg->arch.omit_fp); MONO_INST_NEW (cfg, indir, 0); indir->opcode = OP_REGOFFSET; if (ainfo->pair_storage [0] == ArgInIReg) { indir->inst_basereg = cfg->frame_reg; offset = ALIGN_TO (offset, sizeof (gpointer)); offset += (sizeof (gpointer)); indir->inst_offset = - offset; } else { indir->inst_basereg = cfg->frame_reg; indir->inst_offset = ainfo->offset + ARGS_OFFSET; } ins->opcode = OP_VTARG_ADDR; ins->inst_left = indir; break; } default: NOT_IMPLEMENTED; } if (!inreg && (ainfo->storage != ArgOnStack) && (ainfo->storage != ArgValuetypeAddrInIReg)) { ins->opcode = OP_REGOFFSET; ins->inst_basereg = cfg->frame_reg; /* These arguments are saved to the stack in the prolog */ offset = ALIGN_TO (offset, sizeof (gpointer)); if (cfg->arch.omit_fp) { ins->inst_offset = offset; offset += (ainfo->storage == ArgValuetypeInReg) ? 2 * sizeof (gpointer) : sizeof (gpointer); // Arguments are yet supported by the stack map creation code //cfg->locals_max_stack_offset = MAX (cfg->locals_max_stack_offset, offset); } else { offset += (ainfo->storage == ArgValuetypeInReg) ? 2 * sizeof (gpointer) : sizeof (gpointer); ins->inst_offset = - offset; //cfg->locals_min_stack_offset = MIN (cfg->locals_min_stack_offset, offset); } } } } cfg->stack_offset = offset; } void mono_arch_create_vars (MonoCompile *cfg) { MonoMethodSignature *sig; CallInfo *cinfo; sig = mono_method_signature (cfg->method); if (!cfg->arch.cinfo) cfg->arch.cinfo = get_call_info (cfg->generic_sharing_context, cfg->mempool, sig, FALSE); cinfo = cfg->arch.cinfo; if (cinfo->ret.storage == ArgValuetypeInReg) cfg->ret_var_is_local = TRUE; if ((cinfo->ret.storage != ArgValuetypeInReg) && MONO_TYPE_ISSTRUCT (sig->ret)) { cfg->vret_addr = mono_compile_create_var (cfg, &mono_defaults.int_class->byval_arg, OP_ARG); if (G_UNLIKELY (cfg->verbose_level > 1)) { printf ("vret_addr = "); mono_print_ins (cfg->vret_addr); } } #ifdef MONO_AMD64_NO_PUSHES /* * When this is set, we pass arguments on the stack by moves, and by allocating * a bigger stack frame, instead of pushes. * Pushes complicate exception handling because the arguments on the stack have * to be popped each time a frame is unwound. They also make fp elimination * impossible. * FIXME: This doesn't work inside filter/finally clauses, since those execute * on a new frame which doesn't include a param area. */ cfg->arch.no_pushes = TRUE; #endif } static void add_outarg_reg (MonoCompile *cfg, MonoCallInst *call, ArgStorage storage, int reg, MonoInst *tree) { MonoInst *ins; switch (storage) { case ArgInIReg: MONO_INST_NEW (cfg, ins, OP_MOVE); ins->dreg = mono_alloc_ireg (cfg); ins->sreg1 = tree->dreg; MONO_ADD_INS (cfg->cbb, ins); mono_call_inst_add_outarg_reg (cfg, call, ins->dreg, reg, FALSE); break; case ArgInFloatSSEReg: MONO_INST_NEW (cfg, ins, OP_AMD64_SET_XMMREG_R4); ins->dreg = mono_alloc_freg (cfg); ins->sreg1 = tree->dreg; MONO_ADD_INS (cfg->cbb, ins); mono_call_inst_add_outarg_reg (cfg, call, ins->dreg, reg, TRUE); break; case ArgInDoubleSSEReg: MONO_INST_NEW (cfg, ins, OP_FMOVE); ins->dreg = mono_alloc_freg (cfg); ins->sreg1 = tree->dreg; MONO_ADD_INS (cfg->cbb, ins); mono_call_inst_add_outarg_reg (cfg, call, ins->dreg, reg, TRUE); break; default: g_assert_not_reached (); } } static int arg_storage_to_load_membase (ArgStorage storage) { switch (storage) { case ArgInIReg: return OP_LOAD_MEMBASE; case ArgInDoubleSSEReg: return OP_LOADR8_MEMBASE; case ArgInFloatSSEReg: return OP_LOADR4_MEMBASE; default: g_assert_not_reached (); } return -1; } static void emit_sig_cookie (MonoCompile *cfg, MonoCallInst *call, CallInfo *cinfo) { MonoInst *arg; MonoMethodSignature *tmp_sig; MonoInst *sig_arg; if (call->tail_call) NOT_IMPLEMENTED; /* FIXME: Add support for signature tokens to AOT */ cfg->disable_aot = TRUE; g_assert (cinfo->sig_cookie.storage == ArgOnStack); /* * mono_ArgIterator_Setup assumes the signature cookie is * passed first and all the arguments which were before it are * passed on the stack after the signature. So compensate by * passing a different signature. */ tmp_sig = mono_metadata_signature_dup (call->signature); tmp_sig->param_count -= call->signature->sentinelpos; tmp_sig->sentinelpos = 0; memcpy (tmp_sig->params, call->signature->params + call->signature->sentinelpos, tmp_sig->param_count * sizeof (MonoType*)); MONO_INST_NEW (cfg, sig_arg, OP_ICONST); sig_arg->dreg = mono_alloc_ireg (cfg); sig_arg->inst_p0 = tmp_sig; MONO_ADD_INS (cfg->cbb, sig_arg); if (cfg->arch.no_pushes) { MONO_EMIT_NEW_STORE_MEMBASE (cfg, OP_STORE_MEMBASE_REG, AMD64_RSP, cinfo->sig_cookie.offset, sig_arg->dreg); } else { MONO_INST_NEW (cfg, arg, OP_X86_PUSH); arg->sreg1 = sig_arg->dreg; MONO_ADD_INS (cfg->cbb, arg); } } static inline LLVMArgStorage arg_storage_to_llvm_arg_storage (MonoCompile *cfg, ArgStorage storage) { switch (storage) { case ArgInIReg: return LLVMArgInIReg; case ArgNone: return LLVMArgNone; default: g_assert_not_reached (); return LLVMArgNone; } } #ifdef ENABLE_LLVM LLVMCallInfo* mono_arch_get_llvm_call_info (MonoCompile *cfg, MonoMethodSignature *sig) { int i, n; CallInfo *cinfo; ArgInfo *ainfo; int j; LLVMCallInfo *linfo; n = sig->param_count + sig->hasthis; cinfo = get_call_info (cfg->generic_sharing_context, cfg->mempool, sig, sig->pinvoke); linfo = mono_mempool_alloc0 (cfg->mempool, sizeof (LLVMCallInfo) + (sizeof (LLVMArgInfo) * n)); /* * LLVM always uses the native ABI while we use our own ABI, the * only difference is the handling of vtypes: * - we only pass/receive them in registers in some cases, and only * in 1 or 2 integer registers. */ if (cinfo->ret.storage == ArgValuetypeInReg) { if (sig->pinvoke) { cfg->exception_message = g_strdup ("pinvoke + vtypes"); cfg->disable_llvm = TRUE; return linfo; } linfo->ret.storage = LLVMArgVtypeInReg; for (j = 0; j < 2; ++j) linfo->ret.pair_storage [j] = arg_storage_to_llvm_arg_storage (cfg, cinfo->ret.pair_storage [j]); } if (MONO_TYPE_ISSTRUCT (sig->ret) && cinfo->ret.storage == ArgInIReg) { /* Vtype returned using a hidden argument */ linfo->ret.storage = LLVMArgVtypeRetAddr; } for (i = 0; i < n; ++i) { ainfo = cinfo->args + i; linfo->args [i].storage = LLVMArgNone; switch (ainfo->storage) { case ArgInIReg: linfo->args [i].storage = LLVMArgInIReg; break; case ArgInDoubleSSEReg: case ArgInFloatSSEReg: linfo->args [i].storage = LLVMArgInFPReg; break; case ArgOnStack: if ((i >= sig->hasthis) && (MONO_TYPE_ISSTRUCT(sig->params [i - sig->hasthis]))) { linfo->args [i].storage = LLVMArgVtypeByVal; } else { linfo->args [i].storage = LLVMArgInIReg; if (!sig->params [i - sig->hasthis]->byref) { if (sig->params [i - sig->hasthis]->type == MONO_TYPE_R4) { linfo->args [i].storage = LLVMArgInFPReg; } else if (sig->params [i - sig->hasthis]->type == MONO_TYPE_R8) { linfo->args [i].storage = LLVMArgInFPReg; } } } break; case ArgValuetypeInReg: if (sig->pinvoke) { cfg->exception_message = g_strdup ("pinvoke + vtypes"); cfg->disable_llvm = TRUE; return linfo; } linfo->args [i].storage = LLVMArgVtypeInReg; for (j = 0; j < 2; ++j) linfo->args [i].pair_storage [j] = arg_storage_to_llvm_arg_storage (cfg, ainfo->pair_storage [j]); break; default: cfg->exception_message = g_strdup ("ainfo->storage"); cfg->disable_llvm = TRUE; break; } } return linfo; } #endif void mono_arch_emit_call (MonoCompile *cfg, MonoCallInst *call) { MonoInst *arg, *in; MonoMethodSignature *sig; int i, n, stack_size; CallInfo *cinfo; ArgInfo *ainfo; stack_size = 0; sig = call->signature; n = sig->param_count + sig->hasthis; cinfo = get_call_info (cfg->generic_sharing_context, cfg->mempool, sig, sig->pinvoke); if (COMPILE_LLVM (cfg)) { /* We shouldn't be called in the llvm case */ cfg->disable_llvm = TRUE; return; } if (cinfo->need_stack_align) { if (!cfg->arch.no_pushes) MONO_EMIT_NEW_BIALU_IMM (cfg, OP_SUB_IMM, X86_ESP, X86_ESP, 8); } /* * Emit all arguments which are passed on the stack to prevent register * allocation problems. */ if (cfg->arch.no_pushes) { for (i = 0; i < n; ++i) { MonoType *t; ainfo = cinfo->args + i; in = call->args [i]; if (sig->hasthis && i == 0) t = &mono_defaults.object_class->byval_arg; else t = sig->params [i - sig->hasthis]; if (ainfo->storage == ArgOnStack && !MONO_TYPE_ISSTRUCT (t) && !call->tail_call) { if (!t->byref) { if (t->type == MONO_TYPE_R4) MONO_EMIT_NEW_STORE_MEMBASE (cfg, OP_STORER4_MEMBASE_REG, AMD64_RSP, ainfo->offset, in->dreg); else if (t->type == MONO_TYPE_R8) MONO_EMIT_NEW_STORE_MEMBASE (cfg, OP_STORER8_MEMBASE_REG, AMD64_RSP, ainfo->offset, in->dreg); else MONO_EMIT_NEW_STORE_MEMBASE (cfg, OP_STORE_MEMBASE_REG, AMD64_RSP, ainfo->offset, in->dreg); } else { MONO_EMIT_NEW_STORE_MEMBASE (cfg, OP_STORE_MEMBASE_REG, AMD64_RSP, ainfo->offset, in->dreg); } } } } /* * Emit all parameters passed in registers in non-reverse order for better readability * and to help the optimization in emit_prolog (). */ for (i = 0; i < n; ++i) { ainfo = cinfo->args + i; in = call->args [i]; if (ainfo->storage == ArgInIReg) add_outarg_reg (cfg, call, ainfo->storage, ainfo->reg, in); } for (i = n - 1; i >= 0; --i) { ainfo = cinfo->args + i; in = call->args [i]; switch (ainfo->storage) { case ArgInIReg: /* Already done */ break; case ArgInFloatSSEReg: case ArgInDoubleSSEReg: add_outarg_reg (cfg, call, ainfo->storage, ainfo->reg, in); break; case ArgOnStack: case ArgValuetypeInReg: case ArgValuetypeAddrInIReg: if (ainfo->storage == ArgOnStack && call->tail_call) { MonoInst *call_inst = (MonoInst*)call; cfg->args [i]->flags |= MONO_INST_VOLATILE; EMIT_NEW_ARGSTORE (cfg, call_inst, i, in); } else if ((i >= sig->hasthis) && (MONO_TYPE_ISSTRUCT(sig->params [i - sig->hasthis]))) { guint32 align; guint32 size; if (sig->params [i - sig->hasthis]->type == MONO_TYPE_TYPEDBYREF) { size = sizeof (MonoTypedRef); align = sizeof (gpointer); } else { if (sig->pinvoke) size = mono_type_native_stack_size (&in->klass->byval_arg, &align); else { /* * Other backends use mono_type_stack_size (), but that * aligns the size to 8, which is larger than the size of * the source, leading to reads of invalid memory if the * source is at the end of address space. */ size = mono_class_value_size (in->klass, &align); } } g_assert (in->klass); if (size > 0) { MONO_INST_NEW (cfg, arg, OP_OUTARG_VT); arg->sreg1 = in->dreg; arg->klass = in->klass; arg->backend.size = size; arg->inst_p0 = call; arg->inst_p1 = mono_mempool_alloc (cfg->mempool, sizeof (ArgInfo)); memcpy (arg->inst_p1, ainfo, sizeof (ArgInfo)); MONO_ADD_INS (cfg->cbb, arg); } } else { if (cfg->arch.no_pushes) { /* Already done */ } else { MONO_INST_NEW (cfg, arg, OP_X86_PUSH); arg->sreg1 = in->dreg; if (!sig->params [i - sig->hasthis]->byref) { if (sig->params [i - sig->hasthis]->type == MONO_TYPE_R4) { MONO_EMIT_NEW_BIALU_IMM (cfg, OP_SUB_IMM, X86_ESP, X86_ESP, 8); arg->opcode = OP_STORER4_MEMBASE_REG; arg->inst_destbasereg = X86_ESP; arg->inst_offset = 0; } else if (sig->params [i - sig->hasthis]->type == MONO_TYPE_R8) { MONO_EMIT_NEW_BIALU_IMM (cfg, OP_SUB_IMM, X86_ESP, X86_ESP, 8); arg->opcode = OP_STORER8_MEMBASE_REG; arg->inst_destbasereg = X86_ESP; arg->inst_offset = 0; } } MONO_ADD_INS (cfg->cbb, arg); } } break; default: g_assert_not_reached (); } if (!sig->pinvoke && (sig->call_convention == MONO_CALL_VARARG) && (i == sig->sentinelpos)) /* Emit the signature cookie just before the implicit arguments */ emit_sig_cookie (cfg, call, cinfo); } /* Handle the case where there are no implicit arguments */ if (!sig->pinvoke && (sig->call_convention == MONO_CALL_VARARG) && (n == sig->sentinelpos)) emit_sig_cookie (cfg, call, cinfo); if (sig->ret && MONO_TYPE_ISSTRUCT (sig->ret)) { MonoInst *vtarg; if (cinfo->ret.storage == ArgValuetypeInReg) { if (cinfo->ret.pair_storage [0] == ArgInIReg && cinfo->ret.pair_storage [1] == ArgNone) { /* * Tell the JIT to use a more efficient calling convention: call using * OP_CALL, compute the result location after the call, and save the * result there. */ call->vret_in_reg = TRUE; /* * Nullify the instruction computing the vret addr to enable * future optimizations. */ if (call->vret_var) NULLIFY_INS (call->vret_var); } else { if (call->tail_call) NOT_IMPLEMENTED; /* * The valuetype is in RAX:RDX after the call, need to be copied to * the stack. Push the address here, so the call instruction can * access it. */ if (!cfg->arch.vret_addr_loc) { cfg->arch.vret_addr_loc = mono_compile_create_var (cfg, &mono_defaults.int_class->byval_arg, OP_LOCAL); /* Prevent it from being register allocated or optimized away */ ((MonoInst*)cfg->arch.vret_addr_loc)->flags |= MONO_INST_VOLATILE; } MONO_EMIT_NEW_UNALU (cfg, OP_MOVE, ((MonoInst*)cfg->arch.vret_addr_loc)->dreg, call->vret_var->dreg); } } else { MONO_INST_NEW (cfg, vtarg, OP_MOVE); vtarg->sreg1 = call->vret_var->dreg; vtarg->dreg = mono_alloc_preg (cfg); MONO_ADD_INS (cfg->cbb, vtarg); mono_call_inst_add_outarg_reg (cfg, call, vtarg->dreg, cinfo->ret.reg, FALSE); } } #ifdef HOST_WIN32 if (call->inst.opcode != OP_JMP && OP_TAILCALL != call->inst.opcode) { MONO_EMIT_NEW_BIALU_IMM (cfg, OP_SUB_IMM, X86_ESP, X86_ESP, 0x20); } #endif if (cfg->method->save_lmf) { MONO_INST_NEW (cfg, arg, OP_AMD64_SAVE_SP_TO_LMF); MONO_ADD_INS (cfg->cbb, arg); } call->stack_usage = cinfo->stack_usage; } void mono_arch_emit_outarg_vt (MonoCompile *cfg, MonoInst *ins, MonoInst *src) { MonoInst *arg; MonoCallInst *call = (MonoCallInst*)ins->inst_p0; ArgInfo *ainfo = (ArgInfo*)ins->inst_p1; int size = ins->backend.size; if (ainfo->storage == ArgValuetypeInReg) { MonoInst *load; int part; for (part = 0; part < 2; ++part) { if (ainfo->pair_storage [part] == ArgNone) continue; MONO_INST_NEW (cfg, load, arg_storage_to_load_membase (ainfo->pair_storage [part])); load->inst_basereg = src->dreg; load->inst_offset = part * sizeof (gpointer); switch (ainfo->pair_storage [part]) { case ArgInIReg: load->dreg = mono_alloc_ireg (cfg); break; case ArgInDoubleSSEReg: case ArgInFloatSSEReg: load->dreg = mono_alloc_freg (cfg); break; default: g_assert_not_reached (); } MONO_ADD_INS (cfg->cbb, load); add_outarg_reg (cfg, call, ainfo->pair_storage [part], ainfo->pair_regs [part], load); } } else if (ainfo->storage == ArgValuetypeAddrInIReg) { MonoInst *vtaddr, *load; vtaddr = mono_compile_create_var (cfg, &ins->klass->byval_arg, OP_LOCAL); g_assert (!cfg->arch.no_pushes); MONO_INST_NEW (cfg, load, OP_LDADDR); load->inst_p0 = vtaddr; vtaddr->flags |= MONO_INST_INDIRECT; load->type = STACK_MP; load->klass = vtaddr->klass; load->dreg = mono_alloc_ireg (cfg); MONO_ADD_INS (cfg->cbb, load); mini_emit_memcpy (cfg, load->dreg, 0, src->dreg, 0, size, 4); if (ainfo->pair_storage [0] == ArgInIReg) { MONO_INST_NEW (cfg, arg, OP_X86_LEA_MEMBASE); arg->dreg = mono_alloc_ireg (cfg); arg->sreg1 = load->dreg; arg->inst_imm = 0; MONO_ADD_INS (cfg->cbb, arg); mono_call_inst_add_outarg_reg (cfg, call, arg->dreg, ainfo->pair_regs [0], FALSE); } else { MONO_INST_NEW (cfg, arg, OP_X86_PUSH); arg->sreg1 = load->dreg; MONO_ADD_INS (cfg->cbb, arg); } } else { if (size == 8) { if (cfg->arch.no_pushes) { int dreg = mono_alloc_ireg (cfg); MONO_EMIT_NEW_LOAD_MEMBASE (cfg, dreg, src->dreg, 0); MONO_EMIT_NEW_STORE_MEMBASE (cfg, OP_STORE_MEMBASE_REG, AMD64_RSP, ainfo->offset, dreg); } else { /* Can't use this for < 8 since it does an 8 byte memory load */ MONO_INST_NEW (cfg, arg, OP_X86_PUSH_MEMBASE); arg->inst_basereg = src->dreg; arg->inst_offset = 0; MONO_ADD_INS (cfg->cbb, arg); } } else if (size <= 40) { if (cfg->arch.no_pushes) { mini_emit_memcpy (cfg, AMD64_RSP, ainfo->offset, src->dreg, 0, size, 4); } else { MONO_EMIT_NEW_BIALU_IMM (cfg, OP_SUB_IMM, X86_ESP, X86_ESP, ALIGN_TO (size, 8)); mini_emit_memcpy (cfg, X86_ESP, 0, src->dreg, 0, size, 4); } } else { if (cfg->arch.no_pushes) { // FIXME: Code growth mini_emit_memcpy (cfg, AMD64_RSP, ainfo->offset, src->dreg, 0, size, 4); } else { MONO_INST_NEW (cfg, arg, OP_X86_PUSH_OBJ); arg->inst_basereg = src->dreg; arg->inst_offset = 0; arg->inst_imm = size; MONO_ADD_INS (cfg->cbb, arg); } } } } void mono_arch_emit_setret (MonoCompile *cfg, MonoMethod *method, MonoInst *val) { MonoType *ret = mini_type_get_underlying_type (NULL, mono_method_signature (method)->ret); if (!ret->byref) { if (ret->type == MONO_TYPE_R4) { if (COMPILE_LLVM (cfg)) MONO_EMIT_NEW_UNALU (cfg, OP_FMOVE, cfg->ret->dreg, val->dreg); else MONO_EMIT_NEW_UNALU (cfg, OP_AMD64_SET_XMMREG_R4, cfg->ret->dreg, val->dreg); return; } else if (ret->type == MONO_TYPE_R8) { MONO_EMIT_NEW_UNALU (cfg, OP_FMOVE, cfg->ret->dreg, val->dreg); return; } } MONO_EMIT_NEW_UNALU (cfg, OP_MOVE, cfg->ret->dreg, val->dreg); } #define EMIT_COND_BRANCH(ins,cond,sign) \ if (ins->inst_true_bb->native_offset) { \ x86_branch (code, cond, cfg->native_code + ins->inst_true_bb->native_offset, sign); \ } else { \ mono_add_patch_info (cfg, code - cfg->native_code, MONO_PATCH_INFO_BB, ins->inst_true_bb); \ if ((cfg->opt & MONO_OPT_BRANCH) && \ x86_is_imm8 (ins->inst_true_bb->max_offset - offset)) \ x86_branch8 (code, cond, 0, sign); \ else \ x86_branch32 (code, cond, 0, sign); \ } typedef struct { MonoMethodSignature *sig; CallInfo *cinfo; } ArchDynCallInfo; typedef struct { mgreg_t regs [PARAM_REGS]; mgreg_t res; guint8 *ret; } DynCallArgs; static gboolean dyn_call_supported (MonoMethodSignature *sig, CallInfo *cinfo) { int i; #ifdef HOST_WIN32 return FALSE; #endif switch (cinfo->ret.storage) { case ArgNone: case ArgInIReg: break; case ArgValuetypeInReg: { ArgInfo *ainfo = &cinfo->ret; if (ainfo->pair_storage [0] != ArgNone && ainfo->pair_storage [0] != ArgInIReg) return FALSE; if (ainfo->pair_storage [1] != ArgNone && ainfo->pair_storage [1] != ArgInIReg) return FALSE; break; } default: return FALSE; } for (i = 0; i < cinfo->nargs; ++i) { ArgInfo *ainfo = &cinfo->args [i]; switch (ainfo->storage) { case ArgInIReg: break; case ArgValuetypeInReg: if (ainfo->pair_storage [0] != ArgNone && ainfo->pair_storage [0] != ArgInIReg) return FALSE; if (ainfo->pair_storage [1] != ArgNone && ainfo->pair_storage [1] != ArgInIReg) return FALSE; break; default: return FALSE; } } return TRUE; } /* * mono_arch_dyn_call_prepare: * * Return a pointer to an arch-specific structure which contains information * needed by mono_arch_get_dyn_call_args (). Return NULL if OP_DYN_CALL is not * supported for SIG. * This function is equivalent to ffi_prep_cif in libffi. */ MonoDynCallInfo* mono_arch_dyn_call_prepare (MonoMethodSignature *sig) { ArchDynCallInfo *info; CallInfo *cinfo; cinfo = get_call_info (NULL, NULL, sig, FALSE); if (!dyn_call_supported (sig, cinfo)) { g_free (cinfo); return NULL; } info = g_new0 (ArchDynCallInfo, 1); // FIXME: Preprocess the info to speed up get_dyn_call_args (). info->sig = sig; info->cinfo = cinfo; return (MonoDynCallInfo*)info; } /* * mono_arch_dyn_call_free: * * Free a MonoDynCallInfo structure. */ void mono_arch_dyn_call_free (MonoDynCallInfo *info) { ArchDynCallInfo *ainfo = (ArchDynCallInfo*)info; g_free (ainfo->cinfo); g_free (ainfo); } /* * mono_arch_get_start_dyn_call: * * Convert the arguments ARGS to a format which can be passed to OP_DYN_CALL, and * store the result into BUF. * ARGS should be an array of pointers pointing to the arguments. * RET should point to a memory buffer large enought to hold the result of the * call. * This function should be as fast as possible, any work which does not depend * on the actual values of the arguments should be done in * mono_arch_dyn_call_prepare (). * start_dyn_call + OP_DYN_CALL + finish_dyn_call is equivalent to ffi_call in * libffi. */ void mono_arch_start_dyn_call (MonoDynCallInfo *info, gpointer **args, guint8 *ret, guint8 *buf, int buf_len) { ArchDynCallInfo *dinfo = (ArchDynCallInfo*)info; DynCallArgs *p = (DynCallArgs*)buf; int arg_index, greg, i; MonoMethodSignature *sig = dinfo->sig; g_assert (buf_len >= sizeof (DynCallArgs)); p->res = 0; p->ret = ret; arg_index = 0; greg = 0; if (dinfo->cinfo->vtype_retaddr) p->regs [greg ++] = (mgreg_t)ret; if (sig->hasthis) { p->regs [greg ++] = (mgreg_t)*(args [arg_index ++]); } for (i = 0; i < sig->param_count; i++) { MonoType *t = mono_type_get_underlying_type (sig->params [i]); gpointer *arg = args [arg_index ++]; if (t->byref) { p->regs [greg ++] = (mgreg_t)*(arg); continue; } switch (t->type) { case MONO_TYPE_STRING: case MONO_TYPE_CLASS: case MONO_TYPE_ARRAY: case MONO_TYPE_SZARRAY: case MONO_TYPE_OBJECT: case MONO_TYPE_PTR: case MONO_TYPE_I: case MONO_TYPE_U: case MONO_TYPE_I8: case MONO_TYPE_U8: g_assert (dinfo->cinfo->args [i + sig->hasthis].reg == param_regs [greg]); p->regs [greg ++] = (mgreg_t)*(arg); break; case MONO_TYPE_BOOLEAN: case MONO_TYPE_U1: p->regs [greg ++] = *(guint8*)(arg); break; case MONO_TYPE_I1: p->regs [greg ++] = *(gint8*)(arg); break; case