diff options
Diffstat (limited to 'tcg')
-rw-r--r-- | tcg/arm/tcg-target.c | 53 | ||||
-rw-r--r-- | tcg/hppa/tcg-target.c | 147 | ||||
-rw-r--r-- | tcg/i386/tcg-target.c | 283 | ||||
-rw-r--r-- | tcg/ia64/tcg-target.c | 46 | ||||
-rw-r--r-- | tcg/mips/tcg-target.c | 44 | ||||
-rw-r--r-- | tcg/ppc/tcg-target.c | 45 | ||||
-rw-r--r-- | tcg/ppc/tcg-target.h | 2 | ||||
-rw-r--r-- | tcg/ppc64/tcg-target.c | 45 | ||||
-rw-r--r-- | tcg/s390/tcg-target.c | 44 | ||||
-rw-r--r-- | tcg/sparc/tcg-target.c | 124 | ||||
-rw-r--r-- | tcg/tcg.c | 311 | ||||
-rw-r--r-- | tcg/tcg.h | 11 | ||||
-rw-r--r-- | tcg/tci/tcg-target.c | 6 | ||||
-rw-r--r-- | tcg/tci/tcg-target.h | 2 |
14 files changed, 1051 insertions, 112 deletions
diff --git a/tcg/arm/tcg-target.c b/tcg/arm/tcg-target.c index 5af21b3f5d..4d59a63855 100644 --- a/tcg/arm/tcg-target.c +++ b/tcg/arm/tcg-target.c @@ -929,6 +929,27 @@ static inline void tcg_out_goto_label(TCGContext *s, int cond, int label_index) #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -936,6 +957,8 @@ static void *qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, @@ -943,6 +966,7 @@ static void *qemu_st_helpers[4] = { __stq_mmu, }; #endif +#endif #define TLB_SHIFT (CPU_TLB_ENTRY_BITS + CPU_TLB_BITS) @@ -1075,6 +1099,19 @@ static inline void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc) TCG_REG_R1, 0, addr_reg2, SHIFT_IMM_LSL(0)); tcg_out_dat_imm(s, COND_AL, ARITH_MOV, TCG_REG_R2, 0, mem_index); # endif +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal and incorrect for 64 bit */ + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, + tcg_target_call_iarg_regs[2], 0, + tcg_target_call_iarg_regs[1], SHIFT_IMM_LSL(0)); + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, + tcg_target_call_iarg_regs[1], 0, + tcg_target_call_iarg_regs[0], SHIFT_IMM_LSL(0)); + + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, + tcg_target_call_iarg_regs[0], 0, TCG_AREG0, + SHIFT_IMM_LSL(0)); +#endif tcg_out_call(s, (tcg_target_long) qemu_ld_helpers[s_bits]); switch (opc) { @@ -1341,6 +1378,22 @@ static inline void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc) } # endif +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal and incorrect for 64 bit */ + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, + tcg_target_call_iarg_regs[3], 0, + tcg_target_call_iarg_regs[2], SHIFT_IMM_LSL(0)); + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, + tcg_target_call_iarg_regs[2], 0, + tcg_target_call_iarg_regs[1], SHIFT_IMM_LSL(0)); + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, + tcg_target_call_iarg_regs[1], 0, + tcg_target_call_iarg_regs[0], SHIFT_IMM_LSL(0)); + + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, + tcg_target_call_iarg_regs[0], 0, TCG_AREG0, + SHIFT_IMM_LSL(0)); +#endif tcg_out_call(s, (tcg_target_long) qemu_st_helpers[s_bits]); if (opc == 3) tcg_out_dat_imm(s, COND_AL, ARITH_ADD, TCG_REG_R13, TCG_REG_R13, 0x10); diff --git a/tcg/hppa/tcg-target.c b/tcg/hppa/tcg-target.c index c5a3730a2b..2885212e3a 100644 --- a/tcg/hppa/tcg-target.c +++ b/tcg/hppa/tcg-target.c @@ -336,7 +336,7 @@ static int tcg_target_const_match(tcg_target_long val, #define INSN_COMIBF (INSN_OP(0x23)) /* supplied by libgcc */ -extern void *__canonicalize_funcptr_for_compare(void *); +extern void *__canonicalize_funcptr_for_compare(const void *); static void tcg_out_mov(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg) { @@ -628,7 +628,7 @@ static void tcg_out_bswap32(TCGContext *s, int ret, int arg, int temp) tcg_out_shd(s, ret, arg, temp, 8); /* ret = DCBA */ } -static void tcg_out_call(TCGContext *s, void *func) +static void tcg_out_call(TCGContext *s, const void *func) { tcg_target_long val, hi, lo, disp; @@ -882,6 +882,27 @@ static void tcg_out_setcond2(TCGContext *s, int cond, TCGArg ret, #if defined(CONFIG_SOFTMMU) #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -889,12 +910,15 @@ static void *qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, __stl_mmu, __stq_mmu, }; +#endif /* Load and compare a TLB entry, and branch if TLB miss. OFFSET is set to the offset of the first ADDR_READ or ADDR_WRITE member of the appropriate @@ -1061,6 +1085,15 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc) } tcg_out_movi(s, TCG_TYPE_I32, argreg, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_call(s, qemu_ld_helpers[opc & 3]); switch (opc) { @@ -1212,6 +1245,17 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc) tcg_abort(); } +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_call(s, qemu_st_helpers[opc]); /* label2: */ @@ -1617,23 +1661,18 @@ static int tcg_target_callee_save_regs[] = { TCG_REG_R18 }; +#define FRAME_SIZE ((-TCG_TARGET_CALL_STACK_OFFSET \ + + TCG_TARGET_STATIC_CALL_ARGS_SIZE \ + + ARRAY_SIZE(tcg_target_callee_save_regs) * 4 \ + + CPU_TEMP_BUF_NLONGS * sizeof(long) \ + + TCG_TARGET_STACK_ALIGN - 1) \ + & -TCG_TARGET_STACK_ALIGN) + static void tcg_target_qemu_prologue(TCGContext *s) { int frame_size, i; - /* Allocate space for the fixed frame marker. */ - frame_size = -TCG_TARGET_CALL_STACK_OFFSET; - frame_size += TCG_TARGET_STATIC_CALL_ARGS_SIZE; - - /* Allocate space for the saved registers. */ - frame_size += ARRAY_SIZE(tcg_target_callee_save_regs) * 4; - - /* Allocate