diff options
Diffstat (limited to 'target-cris')
-rw-r--r-- | target-cris/translate.c | 656 |
1 files changed, 312 insertions, 344 deletions
diff --git a/target-cris/translate.c b/target-cris/translate.c index e653afd942..00893a9821 100644 --- a/target-cris/translate.c +++ b/target-cris/translate.c @@ -1,7 +1,7 @@ /* * CRIS emulation for qemu: main translation routines. * - * Copyright (c) 2007 AXIS Communications AB + * Copyright (c) 2008 AXIS Communications AB * Written by Edgar E. Iglesias. * * This library is free software; you can redistribute it and/or @@ -19,28 +19,6 @@ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ -/* - * This file implements a CRIS decoder-stage in SW. The decoder translates the - * guest (CRIS) machine-code into host machine code via dyngen using the - * micro-operations described in op.c - * - * The micro-operations for CRIS translation implement a RISC style ISA. - * Note that the micro-operations typically order their operands - * starting with the dst. CRIS asm, does the opposite. - * - * For example the following CRIS code: - * add.d [$r0], $r1 - * - * translates into: - * - * gen_movl_T0_reg(0); // Fetch $r0 into T0 - * gen_load_T0_T0(); // Load T0, @T0 - * gen_movl_reg_T0(1); // Writeback T0 into $r1 - * - * The actual names for the micro-code generators vary but the example - * illustrates the point. - */ - #include <stdarg.h> #include <stdlib.h> #include <stdio.h> @@ -52,6 +30,7 @@ #include "exec-all.h" #include "disas.h" #include "tcg-op.h" +#include "helper.h" #include "crisv32-decode.h" #define CRIS_STATS 0 @@ -82,6 +61,8 @@ #define CC_MASK_NZVC 0xf #define CC_MASK_RNZV 0x10e +TCGv cpu_env, cpu_T[2]; + /* This is the state at translation time. */ typedef struct DisasContext { CPUState *env; @@ -128,92 +109,6 @@ static void gen_BUG(DisasContext *dc, char *file, int line) cris_prepare_jmp (dc, 0x70000000 + line); } -/* Table to generate quick moves from T0 onto any register. */ -static GenOpFunc *gen_movl_reg_T0[16] = -{ - gen_op_movl_r0_T0, gen_op_movl_r1_T0, - gen_op_movl_r2_T0, gen_op_movl_r3_T0, - gen_op_movl_r4_T0, gen_op_movl_r5_T0, - gen_op_movl_r6_T0, gen_op_movl_r7_T0, - gen_op_movl_r8_T0, gen_op_movl_r9_T0, - gen_op_movl_r10_T0, gen_op_movl_r11_T0, - gen_op_movl_r12_T0, gen_op_movl_r13_T0, - gen_op_movl_r14_T0, gen_op_movl_r15_T0, -}; -static GenOpFunc *gen_movl_T0_reg[16] = -{ - gen_op_movl_T0_r0, gen_op_movl_T0_r1, - gen_op_movl_T0_r2, gen_op_movl_T0_r3, - gen_op_movl_T0_r4, gen_op_movl_T0_r5, - gen_op_movl_T0_r6, gen_op_movl_T0_r7, - gen_op_movl_T0_r8, gen_op_movl_T0_r9, - gen_op_movl_T0_r10, gen_op_movl_T0_r11, - gen_op_movl_T0_r12, gen_op_movl_T0_r13, - gen_op_movl_T0_r14, gen_op_movl_T0_r15, -}; - -static void noop_write(void) { - /* nop. */ -} - -static void gen_vr_read(void) { - gen_op_movl_T0_im(32); -} - -static void gen_movl_T0_p0(void) { - gen_op_movl_T0_im(0); -} - -static void gen_ccs_read(void) { - gen_op_movl_T0_p13(); -} - -static void gen_ccs_write(void) { - gen_op_movl_p13_T0(); -} - -/* Table to generate quick moves from T0 onto any register. */ -static GenOpFunc *gen_movl_preg_T0[16] = -{ - noop_write, /* bz, not writeable. */ - noop_write, /* vr, not writeable. */ - gen_op_movl_p2_T0, gen_op_movl_p3_T0, - noop_write, /* wz, not