MONO_TYPE_I2: p->regs [greg ++] = *(gint16*)(arg); break; case MONO_TYPE_U2: case MONO_TYPE_CHAR: p->regs [greg ++] = *(guint16*)(arg); break; case MONO_TYPE_I4: p->regs [greg ++] = *(gint32*)(arg); break; case MONO_TYPE_U4: p->regs [greg ++] = *(guint32*)(arg); break; case MONO_TYPE_GENERICINST: if (MONO_TYPE_IS_REFERENCE (t)) { p->regs [greg ++] = (mgreg_t)*(arg); break; } else { /* Fall through */ } case MONO_TYPE_VALUETYPE: { ArgInfo *ainfo = &dinfo->cinfo->args [i + sig->hasthis]; g_assert (ainfo->storage == ArgValuetypeInReg); if (ainfo->pair_storage [0] != ArgNone) { g_assert (ainfo->pair_storage [0] == ArgInIReg); p->regs [greg ++] = ((mgreg_t*)(arg))[0]; } if (ainfo->pair_storage [1] != ArgNone) { g_assert (ainfo->pair_storage [1] == ArgInIReg); p->regs [greg ++] = ((mgreg_t*)(arg))[1]; } break; } default: g_assert_not_reached (); } } g_assert (greg <= PARAM_REGS); } /* * mono_arch_finish_dyn_call: * * Store the result of a dyn call into the return value buffer passed to * start_dyn_call (). * This function should be as fast as possible, any work which does not depend * on the actual values of the arguments should be done in * mono_arch_dyn_call_prepare (). */ void mono_arch_finish_dyn_call (MonoDynCallInfo *info, guint8 *buf) { ArchDynCallInfo *dinfo = (ArchDynCallInfo*)info; MonoMethodSignature *sig = dinfo->sig; guint8 *ret = ((DynCallArgs*)buf)->ret; mgreg_t res = ((DynCallArgs*)buf)->res; switch (mono_type_get_underlying_type (sig->ret)->type) { case MONO_TYPE_VOID: *(gpointer*)ret = NULL; break; case MONO_TYPE_STRING: case MONO_TYPE_CLASS: case MONO_TYPE_ARRAY: case MONO_TYPE_SZARRAY: case MONO_TYPE_OBJECT: case MONO_TYPE_I: case MONO_TYPE_U: case MONO_TYPE_PTR: *(gpointer*)ret = (gpointer)res; break; case MONO_TYPE_I1: *(gint8*)ret = res; break; case MONO_TYPE_U1: case MONO_TYPE_BOOLEAN: *(guint8*)ret = res; break; case MONO_TYPE_I2: *(gint16*)ret = res; break; case MONO_TYPE_U2: case MONO_TYPE_CHAR: *(guint16*)ret = res; break; case MONO_TYPE_I4: *(gint32*)ret = res; break; case MONO_TYPE_U4: *(guint32*)ret = res; break; case MONO_TYPE_I8: *(gint64*)ret = res; break; case MONO_TYPE_U8: *(guint64*)ret = res; break; case MONO_TYPE_GENERICINST: if (MONO_TYPE_IS_REFERENCE (sig->ret)) { *(gpointer*)ret = (gpointer)res; break; } else { /* Fall through */ } case MONO_TYPE_VALUETYPE: if (dinfo->cinfo->vtype_retaddr) { /* Nothing to do */ } else { ArgInfo *ainfo = &dinfo->cinfo->ret; g_assert (ainfo->storage == ArgValuetypeInReg); if (ainfo->pair_storage [0] != ArgNone) { g_assert (ainfo->pair_storage [0] == ArgInIReg); ((mgreg_t*)ret)[0] = res; } g_assert (ainfo->pair_storage [1] == ArgNone); } break; default: g_assert_not_reached (); } } /* emit an exception if condition is fail */ #define EMIT_COND_SYSTEM_EXCEPTION(cond,signed,exc_name) \ do { \ MonoInst *tins = mono_branch_optimize_exception_target (cfg, bb, exc_name); \ if (tins == NULL) { \ mono_add_patch_info (cfg, code - cfg->native_code, \ MONO_PATCH_INFO_EXC, exc_name); \ x86_branch32 (code, cond, 0, signed); \ } else { \ EMIT_COND_BRANCH (tins, cond, signed); \ } \ } while (0); #define EMIT_FPCOMPARE(code) do { \ amd64_fcompp (code); \ amd64_fnstsw (code); \ } while (0); #define EMIT_SSE2_FPFUNC(code, op, dreg, sreg1) do { \ amd64_movsd_membase_reg (code, AMD64_RSP, -8, (sreg1)); \ amd64_fld_membase (code, AMD64_RSP, -8, TRUE); \ amd64_ ##op (code); \ amd64_fst_membase (code, AMD64_RSP, -8, TRUE, TRUE); \ amd64_movsd_reg_membase (code, (dreg), AMD64_RSP, -8); \ } while (0); static guint8* emit_call_body (MonoCompile *cfg, guint8 *code, guint32 patch_type, gconstpointer data) { gboolean no_patch = FALSE; /* * FIXME: Add support for thunks */ { gboolean near_call = FALSE; /* * Indirect calls are expensive so try to make a near call if possible. * The caller memory is allocated by the code manager so it is * guaranteed to be at a 32 bit offset. */ if (patch_type != MONO_PATCH_INFO_ABS) { /* The target is in memory allocated using the code manager */ near_call = TRUE; if ((patch_type == MONO_PATCH_INFO_METHOD) || (patch_type == MONO_PATCH_INFO_METHOD_JUMP)) { if (((MonoMethod*)data)->klass->image->aot_module) /* The callee might be an AOT method */ near_call = FALSE; if (((MonoMethod*)data)->dynamic) /* The target is in malloc-ed memory */ near_call = FALSE; } if (patch_type == MONO_PATCH_INFO_INTERNAL_METHOD) { /* * The call might go directly to a native function without * the wrapper. */ MonoJitICallInfo *mi = mono_find_jit_icall_by_name (data); if (mi) { gconstpointer target = mono_icall_get_wrapper (mi); if ((((guint64)target) >> 32) != 0) near_call = FALSE; } } } else { if (cfg->abs_patches && g_hash_table_lookup (cfg->abs_patches, data)) { /* * This is not really an optimization, but required because the * generic class init trampolines use R11 to pass the vtable. */ near_call = TRUE; } else { MonoJitICallInfo *info = mono_find_jit_icall_by_addr (data); if (info) { if ((cfg->method->wrapper_type == MONO_WRAPPER_MANAGED_TO_NATIVE) && strstr (cfg->method->name, info->name)) { /* A call to the wrapped function */ if ((((guint64)data) >> 32) == 0) near_call = TRUE; no_patch = TRUE; } else if (info->func == info->wrapper) { /* No wrapper */ if ((((guint64)info->func) >> 32) == 0) near_call = TRUE; } else { /* See the comment in mono_codegen () */ if ((info->name [0] != 'v') || (strstr (info->name, "ves_array_new_va_") == NULL && strstr (info->name, "ves_array_element_address_") == NULL)) near_call = TRUE; } } else if ((((guint64)data) >> 32) == 0) { near_call = TRUE; no_patch = TRUE; } } } if (cfg->method->dynamic) /* These methods are allocated using malloc */ near_call = FALSE; if (cfg->compile_aot) { near_call = TRUE; no_patch = TRUE; } #ifdef MONO_ARCH_NOMAP32BIT near_call = FALSE; #endif /* The 64bit XEN kernel does not honour the MAP_32BIT flag. (#522894) */ if (optimize_for_xen) near_call = FALSE; if (near_call) { /* * Align the call displacement to an address divisible by 4 so it does * not span cache lines. This is required for code patching to work on SMP * systems. */ if (!no_patch && ((guint32)(code + 1 - cfg->native_code) % 4) != 0) amd64_padding (code, 4 - ((guint32)(code + 1 - cfg->native_code) % 4)); mono_add_patch_info (cfg, code - cfg->native_code, patch_type, data); amd64_call_code (code, 0); } else { mono_add_patch_info (cfg, code - cfg->native_code, patch_type, data); amd64_set_reg_template (code, GP_SCRATCH_REG); amd64_call_reg (code, GP_SCRATCH_REG); } } return code; } static inline guint8* emit_call (MonoCompile *cfg, guint8 *code, guint32 patch_type, gconstpointer data, gboolean win64_adjust_stack) { #ifdef HOST_WIN32 if (win64_adjust_stack) amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 32); #endif code = emit_call_body (cfg, code, patch_type, data); #ifdef HOST_WIN32 if (win64_adjust_stack) amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, 32); #endif return code; } static inline int store_membase_imm_to_store_membase_reg (int opcode) { switch (opcode) { case OP_STORE_MEMBASE_IMM: return OP_STORE_MEMBASE_REG; case OP_STOREI4_MEMBASE_IMM: return OP_STOREI4_MEMBASE_REG; case OP_STOREI8_MEMBASE_IMM: return OP_STOREI8_MEMBASE_REG; } return -1; } #define INST_IGNORES_CFLAGS(opcode) (!(((opcode) == OP_ADC) || ((opcode) == OP_ADC_IMM) || ((opcode) == OP_IADC) || ((opcode) == OP_IADC_IMM) || ((opcode) == OP_SBB) || ((opcode) == OP_SBB_IMM) || ((opcode) == OP_ISBB) || ((opcode) == OP_ISBB_IMM))) /* * mono_arch_peephole_pass_1: * * Perform peephole opts which should/can be performed before local regalloc */ void mono_arch_peephole_pass_1 (MonoCompile *cfg, MonoBasicBlock *bb) { MonoInst *ins, *n; MONO_BB_FOR_EACH_INS_SAFE (bb, n, ins) { MonoInst *last_ins = ins->prev; switch (ins->opcode) { case OP_ADD_IMM: case OP_IADD_IMM: case OP_LADD_IMM: if ((ins->sreg1 < MONO_MAX_IREGS) && (ins->dreg >= MONO_MAX_IREGS) && (ins->inst_imm > 0)) { /* * X86_LEA is like ADD, but doesn't have the * sreg1==dreg restriction. inst_imm > 0 is needed since LEA sign-extends * its operand to 64 bit. */ ins->opcode = OP_X86_LEA_MEMBASE; ins->inst_basereg = ins->sreg1; } break; case OP_LXOR: case OP_IXOR: if ((ins->sreg1 == ins->sreg2) && (ins->sreg1 == ins->dreg)) { MonoInst *ins2; /* * Replace STORE_MEMBASE_IMM 0 with STORE_MEMBASE_REG since * the latter has length 2-3 instead of 6 (reverse constant * propagation). These instruction sequences are very common * in the initlocals bblock. */ for (ins2 = ins->next; ins2; ins2 = ins2->next) { if (((ins2->opcode == OP_STORE_MEMBASE_IMM) || (ins2->opcode == OP_STOREI4_MEMBASE_IMM) || (ins2->opcode == OP_STOREI8_MEMBASE_IMM) || (ins2->opcode == OP_STORE_MEMBASE_IMM)) && (ins2->inst_imm == 0)) { ins2->opcode = store_membase_imm_to_store_membase_reg (ins2->opcode); ins2->sreg1 = ins->dreg; } else if ((ins2->opcode == OP_STOREI1_MEMBASE_IMM) || (ins2->opcode == OP_STOREI2_MEMBASE_IMM) || (ins2->opcode == OP_STOREI8_MEMBASE_REG) || (ins2->opcode == OP_STORE_MEMBASE_REG)) { /* Continue */ } else if (((ins2->opcode == OP_ICONST) || (ins2->opcode == OP_I8CONST)) && (ins2->dreg == ins->dreg) && (ins2->inst_c0 == 0)) { NULLIFY_INS (ins2); /* Continue */ } else { break; } } } break; case OP_COMPARE_IMM: case OP_LCOMPARE_IMM: /* OP_COMPARE_IMM (reg, 0) * --> * OP_AMD64_TEST_NULL (reg) */ if (!ins->inst_imm) ins->opcode = OP_AMD64_TEST_NULL; break; case OP_ICOMPARE_IMM: if (!ins->inst_imm) ins->opcode = OP_X86_TEST_NULL; break; case OP_AMD64_ICOMPARE_MEMBASE_IMM: /* * OP_STORE_MEMBASE_REG reg, offset(basereg) * OP_X86_COMPARE_MEMBASE_IMM offset(basereg), imm * --> * OP_STORE_MEMBASE_REG reg, offset(basereg) * OP_COMPARE_IMM reg, imm * * Note: if imm = 0 then OP_COMPARE_IMM replaced with OP_X86_TEST_NULL */ if (last_ins && (last_ins->opcode == OP_STOREI4_MEMBASE_REG) && ins->inst_basereg == last_ins->inst_destbasereg && ins->inst_offset == last_ins->inst_offset) { ins->opcode = OP_ICOMPARE_IMM; ins->sreg1 = last_ins->sreg1; /* check if we can remove cmp reg,0 with test null */ if (!ins->inst_imm) ins->opcode = OP_X86_TEST_NULL; } break; } mono_peephole_ins (bb, ins); } } void mono_arch_peephole_pass_2 (MonoCompile *cfg, MonoBasicBlock *bb) { MonoInst *ins, *n; MONO_BB_FOR_EACH_INS_SAFE (bb, n, ins) { switch (ins->opcode) { case OP_ICONST: case OP_I8CONST: { /* reg = 0 -> XOR (reg, reg) */ /* XOR sets cflags on x86, so we cant do it always */ if (ins->inst_c0 == 0 && (!ins->next || (ins->next && INST_IGNORES_CFLAGS (ins->next->opcode)))) { ins->opcode = OP_LXOR; ins->sreg1 = ins->dreg; ins->sreg2 = ins->dreg; /* Fall through */ } else { break; } } case OP_LXOR: /* * Use IXOR to avoid a rex prefix if possible. The cpu will sign extend the * 0 result into 64 bits. */ if ((ins->sreg1 == ins->sreg2) && (ins->sreg1 == ins->dreg)) { ins->opcode = OP_IXOR; } /* Fall through */ case OP_IXOR: if ((ins->sreg1 == ins->sreg2) && (ins->sreg1 == ins->dreg)) { MonoInst *ins2; /* * Replace STORE_MEMBASE_IMM 0 with STORE_MEMBASE_REG since * the latter has length 2-3 instead of 6 (reverse constant * propagation). These instruction sequences are very common * in the initlocals bblock. */ for (ins2 = ins->next; ins2; ins2 = ins2->next) { if (((ins2->opcode == OP_STORE_MEMBASE_IMM) || (ins2->opcode == OP_STOREI4_MEMBASE_IMM) || (ins2->opcode == OP_STOREI8_MEMBASE_IMM) || (ins2->opcode == OP_STORE_MEMBASE_IMM)) && (ins2->inst_imm == 0)) { ins2->opcode = store_membase_imm_to_store_membase_reg (ins2->opcode); ins2->sreg1 = ins->dreg; } else if ((ins2->opcode == OP_STOREI1_MEMBASE_IMM) || (ins2->opcode == OP_STOREI2_MEMBASE_IMM) || (ins2->opcode == OP_STOREI4_MEMBASE_REG) || (ins2->opcode == OP_STOREI8_MEMBASE_REG) || (ins2->opcode == OP_STORE_MEMBASE_REG)) { /* Continue */ } else if (((ins2->opcode == OP_ICONST) || (ins2->opcode == OP_I8CONST)) && (ins2->dreg == ins->dreg) && (ins2->inst_c0 == 0)) { NULLIFY_INS (ins2); /* Continue */ } else { break; } } } break; case OP_IADD_IMM: if ((ins->inst_imm == 1) && (ins->dreg == ins->sreg1)) ins->opcode = OP_X86_INC_REG; break; case OP_ISUB_IMM: if ((ins->inst_imm == 1) && (ins->dreg == ins->sreg1)) ins->opcode = OP_X86_DEC_REG; break; } mono_peephole_ins (bb, ins); } } #define NEW_INS(cfg,ins,dest,op) do { \ MONO_INST_NEW ((cfg), (dest), (op)); \ (dest)->cil_code = (ins)->cil_code; \ mono_bblock_insert_before_ins (bb, ins, (dest)); \ } while (0) /* * mono_arch_lowering_pass: * * Converts complex opcodes into simpler ones so that each IR instruction * corresponds to one machine instruction. */ void mono_arch_lowering_pass (MonoCompile *cfg, MonoBasicBlock *bb) { MonoInst *ins, *n, *temp; /* * FIXME: Need to add more instructions, but the current machine * description can't model some parts of the composite instructions like * cdq. */ MONO_BB_FOR_EACH_INS_SAFE (bb, n, ins) { switch (ins->opcode) { case OP_DIV_IMM: case OP_REM_IMM: case OP_IDIV_IMM: case OP_IDIV_UN_IMM: case OP_IREM_UN_IMM: mono_decompose_op_imm (cfg, bb, ins); break; case OP_IREM_IMM: /* Keep the opcode if we can implement it efficiently */ if (!((ins->inst_imm > 0) && (mono_is_power_of_two (ins->inst_imm) != -1))) mono_decompose_op_imm (cfg, bb, ins); break; case OP_COMPARE_IMM: case OP_LCOMPARE_IMM: if (!amd64_is_imm32 (ins->inst_imm)) { NEW_INS (cfg, ins, temp, OP_I8CONST); temp->inst_c0 = ins->inst_imm; temp->dreg = mono_alloc_ireg (cfg); ins->opcode = OP_COMPARE; ins->sreg2 = temp->dreg; } break; case OP_LOAD_MEMBASE: case OP_LOADI8_MEMBASE: if (!amd64_is_imm32 (ins->inst_offset)) { NEW_INS (cfg, ins, temp, OP_I8CONST); temp->inst_c0 = ins->inst_offset; temp->dreg = mono_alloc_ireg (cfg); ins->opcode = OP_AMD64_LOADI8_MEMINDEX; ins->inst_indexreg = temp->dreg; } break; case OP_STORE_MEMBASE_IMM: case OP_STOREI8_MEMBASE_IMM: if (!amd64_is_imm32 (ins->inst_imm)) { NEW_INS (cfg, ins, temp, OP_I8CONST); temp->inst_c0 = ins->inst_imm; temp->dreg = mono_alloc_ireg (cfg); ins->opcode = OP_STOREI8_MEMBASE_REG; ins->sreg1 = temp->dreg; } break; #ifdef MONO_ARCH_SIMD_INTRINSICS case OP_EXPAND_I1: { int temp_reg1 = mono_alloc_ireg (cfg); int temp_reg2 = mono_alloc_ireg (cfg); int original_reg = ins->sreg1; NEW_INS (cfg, ins, temp, OP_ICONV_TO_U1); temp->sreg1 = original_reg; temp->dreg = temp_reg1; NEW_INS (cfg, ins, temp, OP_SHL_IMM); temp->sreg1 = temp_reg1; temp->dreg = temp_reg2; temp->inst_imm = 8; NEW_INS (cfg, ins, temp, OP_LOR); temp->sreg1 = temp->dreg = temp_reg2; temp->sreg2 = temp_reg1; ins->opcode = OP_EXPAND_I2; ins->sreg1 = temp_reg2; } break; #endif default: break; } } bb->max_vreg = cfg->next_vreg; } static const int branch_cc_table [] = { X86_CC_EQ, X86_CC_GE, X86_CC_GT, X86_CC_LE, X86_CC_LT, X86_CC_NE, X86_CC_GE, X86_CC_GT, X86_CC_LE, X86_CC_LT, X86_CC_O, X86_CC_NO, X86_CC_C, X86_CC_NC }; /* Maps CMP_... constants to X86_CC_... constants */ static const int cc_table [] = { X86_CC_EQ, X86_CC_NE, X86_CC_LE, X86_CC_GE, X86_CC_LT, X86_CC_GT, X86_CC_LE, X86_CC_GE, X86_CC_LT, X86_CC_GT }; static const int cc_signed_table [] = { TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, FALSE, FALSE, FALSE, FALSE }; /*#include "cprop.c"*/ static unsigned char* emit_float_to_int (MonoCompile *cfg, guchar *code, int dreg, int sreg, int size, gboolean is_signed) { amd64_sse_cvttsd2si_reg_reg (code, dreg, sreg); if (size == 1) amd64_widen_reg (code, dreg, dreg, is_signed, FALSE); else if (size == 2) amd64_widen_reg (code, dreg, dreg, is_signed, TRUE); return code; } static unsigned char* mono_emit_stack_alloc (MonoCompile *cfg, guchar *code, MonoInst* tree) { int sreg = tree->sreg1; int need_touch = FALSE; #if defined(HOST_WIN32) || defined(MONO_ARCH_SIGSEGV_ON_ALTSTACK) if (!tree->flags & MONO_INST_INIT) need_touch = TRUE; #endif if (need_touch) { guint8* br[5]; /* * Under Windows: * If requested stack size is larger than one page, * perform stack-touch operation */ /* * Generate stack probe code. * Under Windows, it is necessary to allocate one page at a time, * "touching" stack after each successful sub-allocation. This is * because of the way stack growth is implemented - there is a * guard page before the lowest stack page that is currently commited. * Stack normally grows sequentially so OS traps access to the * guard page and commits more pages when needed. */ amd64_test_reg_imm (code, sreg, ~0xFFF); br[0] = code; x86_branch8 (code, X86_CC_Z, 0, FALSE); br[2] = code; /* loop */ amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 0x1000); amd64_test_membase_reg (code, AMD64_RSP, 0, AMD64_RSP); amd64_alu_reg_imm (code, X86_SUB, sreg, 0x1000); amd64_alu_reg_imm (code, X86_CMP, sreg, 0x1000); br[3] = code; x86_branch8 (code, X86_CC_AE, 0, FALSE); amd64_patch (br[3], br[2]); amd64_test_reg_reg (code, sreg, sreg); br[4] = code; x86_branch8 (code, X86_CC_Z, 0, FALSE); amd64_alu_reg_reg (code, X86_SUB, AMD64_RSP, sreg); br[1] = code; x86_jump8 (code, 0); amd64_patch (br[0], code); amd64_alu_reg_reg (code, X86_SUB, AMD64_RSP, sreg); amd64_patch (br[1], code); amd64_patch (br[4], code); } else amd64_alu_reg_reg (code, X86_SUB, AMD64_RSP, tree->sreg1); if (tree->flags & MONO_INST_INIT) { int offset = 0; if (tree->dreg != AMD64_RAX && sreg != AMD64_RAX) { amd64_push_reg (code, AMD64_RAX); offset += 8; } if (tree->dreg != AMD64_RCX && sreg != AMD64_RCX) { amd64_push_reg (code, AMD64_RCX); offset += 8; } if (tree->dreg != AMD64_RDI && sreg != AMD64_RDI) { amd64_push_reg (code, AMD64_RDI); offset += 8; } amd64_shift_reg_imm (code, X86_SHR, sreg, 3); if (sreg != AMD64_RCX) amd64_mov_reg_reg (code, AMD64_RCX, sreg, 8); amd64_alu_reg_reg (code, X86_XOR, AMD64_RAX, AMD64_RAX); amd64_lea_membase (code, AMD64_RDI, AMD64_RSP, offset); if (cfg->param_area && cfg->arch.no_pushes) amd64_alu_reg_imm (code, X86_ADD, AMD64_RDI, cfg->param_area); amd64_cld (code); amd64_prefix (code, X86_REP_PREFIX); amd64_stosl (code); if (tree->dreg != AMD64_RDI && sreg != AMD64_RDI) amd64_pop_reg (code, AMD64_RDI); if (tree->dreg != AMD64_RCX && sreg != AMD64_RCX) amd64_pop_reg (code, AMD64_RCX); if (tree->dreg != AMD64_RAX && sreg != AMD64_RAX) amd64_pop_reg (code, AMD64_RAX); } return code; } static guint8* emit_move_return_value (MonoCompile *cfg, MonoInst *ins, guint8 *code) { CallInfo *cinfo; guint32 quad; /* Move return value to the target register */ /* FIXME: do this in the local reg allocator */ switch (ins->opcode) { case OP_CALL: case OP_CALL_REG: case OP_CALL_MEMBASE: case OP_LCALL: case OP_LCALL_REG: case OP_LCALL_MEMBASE: g_assert (ins->dreg == AMD64_RAX); break; case OP_FCALL: case OP_FCALL_REG: case OP_FCALL_MEMBASE: if (((MonoCallInst*)ins)->signature->ret->type == MONO_TYPE_R4) { amd64_sse_cvtss2sd_reg_reg (code, ins->dreg, AMD64_XMM0); } else { if (ins->dreg != AMD64_XMM0) amd64_sse_movsd_reg_reg (code, ins->dreg, AMD64_XMM0); } break; case OP_VCALL: case OP_VCALL_REG: case OP_VCALL_MEMBASE: case OP_VCALL2: case OP_VCALL2_REG: case OP_VCALL2_MEMBASE: cinfo = get_call_info (cfg->generic_sharing_context, cfg->mempool, ((MonoCallInst*)ins)->signature, FALSE); if (cinfo->ret.storage == ArgValuetypeInReg) { MonoInst *loc = cfg->arch.vret_addr_loc; /* Load the destination address */ g_assert (loc->opcode == OP_REGOFFSET); amd64_mov_reg_membase (code, AMD64_RCX, loc->inst_basereg, loc->inst_offset, 8); for (quad = 0; quad < 2; quad ++) { switch (cinfo->ret.pair_storage [quad]) { case ArgInIReg: amd64_mov_membase_reg (code, AMD64_RCX, (quad * 8), cinfo->ret.pair_regs [quad], 8); break; case ArgInFloatSSEReg: amd64_movss_membase_reg (code, AMD64_RCX, (quad * 8), cinfo->ret.pair_regs [quad]); break; case ArgInDoubleSSEReg: amd64_movsd_membase_reg (code, AMD64_RCX, (quad * 8), cinfo->ret.pair_regs [quad]); break; case ArgNone: break; default: NOT_IMPLEMENTED; } } } break; } return code; } /* * mono_amd64_emit_tls_get: * @code: buffer to store code to * @dreg: hard register where to place the result * @tls_offset: offset info * * mono_amd64_emit_tls_get emits in @code the native code that puts in * the dreg register the item in the thread local storage identified * by tls_offset. * * Returns: a pointer to the end of the stored code */ guint8* mono_amd64_emit_tls_get (guint8* code, int dreg, int tls_offset) { #ifdef HOST_WIN32 g_assert (tls_offset < 64); x86_prefix (code, X86_GS_PREFIX); amd64_mov_reg_mem (code, dreg, (tls_offset * 8) + 0x1480, 8); #else if (optimize_for_xen) { x86_prefix (code, X86_FS_PREFIX); amd64_mov_reg_mem (code, dreg, 0, 8); amd64_mov_reg_membase (code, dreg, dreg, tls_offset, 8); } else { x86_prefix (code, X86_FS_PREFIX); amd64_mov_reg_mem (code, dreg, tls_offset, 8); } #endif return code; } #define REAL_PRINT_REG(text,reg) \ mono_assert (reg >= 0); \ amd64_push_reg (code, AMD64_RAX); \ amd64_push_reg (code, AMD64_RDX); \ amd64_push_reg (code, AMD64_RCX); \ amd64_push_reg (code, reg); \ amd64_push_imm (code, reg); \ amd64_push_imm (code, text " %d %p\n"); \ amd64_mov_reg_imm (code, AMD64_RAX, printf); \ amd64_call_reg (code, AMD64_RAX); \ amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, 3*4); \ amd64_pop_reg (code, AMD64_RCX); \ amd64_pop_reg (code, AMD64_RDX); \ amd64_pop_reg (code, AMD64_RAX); /* benchmark and set based on cpu */ #define LOOP_ALIGNMENT 8 #define bb_is_loop_start(bb) ((bb)->loop_body_start && (bb)->nesting) #ifndef DISABLE_JIT void mono_arch_output_basic_block (MonoCompile *cfg, MonoBasicBlock *bb) { MonoInst *ins; MonoCallInst *call; guint offset; guint8 *code = cfg->native_code + cfg->code_len; MonoInst *last_ins = NULL; guint last_offset = 0; int max_len; /* Fix max_offset estimate for each successor bb */ if (cfg->opt & MONO_OPT_BRANCH) { int current_offset = cfg->code_len; MonoBasicBlock *current_bb; for (current_bb = bb; current_bb != NULL; current_bb = current_bb->next_bb) { current_bb->max_offset = current_offset; current_offset += current_bb->max_length; } } if (cfg->opt & MONO_OPT_LOOP) { int pad, align = LOOP_ALIGNMENT; /* set alignment depending on cpu */ if (bb_is_loop_start (bb) && (pad = (cfg->code_len & (align - 1)))) { pad = align - pad; /*g_print ("adding %d pad at %x to loop in %s\n", pad, cfg->code_len, cfg->method->name);*/ amd64_padding (code, pad); cfg->code_len += pad; bb->native_offset = cfg->code_len; } } if (cfg->verbose_level > 2) g_print ("Basic block %d starting at offset 0x%x\n", bb->block_num, bb->native_offset); if (cfg->prof_options & MONO_PROFILE_COVERAGE) { MonoProfileCoverageInfo *cov = cfg->coverage_info; g_assert (!cfg->compile_aot); cov->data [bb->dfn].cil_code = bb->cil_code; amd64_mov_reg_imm (code, AMD64_R11, (guint64)&cov->data [bb->dfn].count); /* this is not thread save, but good enough */ amd64_inc_membase (code, AMD64_R11, 0); } offset = code - cfg->native_code; mono_debug_open_block (cfg, bb, offset); if (mono_break_at_bb_method && mono_method_desc_full_match (mono_break_at_bb_method, cfg->method) && bb->block_num == mono_break_at_bb_bb_num) x86_breakpoint (code); MONO_BB_FOR_EACH_INS (bb, ins) { offset = code - cfg->native_code; max_len = ((guint8 *)ins_get_spec (ins->opcode))[MONO_INST_LEN]; if (G_UNLIKELY (offset > (cfg->code_size - max_len - 16))) { cfg->code_size *= 2; cfg->native_code = g_realloc (cfg->native_code, cfg->code_size); code = cfg->native_code + offset; mono_jit_stats.code_reallocs++; } if (cfg->debug_info) mono_debug_record_line_number (cfg, ins, offset); switch (ins->opcode) { case OP_BIGMUL: amd64_mul_reg (code, ins->sreg2, TRUE); break; case OP_BIGMUL_UN: amd64_mul_reg (code, ins->sreg2, FALSE); break; case OP_X86_SETEQ_MEMBASE: amd64_set_membase (code, X86_CC_EQ, ins->inst_basereg, ins->inst_offset, TRUE); break; case OP_STOREI1_MEMBASE_IMM: amd64_mov_membase_imm (code, ins->inst_destbasereg, ins->inst_offset, ins->inst_imm, 1); break; case OP_STOREI2_MEMBASE_IMM: amd64_mov_membase_imm (code, ins->inst_destbasereg, ins->inst_offset, ins->inst_imm, 2); break; case OP_STOREI4_MEMBASE_IMM: amd64_mov_membase_imm (code, ins->inst_destbasereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_STOREI1_MEMBASE_REG: amd64_mov_membase_reg (code, ins->inst_destbasereg, ins->inst_offset, ins->sreg1, 1); break; case OP_STOREI2_MEMBASE_REG: amd64_mov_membase_reg (code, ins->inst_destbasereg, ins->inst_offset, ins->sreg1, 2); break; case OP_STORE_MEMBASE_REG: case OP_STOREI8_MEMBASE_REG: amd64_mov_membase_reg (code, ins->inst_destbasereg, ins->inst_offset, ins->sreg1, 8); break; case OP_STOREI4_MEMBASE_REG: amd64_mov_membase_reg (code, ins->inst_destbasereg, ins->inst_offset, ins->sreg1, 4); break; case OP_STORE_MEMBASE_IMM: case OP_STOREI8_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_mov_membase_imm (code, ins->inst_destbasereg, ins->inst_offset, ins->inst_imm, 8); break; case OP_LOAD_MEM: case OP_LOADI8_MEM: // FIXME: Decompose this earlier if (amd64_is_imm32 (ins->inst_imm)) amd64_mov_reg_mem (code, ins->dreg, ins->inst_imm, sizeof (gpointer)); else { amd64_mov_reg_imm (code, ins->dreg, ins->inst_imm); amd64_mov_reg_membase (code, ins->dreg, ins->dreg, 0, 8); } break; case OP_LOADI4_MEM: amd64_mov_reg_imm (code, ins->dreg, ins->inst_imm); amd64_movsxd_reg_membase (code, ins->dreg, ins->dreg, 0); break; case OP_LOADU4_MEM: // FIXME: Decompose this earlier if (amd64_is_imm32 (ins->inst_imm)) amd64_mov_reg_mem (code, ins->dreg, ins->inst_imm, 4); else { amd64_mov_reg_imm (code, ins->dreg, ins->inst_imm); amd64_mov_reg_membase (code, ins->dreg, ins->dreg, 0, 4); } break; case OP_LOADU1_MEM: amd64_mov_reg_imm (code, ins->dreg, ins->inst_imm); amd64_widen_membase (code, ins->dreg, ins->dreg, 0, FALSE, FALSE); break; case OP_LOADU2_MEM: amd64_mov_reg_imm (code, ins->dreg, ins->inst_imm); amd64_widen_membase (code, ins->dreg, ins->dreg, 0, FALSE, TRUE); break; case OP_LOAD_MEMBASE: case OP_LOADI8_MEMBASE: g_assert (amd64_is_imm32 (ins->inst_offset)); amd64_mov_reg_membase (code, ins->dreg, ins->inst_basereg, ins->inst_offset, sizeof (gpointer)); break; case OP_LOADI4_MEMBASE: amd64_movsxd_reg_membase (code, ins->dreg, ins->inst_basereg, ins->inst_offset); break; case OP_LOADU4_MEMBASE: amd64_mov_reg_membase (code, ins->dreg, ins->inst_basereg, ins->inst_offset, 4); break; case OP_LOADU1_MEMBASE: /* The cpu zero extends the result into 64 bits */ amd64_widen_membase_size (code, ins->dreg, ins->inst_basereg, ins->inst_offset, FALSE, FALSE, 4); break; case OP_LOADI1_MEMBASE: amd64_widen_membase (code, ins->dreg, ins->inst_basereg, ins->inst_offset, TRUE, FALSE); break; case OP_LOADU2_MEMBASE: /* The cpu zero extends the result into 64 bits */ amd64_widen_membase_size (code, ins->dreg, ins->inst_basereg, ins->inst_offset, FALSE, TRUE, 4); break; case OP_LOADI2_MEMBASE: amd64_widen_membase (code, ins->dreg, ins->inst_basereg, ins->inst_offset, TRUE, TRUE); break; case OP_AMD64_LOADI8_MEMINDEX: amd64_mov_reg_memindex_size (code, ins->dreg, ins->inst_basereg, 0, ins->inst_indexreg, 0, 8); break; case OP_LCONV_TO_I1: case OP_ICONV_TO_I1: case OP_SEXT_I1: amd64_widen_reg (code, ins->dreg, ins->sreg1, TRUE, FALSE); break; case OP_LCONV_TO_I2: case OP_ICONV_TO_I2: case OP_SEXT_I2: amd64_widen_reg (code, ins->dreg, ins->sreg1, TRUE, TRUE); break; case OP_LCONV_TO_U1: case OP_ICONV_TO_U1: amd64_widen_reg (code, ins->dreg, ins->sreg1, FALSE, FALSE); break; case OP_LCONV_TO_U2: case OP_ICONV_TO_U2: amd64_widen_reg (code, ins->dreg, ins->sreg1, FALSE, TRUE); break; case OP_ZEXT_I4: /* Clean out the upper word */ amd64_mov_reg_reg_size (code, ins->dreg, ins->sreg1, 4); break; case OP_SEXT_I4: amd64_movsxd_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_COMPARE: case OP_LCOMPARE: amd64_alu_reg_reg (code, X86_CMP, ins->sreg1, ins->sreg2); break; case OP_COMPARE_IMM: case OP_LCOMPARE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_CMP, ins->sreg1, ins->inst_imm); break; case OP_X86_COMPARE_REG_MEMBASE: amd64_alu_reg_membase (code, X86_CMP, ins->sreg1, ins->sreg2, ins->inst_offset); break; case OP_X86_TEST_NULL: amd64_test_reg_reg_size (code, ins->sreg1, ins->sreg1, 4); break; case OP_AMD64_TEST_NULL: amd64_test_reg_reg (code, ins->sreg1, ins->sreg1); break; case OP_X86_ADD_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_ADD, ins->sreg1, ins->sreg2, ins->inst_offset, 4); break; case OP_X86_SUB_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_SUB, ins->sreg1, ins->sreg2, ins->inst_offset, 4); break; case OP_X86_AND_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_AND, ins->sreg1, ins->sreg2, ins->inst_offset, 4); break; case OP_X86_OR_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_OR, ins->sreg1, ins->sreg2, ins->inst_offset, 4); break; case OP_X86_XOR_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_XOR, ins->sreg1, ins->sreg2, ins->inst_offset, 4); break; case OP_X86_ADD_MEMBASE_IMM: /* FIXME: Make a 64 version too */ amd64_alu_membase_imm_size (code, X86_ADD, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_X86_SUB_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_SUB, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_X86_AND_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_AND, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_X86_OR_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_OR, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_X86_XOR_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_XOR, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_X86_ADD_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_ADD, ins->inst_basereg, ins->inst_offset, ins->sreg2, 4); break; case OP_X86_SUB_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_SUB, ins->inst_basereg, ins->inst_offset, ins->sreg2, 4); break; case OP_X86_AND_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_AND, ins->inst_basereg, ins->inst_offset, ins->sreg2, 4); break; case OP_X86_OR_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_OR, ins->inst_basereg, ins->inst_offset, ins->sreg2, 4); break; case OP_X86_XOR_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_XOR, ins->inst_basereg, ins->inst_offset, ins->sreg2, 4); break; case OP_X86_INC_MEMBASE: amd64_inc_membase_size (code, ins->inst_basereg, ins->inst_offset, 4); break; case OP_X86_INC_REG: amd64_inc_reg_size (code, ins->dreg, 4); break; case OP_X86_DEC_MEMBASE: amd64_dec_membase_size (code, ins->inst_basereg, ins->inst_offset, 4); break; case OP_X86_DEC_REG: amd64_dec_reg_size (code, ins->dreg, 4); break; case OP_X86_MUL_REG_MEMBASE: case OP_X86_MUL_MEMBASE_REG: amd64_imul_reg_membase_size (code, ins->sreg1, ins->sreg2, ins->inst_offset, 4); break; case OP_AMD64_ICOMPARE_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_CMP, ins->inst_basereg, ins->inst_offset, ins->sreg2, 4); break; case OP_AMD64_ICOMPARE_MEMBASE_IMM: amd64_alu_membase_imm_size (code, X86_CMP, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_AMD64_COMPARE_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_CMP, ins->inst_basereg, ins->inst_offset, ins->sreg2, 8); break; case OP_AMD64_COMPARE_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_CMP, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 8); break; case OP_X86_COMPARE_MEMBASE8_IMM: amd64_alu_membase8_imm_size (code, X86_CMP, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 4); break; case OP_AMD64_ICOMPARE_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_CMP, ins->sreg1, ins->sreg2, ins->inst_offset, 4); break; case OP_AMD64_COMPARE_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_CMP, ins->sreg1, ins->sreg2, ins->inst_offset, 8); break; case OP_AMD64_ADD_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_ADD, ins->sreg1, ins->sreg2, ins->inst_offset, 8); break; case OP_AMD64_SUB_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_SUB, ins->sreg1, ins->sreg2, ins->inst_offset, 8); break; case OP_AMD64_AND_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_AND, ins->sreg1, ins->sreg2, ins->inst_offset, 8); break; case OP_AMD64_OR_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_OR, ins->sreg1, ins->sreg2, ins->inst_offset, 8); break; case OP_AMD64_XOR_REG_MEMBASE: amd64_alu_reg_membase_size (code, X86_XOR, ins->sreg1, ins->sreg2, ins->inst_offset, 8); break; case OP_AMD64_ADD_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_ADD, ins->inst_basereg, ins->inst_offset, ins->sreg2, 8); break; case OP_AMD64_SUB_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_SUB, ins->inst_basereg, ins->inst_offset, ins->sreg2, 8); break; case OP_AMD64_AND_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_AND, ins->inst_basereg, ins->inst_offset, ins->sreg2, 8); break; case OP_AMD64_OR_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_OR, ins->inst_basereg, ins->inst_offset, ins->sreg2, 8); break; case OP_AMD64_XOR_MEMBASE_REG: amd64_alu_membase_reg_size (code, X86_XOR, ins->inst_basereg, ins->inst_offset, ins->sreg2, 8); break; case OP_AMD64_ADD_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_ADD, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 8); break; case OP_AMD64_SUB_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_SUB, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 8); break; case OP_AMD64_AND_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_AND, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 8); break; case OP_AMD64_OR_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_OR, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 8); break; case OP_AMD64_XOR_MEMBASE_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_membase_imm_size (code, X86_XOR, ins->inst_basereg, ins->inst_offset, ins->inst_imm, 8); break; case OP_BREAK: amd64_breakpoint (code); break; case OP_RELAXED_NOP: x86_prefix (code, X86_REP_PREFIX); x86_nop (code); break; case OP_HARD_NOP: x86_nop (code); break; case OP_NOP: case OP_DUMMY_USE: case OP_DUMMY_STORE: case OP_NOT_REACHED: case OP_NOT_NULL: break; case OP_SEQ_POINT: { int i; if (cfg->compile_aot) NOT_IMPLEMENTED; /* * Read from the single stepping trigger page. This will cause a * SIGSEGV when single stepping is enabled. * We do this _before_ the breakpoint, so single stepping after * a breakpoint is hit will step to the next IL offset. */ g_assert (((guint64)ss_trigger_page >> 32) == 0); if (ins->flags & MONO_INST_SINGLE_STEP_LOC) amd64_mov_reg_mem (code, AMD64_R11, (guint64)ss_trigger_page, 4); mono_add_seq_point (cfg, bb, ins, code - cfg->native_code); /* * A placeholder for a possible breakpoint inserted by * mono_arch_set_breakpoint (). */ for (i = 0; i < BREAKPOINT_SIZE; ++i) x86_nop (code); break; } case OP_ADDCC: case OP_LADD: amd64_alu_reg_reg (code, X86_ADD, ins->sreg1, ins->sreg2); break; case OP_ADC: amd64_alu_reg_reg (code, X86_ADC, ins->sreg1, ins->sreg2); break; case OP_ADD_IMM: case OP_LADD_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_ADD, ins->dreg, ins->inst_imm); break; case OP_ADC_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_ADC, ins->dreg, ins->inst_imm); break; case OP_SUBCC: case OP_LSUB: amd64_alu_reg_reg (code, X86_SUB, ins->sreg1, ins->sreg2); break; case OP_SBB: amd64_alu_reg_reg (code, X86_SBB, ins->sreg1, ins->sreg2); break; case OP_SUB_IMM: case OP_LSUB_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_SUB, ins->dreg, ins->inst_imm); break; case OP_SBB_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_SBB, ins->dreg, ins->inst_imm); break; case OP_LAND: amd64_alu_reg_reg (code, X86_AND, ins->sreg1, ins->sreg2); break; case OP_AND_IMM: case OP_LAND_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_AND, ins->sreg1, ins->inst_imm); break; case OP_LMUL: amd64_imul_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_MUL_IMM: case OP_LMUL_IMM: case OP_IMUL_IMM: { guint32 size = (ins->opcode == OP_IMUL_IMM) ? 4 : 8; switch (ins->inst_imm) { case 2: /* MOV r1, r2 */ /* ADD r1, r1 */ if (ins->dreg != ins->sreg1) amd64_mov_reg_reg (code, ins->dreg, ins->sreg1, size); amd64_alu_reg_reg (code, X86_ADD, ins->dreg, ins->dreg); break; case 3: /* LEA r1, [r2 + r2*2] */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 1); break; case 5: /* LEA r1, [r2 + r2*4] */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 2); break; case 6: /* LEA r1, [r2 + r2*2] */ /* ADD r1, r1 */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 1); amd64_alu_reg_reg (code, X86_ADD, ins->dreg, ins->dreg); break; case 9: /* LEA r1, [r2 + r2*8] */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 3); break; case 10: /* LEA r1, [r2 + r2*4] */ /* ADD r1, r1 */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 2); amd64_alu_reg_reg (code, X86_ADD, ins->dreg, ins->dreg); break; case 12: /* LEA r1, [r2 + r2*2] */ /* SHL r1, 2 */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 1); amd64_shift_reg_imm (code, X86_SHL, ins->dreg, 2); break; case 25: /* LEA r1, [r2 + r2*4] */ /* LEA r1, [r1 + r1*4] */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 2); amd64_lea_memindex (code, ins->dreg, ins->dreg, 0, ins->dreg, 2); break; case 100: /* LEA r1, [r2 + r2*4] */ /* SHL r1, 2 */ /* LEA r1, [r1 + r1*4] */ amd64_lea_memindex (code, ins->dreg, ins->sreg1, 0, ins->sreg1, 2); amd64_shift_reg_imm (code, X86_SHL, ins->dreg, 2); amd64_lea_memindex (code, ins->dreg, ins->dreg, 0, ins->dreg, 2); break; default: amd64_imul_reg_reg_imm_size (code, ins->dreg, ins->sreg1, ins->inst_imm, size); break; } break; } case OP_LDIV: case OP_LREM: /* Regalloc magic makes the div/rem cases the same */ if (ins->sreg2 == AMD64_RDX) { amd64_mov_membase_reg (code, AMD64_RSP, -8, AMD64_RDX, 8); amd64_cdq (code); amd64_div_membase (code, AMD64_RSP, -8, TRUE); } else { amd64_cdq (code); amd64_div_reg (code, ins->sreg2, TRUE); } break; case OP_LDIV_UN: case OP_LREM_UN: if (ins->sreg2 == AMD64_RDX) { amd64_mov_membase_reg (code, AMD64_RSP, -8, AMD64_RDX, 8); amd64_alu_reg_reg (code, X86_XOR, AMD64_RDX, AMD64_RDX); amd64_div_membase (code, AMD64_RSP, -8, FALSE); } else { amd64_alu_reg_reg (code, X86_XOR, AMD64_RDX, AMD64_RDX); amd64_div_reg (code, ins->sreg2, FALSE); } break; case OP_IDIV: case OP_IREM: if (ins->sreg2 == AMD64_RDX) { amd64_mov_membase_reg (code, AMD64_RSP, -8, AMD64_RDX, 8); amd64_cdq_size (code, 4); amd64_div_membase_size (code, AMD64_RSP, -8, TRUE, 4); } else { amd64_cdq_size (code, 4); amd64_div_reg_size (code, ins->sreg2, TRUE, 4); } break; case OP_IDIV_UN: case OP_IREM_UN: if (ins->sreg2 == AMD64_RDX) { amd64_mov_membase_reg (code, AMD64_RSP, -8, AMD64_RDX, 8); amd64_alu_reg_reg (code, X86_XOR, AMD64_RDX, AMD64_RDX); amd64_div_membase_size (code, AMD64_RSP, -8, FALSE, 4); } else { amd64_alu_reg_reg (code, X86_XOR, AMD64_RDX, AMD64_RDX); amd64_div_reg_size (code, ins->sreg2, FALSE, 4); } break; case OP_IREM_IMM: { int power = mono_is_power_of_two (ins->inst_imm); g_assert (ins->sreg1 == X86_EAX); g_assert (ins->dreg == X86_EAX); g_assert (power >= 0); if (power == 0) { amd64_mov_reg_imm (code, ins->dreg, 0); break; } /* Based on gcc code */ /* Add compensation for negative dividents */ amd64_mov_reg_reg_size (code, AMD64_RDX, AMD64_RAX, 4); if (power > 1) amd64_shift_reg_imm_size (code, X86_SAR, AMD64_RDX, 31, 4); amd64_shift_reg_imm_size (code, X86_SHR, AMD64_RDX, 32 - power, 4); amd64_alu_reg_reg_size (code, X86_ADD, AMD64_RAX, AMD64_RDX, 4); /* Compute remainder */ amd64_alu_reg_imm_size (code, X86_AND, AMD64_RAX, (1 << power) - 1, 4); /* Remove compensation */ amd64_alu_reg_reg_size (code, X86_SUB, AMD64_RAX, AMD64_RDX, 4); break; } case OP_LMUL_OVF: amd64_imul_reg_reg (code, ins->sreg1, ins->sreg2); EMIT_COND_SYSTEM_EXCEPTION (X86_CC_O, FALSE, "OverflowException"); break; case OP_LOR: amd64_alu_reg_reg (code, X86_OR, ins->sreg1, ins->sreg2); break; case OP_OR_IMM: case OP_LOR_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_OR, ins->sreg1, ins->inst_imm); break; case OP_LXOR: amd64_alu_reg_reg (code, X86_XOR, ins->sreg1, ins->sreg2); break; case OP_XOR_IMM: case OP_LXOR_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_alu_reg_imm (code, X86_XOR, ins->sreg1, ins->inst_imm); break; case OP_LSHL: g_assert (ins->sreg2 == AMD64_RCX); amd64_shift_reg (code, X86_SHL, ins->dreg); break; case OP_LSHR: g_assert (ins->sreg2 == AMD64_RCX); amd64_shift_reg (code, X86_SAR, ins->dreg); break; case OP_SHR_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_shift_reg_imm_size (code, X86_SAR, ins->dreg, ins->inst_imm, 4); break; case OP_LSHR_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_shift_reg_imm (code, X86_SAR, ins->dreg, ins->inst_imm); break; case OP_SHR_UN_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_shift_reg_imm_size (code, X86_SHR, ins->dreg, ins->inst_imm, 4); break; case OP_LSHR_UN_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_shift_reg_imm (code, X86_SHR, ins->dreg, ins->inst_imm); break; case OP_LSHR_UN: g_assert (ins->sreg2 == AMD64_RCX); amd64_shift_reg (code, X86_SHR, ins->dreg); break; case OP_SHL_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_shift_reg_imm_size (code, X86_SHL, ins->dreg, ins->inst_imm, 4); break; case OP_LSHL_IMM: g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_shift_reg_imm (code, X86_SHL, ins->dreg, ins->inst_imm); break; case OP_IADDCC: case OP_IADD: amd64_alu_reg_reg_size (code, X86_ADD, ins->sreg1, ins->sreg2, 4); break; case OP_IADC: amd64_alu_reg_reg_size (code, X86_ADC, ins->sreg1, ins->sreg2, 4); break; case OP_IADD_IMM: amd64_alu_reg_imm_size (code, X86_ADD, ins->dreg, ins->inst_imm, 4); break; case OP_IADC_IMM: amd64_alu_reg_imm_size (code, X86_ADC, ins->dreg, ins->inst_imm, 4); break; case OP_ISUBCC: case OP_ISUB: amd64_alu_reg_reg_size (code, X86_SUB, ins->sreg1, ins->sreg2, 4); break; case OP_ISBB: amd64_alu_reg_reg_size (code, X86_SBB, ins->sreg1, ins->sreg2, 4); break; case OP_ISUB_IMM: amd64_alu_reg_imm_size (code, X86_SUB, ins->dreg, ins->inst_imm, 4); break; case OP_ISBB_IMM: amd64_alu_reg_imm_size (code, X86_SBB, ins->dreg, ins->inst_imm, 4); break; case OP_IAND: amd64_alu_reg_reg_size (code, X86_AND, ins->sreg1, ins->sreg2, 4); break; case OP_IAND_IMM: amd64_alu_reg_imm_size (code, X86_AND, ins->sreg1, ins->inst_imm, 4); break; case OP_IOR: amd64_alu_reg_reg_size (code, X86_OR, ins->sreg1, ins->sreg2, 4); break; case OP_IOR_IMM: amd64_alu_reg_imm_size (code, X86_OR, ins->sreg1, ins->inst_imm, 4); break; case OP_IXOR: amd64_alu_reg_reg_size (code, X86_XOR, ins->sreg1, ins->sreg2, 4); break; case OP_IXOR_IMM: amd64_alu_reg_imm_size (code, X86_XOR, ins->sreg1, ins->inst_imm, 4); break; case OP_INEG: amd64_neg_reg_size (code, ins->sreg1, 4); break; case OP_INOT: amd64_not_reg_size (code, ins->sreg1, 4); break; case OP_ISHL: g_assert (ins->sreg2 == AMD64_RCX); amd64_shift_reg_size (code, X86_SHL, ins->dreg, 4); break; case OP_ISHR: g_assert (ins->sreg2 == AMD64_RCX); amd64_shift_reg_size (code, X86_SAR, ins->dreg, 4); break; case OP_ISHR_IMM: amd64_shift_reg_imm_size (code, X86_SAR, ins->dreg, ins->inst_imm, 4); break; case OP_ISHR_UN_IMM: amd64_shift_reg_imm_size (code, X86_SHR, ins->dreg, ins->inst_imm, 4); break; case OP_ISHR_UN: g_assert (ins->sreg2 == AMD64_RCX); amd64_shift_reg_size (code, X86_SHR, ins->dreg, 4); break; case OP_ISHL_IMM: amd64_shift_reg_imm_size (code, X86_SHL, ins->dreg, ins->inst_imm, 4); break; case OP_IMUL: amd64_imul_reg_reg_size (code, ins->sreg1, ins->sreg2, 4); break; case OP_IMUL_OVF: amd64_imul_reg_reg_size (code, ins->sreg1, ins->sreg2, 4); EMIT_COND_SYSTEM_EXCEPTION (X86_CC_O, FALSE, "OverflowException"); break; case OP_IMUL_OVF_UN: case OP_LMUL_OVF_UN: { /* the mul operation and the exception check should most likely be split */ int non_eax_reg, saved_eax = FALSE, saved_edx = FALSE; int size = (ins->opcode == OP_IMUL_OVF_UN) ? 