space for the TCG temps. */ - frame_size += CPU_TEMP_BUF_NLONGS * sizeof(long); - - /* Align the allocated space. */ - frame_size = ((frame_size + TCG_TARGET_STACK_ALIGN - 1) - & -TCG_TARGET_STACK_ALIGN); + frame_size = FRAME_SIZE; /* The return address is stored in the caller's frame. */ tcg_out_st(s, TCG_TYPE_PTR, TCG_REG_RP, TCG_REG_CALL_STACK, -20); @@ -1708,3 +1747,81 @@ static void tcg_target_init(TCGContext *s) tcg_add_target_add_op_defs(hppa_op_defs); } + +typedef struct { + uint32_t len __attribute__((aligned((sizeof(void *))))); + uint32_t id; + uint8_t version; + char augmentation[1]; + uint8_t code_align; + uint8_t data_align; + uint8_t return_column; +} DebugFrameCIE; + +typedef struct { + uint32_t len __attribute__((aligned((sizeof(void *))))); + uint32_t cie_offset; + tcg_target_long func_start __attribute__((packed)); + tcg_target_long func_len __attribute__((packed)); + uint8_t def_cfa[4]; + uint8_t ret_ofs[3]; + uint8_t reg_ofs[ARRAY_SIZE(tcg_target_callee_save_regs) * 2]; +} DebugFrameFDE; + +typedef struct { + DebugFrameCIE cie; + DebugFrameFDE fde; +} DebugFrame; + +#define ELF_HOST_MACHINE EM_PARISC +#define ELF_HOST_FLAGS EFA_PARISC_1_1 + +/* ??? BFD (and thus GDB) wants very much to distinguish between HPUX + and other extensions. We don't really care, but if we don't set this + to *something* then the object file won't be properly matched. */ +#define ELF_OSABI ELFOSABI_LINUX + +static DebugFrame debug_frame = { + .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */ + .cie.id = -1, + .cie.version = 1, + .cie.code_align = 1, + .cie.data_align = 1, + .cie.return_column = 2, + + .fde.len = sizeof(DebugFrameFDE)-4, /* length after .len member */ + .fde.def_cfa = { + 0x12, 30, /* DW_CFA_def_cfa_sf sp, ... */ + (-FRAME_SIZE & 0x7f) | 0x80, /* ... sleb128 -FRAME_SIZE */ + (-FRAME_SIZE >> 7) & 0x7f + }, + .fde.ret_ofs = { + 0x11, 2, (-20 / 4) & 0x7f /* DW_CFA_offset_extended_sf r2, 20 */ + }, + .fde.reg_ofs = { + /* This must match the ordering in tcg_target_callee_save_regs. */ + 0x80 + 4, 0, /* DW_CFA_offset r4, 0 */ + 0x80 + 5, 4, /* DW_CFA_offset r5, 4 */ + 0x80 + 6, 8, /* DW_CFA_offset r6, 8 */ + 0x80 + 7, 12, /* ... */ + 0x80 + 8, 16, + 0x80 + 9, 20, + 0x80 + 10, 24, + 0x80 + 11, 28, + 0x80 + 12, 32, + 0x80 + 13, 36, + 0x80 + 14, 40, + 0x80 + 15, 44, + 0x80 + 16, 48, + 0x80 + 17, 52, + 0x80 + 18, 56, + } +}; + +void tcg_register_jit(void *buf, size_t buf_size) +{ + debug_frame.fde.func_start = (tcg_target_long) buf; + debug_frame.fde.func_len = buf_size; + + tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame)); +} diff --git a/tcg/i386/tcg-target.c b/tcg/i386/tcg-target.c index fafd900c5a..871a7e74f5 100644 --- a/tcg/i386/tcg-target.c +++ b/tcg/i386/tcg-target.c @@ -116,17 +116,7 @@ static inline int tcg_target_get_call_iarg_regs_count(int flags) return 6; } - flags &= TCG_CALL_TYPE_MASK; - switch(flags) { - case TCG_CALL_TYPE_STD: - return 0; - case TCG_CALL_TYPE_REGPARM_1: - case TCG_CALL_TYPE_REGPARM_2: - case TCG_CALL_TYPE_REGPARM: - return flags - TCG_CALL_TYPE_REGPARM_1 + 1; - default: - tcg_abort(); - } + return 0; } /* parse target specific constraints */ @@ -188,6 +178,9 @@ static int target_parse_constraint(TCGArgConstraint *ct, const char **pct_str) tcg_regset_set32(ct->u.regs, 0, 0xffff); tcg_regset_reset_reg(ct->u.regs, TCG_REG_RSI); tcg_regset_reset_reg(ct->u.regs, TCG_REG_RDI); +#ifdef CONFIG_TCG_PASS_AREG0 + tcg_regset_reset_reg(ct->u.regs, TCG_REG_RDX); +#endif } else { tcg_regset_set32(ct->u.regs, 0, 0xff); tcg_regset_reset_reg(ct->u.regs, TCG_REG_EAX); @@ -967,6 +960,27 @@ static void tcg_out_jmp(TCGContext *s, tcg_target_long dest) #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void *qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void *qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -974,12 +988,15 @@ static void *qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, __stl_mmu, __stq_mmu, }; +#endif /* Perform the TLB load and compare. @@ -1148,7 +1165,12 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int data_reg, data_reg2 = 0; int addrlo_idx; #if defined(CONFIG_SOFTMMU) - int mem_index, s_bits, arg_idx; + int mem_index, s_bits; +#if TCG_TARGET_REG_BITS == 64 + int arg_idx; +#else + int stack_adjust; +#endif uint8_t *label_ptr[3]; #endif @@ -1184,16 +1206,48 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, } /* XXX: move that code at the end of the TB */ +#if TCG_TARGET_REG_BITS == 32 + tcg_out_pushi(s, mem_index); + stack_adjust = 4; + if (TARGET_LONG_BITS == 64) { + tcg_out_push(s, args[addrlo_idx + 1]); + stack_adjust += 4; + } + tcg_out_push(s, args[addrlo_idx]); + stack_adjust += 4; +#ifdef CONFIG_TCG_PASS_AREG0 + tcg_out_push(s, TCG_AREG0); + stack_adjust += 4; +#endif +#else /* The first argument is already loaded with addrlo. */ arg_idx = 1; - if (TCG_TARGET_REG_BITS == 32 && TARGET_LONG_BITS == 64) { - tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[arg_idx++], - args[addrlo_idx + 1]); - } tcg_out_movi(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[arg_idx], mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif +#endif + tcg_out_calli(s, (tcg_target_long)qemu_ld_helpers[s_bits]); +#if TCG_TARGET_REG_BITS == 32 + if (stack_adjust == (TCG_TARGET_REG_BITS / 8)) { + /* Pop and discard. This is 2 bytes smaller than the add. */ + tcg_out_pop(s, TCG_REG_ECX); + } else if (stack_adjust != 0) { + tcg_out_addi(s, TCG_REG_CALL_STACK, stack_adjust); + } +#endif + switch(opc) { case 0 | 4: tcg_out_ext8s(s, data_reg, TCG_REG_EAX, P_REXW); @@ -1359,45 +1413,42 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, } /* XXX: move that code at the end of the TB */ - if (TCG_TARGET_REG_BITS == 64) { - tcg_out_mov(s, (opc == 3 ? TCG_TYPE_I64 : TCG_TYPE_I32), - TCG_REG_RSI, data_reg); - tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_RDX, mem_index); - stack_adjust = 0; - } else if (TARGET_LONG_BITS == 32) { - tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_EDX, data_reg); - if (opc == 3) { - tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_ECX, data_reg2); - tcg_out_pushi(s, mem_index); - stack_adjust = 4; - } else { - tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_ECX, mem_index); - stack_adjust = 0; - } - } else { - if (opc == 3) { - tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_EDX, args[addrlo_idx + 1]); - tcg_out_pushi(s, mem_index); - tcg_out_push(s, data_reg2); - tcg_out_push(s, data_reg); - stack_adjust = 12; - } else { - tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_EDX, args[addrlo_idx + 1]); - switch(opc) { - case 0: - tcg_out_ext8u(s, TCG_REG_ECX, data_reg); - break; - case 1: - tcg_out_ext16u(s, TCG_REG_ECX, data_reg); - break; - case 2: - tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_ECX, data_reg); - break; - } - tcg_out_pushi(s, mem_index); - stack_adjust = 4; - } +#if TCG_TARGET_REG_BITS == 32 + tcg_out_pushi(s, mem_index); + stack_adjust = 4; + if (opc == 3) { + tcg_out_push(s, data_reg2); + stack_adjust += 4; } + tcg_out_push(s, data_reg); + stack_adjust += 4; + if (TARGET_LONG_BITS == 64) { + tcg_out_push(s, args[addrlo_idx + 1]); + stack_adjust += 4; + } + tcg_out_push(s, args[addrlo_idx]); + stack_adjust += 4; +#ifdef CONFIG_TCG_PASS_AREG0 + tcg_out_push(s, TCG_AREG0); + stack_adjust += 4; +#endif +#else + tcg_out_mov(s, (opc == 3 ? TCG_TYPE_I64 : TCG_TYPE_I32), + TCG_REG_RSI, data_reg); + tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_RDX, mem_index); + stack_adjust = 0; +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif +#endif tcg_out_calli(s, (tcg_target_long)qemu_st_helpers[s_bits]); @@ -1938,22 +1989,29 @@ static int tcg_target_callee_save_regs[] = { #endif }; +/* Compute frame size via macros, to share between tcg_target_qemu_prologue + and tcg_register_jit. */ + +#define PUSH_SIZE \ + ((1 + ARRAY_SIZE(tcg_target_callee_save_regs)) \ + * (TCG_TARGET_REG_BITS / 8)) + +#define FRAME_SIZE \ + ((PUSH_SIZE \ + + TCG_STATIC_CALL_ARGS_SIZE \ + + CPU_TEMP_BUF_NLONGS * sizeof(long) \ + + TCG_TARGET_STACK_ALIGN - 1) \ + & ~(TCG_TARGET_STACK_ALIGN - 1)) + /* Generate global QEMU prologue and epilogue code */ static void tcg_target_qemu_prologue(TCGContext *s) { - int i, frame_size, push_size, stack_addend; + int i, stack_addend; /* TB prologue */ /* Reserve some stack space, also for TCG temps. */ - push_size = 1 + ARRAY_SIZE(tcg_target_callee_save_regs); - push_size *= TCG_TARGET_REG_BITS / 8; - - frame_size = push_size + TCG_STATIC_CALL_ARGS_SIZE + - CPU_TEMP_BUF_NLONGS * sizeof(long); - frame_size = (frame_size + TCG_TARGET_STACK_ALIGN - 1) & - ~(TCG_TARGET_STACK_ALIGN - 1); - stack_addend = frame_size - push_size; + stack_addend = FRAME_SIZE - PUSH_SIZE; tcg_set_frame(s, TCG_REG_CALL_STACK, TCG_STATIC_CALL_ARGS_SIZE, CPU_TEMP_BUF_NLONGS * sizeof(long)); @@ -1962,9 +2020,15 @@ static void tcg_target_qemu_prologue(TCGContext *s) tcg_out_push(s, tcg_target_callee_save_regs[i]); } - tcg_out_addi(s, TCG_REG_ESP, -stack_addend); - +#if TCG_TARGET_REG_BITS == 32 + tcg_out_ld(s, TCG_TYPE_PTR, TCG_AREG0, TCG_REG_ESP, + (ARRAY_SIZE(tcg_target_callee_save_regs) + 1) * 4); + tcg_out_ld(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[1], TCG_REG_ESP, + (ARRAY_SIZE(tcg_target_callee_save_regs) + 2) * 4); +#else tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, tcg_target_call_iarg_regs[0]); +#endif + tcg_out_addi(s, TCG_REG_ESP, -stack_addend); /* jmp *tb. */ tcg_out_modrm(s, OPC_GRP5, EXT5_JMPN_Ev, tcg_target_call_iarg_regs[1]); @@ -2013,3 +2077,92 @@ static void tcg_target_init(TCGContext *s) tcg_add_target_add_op_defs(x86_op_defs); } + +typedef struct { + uint32_t len __attribute__((aligned((sizeof(void *))))); + uint32_t id; + uint8_t version; + char augmentation[1]; + uint8_t code_align; + uint8_t data_align; + uint8_t return_column; +} DebugFrameCIE; + +typedef struct { + uint32_t len __attribute__((aligned((sizeof(void *))))); + uint32_t cie_offset; + tcg_target_long func_start __attribute__((packed)); + tcg_target_long func_len __attribute__((packed)); + uint8_t def_cfa[4]; + uint8_t reg_ofs[14]; +} DebugFrameFDE; + +typedef struct { + DebugFrameCIE cie; + DebugFrameFDE fde; +} DebugFrame; + +#if TCG_TARGET_REG_BITS == 64 +#define ELF_HOST_MACHINE EM_X86_64 +static DebugFrame debug_frame = { + .