4 : 8; /*g_assert (ins->sreg2 == X86_EAX); g_assert (ins->dreg == X86_EAX);*/ if (ins->sreg2 == X86_EAX) { non_eax_reg = ins->sreg1; } else if (ins->sreg1 == X86_EAX) { non_eax_reg = ins->sreg2; } else { /* no need to save since we're going to store to it anyway */ if (ins->dreg != X86_EAX) { saved_eax = TRUE; amd64_push_reg (code, X86_EAX); } amd64_mov_reg_reg (code, X86_EAX, ins->sreg1, size); non_eax_reg = ins->sreg2; } if (ins->dreg == X86_EDX) { if (!saved_eax) { saved_eax = TRUE; amd64_push_reg (code, X86_EAX); } } else { saved_edx = TRUE; amd64_push_reg (code, X86_EDX); } amd64_mul_reg_size (code, non_eax_reg, FALSE, size); /* save before the check since pop and mov don't change the flags */ if (ins->dreg != X86_EAX) amd64_mov_reg_reg (code, ins->dreg, X86_EAX, size); if (saved_edx) amd64_pop_reg (code, X86_EDX); if (saved_eax) amd64_pop_reg (code, X86_EAX); EMIT_COND_SYSTEM_EXCEPTION (X86_CC_O, FALSE, "OverflowException"); break; } case OP_ICOMPARE: amd64_alu_reg_reg_size (code, X86_CMP, ins->sreg1, ins->sreg2, 4); break; case OP_ICOMPARE_IMM: amd64_alu_reg_imm_size (code, X86_CMP, ins->sreg1, ins->inst_imm, 4); break; case OP_IBEQ: case OP_IBLT: case OP_IBGT: case OP_IBGE: case OP_IBLE: case OP_LBEQ: case OP_LBLT: case OP_LBGT: case OP_LBGE: case OP_LBLE: case OP_IBNE_UN: case OP_IBLT_UN: case OP_IBGT_UN: case OP_IBGE_UN: case OP_IBLE_UN: case OP_LBNE_UN: case OP_LBLT_UN: case OP_LBGT_UN: case OP_LBGE_UN: case OP_LBLE_UN: EMIT_COND_BRANCH (ins, cc_table [mono_opcode_to_cond (ins->opcode)], cc_signed_table [mono_opcode_to_cond (ins->opcode)]); break; case OP_CMOV_IEQ: case OP_CMOV_IGE: case OP_CMOV_IGT: case OP_CMOV_ILE: case OP_CMOV_ILT: case OP_CMOV_INE_UN: case OP_CMOV_IGE_UN: case OP_CMOV_IGT_UN: case OP_CMOV_ILE_UN: case OP_CMOV_ILT_UN: case OP_CMOV_LEQ: case OP_CMOV_LGE: case OP_CMOV_LGT: case OP_CMOV_LLE: case OP_CMOV_LLT: case OP_CMOV_LNE_UN: case OP_CMOV_LGE_UN: case OP_CMOV_LGT_UN: case OP_CMOV_LLE_UN: case OP_CMOV_LLT_UN: g_assert (ins->dreg == ins->sreg1); /* This needs to operate on 64 bit values */ amd64_cmov_reg (code, cc_table [mono_opcode_to_cond (ins->opcode)], cc_signed_table [mono_opcode_to_cond (ins->opcode)], ins->dreg, ins->sreg2); break; case OP_LNOT: amd64_not_reg (code, ins->sreg1); break; case OP_LNEG: amd64_neg_reg (code, ins->sreg1); break; case OP_ICONST: case OP_I8CONST: if ((((guint64)ins->inst_c0) >> 32) == 0) amd64_mov_reg_imm_size (code, ins->dreg, ins->inst_c0, 4); else amd64_mov_reg_imm_size (code, ins->dreg, ins->inst_c0, 8); break; case OP_AOTCONST: mono_add_patch_info (cfg, offset, (MonoJumpInfoType)ins->inst_i1, ins->inst_p0); amd64_mov_reg_membase (code, ins->dreg, AMD64_RIP, 0, 8); break; case OP_JUMP_TABLE: mono_add_patch_info (cfg, offset, (MonoJumpInfoType)ins->inst_i1, ins->inst_p0); amd64_mov_reg_imm_size (code, ins->dreg, 0, 8); break; case OP_MOVE: amd64_mov_reg_reg (code, ins->dreg, ins->sreg1, sizeof (gpointer)); break; case OP_AMD64_SET_XMMREG_R4: { amd64_sse_cvtsd2ss_reg_reg (code, ins->dreg, ins->sreg1); break; } case OP_AMD64_SET_XMMREG_R8: { if (ins->dreg != ins->sreg1) amd64_sse_movsd_reg_reg (code, ins->dreg, ins->sreg1); break; } case OP_TAILCALL: { /* * Note: this 'frame destruction' logic is useful for tail calls, too. * Keep in sync with the code in emit_epilog. */ int pos = 0, i; /* FIXME: no tracing support... */ if (cfg->prof_options & MONO_PROFILE_ENTER_LEAVE) code = mono_arch_instrument_epilog_full (cfg, mono_profiler_method_leave, code, FALSE, FALSE); g_assert (!cfg->method->save_lmf); if (cfg->arch.omit_fp) { guint32 save_offset = 0; /* Pop callee-saved registers */ for (i = 0; i < AMD64_NREG; ++i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { amd64_mov_reg_membase (code, i, AMD64_RSP, save_offset, 8); save_offset += 8; } amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, cfg->arch.stack_alloc_size); } else { for (i = 0; i < AMD64_NREG; ++i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) pos -= sizeof (gpointer); if (pos) amd64_lea_membase (code, AMD64_RSP, AMD64_RBP, pos); /* Pop registers in reverse order */ for (i = AMD64_NREG - 1; i > 0; --i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { amd64_pop_reg (code, i); } amd64_leave (code); } offset = code - cfg->native_code; mono_add_patch_info (cfg, code - cfg->native_code, MONO_PATCH_INFO_METHOD_JUMP, ins->inst_p0); if (cfg->compile_aot) amd64_mov_reg_membase (code, AMD64_R11, AMD64_RIP, 0, 8); else amd64_set_reg_template (code, AMD64_R11); amd64_jump_reg (code, AMD64_R11); break; } case OP_CHECK_THIS: /* ensure ins->sreg1 is not NULL */ amd64_alu_membase_imm_size (code, X86_CMP, ins->sreg1, 0, 0, 4); break; case OP_ARGLIST: { amd64_lea_membase (code, AMD64_R11, cfg->frame_reg, cfg->sig_cookie); amd64_mov_membase_reg (code, ins->sreg1, 0, AMD64_R11, 8); break; } case OP_CALL: case OP_FCALL: case OP_LCALL: case OP_VCALL: case OP_VCALL2: case OP_VOIDCALL: call = (MonoCallInst*)ins; /* * The AMD64 ABI forces callers to know about varargs. */ if ((call->signature->call_convention == MONO_CALL_VARARG) && (call->signature->pinvoke)) amd64_alu_reg_reg (code, X86_XOR, AMD64_RAX, AMD64_RAX); else if ((cfg->method->wrapper_type == MONO_WRAPPER_MANAGED_TO_NATIVE) && (cfg->method->klass->image != mono_defaults.corlib)) { /* * Since the unmanaged calling convention doesn't contain a * 'vararg' entry, we have to treat every pinvoke call as a * potential vararg call. */ guint32 nregs, i; nregs = 0; for (i = 0; i < AMD64_XMM_NREG; ++i) if (call->used_fregs & (1 << i)) nregs ++; if (!nregs) amd64_alu_reg_reg (code, X86_XOR, AMD64_RAX, AMD64_RAX); else amd64_mov_reg_imm (code, AMD64_RAX, nregs); } if (ins->flags & MONO_INST_HAS_METHOD) code = emit_call (cfg, code, MONO_PATCH_INFO_METHOD, call->method, FALSE); else code = emit_call (cfg, code, MONO_PATCH_INFO_ABS, call->fptr, FALSE); if (call->stack_usage && !CALLCONV_IS_STDCALL (call->signature->call_convention) && !cfg->arch.no_pushes) amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, call->stack_usage); code = emit_move_return_value (cfg, ins, code); break; case OP_FCALL_REG: case OP_LCALL_REG: case OP_VCALL_REG: case OP_VCALL2_REG: case OP_VOIDCALL_REG: case OP_CALL_REG: call = (MonoCallInst*)ins; if (AMD64_IS_ARGUMENT_REG (ins->sreg1)) { amd64_mov_reg_reg (code, AMD64_R11, ins->sreg1, 8); ins->sreg1 = AMD64_R11; } /* * The AMD64 ABI forces callers to know about varargs. */ if ((call->signature->call_convention == MONO_CALL_VARARG) && (call->signature->pinvoke)) { if (ins->sreg1 == AMD64_RAX) { amd64_mov_reg_reg (code, AMD64_R11, AMD64_RAX, 8); ins->sreg1 = AMD64_R11; } amd64_alu_reg_reg (code, X86_XOR, AMD64_RAX, AMD64_RAX); } else if ((cfg->method->wrapper_type == MONO_WRAPPER_MANAGED_TO_NATIVE) && (cfg->method->klass->image != mono_defaults.corlib)) { /* * Since the unmanaged calling convention doesn't contain a * 'vararg' entry, we have to treat every pinvoke call as a * potential vararg call. */ guint32 nregs, i; nregs = 0; for (i = 0; i < AMD64_XMM_NREG; ++i) if (call->used_fregs & (1 << i)) nregs ++; if (ins->sreg1 == AMD64_RAX) { amd64_mov_reg_reg (code, AMD64_R11, AMD64_RAX, 8); ins->sreg1 = AMD64_R11; } if (!nregs) amd64_alu_reg_reg (code, X86_XOR, AMD64_RAX, AMD64_RAX); else amd64_mov_reg_imm (code, AMD64_RAX, nregs); } amd64_call_reg (code, ins->sreg1); if (call->stack_usage && !CALLCONV_IS_STDCALL (call->signature->call_convention) && !cfg->arch.no_pushes) amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, call->stack_usage); code = emit_move_return_value (cfg, ins, code); break; case OP_FCALL_MEMBASE: case OP_LCALL_MEMBASE: case OP_VCALL_MEMBASE: case OP_VCALL2_MEMBASE: case OP_VOIDCALL_MEMBASE: case OP_CALL_MEMBASE: call = (MonoCallInst*)ins; if (AMD64_IS_ARGUMENT_REG (ins->sreg1)) { /* * Can't use R11 because it is clobbered by the trampoline * code, and the reg value is needed by get_vcall_slot_addr. */ amd64_mov_reg_reg (code, AMD64_RAX, ins->sreg1, 8); ins->sreg1 = AMD64_RAX; } /* * Emit a few nops to simplify get_vcall_slot (). */ amd64_nop (code); amd64_nop (code); amd64_nop (code); amd64_call_membase (code, ins->sreg1, ins->inst_offset); if (call->stack_usage && !CALLCONV_IS_STDCALL (call->signature->call_convention) && !cfg->arch.no_pushes) amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, call->stack_usage); code = emit_move_return_value (cfg, ins, code); break; case OP_DYN_CALL: { int i; MonoInst *var = cfg->dyn_call_var; g_assert (var->opcode == OP_REGOFFSET); /* r11 = args buffer filled by mono_arch_get_dyn_call_args () */ amd64_mov_reg_reg (code, AMD64_R11, ins->sreg1, 8); /* r10 = ftn */ amd64_mov_reg_reg (code, AMD64_R10, ins->sreg2, 8); /* Save args buffer */ amd64_mov_membase_reg (code, var->inst_basereg, var->inst_offset, AMD64_R11, 8); /* Set argument registers */ for (i = 0; i < PARAM_REGS; ++i) amd64_mov_reg_membase (code, param_regs [i], AMD64_R11, i * sizeof (gpointer), 8); /* Make the call */ amd64_call_reg (code, AMD64_R10); /* Save result */ amd64_mov_reg_membase (code, AMD64_R11, var->inst_basereg, var->inst_offset, 8); amd64_mov_membase_reg (code, AMD64_R11, G_STRUCT_OFFSET (DynCallArgs, res), AMD64_RAX, 8); break; } case OP_AMD64_SAVE_SP_TO_LMF: amd64_mov_membase_reg (code, cfg->frame_reg, cfg->arch.lmf_offset + G_STRUCT_OFFSET (MonoLMF, rsp), AMD64_RSP, 8); break; case OP_X86_PUSH: g_assert (!cfg->arch.no_pushes); amd64_push_reg (code, ins->sreg1); break; case OP_X86_PUSH_IMM: g_assert (!cfg->arch.no_pushes); g_assert (amd64_is_imm32 (ins->inst_imm)); amd64_push_imm (code, ins->inst_imm); break; case OP_X86_PUSH_MEMBASE: g_assert (!cfg->arch.no_pushes); amd64_push_membase (code, ins->inst_basereg, ins->inst_offset); break; case OP_X86_PUSH_OBJ: { int size = ALIGN_TO (ins->inst_imm, 8); g_assert (!cfg->arch.no_pushes); amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, size); amd64_push_reg (code, AMD64_RDI); amd64_push_reg (code, AMD64_RSI); amd64_push_reg (code, AMD64_RCX); if (ins->inst_offset) amd64_lea_membase (code, AMD64_RSI, ins->inst_basereg, ins->inst_offset); else amd64_mov_reg_reg (code, AMD64_RSI, ins->inst_basereg, 8); amd64_lea_membase (code, AMD64_RDI, AMD64_RSP, (3 * 8)); amd64_mov_reg_imm (code, AMD64_RCX, (size >> 3)); amd64_cld (code); amd64_prefix (code, X86_REP_PREFIX); amd64_movsd (code); amd64_pop_reg (code, AMD64_RCX); amd64_pop_reg (code, AMD64_RSI); amd64_pop_reg (code, AMD64_RDI); break; } case OP_X86_LEA: amd64_lea_memindex (code, ins->dreg, ins->sreg1, ins->inst_imm, ins->sreg2, ins->backend.shift_amount); break; case OP_X86_LEA_MEMBASE: amd64_lea_membase (code, ins->dreg, ins->sreg1, ins->inst_imm); break; case OP_X86_XCHG: amd64_xchg_reg_reg (code, ins->sreg1, ins->sreg2, 4); break; case OP_LOCALLOC: /* keep alignment */ amd64_alu_reg_imm (code, X86_ADD, ins->sreg1, MONO_ARCH_FRAME_ALIGNMENT - 1); amd64_alu_reg_imm (code, X86_AND, ins->sreg1, ~(MONO_ARCH_FRAME_ALIGNMENT - 1)); code = mono_emit_stack_alloc (cfg, code, ins); amd64_mov_reg_reg (code, ins->dreg, AMD64_RSP, 8); if (cfg->param_area && cfg->arch.no_pushes) amd64_alu_reg_imm (code, X86_ADD, ins->dreg, cfg->param_area); break; case OP_LOCALLOC_IMM: { guint32 size = ins->inst_imm; size = (size + (MONO_ARCH_FRAME_ALIGNMENT - 1)) & ~ (MONO_ARCH_FRAME_ALIGNMENT - 1); if (ins->flags & MONO_INST_INIT) { if (size < 64) { int i; amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, size); amd64_alu_reg_reg (code, X86_XOR, ins->dreg, ins->dreg); for (i = 0; i < size; i += 8) amd64_mov_membase_reg (code, AMD64_RSP, i, ins->dreg, 8); amd64_mov_reg_reg (code, ins->dreg, AMD64_RSP, 8); } else { amd64_mov_reg_imm (code, ins->dreg, size); ins->sreg1 = ins->dreg; code = mono_emit_stack_alloc (cfg, code, ins); amd64_mov_reg_reg (code, ins->dreg, AMD64_RSP, 8); } } else { amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, size); amd64_mov_reg_reg (code, ins->dreg, AMD64_RSP, 8); } if (cfg->param_area && cfg->arch.no_pushes) amd64_alu_reg_imm (code, X86_ADD, ins->dreg, cfg->param_area); break; } case OP_THROW: { amd64_mov_reg_reg (code, AMD64_ARG_REG1, ins->sreg1, 8); code = emit_call (cfg, code, MONO_PATCH_INFO_INTERNAL_METHOD, (gpointer)"mono_arch_throw_exception", FALSE); break; } case OP_RETHROW: { amd64_mov_reg_reg (code, AMD64_ARG_REG1, ins->sreg1, 8); code = emit_call (cfg, code, MONO_PATCH_INFO_INTERNAL_METHOD, (gpointer)"mono_arch_rethrow_exception", FALSE); break; } case OP_CALL_HANDLER: /* Align stack */ amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 8); mono_add_patch_info (cfg, code - cfg->native_code, MONO_PATCH_INFO_BB, ins->inst_target_bb); amd64_call_imm (code, 0); /* Restore stack alignment */ amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, 8); break; case OP_START_HANDLER: { MonoInst *spvar = mono_find_spvar_for_region (cfg, bb->region); amd64_mov_membase_reg (code, spvar->inst_basereg, spvar->inst_offset, AMD64_RSP, 8); if ((MONO_BBLOCK_IS_IN_REGION (bb, MONO_REGION_FINALLY) || MONO_BBLOCK_IS_IN_REGION (bb, MONO_REGION_FINALLY)) && cfg->param_area && cfg->arch.no_pushes) { amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, ALIGN_TO (cfg->param_area, MONO_ARCH_FRAME_ALIGNMENT)); } break; } case OP_ENDFINALLY: { MonoInst *spvar = mono_find_spvar_for_region (cfg, bb->region); amd64_mov_reg_membase (code, AMD64_RSP, spvar->inst_basereg, spvar->inst_offset, 8); amd64_ret (code); break; } case OP_ENDFILTER: { MonoInst *spvar = mono_find_spvar_for_region (cfg, bb->region); amd64_mov_reg_membase (code, AMD64_RSP, spvar->inst_basereg, spvar->inst_offset, 8); /* The local allocator will put the result into RAX */ amd64_ret (code); break; } case OP_LABEL: ins->inst_c0 = code - cfg->native_code; break; case OP_BR: //g_print ("target: %p, next: %p, curr: %p, last: %p\n", ins->inst_target_bb, bb->next_bb, ins, bb->last_ins); //if ((ins->inst_target_bb == bb->next_bb) && ins == bb->last_ins) //break; if (ins->inst_target_bb->native_offset) { amd64_jump_code (code, cfg->native_code + ins->inst_target_bb->native_offset); } else { mono_add_patch_info (cfg, offset, MONO_PATCH_INFO_BB, ins->inst_target_bb); if ((cfg->opt & MONO_OPT_BRANCH) && x86_is_imm8 (ins->inst_target_bb->max_offset - offset)) x86_jump8 (code, 0); else x86_jump32 (code, 0); } break; case OP_BR_REG: amd64_jump_reg (code, ins->sreg1); break; case OP_CEQ: case OP_LCEQ: case OP_ICEQ: case OP_CLT: case OP_LCLT: case OP_ICLT: case OP_CGT: case OP_ICGT: case OP_LCGT: case OP_CLT_UN: case OP_LCLT_UN: case OP_ICLT_UN: case OP_CGT_UN: case OP_LCGT_UN: case OP_ICGT_UN: amd64_set_reg (code, cc_table [mono_opcode_to_cond (ins->opcode)], ins->dreg, cc_signed_table [mono_opcode_to_cond (ins->opcode)]); amd64_widen_reg (code, ins->dreg, ins->dreg, FALSE, FALSE); break; case OP_COND_EXC_EQ: case OP_COND_EXC_NE_UN: case OP_COND_EXC_LT: case OP_COND_EXC_LT_UN: case OP_COND_EXC_GT: case OP_COND_EXC_GT_UN: case OP_COND_EXC_GE: case OP_COND_EXC_GE_UN: case OP_COND_EXC_LE: case OP_COND_EXC_LE_UN: case OP_COND_EXC_IEQ: case OP_COND_EXC_INE_UN: case OP_COND_EXC_ILT: case OP_COND_EXC_ILT_UN: case OP_COND_EXC_IGT: case OP_COND_EXC_IGT_UN: case OP_COND_EXC_IGE: case OP_COND_EXC_IGE_UN: case OP_COND_EXC_ILE: case OP_COND_EXC_ILE_UN: EMIT_COND_SYSTEM_EXCEPTION (cc_table [mono_opcode_to_cond (ins->opcode)], cc_signed_table [mono_opcode_to_cond (ins->opcode)], ins->inst_p1); break; case OP_COND_EXC_OV: case OP_COND_EXC_NO: case OP_COND_EXC_C: case OP_COND_EXC_NC: EMIT_COND_SYSTEM_EXCEPTION (branch_cc_table [ins->opcode - OP_COND_EXC_EQ], (ins->opcode < OP_COND_EXC_NE_UN), ins->inst_p1); break; case OP_COND_EXC_IOV: case OP_COND_EXC_INO: case OP_COND_EXC_IC: case OP_COND_EXC_INC: EMIT_COND_SYSTEM_EXCEPTION (branch_cc_table [ins->opcode - OP_COND_EXC_IEQ], (ins->opcode < OP_COND_EXC_INE_UN), ins->inst_p1); break; /* floating point opcodes */ case OP_R8CONST: { double d = *(double *)ins->inst_p0; if ((d == 0.0) && (mono_signbit (d) == 0)) { amd64_sse_xorpd_reg_reg (code, ins->dreg, ins->dreg); } else { mono_add_patch_info (cfg, offset, MONO_PATCH_INFO_R8, ins->inst_p0); amd64_sse_movsd_reg_membase (code, ins->dreg, AMD64_RIP, 0); } break; } case OP_R4CONST: { float f = *(float *)ins->inst_p0; if ((f == 0.0) && (mono_signbit (f) == 0)) { amd64_sse_xorpd_reg_reg (code, ins->dreg, ins->dreg); } else { mono_add_patch_info (cfg, offset, MONO_PATCH_INFO_R4, ins->inst_p0); amd64_sse_movss_reg_membase (code, ins->dreg, AMD64_RIP, 0); amd64_sse_cvtss2sd_reg_reg (code, ins->dreg, ins->dreg); } break; } case OP_STORER8_MEMBASE_REG: amd64_sse_movsd_membase_reg (code, ins->inst_destbasereg, ins->inst_offset, ins->sreg1); break; case OP_LOADR8_MEMBASE: amd64_sse_movsd_reg_membase (code, ins->dreg, ins->inst_basereg, ins->inst_offset); break; case OP_STORER4_MEMBASE_REG: /* This requires a double->single conversion */ amd64_sse_cvtsd2ss_reg_reg (code, AMD64_XMM15, ins->sreg1); amd64_sse_movss_membase_reg (code, ins->inst_destbasereg, ins->inst_offset, AMD64_XMM15); break; case OP_LOADR4_MEMBASE: amd64_sse_movss_reg_membase (code, ins->dreg, ins->inst_basereg, ins->inst_offset); amd64_sse_cvtss2sd_reg_reg (code, ins->dreg, ins->dreg); break; case OP_ICONV_TO_R4: /* FIXME: change precision */ case OP_ICONV_TO_R8: amd64_sse_cvtsi2sd_reg_reg_size (code, ins->dreg, ins->sreg1, 4); break; case OP_LCONV_TO_R4: /* FIXME: change precision */ case OP_LCONV_TO_R8: amd64_sse_cvtsi2sd_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_FCONV_TO_R4: /* FIXME: nothing to do ?? */ break; case OP_FCONV_TO_I1: code = emit_float_to_int (cfg, code, ins->dreg, ins->sreg1, 1, TRUE); break; case OP_FCONV_TO_U1: code = emit_float_to_int (cfg, code, ins->dreg, ins->sreg1, 1, FALSE); break; case OP_FCONV_TO_I2: code = emit_float_to_int (cfg, code, ins->dreg, ins->sreg1, 2, TRUE); break; case OP_FCONV_TO_U2: code = emit_float_to_int (cfg, code, ins->dreg, ins->sreg1, 2, FALSE); break; case OP_FCONV_TO_U4: code = emit_float_to_int (cfg, code, ins->dreg, ins->sreg1, 4, FALSE); break; case OP_FCONV_TO_I4: case OP_FCONV_TO_I: code = emit_float_to_int (cfg, code, ins->dreg, ins->sreg1, 4, TRUE); break; case OP_FCONV_TO_I8: code = emit_float_to_int (cfg, code, ins->dreg, ins->sreg1, 8, TRUE); break; case OP_LCONV_TO_R_UN: { guint8 *br [2]; /* Based on gcc code */ amd64_test_reg_reg (code, ins->sreg1, ins->sreg1); br [0] = code; x86_branch8 (code, X86_CC_S, 0, TRUE); /* Positive case */ amd64_sse_cvtsi2sd_reg_reg (code, ins->dreg, ins->sreg1); br [1] = code; x86_jump8 (code, 0); amd64_patch (br [0], code); /* Negative case */ /* Save to the red zone */ amd64_mov_membase_reg (code, AMD64_RSP, -8, AMD64_RAX, 8); amd64_mov_membase_reg (code, AMD64_RSP, -16, AMD64_RCX, 8); amd64_mov_reg_reg (code, AMD64_RCX, ins->sreg1, 8); amd64_mov_reg_reg (code, AMD64_RAX, ins->sreg1, 8); amd64_alu_reg_imm (code, X86_AND, AMD64_RCX, 1); amd64_shift_reg_imm (code, X86_SHR, AMD64_RAX, 1); amd64_alu_reg_imm (code, X86_OR, AMD64_RAX, AMD64_RCX); amd64_sse_cvtsi2sd_reg_reg (code, ins->dreg, AMD64_RAX); amd64_sse_addsd_reg_reg (code, ins->dreg, ins->dreg); /* Restore */ amd64_mov_reg_membase (code, AMD64_RCX, AMD64_RSP, -16, 8); amd64_mov_reg_membase (code, AMD64_RAX, AMD64_RSP, -8, 8); amd64_patch (br [1], code); break; } case OP_LCONV_TO_OVF_U4: amd64_alu_reg_imm (code, X86_CMP, ins->sreg1, 0); EMIT_COND_SYSTEM_EXCEPTION (X86_CC_LT, TRUE, "OverflowException"); amd64_mov_reg_reg (code, ins->dreg, ins->sreg1, 8); break; case OP_LCONV_TO_OVF_I4_UN: amd64_alu_reg_imm (code, X86_CMP, ins->sreg1, 0x7fffffff); EMIT_COND_SYSTEM_EXCEPTION (X86_CC_GT, FALSE, "OverflowException"); amd64_mov_reg_reg (code, ins->dreg, ins->sreg1, 8); break; case OP_FMOVE: if (ins->dreg != ins->sreg1) amd64_sse_movsd_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_FADD: amd64_sse_addsd_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_FSUB: amd64_sse_subsd_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_FMUL: amd64_sse_mulsd_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_FDIV: amd64_sse_divsd_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_FNEG: { static double r8_0 = -0.0; g_assert (ins->sreg1 == ins->dreg); mono_add_patch_info (cfg, offset, MONO_PATCH_INFO_R8, &r8_0); amd64_sse_xorpd_reg_membase (code, ins->dreg, AMD64_RIP, 0); break; } case OP_SIN: EMIT_SSE2_FPFUNC (code, fsin, ins->dreg, ins->sreg1); break; case OP_COS: EMIT_SSE2_FPFUNC (code, fcos, ins->dreg, ins->sreg1); break; case OP_ABS: { static guint64 d = 0x7fffffffffffffffUL; g_assert (ins->sreg1 == ins->dreg); mono_add_patch_info (cfg, offset, MONO_PATCH_INFO_R8, &d); amd64_sse_andpd_reg_membase (code, ins->dreg, AMD64_RIP, 0); break; } case OP_SQRT: EMIT_SSE2_FPFUNC (code, fsqrt, ins->dreg, ins->sreg1); break; case OP_IMIN: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg_size (code, X86_CMP, ins->sreg1, ins->sreg2, 4); amd64_cmov_reg_size (code, X86_CC_GT, TRUE, ins->dreg, ins->sreg2, 4); break; case OP_IMIN_UN: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg_size (code, X86_CMP, ins->sreg1, ins->sreg2, 4); amd64_cmov_reg_size (code, X86_CC_GT, FALSE, ins->dreg, ins->sreg2, 4); break; case OP_IMAX: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg_size (code, X86_CMP, ins->sreg1, ins->sreg2, 4); amd64_cmov_reg_size (code, X86_CC_LT, TRUE, ins->dreg, ins->sreg2, 4); break; case OP_IMAX_UN: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg_size (code, X86_CMP, ins->sreg1, ins->sreg2, 4); amd64_cmov_reg_size (code, X86_CC_LT, FALSE, ins->dreg, ins->sreg2, 4); break; case OP_LMIN: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg (code, X86_CMP, ins->sreg1, ins->sreg2); amd64_cmov_reg (code, X86_CC_GT, TRUE, ins->dreg, ins->sreg2); break; case OP_LMIN_UN: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg (code, X86_CMP, ins->sreg1, ins->sreg2); amd64_cmov_reg (code, X86_CC_GT, FALSE, ins->dreg, ins->sreg2); break; case OP_LMAX: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg (code, X86_CMP, ins->sreg1, ins->sreg2); amd64_cmov_reg (code, X86_CC_LT, TRUE, ins->dreg, ins->sreg2); break; case OP_LMAX_UN: g_assert (cfg->opt & MONO_OPT_CMOV); g_assert (ins->dreg == ins->sreg1); amd64_alu_reg_reg (code, X86_CMP, ins->sreg1, ins->sreg2); amd64_cmov_reg (code, X86_CC_LT, FALSE, ins->dreg, ins->sreg2); break; case OP_X86_FPOP: break; case OP_FCOMPARE: /* * The two arguments are swapped because the fbranch instructions * depend on this for the non-sse case to work. */ amd64_sse_comisd_reg_reg (code, ins->sreg2, ins->sreg1); break; case OP_FCEQ: { /* zeroing the register at the start results in * shorter and faster code (we can also remove the widening op) */ guchar *unordered_check; amd64_alu_reg_reg (code, X86_XOR, ins->dreg, ins->dreg); amd64_sse_comisd_reg_reg (code, ins->sreg1, ins->sreg2); unordered_check = code; x86_branch8 (code, X86_CC_P, 0, FALSE); amd64_set_reg (code, X86_CC_EQ, ins->dreg, FALSE); amd64_patch (unordered_check, code); break; } case OP_FCLT: case OP_FCLT_UN: /* zeroing the register at the start results in * shorter and faster code (we can also remove the widening op) */ amd64_alu_reg_reg (code, X86_XOR, ins->dreg, ins->dreg); amd64_sse_comisd_reg_reg (code, ins->sreg2, ins->sreg1); if (ins->opcode == OP_FCLT_UN) { guchar *unordered_check = code; guchar *jump_to_end; x86_branch8 (code, X86_CC_P, 0, FALSE); amd64_set_reg (code, X86_CC_GT, ins->dreg, FALSE); jump_to_end = code; x86_jump8 (code, 0); amd64_patch (unordered_check, code); amd64_inc_reg (code, ins->dreg); amd64_patch (jump_to_end, code); } else { amd64_set_reg (code, X86_CC_GT, ins->dreg, FALSE); } break; case OP_FCGT: case OP_FCGT_UN: { /* zeroing the register at the start results in * shorter and faster code (we can also remove the widening op) */ guchar *unordered_check; amd64_alu_reg_reg (code, X86_XOR, ins->dreg, ins->dreg); amd64_sse_comisd_reg_reg (code, ins->sreg2, ins->sreg1); if (ins->opcode == OP_FCGT) { unordered_check = code; x86_branch8 (code, X86_CC_P, 0, FALSE); amd64_set_reg (code, X86_CC_LT, ins->dreg, FALSE); amd64_patch (unordered_check, code); } else { amd64_set_reg (code, X86_CC_LT, ins->dreg, FALSE); } break; } case OP_FCLT_MEMBASE: case OP_FCGT_MEMBASE: case OP_FCLT_UN_MEMBASE: case OP_FCGT_UN_MEMBASE: case OP_FCEQ_MEMBASE: { guchar *unordered_check, *jump_to_end; int x86_cond; amd64_alu_reg_reg (code, X86_XOR, ins->dreg, ins->dreg); amd64_sse_comisd_reg_membase (code, ins->sreg1, ins->sreg2, ins->inst_offset); switch (ins->opcode) { case OP_FCEQ_MEMBASE: x86_cond = X86_CC_EQ; break; case OP_FCLT_MEMBASE: case OP_FCLT_UN_MEMBASE: x86_cond = X86_CC_LT; break; case OP_FCGT_MEMBASE: case OP_FCGT_UN_MEMBASE: x86_cond = X86_CC_GT; break; default: g_assert_not_reached (); } unordered_check = code; x86_branch8 (code, X86_CC_P, 0, FALSE); amd64_set_reg (code, x86_cond, ins->dreg, FALSE); switch (ins->opcode) { case OP_FCEQ_MEMBASE: case OP_FCLT_MEMBASE: case OP_FCGT_MEMBASE: amd64_patch (unordered_check, code); break; case OP_FCLT_UN_MEMBASE: case OP_FCGT_UN_MEMBASE: jump_to_end = code; x86_jump8 (code, 0); amd64_patch (unordered_check, code); amd64_inc_reg (code, ins->dreg); amd64_patch (jump_to_end, code); break; default: break; } break; } case OP_FBEQ: { guchar *jump = code; x86_branch8 (code, X86_CC_P, 0, TRUE); EMIT_COND_BRANCH (ins, X86_CC_EQ, FALSE); amd64_patch (jump, code); break; } case OP_FBNE_UN: /* Branch if C013 != 100 */ /* branch if !ZF or (PF|CF) */ EMIT_COND_BRANCH (ins, X86_CC_NE, FALSE); EMIT_COND_BRANCH (ins, X86_CC_P, FALSE); EMIT_COND_BRANCH (ins, X86_CC_B, FALSE); break; case OP_FBLT: EMIT_COND_BRANCH (ins, X86_CC_GT, FALSE); break; case OP_FBLT_UN: EMIT_COND_BRANCH (ins, X86_CC_P, FALSE); EMIT_COND_BRANCH (ins, X86_CC_GT, FALSE); break; case OP_FBGT: case OP_FBGT_UN: if (ins->opcode == OP_FBGT) { guchar *br1; /* skip branch if C1=1 */ br1 = code; x86_branch8 (code, X86_CC_P, 0, FALSE); /* branch if (C0 | C3) = 1 */ EMIT_COND_BRANCH (ins, X86_CC_LT, FALSE); amd64_patch (br1, code); break; } else { EMIT_COND_BRANCH (ins, X86_CC_LT, FALSE); } break; case OP_FBGE: { /* Branch if C013 == 100 or 001 */ guchar *br1; /* skip branch if C1=1 */ br1 = code; x86_branch8 (code, X86_CC_P, 0, FALSE); /* branch if (C0 | C3) = 1 */ EMIT_COND_BRANCH (ins, X86_CC_BE, FALSE); amd64_patch (br1, code); break; } case OP_FBGE_UN: /* Branch if C013 == 000 */ EMIT_COND_BRANCH (ins, X86_CC_LE, FALSE); break; case OP_FBLE: { /* Branch if C013=000 or 100 */ guchar *br1; /* skip branch if C1=1 */ br1 = code; x86_branch8 (code, X86_CC_P, 0, FALSE); /* branch if C0=0 */ EMIT_COND_BRANCH (ins, X86_CC_NB, FALSE); amd64_patch (br1, code); break; } case OP_FBLE_UN: /* Branch if C013 != 001 */ EMIT_COND_BRANCH (ins, X86_CC_P, FALSE); EMIT_COND_BRANCH (ins, X86_CC_GE, FALSE); break; case OP_CKFINITE: /* Transfer value to the fp stack */ amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 16); amd64_movsd_membase_reg (code, AMD64_RSP, 0, ins->sreg1); amd64_fld_membase (code, AMD64_RSP, 0, TRUE); amd64_push_reg (code, AMD64_RAX); amd64_fxam (code); amd64_fnstsw (code); amd64_alu_reg_imm (code, X86_AND, AMD64_RAX, 0x4100); amd64_alu_reg_imm (code, X86_CMP, AMD64_RAX, X86_FP_C0); amd64_pop_reg (code, AMD64_RAX); amd64_fstp (code, 0); EMIT_COND_SYSTEM_EXCEPTION (X86_CC_EQ, FALSE, "ArithmeticException"); amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, 16); break; case OP_TLS_GET: { code = mono_amd64_emit_tls_get (code, ins->dreg, ins->inst_offset); break; } case OP_MEMORY_BARRIER: { /* Not needed on amd64 */ break; } case OP_ATOMIC_ADD_I4: case OP_ATOMIC_ADD_I8: { int dreg = ins->dreg; guint32 size = (ins->opcode == OP_ATOMIC_ADD_I4) ? 4 : 8; if (dreg == ins->inst_basereg) dreg = AMD64_R11; if (dreg != ins->sreg2) amd64_mov_reg_reg (code, ins->dreg, ins->sreg2, size); x86_prefix (code, X86_LOCK_PREFIX); amd64_xadd_membase_reg (code, ins->inst_basereg, ins->inst_offset, dreg, size); if (dreg != ins->dreg) amd64_mov_reg_reg (code, ins->dreg, dreg, size); break; } case OP_ATOMIC_ADD_NEW_I4: case OP_ATOMIC_ADD_NEW_I8: { int dreg = ins->dreg; guint32 size = (ins->opcode == OP_ATOMIC_ADD_NEW_I4) ? 4 : 8; if ((dreg == ins->sreg2) || (dreg == ins->inst_basereg)) dreg = AMD64_R11; amd64_mov_reg_reg (code, dreg, ins->sreg2, size); amd64_prefix (code, X86_LOCK_PREFIX); amd64_xadd_membase_reg (code, ins->inst_basereg, ins->inst_offset, dreg, size); /* dreg contains the old value, add with sreg2 value */ amd64_alu_reg_reg_size (code, X86_ADD, dreg, ins->sreg2, size); if (ins->dreg != dreg) amd64_mov_reg_reg (code, ins->dreg, dreg, size); break; } case OP_ATOMIC_EXCHANGE_I4: case OP_ATOMIC_EXCHANGE_I8: { guchar *br[2]; int sreg2 = ins->sreg2; int breg = ins->inst_basereg; guint32 size; gboolean need_push = FALSE, rdx_pushed = FALSE; if (ins->opcode == OP_ATOMIC_EXCHANGE_I8) size = 8; else size = 4; /* * See http://msdn.microsoft.com/en-us/magazine/cc302329.aspx for * an explanation of how this works. */ /* cmpxchg uses eax as comperand, need to make sure we can use it * hack to overcome limits in x86 reg allocator * (req: dreg == eax and sreg2 != eax and breg != eax) */ g_assert (ins->dreg == AMD64_RAX); if (breg == AMD64_RAX && ins->sreg2 == AMD64_RAX) /* Highly unlikely, but possible */ need_push = TRUE; /* The pushes invalidate rsp */ if ((breg == AMD64_RAX) || need_push) { amd64_mov_reg_reg (code, AMD64_R11, breg, 8); breg = AMD64_R11; } /* We need the EAX reg for the comparand */ if (ins->sreg2 == AMD64_RAX) { if (breg != AMD64_R11) { amd64_mov_reg_reg (code, AMD64_R11, AMD64_RAX, 8); sreg2 = AMD64_R11; } else { g_assert (need_push); amd64_push_reg (code, AMD64_RDX); amd64_mov_reg_reg (code, AMD64_RDX, AMD64_RAX, size); sreg2 = AMD64_RDX; rdx_pushed = TRUE; } } amd64_mov_reg_membase (code, AMD64_RAX, breg, ins->inst_offset, size); br [0] = code; amd64_prefix (code, X86_LOCK_PREFIX); amd64_cmpxchg_membase_reg_size (code, breg, ins->inst_offset, sreg2, size); br [1] = code; amd64_branch8 (code, X86_CC_NE, -1, FALSE); amd64_patch (br [1], br [0]); if (rdx_pushed) amd64_pop_reg (code, AMD64_RDX); break; } case OP_ATOMIC_CAS_I4: case OP_ATOMIC_CAS_I8: { guint32 size; if (ins->opcode == OP_ATOMIC_CAS_I8) size = 8; else size = 4; /* * See http://msdn.microsoft.com/en-us/magazine/cc302329.aspx for * an explanation of how this works. */ g_assert (ins->sreg3 == AMD64_RAX); g_assert (ins->sreg1 != AMD64_RAX); g_assert (ins->sreg1 != ins->sreg2); amd64_prefix (code, X86_LOCK_PREFIX); amd64_cmpxchg_membase_reg_size (code, ins->sreg1, ins->inst_offset, ins->sreg2, size); if (ins->dreg != AMD64_RAX) amd64_mov_reg_reg (code, ins->dreg, AMD64_RAX, size); break; } #ifdef MONO_ARCH_SIMD_INTRINSICS /* TODO: Some of these IR opcodes are marked as no clobber when they indeed do. */ case OP_ADDPS: amd64_sse_addps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_DIVPS: amd64_sse_divps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_MULPS: amd64_sse_mulps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_SUBPS: amd64_sse_subps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_MAXPS: amd64_sse_maxps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_MINPS: amd64_sse_minps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_COMPPS: g_assert (ins->inst_c0 >= 0 && ins->inst_c0 <= 7); amd64_sse_cmpps_reg_reg_imm (code, ins->sreg1, ins->sreg2, ins->inst_c0); break; case OP_ANDPS: amd64_sse_andps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_ANDNPS: amd64_sse_andnps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_ORPS: amd64_sse_orps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_XORPS: amd64_sse_xorps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_SQRTPS: amd64_sse_sqrtps_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_RSQRTPS: amd64_sse_rsqrtps_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_RCPPS: amd64_sse_rcpps_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_ADDSUBPS: amd64_sse_addsubps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_HADDPS: amd64_sse_haddps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_HSUBPS: amd64_sse_hsubps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_DUPPS_HIGH: amd64_sse_movshdup_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_DUPPS_LOW: amd64_sse_movsldup_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_PSHUFLEW_HIGH: g_assert (ins->inst_c0 >= 0 && ins->inst_c0 <= 0xFF); amd64_sse_pshufhw_reg_reg_imm (code, ins->dreg, ins->sreg1, ins->inst_c0); break; case OP_PSHUFLEW_LOW: g_assert (ins->inst_c0 >= 0 && ins->inst_c0 <= 0xFF); amd64_sse_pshuflw_reg_reg_imm (code, ins->dreg, ins->sreg1, ins->inst_c0); break; case OP_PSHUFLED: g_assert (ins->inst_c0 >= 0 && ins->inst_c0 <= 0xFF); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->sreg1, ins->inst_c0); break; case OP_ADDPD: amd64_sse_addpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_DIVPD: amd64_sse_divpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_MULPD: amd64_sse_mulpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_SUBPD: amd64_sse_subpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_MAXPD: amd64_sse_maxpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_MINPD: amd64_sse_minpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_COMPPD: g_assert (ins->inst_c0 >= 0 && ins->inst_c0 <= 7); amd64_sse_cmppd_reg_reg_imm (code, ins->sreg1, ins->sreg2, ins->inst_c0); break; case OP_ANDPD: amd64_sse_andpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_ANDNPD: amd64_sse_andnpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_ORPD: amd64_sse_orpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_XORPD: amd64_sse_xorpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_SQRTPD: amd64_sse_sqrtpd_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_ADDSUBPD: amd64_sse_addsubpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_HADDPD: amd64_sse_haddpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_HSUBPD: amd64_sse_hsubpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_DUPPD: amd64_sse_movddup_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_EXTRACT_MASK: amd64_sse_pmovmskb_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_PAND: amd64_sse_pand_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_POR: amd64_sse_por_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PXOR: amd64_sse_pxor_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDB: amd64_sse_paddb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDW: amd64_sse_paddw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDD: amd64_sse_paddd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDQ: amd64_sse_paddq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBB: amd64_sse_psubb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBW: amd64_sse_psubw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBD: amd64_sse_psubd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBQ: amd64_sse_psubq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMAXB_UN: amd64_sse_pmaxub_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMAXW_UN: amd64_sse_pmaxuw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMAXD_UN: amd64_sse_pmaxud_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMAXB: amd64_sse_pmaxsb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMAXW: amd64_sse_pmaxsw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMAXD: amd64_sse_pmaxsd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PAVGB_UN: amd64_sse_pavgb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PAVGW_UN: amd64_sse_pavgw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMINB_UN: amd64_sse_pminub_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMINW_UN: amd64_sse_pminuw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMIND_UN: amd64_sse_pminud_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMINB: amd64_sse_pminsb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMINW: amd64_sse_pminsw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMIND: amd64_sse_pminsd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPEQB: amd64_sse_pcmpeqb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPEQW: amd64_sse_pcmpeqw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPEQD: amd64_sse_pcmpeqd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPEQQ: amd64_sse_pcmpeqq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPGTB: amd64_sse_pcmpgtb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPGTW: amd64_sse_pcmpgtw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPGTD: amd64_sse_pcmpgtd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PCMPGTQ: amd64_sse_pcmpgtq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUM_ABS_DIFF: amd64_sse_psadbw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_LOWB: amd64_sse_punpcklbw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_LOWW: amd64_sse_punpcklwd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_LOWD: amd64_sse_punpckldq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_LOWQ: amd64_sse_punpcklqdq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_LOWPS: amd64_sse_unpcklps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_LOWPD: amd64_sse_unpcklpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_HIGHB: amd64_sse_punpckhbw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_HIGHW: amd64_sse_punpckhwd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_HIGHD: amd64_sse_punpckhdq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_HIGHQ: amd64_sse_punpckhqdq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_HIGHPS: amd64_sse_unpckhps_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_UNPACK_HIGHPD: amd64_sse_unpckhpd_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PACKW: amd64_sse_packsswb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PACKD: amd64_sse_packssdw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PACKW_UN: amd64_sse_packuswb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PACKD_UN: amd64_sse_packusdw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDB_SAT_UN: amd64_sse_paddusb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBB_SAT_UN: amd64_sse_psubusb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDW_SAT_UN: amd64_sse_paddusw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBW_SAT_UN: amd64_sse_psubusw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDB_SAT: amd64_sse_paddsb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBB_SAT: amd64_sse_psubsb_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PADDW_SAT: amd64_sse_paddsw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSUBW_SAT: amd64_sse_psubsw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMULW: amd64_sse_pmullw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMULD: amd64_sse_pmulld_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMULQ: amd64_sse_pmuludq_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMULW_HIGH_UN: amd64_sse_pmulhuw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PMULW_HIGH: amd64_sse_pmulhw_reg_reg (code, ins->sreg1, ins->sreg2); break; case OP_PSHRW: amd64_sse_psrlw_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSHRW_REG: amd64_sse_psrlw_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_PSARW: amd64_sse_psraw_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSARW_REG: amd64_sse_psraw_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_PSHLW: amd64_sse_psllw_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSHLW_REG: amd64_sse_psllw_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_PSHRD: amd64_sse_psrld_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSHRD_REG: amd64_sse_psrld_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_PSARD: amd64_sse_psrad_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSARD_REG: amd64_sse_psrad_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_PSHLD: amd64_sse_pslld_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSHLD_REG: amd64_sse_pslld_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_PSHRQ: amd64_sse_psrlq_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSHRQ_REG: amd64_sse_psrlq_reg_reg (code, ins->dreg, ins->sreg2); break; /*TODO: This is appart of the sse spec but not added case OP_PSARQ: amd64_sse_psraq_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSARQ_REG: amd64_sse_psraq_reg_reg (code, ins->dreg, ins->sreg2); break; */ case OP_PSHLQ: amd64_sse_psllq_reg_imm (code, ins->dreg, ins->inst_imm); break; case OP_PSHLQ_REG: amd64_sse_psllq_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_ICONV_TO_X: amd64_movd_xreg_reg_size (code, ins->dreg, ins->sreg1, 4); break; case OP_EXTRACT_I4: amd64_movd_reg_xreg_size (code, ins->dreg, ins->sreg1, 4); break; case OP_EXTRACT_I8: if (ins->inst_c0) { amd64_movhlps_reg_reg (code, AMD64_XMM15, ins->sreg1); amd64_movd_reg_xreg_size (code, ins->dreg, AMD64_XMM15, 8); } else { amd64_movd_reg_xreg_size (code, ins->dreg, ins->sreg1, 8); } break; case OP_EXTRACT_I1: case OP_EXTRACT_U1: amd64_movd_reg_xreg_size (code, ins->dreg, ins->sreg1, 4); if (ins->inst_c0) amd64_shift_reg_imm (code, X86_SHR, ins->dreg, ins->inst_c0 * 8); amd64_widen_reg (code, ins->dreg, ins->dreg, ins->opcode == OP_EXTRACT_I1, FALSE); break; case OP_EXTRACT_I2: case OP_EXTRACT_U2: /*amd64_movd_reg_xreg_size (code, ins->dreg, ins->sreg1, 4); if (ins->inst_c0) amd64_shift_reg_imm_size (code, X86_SHR, ins->dreg, 16, 4);*/ amd64_sse_pextrw_reg_reg_imm (code, ins->dreg, ins->sreg1, ins->inst_c0); amd64_widen_reg_size (code, ins->dreg, ins->dreg, ins->opcode == OP_EXTRACT_I2, TRUE, 4); break; case OP_EXTRACT_R8: if (ins->inst_c0) amd64_movhlps_reg_reg (code, ins->dreg, ins->sreg1); else amd64_sse_movsd_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_INSERT_I2: amd64_sse_pinsrw_reg_reg_imm (code, ins->sreg1, ins->sreg2, ins->inst_c0); break; case OP_EXTRACTX_U2: amd64_sse_pextrw_reg_reg_imm (code, ins->dreg, ins->sreg1, ins->inst_c0); break; case OP_INSERTX_U1_SLOW: /*sreg1 is the extracted ireg (scratch) /sreg2 is the to be inserted ireg (scratch) /dreg is the xreg to receive the value*/ /*clear the bits from the extracted word*/ amd64_alu_reg_imm (code, X86_AND, ins->sreg1, ins->inst_c0 & 1 ? 