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */ + .cie.id = -1, + .cie.version = 1, + .cie.code_align = 1, + .cie.data_align = 0x78, /* sleb128 -8 */ + .cie.return_column = 16, + + .fde.len = sizeof(DebugFrameFDE)-4, /* length after .len member */ + .fde.def_cfa = { + 12, 7, /* DW_CFA_def_cfa %rsp, ... */ + (FRAME_SIZE & 0x7f) | 0x80, /* ... uleb128 FRAME_SIZE */ + (FRAME_SIZE >> 7) + }, + .fde.reg_ofs = { + 0x90, 1, /* DW_CFA_offset, %rip, -8 */ + /* The following ordering must match tcg_target_callee_save_regs. */ + 0x86, 2, /* DW_CFA_offset, %rbp, -16 */ + 0x83, 3, /* DW_CFA_offset, %rbx, -24 */ + 0x8c, 4, /* DW_CFA_offset, %r12, -32 */ + 0x8d, 5, /* DW_CFA_offset, %r13, -40 */ + 0x8e, 6, /* DW_CFA_offset, %r14, -48 */ + 0x8f, 7, /* DW_CFA_offset, %r15, -56 */ + } +}; +#else +#define ELF_HOST_MACHINE EM_386 +static DebugFrame debug_frame = { + .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */ + .cie.id = -1, + .cie.version = 1, + .cie.code_align = 1, + .cie.data_align = 0x7c, /* sleb128 -4 */ + .cie.return_column = 8, + + .fde.len = sizeof(DebugFrameFDE)-4, /* length after .len member */ + .fde.def_cfa = { + 12, 4, /* DW_CFA_def_cfa %esp, ... */ + (FRAME_SIZE & 0x7f) | 0x80, /* ... uleb128 FRAME_SIZE */ + (FRAME_SIZE >> 7) + }, + .fde.reg_ofs = { + 0x88, 1, /* DW_CFA_offset, %eip, -4 */ + /* The following ordering must match tcg_target_callee_save_regs. */ + 0x85, 2, /* DW_CFA_offset, %ebp, -8 */ + 0x83, 3, /* DW_CFA_offset, %ebx, -12 */ + 0x86, 4, /* DW_CFA_offset, %esi, -16 */ + 0x87, 5, /* DW_CFA_offset, %edi, -20 */ + } +}; +#endif + +void tcg_register_jit(void *buf, size_t buf_size) +{ + /* We're expecting a 2 byte uleb128 encoded value. */ + assert(FRAME_SIZE >> 14 == 0); + + debug_frame.fde.func_start = (tcg_target_long) buf; + debug_frame.fde.func_len = buf_size; + + tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame)); +} diff --git a/tcg/ia64/tcg-target.c b/tcg/ia64/tcg-target.c index f90252a443..e02dacc84f 100644 --- a/tcg/ia64/tcg-target.c +++ b/tcg/ia64/tcg-target.c @@ -1452,12 +1452,25 @@ static inline void tcg_out_qemu_tlb(TCGContext *s, TCGArg addr_reg, TCG_REG_P7, TCG_REG_R3, TCG_REG_R57)); } +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, __ldl_mmu, __ldq_mmu, }; +#endif static inline void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc) { @@ -1517,6 +1530,15 @@ static inline void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc) tcg_opc_m1 (TCG_REG_P7, OPC_LD8_M1, TCG_REG_R1, TCG_REG_R2), tcg_opc_i18(TCG_REG_P0, OPC_NOP_I18, 0)); } +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif if (!bswap || s_bits == 0) { tcg_out_bundle(s, miB, tcg_opc_m48(TCG_REG_P0, OPC_NOP_M48, 0), @@ -1547,12 +1569,25 @@ static inline void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc) } } +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, __stl_mmu, __stq_mmu, }; +#endif static inline void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc) { @@ -1622,6 +1657,17 @@ static inline void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc) data_reg = TCG_REG_R2; } +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_bundle(s, miB, tcg_opc_m4 (TCG_REG_P6, opc_st_m4[opc], data_reg, TCG_REG_R3), diff --git a/tcg/mips/tcg-target.c b/tcg/mips/tcg-target.c index c6aa5bced5..393ba07f25 100644 --- a/tcg/mips/tcg-target.c +++ b/tcg/mips/tcg-target.c @@ -750,6 +750,27 @@ static void tcg_out_setcond2(TCGContext *s, TCGCond cond, int ret, #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -757,6 +778,8 @@ static void *qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, @@ -764,6 +787,7 @@ static void *qemu_st_helpers[4] = { __stq_mmu, }; #endif +#endif static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc) @@ -858,6 +882,15 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, # endif tcg_out_movi(s, TCG_TYPE_I32, sp_args++, mem_index); tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_T9, (tcg_target_long)qemu_ld_helpers[s_bits]); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal and incorrect for 64 on 32 bit */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_opc_reg(s, OPC_JALR, TCG_REG_RA, TCG_REG_T9, 0); tcg_out_nop(s); @@ -1069,6 +1102,17 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, } tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_T9, (tcg_target_long)qemu_st_helpers[s_bits]); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal and incorrect for 64 on 32 bit */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_opc_reg(s, OPC_JALR, TCG_REG_RA, TCG_REG_T9, 0); tcg_out_nop(s); diff --git a/tcg/ppc/tcg-target.c b/tcg/ppc/tcg-target.c index 6a34cab5f9..b0aa914798 100644 --- a/tcg/ppc/tcg-target.c +++ b/tcg/ppc/tcg-target.c @@ -508,6 +508,27 @@ static void tcg_out_call (TCGContext *s, tcg_target_long arg, int const_arg) #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -515,6 +536,8 @@ static void *qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, @@ -522,6 +545,7 @@ static void *qemu_st_helpers[4] = { __stq_mmu, }; #endif +#endif static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc) { @@ -598,6 +622,16 @@ static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc) tcg_out_movi (s, TCG_TYPE_I32, 5, mem_index); #endif +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif + tcg_out_call (s, (tcg_target_long) qemu_ld_helpers[s_bits], 1); switch (opc) { case 0|4: @@ -829,6 +863,17 @@ static void tcg_out_qemu_st (TCGContext *s, const TCGArg *args, int opc) ir++; tcg_out_movi (s, TCG_TYPE_I32, ir, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_call (s, (tcg_target_long) qemu_st_helpers[opc], 1); label2_ptr = s->code_ptr; tcg_out32 (s, B); diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index 3f22aaac9d..2f37fd289b 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -98,5 +98,5 @@ typedef enum { #define TCG_TARGET_HAS_GUEST_BASE #define tcg_qemu_tb_exec(env, tb_ptr) \ - ((long REGPARM __attribute__ ((longcall)) \ + ((long __attribute__ ((longcall)) \ (*)(void *, void *))code_gen_prologue)(env, tb_ptr) diff --git a/tcg/ppc64/tcg-target.c b/tcg/ppc64/tcg-target.c index 7f723b5c2c..f2ad9e3d85 100644 --- a/tcg/ppc64/tcg-target.c +++ b/tcg/ppc64/tcg-target.c @@ -552,6 +552,27 @@ static void tcg_out_ldsta (TCGContext *s, int ret, int addr, #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -559,12 +580,15 @@ static void *qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, __stl_mmu, __stq_mmu, }; +#endif static void tcg_out_tlb_read (TCGContext *s, int r0, int r1, int r2, int addr_reg, int s_bits, int offset) @@ -648,6 +672,15 @@ static void tcg_out_qemu_ld (TCGContext *s, const TCGArg *args, int opc) tcg_out_mov (s, TCG_TYPE_I64, 3, addr_reg); tcg_out_movi (s, TCG_TYPE_I64, 4, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_call (s, (tcg_target_long) qemu_ld_helpers[s_bits], 1); switch (opc) { @@ -796,6 +829,17 @@ static void tcg_out_qemu_st (TCGContext *s, const TCGArg *args, int opc) tcg_out_rld (s, RLDICL, 4, data_reg, 0, 64 - (1 << (3 + opc))); tcg_out_movi (s, TCG_TYPE_I64, 5, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tcg_out_call (s, (tcg_target_long) qemu_st_helpers[opc], 1); label2_ptr = s->code_ptr; @@ -1612,7 +1656,6 @@ static const TCGTargetOpDef ppc_op_defs[] = { { INDEX_op_ld16s_i64, { "r", "r" } }, { INDEX_op_ld32u_i64, { "r", "r" } }, { INDEX_op_ld32s_i64, { "r", "r" } }, - { INDEX_op_ld_i64, { "r", "r" } }, { INDEX_op_add_i32, { "r", "r", "ri" } }, { INDEX_op_mul_i32, { "r", "r", "ri" } }, diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c index 47ffcc1f51..04662c15fd 100644 --- a/tcg/s390/tcg-target.c +++ b/tcg/s390/tcg-target.c @@ -301,6 +301,27 @@ static const uint8_t tcg_cond_to_ltr_cond[10] = { #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static void *qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -308,6 +329,8 @@ static void *qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static void *qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, @@ -315,6 +338,7 @@ static void *qemu_st_helpers[4] = { __stq_mmu, }; #endif +#endif static uint8_t *tb_ret_addr; @@ -1483,9 +1507,29 @@ static void tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg, tcg_abort(); } tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_R4, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tgen_calli(s, (tcg_target_ulong)qemu_st_helpers[s_bits]); } else { tcg_out_movi(s, TCG_TYPE_I32, arg1, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif tgen_calli(s, (tcg_target_ulong)qemu_ld_helpers[s_bits]); /* sign extension */ diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index b287122df5..247a27814a 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -59,6 +59,12 @@ static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = { }; #endif +#ifdef CONFIG_TCG_PASS_AREG0 +#define ARG_OFFSET 1 +#else +#define ARG_OFFSET 0 +#endif + static const int tcg_target_reg_alloc_order[] = { TCG_REG_L0, TCG_REG_L1, @@ -86,9 +92,9 @@ static const int tcg_target_call_iarg_regs[6] = { static const int tcg_target_call_oarg_regs[] = { TCG_REG_O0, -#if TCG_TARGET_REG_BITS == 32 - TCG_REG_O1 -#endif + TCG_REG_O1, + TCG_REG_O2, + TCG_REG_O3, }; static inline int check_fit_tl(tcg_target_long val, unsigned int bits) @@ -155,6 +161,9 @@ static int target_parse_constraint(TCGArgConstraint *ct, const char **pct_str) tcg_regset_reset_reg(ct->u.regs, TCG_REG_O0); tcg_regset_reset_reg(ct->u.regs, TCG_REG_O1); tcg_regset_reset_reg(ct->u.regs, TCG_REG_O2); +#ifdef CONFIG_TCG_PASS_AREG0 + tcg_regset_reset_reg(ct->u.regs, TCG_REG_O3); +#endif break; case 'I': ct->ct |= TCG_CT_CONST_S11; @@ -706,6 +715,27 @@ static void tcg_target_qemu_prologue(TCGContext *s) #include "../../softmmu_defs.h" +#ifdef CONFIG_TCG_PASS_AREG0 +/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, + int mmu_idx) */ +static const void * const qemu_ld_helpers[4] = { + helper_ldb_mmu, + helper_ldw_mmu, + helper_ldl_mmu, + helper_ldq_mmu, +}; + +/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr, + uintxx_t val, int mmu_idx) */ +static const void * const qemu_st_helpers[4] = { + helper_stb_mmu, + helper_stw_mmu, + helper_stl_mmu, + helper_stq_mmu, +}; +#else +/* legacy helper signature: __ld_mmu(target_ulong addr, int + mmu_idx) */ static const void * const qemu_ld_helpers[4] = { __ldb_mmu, __ldw_mmu, @@ -713,6 +743,8 @@ static const void * const qemu_ld_helpers[4] = { __ldq_mmu, }; +/* legacy helper signature: __st_mmu(target_ulong addr, uintxx_t val, + int mmu_idx) */ static const void * const qemu_st_helpers[4] = { __stb_mmu, __stw_mmu, @@ -720,6 +752,7 @@ static const void * const qemu_st_helpers[4] = { __stq_mmu, }; #endif +#endif #if