0x00FF : 0xFF00); /*shift the value to insert if needed*/ if (ins->inst_c0 & 1) amd64_shift_reg_imm_size (code, X86_SHL, ins->sreg2, 8, 4); /*join them together*/ amd64_alu_reg_reg (code, X86_OR, ins->sreg1, ins->sreg2); amd64_sse_pinsrw_reg_reg_imm (code, ins->dreg, ins->sreg1, ins->inst_c0 / 2); break; case OP_INSERTX_I4_SLOW: amd64_sse_pinsrw_reg_reg_imm (code, ins->dreg, ins->sreg2, ins->inst_c0 * 2); amd64_shift_reg_imm (code, X86_SHR, ins->sreg2, 16); amd64_sse_pinsrw_reg_reg_imm (code, ins->dreg, ins->sreg2, ins->inst_c0 * 2 + 1); break; case OP_INSERTX_I8_SLOW: amd64_movd_xreg_reg_size(code, AMD64_XMM15, ins->sreg2, 8); if (ins->inst_c0) amd64_movlhps_reg_reg (code, ins->dreg, AMD64_XMM15); else amd64_sse_movsd_reg_reg (code, ins->dreg, AMD64_XMM15); break; case OP_INSERTX_R4_SLOW: switch (ins->inst_c0) { case 0: amd64_sse_cvtsd2ss_reg_reg (code, ins->dreg, ins->sreg2); break; case 1: amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, mono_simd_shuffle_mask(1, 0, 2, 3)); amd64_sse_cvtsd2ss_reg_reg (code, ins->dreg, ins->sreg2); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, mono_simd_shuffle_mask(1, 0, 2, 3)); break; case 2: amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, mono_simd_shuffle_mask(2, 1, 0, 3)); amd64_sse_cvtsd2ss_reg_reg (code, ins->dreg, ins->sreg2); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, mono_simd_shuffle_mask(2, 1, 0, 3)); break; case 3: amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, mono_simd_shuffle_mask(3, 1, 2, 0)); amd64_sse_cvtsd2ss_reg_reg (code, ins->dreg, ins->sreg2); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, mono_simd_shuffle_mask(3, 1, 2, 0)); break; } break; case OP_INSERTX_R8_SLOW: if (ins->inst_c0) amd64_movlhps_reg_reg (code, ins->dreg, ins->sreg2); else amd64_sse_movsd_reg_reg (code, ins->dreg, ins->sreg2); break; case OP_STOREX_MEMBASE_REG: case OP_STOREX_MEMBASE: amd64_sse_movups_membase_reg (code, ins->dreg, ins->inst_offset, ins->sreg1); break; case OP_LOADX_MEMBASE: amd64_sse_movups_reg_membase (code, ins->dreg, ins->sreg1, ins->inst_offset); break; case OP_LOADX_ALIGNED_MEMBASE: amd64_sse_movaps_reg_membase (code, ins->dreg, ins->sreg1, ins->inst_offset); break; case OP_STOREX_ALIGNED_MEMBASE_REG: amd64_sse_movaps_membase_reg (code, ins->dreg, ins->inst_offset, ins->sreg1); break; case OP_STOREX_NTA_MEMBASE_REG: amd64_sse_movntps_reg_membase (code, ins->dreg, ins->sreg1, ins->inst_offset); break; case OP_PREFETCH_MEMBASE: amd64_sse_prefetch_reg_membase (code, ins->backend.arg_info, ins->sreg1, ins->inst_offset); break; case OP_XMOVE: /*FIXME the peephole pass should have killed this*/ if (ins->dreg != ins->sreg1) amd64_sse_movaps_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_XZERO: amd64_sse_pxor_reg_reg (code, ins->dreg, ins->dreg); break; case OP_ICONV_TO_R8_RAW: amd64_movd_xreg_reg_size (code, ins->dreg, ins->sreg1, 4); amd64_sse_cvtss2sd_reg_reg (code, ins->dreg, ins->dreg); break; case OP_FCONV_TO_R8_X: amd64_sse_movsd_reg_reg (code, ins->dreg, ins->sreg1); break; case OP_XCONV_R8_TO_I4: amd64_sse_cvttsd2si_reg_xreg_size (code, ins->dreg, ins->sreg1, 4); switch (ins->backend.source_opcode) { case OP_FCONV_TO_I1: amd64_widen_reg (code, ins->dreg, ins->dreg, TRUE, FALSE); break; case OP_FCONV_TO_U1: amd64_widen_reg (code, ins->dreg, ins->dreg, FALSE, FALSE); break; case OP_FCONV_TO_I2: amd64_widen_reg (code, ins->dreg, ins->dreg, TRUE, TRUE); break; case OP_FCONV_TO_U2: amd64_widen_reg (code, ins->dreg, ins->dreg, FALSE, TRUE); break; } break; case OP_EXPAND_I2: amd64_sse_pinsrw_reg_reg_imm (code, ins->dreg, ins->sreg1, 0); amd64_sse_pinsrw_reg_reg_imm (code, ins->dreg, ins->sreg1, 1); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, 0); break; case OP_EXPAND_I4: amd64_movd_xreg_reg_size (code, ins->dreg, ins->sreg1, 4); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, 0); break; case OP_EXPAND_I8: amd64_movd_xreg_reg_size (code, ins->dreg, ins->sreg1, 8); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, 0x44); break; case OP_EXPAND_R4: amd64_sse_movsd_reg_reg (code, ins->dreg, ins->sreg1); amd64_sse_cvtsd2ss_reg_reg (code, ins->dreg, ins->dreg); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, 0); break; case OP_EXPAND_R8: amd64_sse_movsd_reg_reg (code, ins->dreg, ins->sreg1); amd64_sse_pshufd_reg_reg_imm (code, ins->dreg, ins->dreg, 0x44); break; #endif case OP_LIVERANGE_START: { if (cfg->verbose_level > 1) printf ("R%d START=0x%x\n", MONO_VARINFO (cfg, ins->inst_c0)->vreg, (int)(code - cfg->native_code)); MONO_VARINFO (cfg, ins->inst_c0)->live_range_start = code - cfg->native_code; break; } case OP_LIVERANGE_END: { if (cfg->verbose_level > 1) printf ("R%d END=0x%x\n", MONO_VARINFO (cfg, ins->inst_c0)->vreg, (int)(code - cfg->native_code)); MONO_VARINFO (cfg, ins->inst_c0)->live_range_end = code - cfg->native_code; break; } default: g_warning ("unknown opcode %s in %s()\n", mono_inst_name (ins->opcode), __FUNCTION__); g_assert_not_reached (); } if ((code - cfg->native_code - offset) > max_len) { g_warning ("wrong maximal instruction length of instruction %s (expected %d, got %ld)", mono_inst_name (ins->opcode), max_len, code - cfg->native_code - offset); g_assert_not_reached (); } last_ins = ins; last_offset = offset; } cfg->code_len = code - cfg->native_code; } #endif /* DISABLE_JIT */ void mono_arch_register_lowlevel_calls (void) { /* The signature doesn't matter */ mono_register_jit_icall (mono_amd64_throw_exception, "mono_amd64_throw_exception", mono_create_icall_signature ("void"), TRUE); } void mono_arch_patch_code (MonoMethod *method, MonoDomain *domain, guint8 *code, MonoJumpInfo *ji, gboolean run_cctors) { MonoJumpInfo *patch_info; gboolean compile_aot = !run_cctors; for (patch_info = ji; patch_info; patch_info = patch_info->next) { unsigned char *ip = patch_info->ip.i + code; unsigned char *target; target = mono_resolve_patch_target (method, domain, code, patch_info, run_cctors); if (compile_aot) { switch (patch_info->type) { case MONO_PATCH_INFO_BB: case MONO_PATCH_INFO_LABEL: break; default: /* No need to patch these */ continue; } } switch (patch_info->type) { case MONO_PATCH_INFO_NONE: continue; case MONO_PATCH_INFO_METHOD_REL: case MONO_PATCH_INFO_R8: case MONO_PATCH_INFO_R4: g_assert_not_reached (); continue; case MONO_PATCH_INFO_BB: break; default: break; } /* * Debug code to help track down problems where the target of a near call is * is not valid. */ if (amd64_is_near_call (ip)) { gint64 disp = (guint8*)target - (guint8*)ip; if (!amd64_is_imm32 (disp)) { printf ("TYPE: %d\n", patch_info->type); switch (patch_info->type) { case MONO_PATCH_INFO_INTERNAL_METHOD: printf ("V: %s\n", patch_info->data.name); break; case MONO_PATCH_INFO_METHOD_JUMP: case MONO_PATCH_INFO_METHOD: printf ("V: %s\n", patch_info->data.method->name); break; default: break; } } } amd64_patch (ip, (gpointer)target); } } static int get_max_epilog_size (MonoCompile *cfg) { int max_epilog_size = 16; if (cfg->method->save_lmf) max_epilog_size += 256; if (mono_jit_trace_calls != NULL) max_epilog_size += 50; if (cfg->prof_options & MONO_PROFILE_ENTER_LEAVE) max_epilog_size += 50; max_epilog_size += (AMD64_NREG * 2); return max_epilog_size; } /* * This macro is used for testing whenever the unwinder works correctly at every point * where an async exception can happen. */ /* This will generate a SIGSEGV at the given point in the code */ #define async_exc_point(code) do { \ if (mono_inject_async_exc_method && mono_method_desc_full_match (mono_inject_async_exc_method, cfg->method)) { \ if (cfg->arch.async_point_count == mono_inject_async_exc_pos) \ amd64_mov_reg_mem (code, AMD64_RAX, 0, 4); \ cfg->arch.async_point_count ++; \ } \ } while (0) guint8 * mono_arch_emit_prolog (MonoCompile *cfg) { MonoMethod *method = cfg->method; MonoBasicBlock *bb; MonoMethodSignature *sig; MonoInst *ins; int alloc_size, pos, i, cfa_offset, quad, max_epilog_size; guint8 *code; CallInfo *cinfo; gint32 lmf_offset = cfg->arch.lmf_offset; gboolean args_clobbered = FALSE; gboolean trace = FALSE; cfg->code_size = MAX (((MonoMethodNormal *)method)->header->code_size * 4, 10240); code = cfg->native_code = g_malloc (cfg->code_size); if (mono_jit_trace_calls != NULL && mono_trace_eval (method)) trace = TRUE; /* Amount of stack space allocated by register saving code */ pos = 0; /* Offset between RSP and the CFA */ cfa_offset = 0; /* * The prolog consists of the following parts: * FP present: * - push rbp, mov rbp, rsp * - save callee saved regs using pushes * - allocate frame * - save rgctx if needed * - save lmf if needed * FP not present: * - allocate frame * - save rgctx if needed * - save lmf if needed * - save callee saved regs using moves */ // CFA = sp + 8 cfa_offset = 8; mono_emit_unwind_op_def_cfa (cfg, code, AMD64_RSP, 8); // IP saved at CFA - 8 mono_emit_unwind_op_offset (cfg, code, AMD64_RIP, -cfa_offset); async_exc_point (code); if (!cfg->arch.omit_fp) { amd64_push_reg (code, AMD64_RBP); cfa_offset += 8; mono_emit_unwind_op_def_cfa_offset (cfg, code, cfa_offset); mono_emit_unwind_op_offset (cfg, code, AMD64_RBP, - cfa_offset); async_exc_point (code); #ifdef HOST_WIN32 mono_arch_unwindinfo_add_push_nonvol (&cfg->arch.unwindinfo, cfg->native_code, code, AMD64_RBP); #endif amd64_mov_reg_reg (code, AMD64_RBP, AMD64_RSP, sizeof (gpointer)); mono_emit_unwind_op_def_cfa_reg (cfg, code, AMD64_RBP); async_exc_point (code); #ifdef HOST_WIN32 mono_arch_unwindinfo_add_set_fpreg (&cfg->arch.unwindinfo, cfg->native_code, code, AMD64_RBP); #endif } /* Save callee saved registers */ if (!cfg->arch.omit_fp && !method->save_lmf) { int offset = cfa_offset; for (i = 0; i < AMD64_NREG; ++i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { amd64_push_reg (code, i); pos += sizeof (gpointer); offset += 8; mono_emit_unwind_op_offset (cfg, code, i, - offset); async_exc_point (code); } } /* The param area is always at offset 0 from sp */ /* This needs to be allocated here, since it has to come after the spill area */ if (cfg->arch.no_pushes && cfg->param_area) { if (cfg->arch.omit_fp) // FIXME: g_assert_not_reached (); cfg->stack_offset += ALIGN_TO (cfg->param_area, sizeof (gpointer)); } if (cfg->arch.omit_fp) { /* * On enter, the stack is misaligned by the the pushing of the return * address. It is either made aligned by the pushing of %rbp, or by * this. */ alloc_size = ALIGN_TO (cfg->stack_offset, 8); if ((alloc_size % 16) == 0) alloc_size += 8; } else { alloc_size = ALIGN_TO (cfg->stack_offset, MONO_ARCH_FRAME_ALIGNMENT); alloc_size -= pos; } cfg->arch.stack_alloc_size = alloc_size; /* Allocate stack frame */ if (alloc_size) { /* See mono_emit_stack_alloc */ #if defined(HOST_WIN32) || defined(MONO_ARCH_SIGSEGV_ON_ALTSTACK) guint32 remaining_size = alloc_size; /*FIXME handle unbounded code expansion, we should use a loop in case of more than X interactions*/ guint32 required_code_size = ((remaining_size / 0x1000) + 1) * 10; /*10 is the max size of amd64_alu_reg_imm + amd64_test_membase_reg*/ guint32 offset = code - cfg->native_code; if (G_UNLIKELY (required_code_size >= (cfg->code_size - offset))) { while (required_code_size >= (cfg->code_size - offset)) cfg->code_size *= 2; cfg->native_code = g_realloc (cfg->native_code, cfg->code_size); code = cfg->native_code + offset; mono_jit_stats.code_reallocs++; } while (remaining_size >= 0x1000) { amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 0x1000); if (cfg->arch.omit_fp) { cfa_offset += 0x1000; mono_emit_unwind_op_def_cfa_offset (cfg, code, cfa_offset); } async_exc_point (code); #ifdef HOST_WIN32 if (cfg->arch.omit_fp) mono_arch_unwindinfo_add_alloc_stack (&cfg->arch.unwindinfo, cfg->native_code, code, 0x1000); #endif amd64_test_membase_reg (code, AMD64_RSP, 0, AMD64_RSP); remaining_size -= 0x1000; } if (remaining_size) { amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, remaining_size); if (cfg->arch.omit_fp) { cfa_offset += remaining_size; mono_emit_unwind_op_def_cfa_offset (cfg, code, cfa_offset); async_exc_point (code); } #ifdef HOST_WIN32 if (cfg->arch.omit_fp) mono_arch_unwindinfo_add_alloc_stack (&cfg->arch.unwindinfo, cfg->native_code, code, remaining_size); #endif } #else amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, alloc_size); if (cfg->arch.omit_fp) { cfa_offset += alloc_size; mono_emit_unwind_op_def_cfa_offset (cfg, code, cfa_offset); async_exc_point (code); } #endif } /* Stack alignment check */ #if 0 { amd64_mov_reg_reg (code, AMD64_RAX, AMD64_RSP, 8); amd64_alu_reg_imm (code, X86_AND, AMD64_RAX, 0xf); amd64_alu_reg_imm (code, X86_CMP, AMD64_RAX, 0); x86_branch8 (code, X86_CC_EQ, 2, FALSE); amd64_breakpoint (code); } #endif /* Save LMF */ if (method->save_lmf) { /* * The ip field is not set, the exception handling code will obtain it from the stack location pointed to by the sp field. */ /* sp is saved right before calls */ /* Skip method (only needed for trampoline LMF frames) */ /* Save callee saved regs */ for (i = 0; i < MONO_MAX_IREGS; ++i) { int offset; switch (i) { case AMD64_RBX: offset = G_STRUCT_OFFSET (MonoLMF, rbx); break; case AMD64_RBP: offset = G_STRUCT_OFFSET (MonoLMF, rbp); break; case AMD64_R12: offset = G_STRUCT_OFFSET (MonoLMF, r12); break; case AMD64_R13: offset = G_STRUCT_OFFSET (MonoLMF, r13); break; case AMD64_R14: offset = G_STRUCT_OFFSET (MonoLMF, r14); break; case AMD64_R15: offset = G_STRUCT_OFFSET (MonoLMF, r15); break; #ifdef HOST_WIN32 case AMD64_RDI: offset = G_STRUCT_OFFSET (MonoLMF, rdi); break; case AMD64_RSI: offset = G_STRUCT_OFFSET (MonoLMF, rsi); break; #endif default: offset = -1; break; } if (offset != -1) { amd64_mov_membase_reg (code, cfg->frame_reg, lmf_offset + offset, i, 8); if (cfg->arch.omit_fp || (i != AMD64_RBP)) mono_emit_unwind_op_offset (cfg, code, i, - (cfa_offset - (lmf_offset + offset))); } } } /* Save callee saved registers */ if (cfg->arch.omit_fp && !method->save_lmf) { gint32 save_area_offset = cfg->arch.reg_save_area_offset; /* Save caller saved registers after sp is adjusted */ /* The registers are saved at the bottom of the frame */ /* FIXME: Optimize this so the regs are saved at the end of the frame in increasing order */ for (i = 0; i < AMD64_NREG; ++i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { amd64_mov_membase_reg (code, AMD64_RSP, save_area_offset, i, 8); mono_emit_unwind_op_offset (cfg, code, i, - (cfa_offset - save_area_offset)); save_area_offset += 8; async_exc_point (code); } } /* store runtime generic context */ if (cfg->rgctx_var) { g_assert (cfg->rgctx_var->opcode == OP_REGOFFSET && (cfg->rgctx_var->inst_basereg == AMD64_RBP || cfg->rgctx_var->inst_basereg == AMD64_RSP)); amd64_mov_membase_reg (code, cfg->rgctx_var->inst_basereg, cfg->rgctx_var->inst_offset, MONO_ARCH_RGCTX_REG, 8); } /* compute max_length in order to use short forward jumps */ max_epilog_size = get_max_epilog_size (cfg); if (cfg->opt & MONO_OPT_BRANCH) { for (bb = cfg->bb_entry; bb; bb = bb->next_bb) { MonoInst *ins; int max_length = 0; if (cfg->prof_options & MONO_PROFILE_COVERAGE) max_length += 6; /* max alignment for loops */ if ((cfg->opt & MONO_OPT_LOOP) && bb_is_loop_start (bb)) max_length += LOOP_ALIGNMENT; MONO_BB_FOR_EACH_INS (bb, ins) { max_length += ((guint8 *)ins_get_spec (ins->opcode))[MONO_INST_LEN]; } /* Take prolog and epilog instrumentation into account */ if (bb == cfg->bb_entry || bb == cfg->bb_exit) max_length += max_epilog_size; bb->max_length = max_length; } } sig = mono_method_signature (method); pos = 0; cinfo = cfg->arch.cinfo; if (sig->ret->type != MONO_TYPE_VOID) { /* Save volatile arguments to the stack */ if (cfg->vret_addr && (cfg->vret_addr->opcode != OP_REGVAR)) amd64_mov_membase_reg (code, cfg->vret_addr->inst_basereg, cfg->vret_addr->inst_offset, cinfo->ret.reg, 8); } /* Keep this in sync with emit_load_volatile_arguments */ for (i = 0; i < sig->param_count + sig->hasthis; ++i) { ArgInfo *ainfo = cinfo->args + i; gint32 stack_offset; MonoType *arg_type; ins = cfg->args [i]; if ((ins->flags & MONO_INST_IS_DEAD) && !trace) /* Unused arguments */ continue; if (sig->hasthis && (i == 0)) arg_type = &mono_defaults.object_class->byval_arg; else arg_type = sig->params [i - sig->hasthis]; stack_offset = ainfo->offset + ARGS_OFFSET; if (cfg->globalra) { /* All the other moves are done by the register allocator */ switch (ainfo->storage) { case ArgInFloatSSEReg: amd64_sse_cvtss2sd_reg_reg (code, ainfo->reg, ainfo->reg); break; case ArgValuetypeInReg: for (quad = 0; quad < 2; quad ++) { switch (ainfo->pair_storage [quad]) { case ArgInIReg: amd64_mov_membase_reg (code, ins->inst_basereg, ins->inst_offset + (quad * sizeof (gpointer)), ainfo->pair_regs [quad], sizeof (gpointer)); break; case ArgInFloatSSEReg: amd64_movss_membase_reg (code, ins->inst_basereg, ins->inst_offset + (quad * sizeof (gpointer)), ainfo->pair_regs [quad]); break; case ArgInDoubleSSEReg: amd64_movsd_membase_reg (code, ins->inst_basereg, ins->inst_offset + (quad * sizeof (gpointer)), ainfo->pair_regs [quad]); break; case ArgNone: break; default: g_assert_not_reached (); } } break; default: break; } continue; } /* Save volatile arguments to the stack */ if (ins->opcode != OP_REGVAR) { switch (ainfo->storage) { case ArgInIReg: { guint32 size = 8; /* FIXME: I1 etc */ /* if (stack_offset & 0x1) size = 1; else if (stack_offset & 0x2) size = 2; else if (stack_offset & 0x4) size = 4; else size = 8; */ amd64_mov_membase_reg (code, ins->inst_basereg, ins->inst_offset, ainfo->reg, size); break; } case ArgInFloatSSEReg: amd64_movss_membase_reg (code, ins->inst_basereg, ins->inst_offset, ainfo->reg); break; case ArgInDoubleSSEReg: amd64_movsd_membase_reg (code, ins->inst_basereg, ins->inst_offset, ainfo->reg); break; case ArgValuetypeInReg: for (quad = 0; quad < 2; quad ++) { switch (ainfo->pair_storage [quad]) { case ArgInIReg: amd64_mov_membase_reg (code, ins->inst_basereg, ins->inst_offset + (quad * sizeof (gpointer)), ainfo->pair_regs [quad], sizeof (gpointer)); break; case ArgInFloatSSEReg: amd64_movss_membase_reg (code, ins->inst_basereg, ins->inst_offset + (quad * sizeof (gpointer)), ainfo->pair_regs [quad]); break; case ArgInDoubleSSEReg: amd64_movsd_membase_reg (code, ins->inst_basereg, ins->inst_offset + (quad * sizeof (gpointer)), ainfo->pair_regs [quad]); break; case ArgNone: break; default: g_assert_not_reached (); } } break; case ArgValuetypeAddrInIReg: if (ainfo->pair_storage [0] == ArgInIReg) amd64_mov_membase_reg (code, ins->inst_left->inst_basereg, ins->inst_left->inst_offset, ainfo->pair_regs [0], sizeof (gpointer)); break; default: break; } } else { /* Argument allocated to (non-volatile) register */ switch (ainfo->storage) { case ArgInIReg: amd64_mov_reg_reg (code, ins->dreg, ainfo->reg, 8); break; case ArgOnStack: amd64_mov_reg_membase (code, ins->dreg, AMD64_RBP, ARGS_OFFSET + ainfo->offset, 8); break; default: g_assert_not_reached (); } } } /* Might need to attach the thread to the JIT or change the domain for the callback */ if (method->wrapper_type == MONO_WRAPPER_NATIVE_TO_MANAGED) { guint64 domain = (guint64)cfg->domain; args_clobbered = TRUE; /* * The call might clobber argument registers, but they are already * saved to the stack/global regs. */ if (appdomain_tls_offset != -1 && lmf_tls_offset != -1) { guint8 *buf, *no_domain_branch; code = mono_amd64_emit_tls_get (code, AMD64_RAX, appdomain_tls_offset); if (cfg->compile_aot) { /* AOT code is only used in the root domain */ amd64_mov_reg_imm (code, AMD64_ARG_REG1, 0); } else { if ((domain >> 32) == 0) amd64_mov_reg_imm_size (code, AMD64_ARG_REG1, domain, 4); else amd64_mov_reg_imm_size (code, AMD64_ARG_REG1, domain, 8); } amd64_alu_reg_reg (code, X86_CMP, AMD64_RAX, AMD64_ARG_REG1); no_domain_branch = code; x86_branch8 (code, X86_CC_NE, 0, 0); code = mono_amd64_emit_tls_get ( code, AMD64_RAX, lmf_addr_tls_offset); amd64_test_reg_reg (code, AMD64_RAX, AMD64_RAX); buf = code; x86_branch8 (code, X86_CC_NE, 0, 0); amd64_patch (no_domain_branch, code); code = emit_call (cfg, code, MONO_PATCH_INFO_INTERNAL_METHOD, (gpointer)"mono_jit_thread_attach", TRUE); amd64_patch (buf, code); #ifdef HOST_WIN32 /* The TLS key actually contains a pointer to the MonoJitTlsData structure */ /* FIXME: Add a separate key for LMF to avoid this */ amd64_alu_reg_imm (code, X86_ADD, AMD64_RAX, G_STRUCT_OFFSET (MonoJitTlsData, lmf)); #endif } else { g_assert (!cfg->compile_aot); if (cfg->compile_aot) { /* AOT code is only used in the root domain */ amd64_mov_reg_imm (code, AMD64_ARG_REG1, 0); } else { if ((domain >> 32) == 0) amd64_mov_reg_imm_size (code, AMD64_ARG_REG1, domain, 4); else amd64_mov_reg_imm_size (code, AMD64_ARG_REG1, domain, 8); } code = emit_call (cfg, code, MONO_PATCH_INFO_INTERNAL_METHOD, (gpointer)"mono_jit_thread_attach", TRUE); } } if (method->save_lmf) { if ((lmf_tls_offset != -1) && !optimize_for_xen) { /* * Optimized version which uses the mono_lmf TLS variable instead of * indirection through the mono_lmf_addr TLS variable. */ /* %rax = previous_lmf */ x86_prefix (code, X86_FS_PREFIX); amd64_mov_reg_mem (code, AMD64_RAX, lmf_tls_offset, 8); /* Save previous_lmf */ amd64_mov_membase_reg (code, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, previous_lmf), AMD64_RAX, 8); /* Set new lmf */ if (lmf_offset == 0) { x86_prefix (code, X86_FS_PREFIX); amd64_mov_mem_reg (code, lmf_tls_offset, cfg->frame_reg, 8); } else { amd64_lea_membase (code, AMD64_R11, cfg->frame_reg, lmf_offset); x86_prefix (code, X86_FS_PREFIX); amd64_mov_mem_reg (code, lmf_tls_offset, AMD64_R11, 8); } } else { if (lmf_addr_tls_offset != -1) { /* Load lmf quicky using the FS register */ code = mono_amd64_emit_tls_get (code, AMD64_RAX, lmf_addr_tls_offset); #ifdef HOST_WIN32 /* The TLS key actually contains a pointer to the MonoJitTlsData structure */ /* FIXME: Add a separate key for LMF to avoid this */ amd64_alu_reg_imm (code, X86_ADD, AMD64_RAX, G_STRUCT_OFFSET (MonoJitTlsData, lmf)); #endif } else { /* * The call might clobber argument registers, but they are already * saved to the stack/global regs. */ args_clobbered = TRUE; code = emit_call (cfg, code, MONO_PATCH_INFO_INTERNAL_METHOD, (gpointer)"mono_get_lmf_addr", TRUE); } /* Save lmf_addr */ amd64_mov_membase_reg (code, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, lmf_addr), AMD64_RAX, 8); /* Save previous_lmf */ amd64_mov_reg_membase (code, AMD64_R11, AMD64_RAX, 0, 8); amd64_mov_membase_reg (code, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, previous_lmf), AMD64_R11, 8); /* Set new lmf */ amd64_lea_membase (code, AMD64_R11, cfg->frame_reg, lmf_offset); amd64_mov_membase_reg (code, AMD64_RAX, 0, AMD64_R11, 8); } } if (trace) { args_clobbered = TRUE; code = mono_arch_instrument_prolog (cfg, mono_trace_enter_method, code, TRUE); } if (cfg->prof_options & MONO_PROFILE_ENTER_LEAVE) args_clobbered = TRUE; /* * Optimize the common case of the first bblock making a call with the same * arguments as the method. This works because the arguments are still in their * original argument registers. * FIXME: Generalize this */ if (!args_clobbered) { MonoBasicBlock *first_bb = cfg->bb_entry; MonoInst *next; next = mono_bb_first_ins (first_bb); if (!next && first_bb->next_bb) { first_bb = first_bb->next_bb; next = mono_bb_first_ins (first_bb); } if (first_bb->in_count > 1) next = NULL; for (i = 0; next && i < sig->param_count + sig->hasthis; ++i) { ArgInfo *ainfo = cinfo->args + i; gboolean match = FALSE; ins = cfg->args [i]; if (ins->opcode != OP_REGVAR) { switch (ainfo->storage) { case ArgInIReg: { if (((next->opcode == OP_LOAD_MEMBASE) || (next->opcode == OP_LOADI4_MEMBASE)) && next->inst_basereg == ins->inst_basereg && next->inst_offset == ins->inst_offset) { if (next->dreg == ainfo->reg) { NULLIFY_INS (next); match = TRUE; } else { next->opcode = OP_MOVE; next->sreg1 = ainfo->reg; /* Only continue if the instruction doesn't change argument regs */ if (next->dreg == ainfo->reg || next->dreg == AMD64_RAX) match = TRUE; } } break; } default: break; } } else { /* Argument allocated to (non-volatile) register */ switch (ainfo->storage) { case ArgInIReg: if (next->opcode == OP_MOVE && next->sreg1 == ins->dreg && next->dreg == ainfo->reg) { NULLIFY_INS (next); match = TRUE; } break; default: break; } } if (match) { next = next->next; //next = mono_inst_list_next (&next->node, &first_bb->ins_list); if (!next) break; } } } cfg->code_len = code - cfg->native_code; g_assert (cfg->code_len < cfg->code_size); return code; } void mono_arch_emit_epilog (MonoCompile *cfg) { MonoMethod *method = cfg->method; int quad, pos, i; guint8 *code; int max_epilog_size; CallInfo *cinfo; gint32 lmf_offset = cfg->arch.lmf_offset; max_epilog_size = get_max_epilog_size (cfg); while (cfg->code_len + max_epilog_size > (cfg->code_size - 16)) { cfg->code_size *= 2; cfg->native_code = g_realloc (cfg->native_code, cfg->code_size); mono_jit_stats.code_reallocs++; } code = cfg->native_code + cfg->code_len; if (mono_jit_trace_calls != NULL && mono_trace_eval (method)) code = mono_arch_instrument_epilog (cfg, mono_trace_leave_method, code, TRUE); /* the code restoring the registers must be kept in sync with OP_JMP */ pos = 0; if (method->save_lmf) { /* check if we need to restore protection of the stack after a stack overflow */ if (mono_get_jit_tls_offset () != -1) { guint8 *patch; code = mono_amd64_emit_tls_get (code, X86_ECX, mono_get_jit_tls_offset ()); /* we load the value in a separate instruction: this mechanism may be * used later as a safer way to do thread interruption */ amd64_mov_reg_membase (code, X86_ECX, X86_ECX, G_STRUCT_OFFSET (MonoJitTlsData, restore_stack_prot), 8); x86_alu_reg_imm (code, X86_CMP, X86_ECX, 0); patch = code; x86_branch8 (code, X86_CC_Z, 0, FALSE); /* note that the call trampoline will preserve eax/edx */ x86_call_reg (code, X86_ECX); x86_patch (patch, code); } else { /* FIXME: maybe save the jit tls in the prolog */ } if ((lmf_tls_offset != -1) && !optimize_for_xen) { /* * Optimized version which uses the mono_lmf TLS variable instead of indirection * through the mono_lmf_addr TLS variable. */ /* reg = previous_lmf */ amd64_mov_reg_membase (code, AMD64_R11, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, previous_lmf), 8); x86_prefix (code, X86_FS_PREFIX); amd64_mov_mem_reg (code, lmf_tls_offset, AMD64_R11, 8); } else { /* Restore previous lmf */ amd64_mov_reg_membase (code, AMD64_RCX, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, previous_lmf), 8); amd64_mov_reg_membase (code, AMD64_R11, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, lmf_addr), 8); amd64_mov_membase_reg (code, AMD64_R11, 0, AMD64_RCX, 8); } /* Restore caller saved regs */ if (cfg->used_int_regs & (1 << AMD64_RBP)) { amd64_mov_reg_membase (code, AMD64_RBP, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, rbp), 8); } if (cfg->used_int_regs & (1 << AMD64_RBX)) { amd64_mov_reg_membase (code, AMD64_RBX, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, rbx), 8); } if (cfg->used_int_regs & (1 << AMD64_R12)) { amd64_mov_reg_membase (code, AMD64_R12, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, r12), 8); } if (cfg->used_int_regs & (1 << AMD64_R13)) { amd64_mov_reg_membase (code, AMD64_R13, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, r13), 8); } if (cfg->used_int_regs & (1 << AMD64_R14)) { amd64_mov_reg_membase (code, AMD64_R14, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, r14), 8); } if (cfg->used_int_regs & (1 << AMD64_R15)) { amd64_mov_reg_membase (code, AMD64_R15, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, r15), 8); } #ifdef HOST_WIN32 if (cfg->used_int_regs & (1 << AMD64_RDI)) { amd64_mov_reg_membase (code, AMD64_RDI, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, rdi), 8); } if (cfg->used_int_regs & (1 << AMD64_RSI)) { amd64_mov_reg_membase (code, AMD64_RSI, cfg->frame_reg, lmf_offset + G_STRUCT_OFFSET (MonoLMF, rsi), 8); } #endif } else { if (cfg->arch.omit_fp) { gint32 save_area_offset = cfg->arch.reg_save_area_offset; for (i = 0; i < AMD64_NREG; ++i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { amd64_mov_reg_membase (code, i, AMD64_RSP, save_area_offset, 8); save_area_offset += 8; } } else { for (i = 0; i < AMD64_NREG; ++i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) pos -= sizeof (gpointer); if (pos) { if (pos == - sizeof (gpointer)) { /* Only one register, so avoid lea */ for (i = AMD64_NREG - 1; i > 0; --i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { amd64_mov_reg_membase (code, i, AMD64_RBP, pos, 8); } } else { amd64_lea_membase (code, AMD64_RSP, AMD64_RBP, pos); /* Pop registers in reverse order */ for (i = AMD64_NREG - 1; i > 0; --i) if (AMD64_IS_CALLEE_SAVED_REG (i) && (cfg->used_int_regs & (1 << i))) { amd64_pop_reg (code, i); } } } } } /* Load returned vtypes into registers if needed */ cinfo = cfg->arch.cinfo; if (cinfo->ret.storage == ArgValuetypeInReg) { ArgInfo *ainfo = &cinfo->ret; MonoInst *inst = cfg->ret; for (quad = 0; quad < 2; quad ++) { switch (ainfo->pair_storage [quad]) { case ArgInIReg: amd64_mov_reg_membase (code, ainfo->pair_regs [quad], inst->inst_basereg, inst->inst_offset + (quad * sizeof (gpointer)), sizeof (gpointer)); break; case ArgInFloatSSEReg: amd64_movss_reg_membase (code, ainfo->pair_regs [quad], inst->inst_basereg, inst->inst_offset + (quad * sizeof (gpointer))); break; case ArgInDoubleSSEReg: amd64_movsd_reg_membase (code, ainfo->pair_regs [quad], inst->inst_basereg, inst->inst_offset + (quad * sizeof (gpointer))); break; case ArgNone: break; default: g_assert_not_reached (); } } } if (cfg->arch.omit_fp) { if (cfg->arch.stack_alloc_size) amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, cfg->arch.stack_alloc_size); } else { amd64_leave (code); } async_exc_point (code); amd64_ret (code); cfg->code_len = code - cfg->native_code; g_assert (cfg->code_len < cfg->code_size); } void mono_arch_emit_exceptions (MonoCompile *cfg) { MonoJumpInfo *patch_info; int nthrows, i; guint8 *code; MonoClass *exc_classes [16]; guint8 *exc_throw_start [16], *exc_throw_end [16]; guint32 code_size = 0; /* Compute needed space */ for (patch_info = cfg->patch_info; patch_info; patch_info = patch_info->next) { if (patch_info->type == MONO_PATCH_INFO_EXC) code_size += 40; if (patch_info->type == MONO_PATCH_INFO_R8) code_size += 8 + 15; /* sizeof (double) + alignment */ if (patch_info->type == MONO_PATCH_INFO_R4) code_size += 4 + 15; /* sizeof (float) + alignment */ } while (cfg->code_len + code_size > (cfg->code_size - 16)) { cfg->code_size *= 2; cfg->native_code = g_realloc (cfg->native_code, cfg->code_size); mono_jit_stats.code_reallocs++; } code = cfg->native_code + cfg->code_len; /* add code to raise exceptions */ nthrows = 0; for (patch_info = cfg->patch_info; patch_info; patch_info = patch_info->next) { switch (patch_info->type) { case MONO_PATCH_INFO_EXC: { MonoClass *exc_class; guint8 *buf, *buf2; guint32 throw_ip; amd64_patch (patch_info->ip.i + cfg->native_code, code); exc_class = mono_class_from_name (mono_defaults.corlib, "System", patch_info->data.name); g_assert (exc_class); throw_ip = patch_info->ip.i; //x86_breakpoint (code); /* Find a throw sequence for the same exception class */ for (i = 0; i < nthrows; ++i) if (exc_classes [i] == exc_class) break; if (i < nthrows) { amd64_mov_reg_imm (code, AMD64_ARG_REG2, (exc_throw_end [i] - cfg->native_code) - throw_ip); x86_jump_code (code, exc_throw_start [i]); patch_info->type = MONO_PATCH_INFO_NONE; } else { buf = code; amd64_mov_reg_imm_size (code, AMD64_ARG_REG2, 0xf0f0f0f0, 4); buf2 = code; if (nthrows < 16) { exc_classes [nthrows] = exc_class; exc_throw_start [nthrows] = code; } amd64_mov_reg_imm (code, AMD64_ARG_REG1, exc_class->type_token); patch_info->type = MONO_PATCH_INFO_NONE; code = emit_call_body (cfg, code, MONO_PATCH_INFO_INTERNAL_METHOD, "mono_arch_throw_corlib_exception"); amd64_mov_reg_imm (buf, AMD64_ARG_REG2, (code - cfg->native_code) - throw_ip); while (buf < buf2) x86_nop (buf); if (nthrows < 16) { exc_throw_end [nthrows] = code; nthrows ++; } } break; } default: /* do nothing */ break; } } /* Handle relocations with RIP relative addressing */ for (patch_info = cfg->patch_info; patch_info; patch_info = patch_info->next) { gboolean remove = FALSE; switch (patch_info->type) { case MONO_PATCH_INFO_R8: case MONO_PATCH_INFO_R4: { guint8 *pos; /* The SSE opcodes require a 16 byte alignment */ code = (guint8*)ALIGN_TO (code, 16); pos = cfg->native_code + patch_info->ip.i; if (IS_REX (pos [1])) *(guint32*)(pos + 5) = (guint8*)code - pos - 9; else *(guint32*)(pos + 4) = (guint8*)code - pos - 8; if (patch_info->type == MONO_PATCH_INFO_R8) { *(double*)code = *(double*)patch_info->data.target; code += sizeof (double); } else { *(float*)code = *(float*)patch_info->data.target; code += sizeof (float); } remove = TRUE; break; } default: break; } if (remove) { if (patch_info == cfg->patch_info) cfg->patch_info = patch_info->next; else { MonoJumpInfo *tmp; for (tmp = cfg->patch_info; tmp->next != patch_info; tmp = tmp->next) ; tmp->next = patch_info->next; } } } cfg->code_len = code - cfg->native_code; g_assert (cfg->code_len < cfg->code_size); } void* mono_arch_instrument_prolog (MonoCompile *cfg, void *func, void *p, gboolean enable_arguments) { guchar *code = p; CallInfo *cinfo = NULL; MonoMethodSignature *sig; MonoInst *inst; int i, n, stack_area = 0; /* Keep this in sync with mono_arch_get_argument_info */ if (enable_arguments) { /* Allocate a new area on the stack and save arguments there */ sig = mono_method_signature (cfg->method); cinfo = get_call_info (cfg->generic_sharing_context, cfg->mempool, sig, FALSE); n = sig->param_count + sig->hasthis; stack_area = ALIGN_TO (n * 8, 16); amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, stack_area); for (i = 0; i < n; ++i) { inst = cfg->args [i]; if (inst->opcode == OP_REGVAR) amd64_mov_membase_reg (code, AMD64_RSP, (i * 8), inst->dreg, 8); else { amd64_mov_reg_membase (code, AMD64_R11, inst->inst_basereg, inst->inst_offset, 8); amd64_mov_membase_reg (code, AMD64_RSP, (i * 8), AMD64_R11, 8); } } } mono_add_patch_info (cfg, code-cfg->native_code, MONO_PATCH_INFO_METHODCONST, cfg->method); amd64_set_reg_template (code, AMD64_ARG_REG1); amd64_mov_reg_reg (code, AMD64_ARG_REG2, AMD64_RSP, 8); code = emit_call (cfg, code, MONO_PATCH_INFO_ABS, (gpointer)func, TRUE); if (enable_arguments) amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, stack_area); return code; } enum { SAVE_NONE, SAVE_STRUCT, SAVE_EAX, SAVE_EAX_EDX, SAVE_XMM }; void* mono_arch_instrument_epilog_full (MonoCompile *cfg, void *func, void *p, gboolean enable_arguments, gboolean preserve_argument_registers) { guchar *code = p; int save_mode = SAVE_NONE; MonoMethod *method = cfg->method; MonoType *ret_type = mini_type_get_underlying_type (NULL, mono_method_signature (method)->ret); switch (ret_type->type) { case MONO_TYPE_VOID: /* special case string .ctor icall */ if (strcmp (".ctor", method->name) && method->klass == mono_defaults.string_class) save_mode = SAVE_EAX; else save_mode = SAVE_NONE; break; case MONO_TYPE_I8: case MONO_TYPE_U8: save_mode = SAVE_EAX; break; case MONO_TYPE_R4: case MONO_TYPE_R8: save_mode = SAVE_XMM; break; case MONO_TYPE_GENERICINST: if (!mono_type_generic_inst_is_valuetype (ret_type)) { save_mode = SAVE_EAX; break; } /* Fall through */ case MONO_TYPE_VALUETYPE: save_mode = SAVE_STRUCT; break; default: save_mode = SAVE_EAX; break; } /* Save the result and copy it into the proper argument register */ switch (save_mode) { case SAVE_EAX: amd64_push_reg (code, AMD64_RAX); /* Align stack */ amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 8); if (enable_arguments) amd64_mov_reg_reg (code, AMD64_ARG_REG2, AMD64_RAX, 8); break; case SAVE_STRUCT: /* FIXME: */ if (enable_arguments) amd64_mov_reg_imm (code, AMD64_ARG_REG2, 0); break; case SAVE_XMM: amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 8); amd64_movsd_membase_reg (code, AMD64_RSP, 0, AMD64_XMM0); /* Align stack */ amd64_alu_reg_imm (code, X86_SUB, AMD64_RSP, 8); /* * The result is already in the proper argument register so no copying * needed. */ break; case SAVE_NONE: break; default: g_assert_not_reached (); } /* Set %al since this is a varargs call */ if (save_mode == SAVE_XMM) amd64_mov_reg_imm (code, AMD64_RAX, 1); else amd64_mov_reg_imm (code, AMD64_RAX, 0); if (preserve_argument_registers) { amd64_push_reg (code, MONO_AMD64_ARG_REG1); amd64_push_reg (code, MONO_AMD64_ARG_REG2); } mono_add_patch_info (cfg, code-cfg->native_code, MONO_PATCH_INFO_METHODCONST, method); amd64_set_reg_template (code, AMD64_ARG_REG1); code = emit_call (cfg, code, MONO_PATCH_INFO_ABS, (gpointer)func, TRUE); if (preserve_argument_registers) { amd64_pop_reg (code, MONO_AMD64_ARG_REG2); amd64_pop_reg (code, MONO_AMD64_ARG_REG1); } /* Restore result */ switch (save_mode) { case SAVE_EAX: amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, 8); amd64_pop_reg (code, AMD64_RAX); break; case SAVE_STRUCT: /* FIXME: */ break; case SAVE_XMM: amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, 8); amd64_movsd_reg_membase (code, AMD64_XMM0, AMD64_RSP, 0); amd64_alu_reg_imm (code, X86_ADD, AMD64_RSP, 8); break; case SAVE_NONE: break; default: g_assert_not_reached (); } return code; } void mono_arch_flush_icache (guint8 *code, gint size) { /* Not needed */ } void mono_arch_flush_register_windows (void) { } gboolean mono_arch_is_inst_imm (gint64 imm) { return amd64_is_imm32 (imm); } /* * Determine whenever the trap whose info is in SIGINFO is caused by * integer overflow. */ gboolean mono_arch_is_int_overflow (void *sigctx, void *info) { MonoContext ctx; guint8* rip; int reg; gint64 value; mono_arch_sigctx_to_monoctx (sigctx, &ctx); rip = (guint8*)ctx.rip; if (IS_REX (rip [0])) { reg = amd64_rex_b (rip [0]); rip ++; } else reg = 0; if ((rip [0] == 0xf7) && (x86_modrm_mod (rip [1]) == 0x3) && (x86_modrm_reg (rip [1]) == 0x7)) { /* idiv REG */ reg += x86_modrm_rm (rip [1]); switch (reg) { case AMD64_RAX: value = ctx.rax; break; case AMD64_RBX: value = ctx.rbx; break; case AMD64_RCX: value = ctx.rcx; break; case AMD64_RDX: value = ctx.rdx; break; case AMD64_RBP: value = ctx.rbp; break; case AMD64_RSP: value = ctx.rsp; break; case AMD64_RSI: value = ctx.rsi; break; case AMD64_RDI: value = ctx.rdi; break; case AMD64_R12: value = ctx.r12; break; case AMD64_R13: value = ctx.r13; break; case AMD64_R14: value = ctx.r14; break; case AMD64_R15: value = ctx.r15; break; default: g_assert_not_reached (); reg = -1; } if (value == -1) return TRUE; } return FALSE; } guint32 mono_arch_get_patch_offset (guint8 *code) { return 3; } /** * mono_breakpoint_clean_code: * * Copy @size bytes from @code - @offset to the buffer @buf. If the debugger inserted software * breakpoints in the original code, they are removed in the copy. * * Returns TRUE if no sw breakpoint was present. */ gboolean mono_breakpoint_clean_code (guint8 *method_start, guint8 *code, int offset, guint8 *buf, int size) { int i; gboolean can_write = TRUE; /* * If method_start is non-NULL we need to perform bound checks, since we access memory * at code - offset we could go before the start of the method and end up in a different * page of memory that is not mapped or read incorrect data anyway. We zero-fill the bytes * instead. */ if (!method_start || code - offset >= method_start) { memcpy (buf, code - offset, size); } else { int diff = code - method_start; memset (buf, 0, size); memcpy (buf + offset - diff, method_start, diff + size - offset); } code -= offset; for (i = 0; i < MONO_BREAKPOINT_ARRAY_SIZE; ++i) { int idx = mono_breakpoint_info_index [i]; guint8 *ptr; if (idx < 1) continue; ptr = mono_breakpoint_info [idx].address; if (ptr >= code && ptr < code + size) { guint8 saved_byte = mono_breakpoint_info [idx].saved_byte; can_write = FALSE; /*g_print ("patching %p with 0x%02x (was: 0x%02x)\n", ptr, saved_byte, buf [ptr - code]);*/ buf [ptr - code] = saved_byte; } } return can_write; } gpointer mono_arch_get_vcall_slot (guint8 *code, mgreg_t *regs, int *displacement) { guint8 buf [10]; guint32 reg; gint32 disp; guint8 rex = 0; MonoJitInfo *ji = NULL; #ifdef ENABLE_LLVM /* code - 9 might be before the start of the method */ /* FIXME: Avoid this expensive call somehow */ ji = mono_jit_info_table_find (mono_domain_get (), (char*)code); #endif mono_breakpoint_clean_code (ji ? ji->code_start : NULL, code, 9, buf, sizeof (buf)); code = buf + 9; *displacement = 0; code -= 7; /* * A given byte sequence can match more than case here, so we have to be * really careful about the ordering of the cases. Longer sequences * come first. * There are two types of calls: * - direct calls: 0xff address_byte 8/32 bits displacement * - indirect calls: nop nop nop * The nops make sure we don't confuse the instruction preceeding an indirect * call with a direct call. */ if ((code [0] == 0x41) && (code [1] == 0xff) && (code [2] == 0x15)) { /* call OFFSET(%rip) */ disp = *(guint32*)(code + 3); return (gpointer*)(code + disp + 7); } else if ((code [0] == 0xff) && (amd64_modrm_reg (code [1]) == 0x2) && (amd64_modrm_mod (code [1]) == 0x2) && (amd64_sib_index (code [2]) == 4) && (amd64_sib_scale (code [2]) == 0)) { /* call *[reg+disp32] using indexed addressing */ /* The LLVM JIT emits this, and we emit it too for %r12 */ if (IS_REX (code [-1])) { rex = code [-1]; g_assert (amd64_rex_x (rex) == 0); } reg = amd64_sib_base (code [2]); disp = *(gint32*)(code + 3); } else if ((code [1] == 0xff) && (amd64_modrm_reg (code [2]) == 0x2) && (amd64_modrm_mod (code [2]) == 0x2)) { /* call *[reg+disp32] */ if (IS_REX (code [0])) rex = code [0]; reg = amd64_modrm_rm (code [2]); disp = *(gint32*)(code + 3); /* R10 is clobbered by the IMT thunk code */ g_assert (reg != AMD64_R10); } else if (code [2] == 0xe8) { /* call */ return NULL; } else if ((code [3] == 0xff) && (amd64_modrm_reg (code [4]) == 0x2) && (amd64_modrm_mod (code [4]) == 0x1) && (amd64_sib_index (code [5]) == 4) && (amd64_sib_scale (code [5]) == 0)) { /* call *[r12+disp8] using indexed addressing */ if (IS_REX (code [2])) rex = code [2]; reg = amd64_sib_base (code [5]); disp = *(gint8*)(code + 6); } else if (IS_REX (code [4]) && (code [5] == 0xff) && (amd64_modrm_reg (code [6]) == 0x2) && (amd64_modrm_mod (code [6]) == 0x3)) { /* call *%reg */ return NULL; } else if ((code [4] == 0xff) && (amd64_modrm_reg (code [5]) == 0x2) && (amd64_modrm_mod (code [5]) == 0x1)) { /* call *[reg+disp8] */ if (IS_REX (code [3])) rex = code [3]; reg = amd64_modrm_rm (code [5]); disp = *(gint8*)(code + 6); //printf ("B: [%%r%d+0x%x]\n", reg, disp); } else if ((code [5] == 0xff) && (amd64_modrm_reg (code [6]) == 0x2) && (amd64_modrm_mod (code [6]) == 0x0)) { /* call *%reg */ if (IS_REX (code [4])) rex = code [4]; reg = amd64_modrm_rm (code [6]); disp = 0; } else g_assert_not_reached (); reg += amd64_rex_b (rex); /* R11 is clobbered by the trampoline code */ g_assert (reg != AMD64_R11); *displacement = disp; return (gpointer)regs [reg]; } int mono_arch_get_this_arg_reg (MonoMethodSignature *sig, MonoGenericSharingContext *gsctx, guint8 *code) { int this_reg = AMD64_ARG_REG1; if (MONO_TYPE_ISSTRUCT (sig->ret)) { CallInfo *cinfo; if (!gsctx && code) gsctx = mono_get_generic_context_from_code (code); cinfo = get_call_info (gsctx, NULL, sig, FALSE); if (cinfo->ret.storage != ArgValuetypeInReg) this_reg = AMD64_ARG_REG2; g_free (cinfo); } return this_reg; } gpointer mono_arch_get_this_arg_from_call (MonoGenericSharingContext *gsctx, MonoMethodSignature *sig, mgreg_t *regs, guint8 *code) { return (gpointer)regs [mono_arch_get_this_arg_reg (sig, gsctx, code)]; } #define MAX_ARCH_DELEGATE_PARAMS 10 static gpointer get_delegate_invoke_impl (gboolean has_target, guint32 param_count, guint32 *code_len) { guint8 *code, *start; int i; if (has_target) { start = code = mono_global_codeman_reserve (64); /* Replace the this argument with the target */ amd64_mov_reg_reg (code, AMD64_RAX, AMD64_ARG_REG1, 8); amd64_mov_reg_membase (code, AMD64_ARG_REG1, AMD64_RAX, G_STRUCT_OFFSET (MonoDelegate, target), 8); amd64_jump_membase (code, AMD64_RAX, G_STRUCT_OFFSET (MonoDelegate, method_ptr)); g_assert ((code - start) < 64); } else { start = code = mono_global_codeman_reserve (64); if (param_count == 0) { amd64_jump_membase (code, AMD64_ARG_REG1, G_STRUCT_OFFSET (MonoDelegate, method_ptr)); } else { /* We have to shift the arguments left */ amd64_mov_reg_reg (code, AMD64_RAX, AMD64_ARG_REG1, 8); for (i = 0; i < param_count; ++i) { #ifdef HOST_WIN32 if (i < 3) amd64_mov_reg_reg (code, param_regs [i], param_regs [i + 1], 8); else amd64_mov_reg_membase (code, param_regs [i], AMD64_RSP, 0x28, 8); #else amd64_mov_reg_reg (code, param_regs [i], param_regs [i + 1], 8); #endif } amd64_jump_membase (code, AMD64_RAX, G_STRUCT_OFFSET (MonoDelegate, method_ptr)); } g_assert ((code - start) < 64); } mono_debug_add_delegate_trampoline (start, code - start); if (code_len) *code_len = code - start; return start; } /* * mono_arch_get_delegate_invoke_impls: * * Return a list of MonoAotTrampInfo structures for the delegate invoke impl * trampolines. */ GSList* mono_arch_get_delegate_invoke_impls (void) { GSList *res = NULL; guint8 *code; guint32 code_len; int i; code = get_delegate_invoke_impl (TRUE, 0, &code_len); res = g_slist_prepend (res, mono_aot_tramp_info_create (g_strdup ("delegate_invoke_impl_has_target"), code, code_len)); for (i = 0; i < MAX_ARCH_DELEGATE_PARAMS; ++i) { code = get_delegate_invoke_impl (FALSE, i, &code_len); res = g_slist_prepend (res, mono_aot_tramp_info_create (g_strdup_printf ("delegate_invoke_impl_target_%d", i), code, code_len)); } return res; } gpointer mono_arch_get_delegate_invoke_impl (MonoMethodSignature *sig, gboolean has_target) { guint8 *code, *start; int i; if (sig->param_count > MAX_ARCH_DELEGATE_PARAMS) return NULL; /* FIXME: Support more cases */ if (MONO_TYPE_ISSTRUCT (sig->ret)) return NULL; if (has_target) { static guint8* cached = NULL; if (cached) return cached; if (mono_aot_only) start = mono_aot_get_named_code ("delegate_invoke_impl_has_target"); else start = get_delegate_invoke_impl (TRUE, 0, NULL); mono_memory_barrier (); cached = start; } else { static guint8* cache [MAX_ARCH_DELEGATE_PARAMS + 1] = {NULL}; for (i = 0; i < sig->param_count; ++i) if (!mono_is_regsize_var (sig->params [i])) return NULL; if (sig->param_count > 4) return NULL; code = cache [sig->param_count]; if (code) return code; if (mono_aot_only) { char *name = g_strdup_printf ("delegate_invoke_impl_target_%d", sig->param_count); start = mono_aot_get_named_code (name); g_free (name); } else { start = get_delegate_invoke_impl (FALSE, sig->param_count, NULL); } mono_memory_barrier (); cache [sig->param_count] = start; } return start; } /* * Support for fast access to the thread-local lmf structure using the GS * segment register on NPTL + kernel 2.6.x. */ static gboolean tls_offset_inited = FALSE; void mono_arch_setup_jit_tls_data (MonoJitTlsData *tls) { if (!tls_offset_inited) { #ifdef HOST_WIN32 /* * We need to init this multiple times, since when we are first called, the key might not * be initialized yet. */ appdomain_tls_offset = mono_domain_get_tls_key (); lmf_tls_offset = mono_get_jit_tls_key (); lmf_addr_tls_offset = mono_get_jit_tls_key (); /* Only 64 tls entries can be accessed using inline code */ if (appdomain_tls_offset >= 64) appdomain_tls_offset = -1; if (lmf_tls_offset >= 64) lmf_tls_offset = -1; #else tls_offset_inited = TRUE; #ifdef MONO_XEN_OPT optimize_for_xen = access ("/proc/xen", F_OK) == 0; #endif appdomain_tls_offset = mono_domain_get_tls_offset (); lmf_tls_offset = mono_get_lmf_tls_offset (); lmf_addr_tls_offset = mono_get_lmf_addr_tls_offset (); #endif } } void mono_arch_free_jit_tls_data (MonoJitTlsData *tls) { } #ifdef MONO_ARCH_HAVE_IMT #define CMP_SIZE (6 + 1) #define CMP_REG_REG_SIZE (4 + 1) #define BR_SMALL_SIZE 2 #define BR_LARGE_SIZE 6 #define MOV_REG_IMM_SIZE 10 #define MOV_REG_IMM_32BIT_SIZE 6 #define JUMP_REG_SIZE (2 + 1) static int imt_branch_distance (MonoIMTCheckItem **imt_entries, int start, int target) { int i, distance = 0; for (i = start; i < target; ++i) distance += imt_entries [i]->chunk_size; return distance; } /* * LOCKING: called with the domain lock held */ gpointer mono_arch_build_imt_thunk (MonoVTable *vtable, MonoDomain *domain, MonoIMTCheckItem **imt_entries, int count, gpointer fail_tramp) { int i; int size = 0; guint8 *code, *start; gboolean vtable_is_32bit = ((gsize)(vtable) == (gsize)(int)(gsize)(vtable)); for (i = 0; i < count; ++i) { MonoIMTCheckItem *item = imt_entries [i]; if (item->is_equals) { if (item->check_target_idx) { if (!item->compare_done) { if (amd64_is_imm32 (item->key)) item->chunk_size += CMP_SIZE; else item->chunk_size += MOV_REG_IMM_SIZE + CMP_REG_REG_SIZE; } if (item->has_target_code) { item->chunk_size += MOV_REG_IMM_SIZE; } else { if (vtable_is_32bit) item->chunk_size += MOV_REG_IMM_32BIT_SIZE; else item->chunk_size += MOV_REG_IMM_SIZE; } item->chunk_size += BR_SMALL_SIZE + JUMP_REG_SIZE; } else { if (fail_tramp) { item->chunk_size += MOV_REG_IMM_SIZE * 3 + CMP_REG_REG_SIZE + BR_SMALL_SIZE + JUMP_REG_SIZE * 2; } else { if (vtable_is_32bit) item->chunk_size += MOV_REG_IMM_32BIT_SIZE; else item->chunk_size += MOV_REG_IMM_SIZE; item->chunk_size += JUMP_REG_SIZE; /* with assert below: * item->chunk_size += CMP_SIZE + BR_SMALL_SIZE + 1; */ } } } else { if (amd64_is_imm32 (item->key)) item->chunk_size += CMP_SIZE; else item->chunk_size += MOV_REG_IMM_SIZE + CMP_REG_REG_SIZE; item->chunk_size += BR_LARGE_SIZE; imt_entries [item->check_target_idx]->compare_done = TRUE; } size += item->chunk_size; } if (fail_tramp) code = mono_method_alloc_generic_virtual_thunk (domain, size); else code = mono_domain_code_reserve (domain, size); start = code; for (i = 0; i < count; ++i) { MonoIMTCheckItem *item = imt_entries [i]; item->code_target = code; if (item->is_equals) { gboolean fail_case = !item->check_target_idx && fail_tramp; if (item->check_target_idx || fail_case) { if (!item->compare_done || fail_case) { if (amd64_is_imm32 (item->key)) amd64_alu_reg_imm (code, X86_CMP, MONO_ARCH_IMT_REG, (guint32)(gssize)item->key); else { amd64_mov_reg_imm (code, AMD64_R10, item->key); amd64_alu_reg_reg (code, X86_CMP, MONO_ARCH_IMT_REG, AMD64_R10); } } item->jmp_code = code; amd64_branch8 (code, X86_CC_NE, 0, FALSE); /* See the comment below about R10 */ if (item->has_target_code) { amd64_mov_reg_imm (code, AMD64_R10, item->value.target_code); amd64_jump_reg (code, AMD64_R10); } else { amd64_mov_reg_imm (code, AMD64_R10, & (vtable->vtable [item->value.vtable_slot])); amd64_jump_membase (code, AMD64_R10, 0); } if (fail_case) { amd64_patch (item->jmp_code, code); amd64_mov_reg_imm (code, AMD64_R10, fail_tramp); amd64_jump_reg (code, AMD64_R10); item->jmp_code = NULL; } } else { /* enable the commented code to assert on wrong method */ #if 0 if (amd64_is_imm32 (item->key)) amd64_alu_reg_imm (code, X86_CMP, MONO_ARCH_IMT_REG, (guint32)(gssize)item->key); else { amd64_mov_reg_imm (code, AMD64_R10, item->key); amd64_alu_reg_reg (code, X86_CMP, MONO_ARCH_IMT_REG, AMD64_R10); } item->jmp_code = code; amd64_branch8 (code, X86_CC_NE, 0, FALSE); /* See the comment below about R10 */ amd64_mov_reg_imm (code, AMD64_R10, & (vtable->vtable [item->value.vtable_slot])); amd64_jump_membase (code, AMD64_R10, 0); amd64_patch (item->jmp_code, code); amd64_breakpoint (code); item->jmp_code = NULL; #else /* We're using R10 here because R11 needs to be preserved. R10 needs to be preserved for calls which require a runtime generic context, but interface calls don't. */ amd64_mov_reg_imm (code, AMD64_R10, & (vtable->vtable [item->value.vtable_slot])); amd64_jump_membase (code, AMD64_R10, 0); #endif } } else { if (amd64_is_imm32 (item->key)) amd64_alu_reg_imm (code, X86_CMP, MONO_ARCH_IMT_REG, (guint32)(gssize)item->key); else { amd64_mov_reg_imm (code, AMD64_R10, item->key); amd64_alu_reg_reg (code, X86_CMP, MONO_ARCH_IMT_REG, AMD64_R10); } item->jmp_code = code; if (x86_is_imm8 (imt_branch_distance (imt_entries, i, item->check_target_idx))) x86_branch8 (code, X86_CC_GE, 0, FALSE); else x86_branch32 (code, X86_CC_GE, 0, FALSE); } g_assert (code - item->code_target <= item->chunk_size); } /* patch the branches to get to the target items */ for (i = 0; i < count; ++i) { MonoIMTCheckItem *item = imt_entries [i]; if (item->jmp_code) { if (item->check_target_idx) { amd64_patch (item->jmp_code, imt_entries [item->check_target_idx]->code_target); } } } if (!fail_tramp) mono_stats.imt_thunks_size += code - start; g_assert (code - start <= size); return start; } MonoMethod* mono_arch_find_imt_method (mgreg_t *regs, guint8 *code) { return (MonoMethod*)regs [MONO_ARCH_IMT_REG]; } #endif MonoVTable* mono_arch_find_static_call_vtable (mgreg_t *regs, guint8 *code) { return (MonoVTable*) regs [MONO_ARCH_RGCTX_REG]; } MonoInst* mono_arch_emit_inst_for_method (MonoCompile *cfg, MonoMethod *cmethod, MonoMethodSignature *fsig, MonoInst **args) { MonoInst *ins = NULL; int opcode = 0; if (cmethod->klass == mono_defaults.math_class) { if (strcmp (cmethod->name, "Sin") == 0) { opcode = OP_SIN; } else if (strcmp (cmethod->name, "Cos") == 0) { opcode = OP_COS; } else if (strcmp (cmethod->name, "Sqrt") == 0) { opcode = OP_SQRT; } else if (strcmp (cmethod->name, "Abs") == 0 && fsig->params [0]->type == MONO_TYPE_R8) { opcode = OP_ABS; } if (opcode) { MONO_INST_NEW (cfg, ins, opcode); ins->type = STACK_R8; ins->dreg = mono_alloc_freg (cfg); ins->sreg1 = args [0]->dreg; MONO_ADD_INS (cfg->cbb, ins); } opcode = 0; if (cfg->opt & MONO_OPT_CMOV) { if (strcmp (cmethod->name, "Min") == 0) { if (fsig->params [0]->type == MONO_TYPE_I4) opcode = OP_IMIN; if (fsig->params [0]->type == MONO_TYPE_U4) opcode = OP_IMIN_UN; else if (fsig->params [0]->type == MONO_TYPE_I8) opcode = OP_LMIN; else if (fsig->params [0]->type == MONO_TYPE_U8) opcode = OP_LMIN_UN; } else if (strcmp (cmethod->name, "Max") == 0) { if (fsig->params [0]->type == MONO_TYPE_I4) opcode = OP_IMAX; if (fsig->params [0]->type == MONO_TYPE_U4) opcode = OP_IMAX_UN; else if (fsig->params [0]->type == MONO_TYPE_I8) opcode = OP_LMAX; else if (fsig->params [0]->type == MONO_TYPE_U8) opcode = OP_LMAX_UN; } } if (opcode) { MONO_INST_NEW (cfg, ins, opcode); ins->type = fsig->params [0]->type == MONO_TYPE_I4 ? STACK_I4 : STACK_I8; ins->dreg = mono_alloc_ireg (cfg); ins->sreg1 = args [0]->dreg; ins->sreg2 = args [1]->dreg; MONO_ADD_INS (cfg->cbb, ins); } #if 0 /* OP_FREM is not IEEE compatible */ else if (strcmp (cmethod->name, "IEEERemainder") == 0) { MONO_INST_NEW (cfg, ins, OP_FREM); ins->inst_i0 = args [0]; ins->inst_i1 = args [1]; } #endif } /* * Can't implement CompareExchange methods this way since they have * three arguments. */ return ins; } gboolean mono_arch_print_tree (MonoInst *tree, int arity) { return 0; } MonoInst* mono_arch_get_domain_intrinsic (MonoCompile* cfg) { MonoInst* ins; if (appdomain_tls_offset == -1) return NULL; MONO_INST_NEW (cfg, ins, OP_TLS_GET); ins->inst_offset = appdomain_tls_offset; return ins; } #define _CTX_REG(ctx,fld,i) ((gpointer)((&ctx->fld)[i])) gpointer mono_arch_context_get_int_reg (MonoContext *ctx, int reg) { switch (reg) { case AMD64_RCX: return (gpointer)ctx->rcx; case AMD64_RDX: return (gpointer)ctx->rdx; case AMD64_RBX: return (gpointer)ctx->rbx; case AMD64_RBP: return (gpointer)ctx->rbp; case AMD64_RSP: return (gpointer)ctx->rsp; default: if (reg < 8) return _CTX_REG (ctx, rax, reg); else if (reg >= 12) return _CTX_REG (ctx, r12, reg - 12); else g_assert_not_reached (); } } /* Soft Debug support */ #ifdef MONO_ARCH_SOFT_DEBUG_SUPPORTED /* * mono_arch_set_breakpoint: * * Set a breakpoint at the native code corresponding to JI at NATIVE_OFFSET. * The location should contain code emitted by OP_SEQ_POINT. */ void mono_arch_set_breakpoint (MonoJitInfo *ji, guint8 *ip) { guint8 *code = ip; guint8 *orig_code = code; /* * In production, we will use int3 (has to fix the size in the md * file). But that could confuse gdb, so during development, we emit a SIGSEGV * instead. */ g_assert (code [0] == 0x90); g_assert (((guint64)bp_trigger_page >> 32) == 0); amd64_mov_reg_mem (code, AMD64_R11, (guint64)bp_trigger_page, 4); g_assert (code - orig_code == BREAKPOINT_SIZE); } /* * mono_arch_clear_breakpoint: * * Clear the breakpoint at IP. */ void mono_arch_clear_breakpoint (MonoJitInfo *ji, guint8 *ip) { guint8 *code = ip; int i; for (i = 0; i < BREAKPOINT_SIZE; ++i) x86_nop (code); } /* * mono_arch_start_single_stepping: * * Start single stepping. */ void mono_arch_start_single_stepping (void) { mono_mprotect (ss_trigger_page, mono_pagesize (), 0); } /* * mono_arch_stop_single_stepping: * * Stop single stepping. */ void mono_arch_stop_single_stepping (void) { mono_mprotect (ss_trigger_page, mono_pagesize (), MONO_MMAP_READ); } /* * mono_arch_is_single_step_event: * * Return whenever the machine state in SIGCTX corresponds to a single * step event. */ gboolean mono_arch_is_single_step_event (void *info, void *sigctx) { #ifdef HOST_WIN32 EXCEPTION_RECORD* einfo = (EXCEPTION_RECORD*)info; return FALSE; #else siginfo_t* sinfo = (siginfo_t*) info; /* Sometimes the address is off by 4 */ if (sinfo->si_addr >= ss_trigger_page && (guint8*)sinfo->si_addr <= (guint8*)ss_trigger_page + 128) return TRUE; else return FALSE; #endif } gboolean mono_arch_is_breakpoint_event (void *info, void *sigctx) { #ifdef HOST_WIN32 EXCEPTION_RECORD* einfo = (EXCEPTION_RECORD*)info; return FALSE; #else siginfo_t* sinfo = (siginfo_t*) info; /* Sometimes the address is off by 4 */ if (sinfo->si_addr >= bp_trigger_page && (guint8*)sinfo->si_addr <= (guint8*)bp_trigger_page + 128) return TRUE; else return FALSE; #endif } /* * mono_arch_get_ip_for_breakpoint: * * Convert the ip in CTX to the address where a breakpoint was placed. */ guint8* mono_arch_get_ip_for_breakpoint (MonoJitInfo *ji, MonoContext *ctx) { guint8 *ip = MONO_CONTEXT_GET_IP (ctx); /* size of xor r11, r11 */ ip -= 0; return ip; } /* * mono_arch_get_ip_for_single_step: * * Convert the ip in CTX to the address stored in seq_points. */ guint8* mono_arch_get_ip_for_single_step (MonoJitInfo *ji, MonoContext *ctx) { guint8 *ip = MONO_CONTEXT_GET_IP (ctx); /* Size of amd64_mov_reg_mem (r11) */ ip += 8; return ip; } /* * mono_arch_skip_breakpoint: * * Modify CTX so the ip is placed after the breakpoint instruction, so when * we resume, the instruction is not executed again. */ void mono_arch_skip_breakpoint (MonoContext *ctx) { MONO_CONTEXT_SET_IP (ctx, (guint8*)MONO_CONTEXT_GET_IP (ctx) + BREAKPOINT_SIZE); } /* * mono_arch_skip_single_step: * * Modify CTX so the ip is placed after the single step trigger instruction, * we resume, the instruction is not executed again. */ void mono_arch_skip_single_step (MonoContext *ctx) { MONO_CONTEXT_SET_IP (ctx, (guint8*)MONO_CONTEXT_GET_IP (ctx) + 8); } /* * mono_arch_create_seq_point_info: * * Return a pointer to a data structure which is used by the sequence * point implementation in AOTed code. */ gpointer mono_arch_get_seq_point_info (MonoDomain *domain, guint8 *code) { NOT_IMPLEMENTED; return NULL; } #endif