TARGET_LONG_BITS == 32 #define TARGET_LD_OP LDUW @@ -801,6 +834,17 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, /* mov */ tcg_out_movi(s, TCG_TYPE_I32, arg1, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif /* XXX: move that code at the end of the TB */ /* qemu_ld_helper[s_bits](arg0, arg1) */ @@ -1017,6 +1061,17 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, /* mov */ tcg_out_movi(s, TCG_TYPE_I32, arg2, mem_index); +#ifdef CONFIG_TCG_PASS_AREG0 + /* XXX/FIXME: suboptimal */ + tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], + tcg_target_call_iarg_regs[2]); + tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], + tcg_target_call_iarg_regs[1]); + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + tcg_target_call_iarg_regs[0]); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], + TCG_AREG0); +#endif /* XXX: move that code at the end of the TB */ /* qemu_st_helper[s_bits](arg0, arg1, arg2) */ tcg_out32(s, CALL | ((((tcg_target_ulong)qemu_st_helpers[s_bits] @@ -1569,3 +1624,66 @@ static void tcg_target_init(TCGContext *s) tcg_regset_set_reg(s->reserved_regs, TCG_REG_O7); tcg_add_target_add_op_defs(sparc_op_defs); } + +#if TCG_TARGET_REG_BITS == 64 +# define ELF_HOST_MACHINE EM_SPARCV9 +#elif defined(__sparc_v8plus__) +# define ELF_HOST_MACHINE EM_SPARC32PLUS +# define ELF_HOST_FLAGS EF_SPARC_32PLUS +#else +# define ELF_HOST_MACHINE EM_SPARC +#endif + +typedef struct { + uint32_t len __attribute__((aligned((sizeof(void *))))); + uint32_t id; + uint8_t version; + char augmentation[1]; + uint8_t code_align; + uint8_t data_align; + uint8_t return_column; +} DebugFrameCIE; + +typedef struct { + uint32_t len __attribute__((aligned((sizeof(void *))))); + uint32_t cie_offset; + tcg_target_long func_start __attribute__((packed)); + tcg_target_long func_len __attribute__((packed)); + uint8_t def_cfa[TCG_TARGET_REG_BITS == 64 ? 4 : 2]; + uint8_t win_save; + uint8_t ret_save[3]; +} DebugFrameFDE; + +typedef struct { + DebugFrameCIE cie; + DebugFrameFDE fde; +} DebugFrame; + +static DebugFrame debug_frame = { + .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */ + .cie.id = -1, + .cie.version = 1, + .cie.code_align = 1, + .cie.data_align = -sizeof(void *) & 0x7f, + .cie.return_column = 15, /* o7 */ + + .fde.len = sizeof(DebugFrameFDE)-4, /* length after .len member */ + .fde.def_cfa = { +#if TCG_TARGET_REG_BITS == 64 + 12, 30, /* DW_CFA_def_cfa i6, 2047 */ + (2047 & 0x7f) | 0x80, (2047 >> 7) +#else + 13, 30 /* DW_CFA_def_cfa_register i6 */ +#endif + }, + .fde.win_save = 0x2d, /* DW_CFA_GNU_window_save */ + .fde.ret_save = { 9, 15, 31 }, /* DW_CFA_register o7, i7 */ +}; + +void tcg_register_jit(void *buf, size_t buf_size) +{ + debug_frame.fde.func_start = (tcg_target_long) buf; + debug_frame.fde.func_len = buf_size; + + tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame)); +} @@ -28,6 +28,9 @@ #include "config.h" +/* Define to jump the ELF file used to communicate with GDB. */ +#undef DEBUG_JIT + #if !defined(CONFIG_DEBUG_TCG) && !defined(NDEBUG) /* define it to suppress various consistency checks (faster) */ #define NDEBUG @@ -45,6 +48,18 @@ #include "cpu.h" #include "tcg-op.h" + +#if TCG_TARGET_REG_BITS == 64 +# define ELF_CLASS ELFCLASS64 +#else +# define ELF_CLASS ELFCLASS32 +#endif +#ifdef HOST_WORDS_BIGENDIAN +# define ELF_DATA ELFDATA2MSB +#else +# define ELF_DATA ELFDATA2LSB +#endif + #include "elf.h" #if defined(CONFIG_USE_GUEST_BASE) && !defined(TCG_TARGET_HAS_GUEST_BASE) @@ -57,6 +72,10 @@ static void tcg_target_qemu_prologue(TCGContext *s); static void patch_reloc(uint8_t *code_ptr, int type, tcg_target_long value, tcg_target_long addend); +static void tcg_register_jit_int(void *buf, size_t size, + void *debug_frame, size_t debug_frame_size) + __attribute__((unused)); + /* Forward declarations for functions declared and used in tcg-target.c. */ static int target_parse_constraint(TCGArgConstraint *ct, const char **pct_str); static void tcg_out_ld(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg1, @@ -173,11 +192,9 @@ void *tcg_malloc_internal(TCGContext *s, int size) /* big malloc: insert a new pool (XXX: could optimize) */ p = g_malloc(sizeof(TCGPool) + size); p->size = size; - if (s->pool_current) - s->pool_current->next = p; - else - s->pool_first = p; - p->next = s->pool_current; + p->next = s->pool_first_large; + s->pool_first_large = p; + return p->data; } else { p = s->pool_current; if (!p) { @@ -208,6 +225,12 @@ void *tcg_malloc_internal(TCGContext *s, int size) void tcg_pool_reset(TCGContext *s) { + TCGPool *p, *t; + for (p = s->pool_first_large; p; p = t) { + t = p->next; + g_free(p); + } + s->pool_first_large = NULL; s->pool_cur = s->pool_end = NULL; s->pool_current = NULL; } @@ -590,9 +613,6 @@ void tcg_register_helper(void *func, const char *name) void tcg_gen_callN(TCGContext *s, TCGv_ptr func, unsigned int flags, int sizemask, TCGArg ret, int nargs, TCGArg *args) { -#if defined(TCG_TARGET_I386) && TCG_TARGET_REG_BITS < 64 - int call_type; -#endif int i; int real_args; int nb_rets; @@ -617,9 +637,6 @@ void tcg_gen_callN(TCGContext *s, TCGv_ptr func, unsigned int flags, *gen_opc_ptr++ = INDEX_op_call; nparam = gen_opparam_ptr++; -#if defined(TCG_TARGET_I386) && TCG_TARGET_REG_BITS < 64 - call_type = (flags & TCG_CALL_TYPE_MASK); -#endif if (ret != TCG_CALL_DUMMY_ARG) { #if TCG_TARGET_REG_BITS < 64 if (sizemask & 1) { @@ -645,14 +662,6 @@ void tcg_gen_callN(TCGContext *s, TCGv_ptr func, unsigned int flags, #if TCG_TARGET_REG_BITS < 64 int is_64bit = sizemask & (1 << (i+1)*2); if (is_64bit) { -#ifdef TCG_TARGET_I386 - /* REGPARM case: if the third parameter is 64 bit, it is - allocated on the stack */ - if (i == 2 && call_type == TCG_CALL_TYPE_REGPARM) { - call_type = TCG_CALL_TYPE_REGPARM_2; - flags = (flags & ~TCG_CALL_TYPE_MASK) | call_type; - } -#endif #ifdef TCG_TARGET_CALL_ALIGN_ARGS /* some targets want aligned 64 bit args */ if (real_args & 1) { @@ -2241,3 +2250,267 @@ void tcg_dump_info(FILE *f, fprintf_function cpu_fprintf) cpu_fprintf(f, "[TCG profiler not compiled]\n"); } #endif + +#ifdef ELF_HOST_MACHINE +/* In order to use this feature, the backend needs to do three things: + + (1) Define ELF_HOST_MACHINE to indicate both what value to + put into the ELF image and to indicate support for the feature. + + (2) Define tcg_register_jit. This should create a buffer containing + the contents of a .debug_frame section that describes the post- + prologue unwind info for the tcg machine. + + (3) Call tcg_register_jit_int, with the constructed .debug_frame. +*/ + +/* Begin GDB interface. THE FOLLOWING MUST MATCH GDB DOCS. */ +typedef enum { + JIT_NOACTION = 0, + JIT_REGISTER_FN, + JIT_UNREGISTER_FN +} jit_actions_t; + +struct jit_code_entry { + struct jit_code_entry *next_entry; + struct jit_code_entry *prev_entry; + const void *symfile_addr; + uint64_t symfile_size; +}; + +struct jit_descriptor { + uint32_t version; + uint32_t action_flag; + struct jit_code_entry *relevant_entry; + struct jit_code_entry *first_entry; +}; + +void __jit_debug_register_code(void) __attribute__((noinline)); +void __jit_debug_register_code(void) +{ + asm(""); +} + +/* Must statically initialize the version, because GDB may check + the version before we can set it. */ +struct jit_descriptor __jit_debug_descriptor = { 1, 0, 0, 0 }; + +/* End GDB interface. */ + +static int find_string(const char *strtab, const char *str) +{ + const char *p = strtab + 1; + + while (1) { + if (strcmp(p, str) == 0) { + return p - strtab; + } + p += strlen(p) + 1; + } +} + +static void tcg_register_jit_int(void *buf_ptr, size_t buf_size, + void *debug_frame, size_t debug_frame_size) +{ + struct __attribute__((packed)) DebugInfo { + uint32_t len; + uint16_t version; + uint32_t abbrev; + uint8_t ptr_size; + uint8_t cu_die; + uint16_t cu_lang; + uintptr_t cu_low_pc; + uintptr_t cu_high_pc; + uint8_t fn_die; + char fn_name[16]; + uintptr_t fn_low_pc; + uintptr_t fn_high_pc; + uint8_t cu_eoc; + }; + + struct ElfImage { + ElfW(Ehdr) ehdr; + ElfW(Phdr) phdr; + ElfW(Shdr) shdr[7]; + ElfW(Sym) sym[2]; + struct DebugInfo di; + uint8_t da[24]; + char str[80]; + }; + + struct ElfImage *img; + + static const struct ElfImage img_template = { + .ehdr = { + .e_ident[EI_MAG0] = ELFMAG0, + .e_ident[EI_MAG1] = ELFMAG1, + .e_ident[EI_MAG2] = ELFMAG2, + .e_ident[EI_MAG3] = ELFMAG3, + .e_ident[EI_CLASS] = ELF_CLASS, + .e_ident[EI_DATA] = ELF_DATA, + .e_ident[EI_VERSION] = EV_CURRENT, + .e_type = ET_EXEC, + .e_machine = ELF_HOST_MACHINE, + .e_version = EV_CURRENT, + .e_phoff = offsetof(struct ElfImage, phdr), + .e_shoff = offsetof(struct ElfImage, shdr), + .e_ehsize = sizeof(ElfW(Shdr)), + .e_phentsize = sizeof(ElfW(Phdr)), + .e_phnum = 1, + .e_shentsize = sizeof(ElfW(Shdr)), + .e_shnum = ARRAY_SIZE(img->shdr), + .e_shstrndx = ARRAY_SIZE(img->shdr) - 1, +#ifdef ELF_HOST_FLAGS + .e_flags = ELF_HOST_FLAGS, +#endif +#ifdef ELF_OSABI + .e_ident[EI_OSABI] = ELF_OSABI, +#endif + }, + .phdr = { + .p_type = PT_LOAD, + .p_flags = PF_X, + }, + .shdr = { + [0] = { .sh_type = SHT_NULL }, + /* Trick: The contents of code_gen_buffer are not present in + this fake ELF file; that got allocated elsewhere. Therefore + we mark .text as SHT_NOBITS (similar to .bss) so that readers + will not look for contents. We can record any address. */ + [1] = { /* .text */ + .sh_type = SHT_NOBITS, + .sh_flags = SHF_EXECINSTR | SHF_ALLOC, + }, + [2] = { /* .debug_info */ + .sh_type = SHT_PROGBITS, + .sh_offset = offsetof(struct ElfImage, di), + .sh_size = sizeof(struct DebugInfo), + }, + [3] = { /* .debug_abbrev */ + .sh_type = SHT_PROGBITS, + .sh_offset = offsetof(struct ElfImage, da), + .sh_size = sizeof(img->da), + }, + [4] = { /* .debug_frame */ + .sh_type = SHT_PROGBITS, + .sh_offset = sizeof(struct ElfImage), + }, + [5] = { /* .symtab */ + .sh_type = SHT_SYMTAB, + .sh_offset = offsetof(struct ElfImage, sym), + .sh_size = sizeof(img->sym), + .sh_info = 1, + .sh_link = ARRAY_SIZE(img->shdr) - 1, + .sh_entsize = sizeof(ElfW(Sym)), + }, + [6] = { /* .strtab */ + .sh_type = SHT_STRTAB, + .sh_offset = offsetof(struct ElfImage, str), + .sh_size = sizeof(img->str), + } + }, + .sym = { + [1] = { /* code_gen_buffer */ + .st_info = ELF_ST_INFO(STB_GLOBAL, STT_FUNC), + .st_shndx = 1, + } + }, + .di = { + .len = sizeof(struct DebugInfo) - 4, + .version = 2, + .ptr_size = sizeof(void *), + .cu_die = 1, + .cu_lang = 0x8001, /* DW_LANG_Mips_Assembler */ + .fn_die = 2, + .fn_name = "code_gen_buffer" + }, + .da = { + 1, /* abbrev number (the cu) */ + 0x11, 1, /* DW_TAG_compile_unit, has children */ + 0x13, 0x5, /* DW_AT_language, DW_FORM_data2 */ + 0x11, 0x1, /* DW_AT_low_pc, DW_FORM_addr */ + 0x12, 0x1, /* DW_AT_high_pc, DW_FORM_addr */ + 0, 0, /* end of abbrev */ + 2, /* abbrev number (the fn) */ + 0x2e, 0, /* DW_TAG_subprogram, no children */ + 0x3, 0x8, /* DW_AT_name, DW_FORM_string */ + 0x11, 0x1, /* DW_AT_low_pc, DW_FORM_addr */ + 0x12, 0x1, /* DW_AT_high_pc, DW_FORM_addr */ + 0, 0, /* end of abbrev */ + 0 /* no more abbrev */ + }, + .str = "\0" ".text\0" ".debug_info\0" ".debug_abbrev\0" + ".debug_frame\0" ".symtab\0" ".strtab\0" "code_gen_buffer", + }; + + /* We only need a single jit entry; statically allocate it. */ + static struct jit_code_entry one_entry; + + uintptr_t buf = (uintptr_t)buf_ptr; + size_t img_size = sizeof(struct ElfImage) + debug_frame_size; + + img = g_malloc(img_size); + *img = img_template; + memcpy(img + 1, debug_frame, debug_frame_size); + + img->phdr.p_vaddr = buf; + img->phdr.p_paddr = buf; + img->phdr.p_memsz = buf_size; + + img->shdr[1].sh_name = find_string(img->str, ".text"); + img->shdr[1].sh_addr = buf; + img->shdr[1].sh_size = buf_size; + + img->shdr[2].sh_name = find_string(img->str, ".debug_info"); + img->shdr[3].sh_name = find_string(img->str, ".debug_abbrev"); + + img->shdr[4].sh_name = find_string(img->str, ".debug_frame"); + img->shdr[4].sh_size = debug_frame_size; + + img->shdr[5].sh_name = find_string(img->str, ".symtab"); + img->shdr[6].sh_name = find_string(img->str, ".strtab"); + + img->sym[1].st_name = find_string(img->str, "code_gen_buffer"); + img->sym[1].st_value = buf; + img->sym[1].st_size = buf_size; + + img->di.cu_low_pc = buf; + img->di.cu_high_pc = buf_size; + img->di.fn_low_pc = buf; + img->di.fn_high_pc = buf_size; + +#ifdef DEBUG_JIT + /* Enable this block to be able to debug the ELF image file creation. + One can use readelf, objdump, or other inspection utilities. */ + { + FILE *f = fopen("/tmp/qemu.jit", "w+b"); + if (f) { + if (fwrite(img, img_size, 1, f) != img_size) { + /* Avoid stupid unused return value warning for fwrite. */ + } + fclose(f); + } + } +#endif + + one_entry.symfile_addr = img; + one_entry.symfile_size = img_size; + + __jit_debug_descriptor.action_flag = JIT_REGISTER_FN; + __jit_debug_descriptor.relevant_entry = &one_entry; + __jit_debug_descriptor.first_entry = &one_entry; + __jit_debug_register_code(); +} +#else +/* No support for the feature. Provide the entry point expected by exec.c, + and implement the internal function we declared earlier. */ + +static void tcg_register_jit_int(void *buf, size_t size, + void *debug_frame, size_t debug_frame_size) +{ +} + +void tcg_register_jit(void *buf, size_t buf_size) +{ +} +#endif /* ELF_HOST_MACHINE */ @@ -252,11 +252,6 @@ typedef int TCGv_i64; #define TCGV_UNUSED_I64(x) x = MAKE_TCGV_I64(-1) /* call flags */ -#define TCG_CALL_TYPE_MASK 0x000f -#define TCG_CALL_TYPE_STD 0x0000 /* standard C call */ -#define TCG_CALL_TYPE_REGPARM_1 0x0001 /* i386 style regparm call (1 reg) */ -#define TCG_CALL_TYPE_REGPARM_2 0x0002 /* i386 style regparm call (2 regs) */ -#define TCG_CALL_TYPE_REGPARM 0x0003 /* i386 style regparm call (3 regs) */ /* A pure function only reads its arguments and TCG global variables and cannot raise exceptions. Hence a call to a pure function can be safely suppressed if the return value is not used. */ @@ -337,7 +332,7 @@ typedef struct TCGContext TCGContext; struct TCGContext { uint8_t *pool_cur, *pool_end; - TCGPool *pool_first, *pool_current; + TCGPool *pool_first, *pool_current, *pool_first_large; TCGLabel *labels; int nb_labels; TCGTemp *temps; /* globals first, temps after */ @@ -589,5 +584,7 @@ extern uint8_t code_gen_prologue[]; /* TCG targets may use a different definition of tcg_qemu_tb_exec. */ #if !defined(tcg_qemu_tb_exec) # define tcg_qemu_tb_exec(env, tb_ptr) \ - ((long REGPARM (*)(void *, void *))code_gen_prologue)(env, tb_ptr) + ((tcg_target_ulong (*)(void *, void *))code_gen_prologue)(env, tb_ptr) #endif + +void tcg_register_jit(void *buf, size_t buf_size); diff --git a/tcg/tci/tcg-target.c b/tcg/tci/tcg-target.c index bd85073662..453f1875e2 100644 --- a/tcg/tci/tcg-target.c +++ b/tcg/tci/tcg-target.c @@ -798,6 +798,9 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, case INDEX_op_qemu_st8: case INDEX_op_qemu_st16: case INDEX_op_qemu_st32: +#ifdef CONFIG_TCG_PASS_AREG0 + tcg_out_r(s, TCG_AREG0); +#endif tcg_out_r(s, *args++); tcg_out_r(s, *args++); #if TARGET_LONG_BITS > TCG_TARGET_REG_BITS @@ -808,6 +811,9 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, #endif break; case INDEX_op_qemu_st64: +#ifdef CONFIG_TCG_PASS_AREG0 + tcg_out_r(s, TCG_AREG0); +#endif tcg_out_r(s, *args++); #if TCG_TARGET_REG_BITS == 32 tcg_out_r(s, *args++); diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h index b61e99aff1..30a0f21596 100644 --- a/tcg/tci/tcg-target.h +++ b/tcg/tci/tcg-target.h @@ -154,7 +154,7 @@ typedef enum { void tci_disas(uint8_t opc); -unsigned long tcg_qemu_tb_exec(CPUArchState *env, uint8_t *tb_ptr); +tcg_target_ulong tcg_qemu_tb_exec(CPUArchState *env, uint8_t *tb_ptr); #define tcg_qemu_tb_exec tcg_qemu_tb_exec static inline void flush_icache_range(tcg_target_ulong start, |