writeable. */ - gen_op_movl_p5_T0, - gen_op_movl_p6_T0, gen_op_movl_p7_T0, - noop_write, /* dz, not writeable. */ - gen_op_movl_p9_T0, - gen_op_movl_p10_T0, gen_op_movl_p11_T0, - gen_op_movl_p12_T0, - gen_ccs_write, /* ccs needs special treatment. */ - gen_op_movl_p14_T0, gen_op_movl_p15_T0, -}; -static GenOpFunc *gen_movl_T0_preg[16] = -{ - gen_movl_T0_p0, - gen_vr_read, - gen_op_movl_T0_p2, gen_op_movl_T0_p3, - gen_op_movl_T0_p4, gen_op_movl_T0_p5, - gen_op_movl_T0_p6, gen_op_movl_T0_p7, - gen_op_movl_T0_p8, gen_op_movl_T0_p9, - gen_op_movl_T0_p10, gen_op_movl_T0_p11, - gen_op_movl_T0_p12, - gen_ccs_read, /* ccs needs special treatment. */ - gen_op_movl_T0_p14, gen_op_movl_T0_p15, -}; - -/* We need this table to handle moves with implicit width. */ -int preg_sizes[] = { - 1, /* bz. */ - 1, /* vr. */ - 4, /* pid. */ - 1, /* srs. */ - 2, /* wz. */ - 4, 4, 4, - 4, 4, 4, 4, - 4, 4, 4, 4, -}; - #ifdef CONFIG_USER_ONLY #define GEN_OP_LD(width, reg) \ void gen_op_ld##width##_T0_##reg (DisasContext *dc) { \ @@ -245,18 +140,112 @@ GEN_OP_ST(w, T0) GEN_OP_LD(l, T0) GEN_OP_ST(l, T0) +/* We need this table to handle preg-moves with implicit width. */ +int preg_sizes[] = { + 1, /* bz. */ + 1, /* vr. */ + 4, /* pid. */ + 1, /* srs. */ + 2, /* wz. */ + 4, 4, 4, + 4, 4, 4, 4, + 4, 4, 4, 4, +}; + +#define t_gen_mov_TN_env(tn, member) \ + _t_gen_mov_TN_env((tn), offsetof(CPUState, (member))) +#define t_gen_mov_env_TN(member, tn) \ + _t_gen_mov_env_TN(offsetof(CPUState, (member)), (tn)) + +#define t_gen_mov_TN_reg(tn, regno) \ + _t_gen_mov_TN_env((tn), offsetof(CPUState, regs[regno])) +#define t_gen_mov_reg_TN(regno, tn) \ + _t_gen_mov_env_TN(offsetof(CPUState, regs[regno]), (tn)) + +static inline void _t_gen_mov_TN_env(TCGv tn, int offset) +{ + tcg_gen_ld_tl(tn, cpu_env, offset); +} +static inline void _t_gen_mov_env_TN(int offset, TCGv tn) +{ + tcg_gen_st_tl(tn, cpu_env, offset); +} + +static inline void t_gen_mov_TN_preg(TCGv tn, int r) +{ + if (r == PR_BZ || r == PR_WZ || r == PR_DZ) + tcg_gen_mov_tl(tn, tcg_const_i32(0)); + else if (r == PR_VR) + tcg_gen_mov_tl(tn, tcg_const_i32(32)); + else + tcg_gen_ld_tl(tn, cpu_env, offsetof(CPUState, pregs[r])); +} +static inline void t_gen_mov_preg_TN(int r, TCGv tn) +{ + if (r == PR_BZ || r == PR_WZ || r == PR_DZ) + return; + else + tcg_gen_st_tl(tn, cpu_env, offsetof(CPUState, pregs[r])); +} + +static inline void t_gen_mov_TN_im(TCGv tn, int32_t val) +{ + tcg_gen_movi_tl(tn, val); +} + +static void t_gen_lsl(TCGv d, TCGv a, TCGv b) +{ + int l1; + + l1 = gen_new_label(); + /* Speculative shift. */ + tcg_gen_shl_tl(d, a, b); + tcg_gen_brcond_tl(TCG_COND_LE, b, tcg_const_i32(31), l1); + /* Clear dst if shift operands were to large. */ + tcg_gen_movi_tl(d, 0); + gen_set_label(l1); +} + +static void t_gen_lsr(TCGv d, TCGv a, TCGv b) +{ + int l1; + + l1 = gen_new_label(); + /* Speculative shift. */ + tcg_gen_shr_tl(d, a, b); + tcg_gen_brcond_tl(TCG_COND_LE, b, tcg_const_i32(31), l1); + /* Clear dst if shift operands were to large. */ + tcg_gen_movi_tl(d, 0); + gen_set_label(l1); +} + +static void t_gen_asr(TCGv d, TCGv a, TCGv b) +{ + int l1; + + l1 = gen_new_label(); + /* Speculative shift. */ + tcg_gen_sar_tl(d, a, b); + tcg_gen_brcond_tl(TCG_COND_LE, b, tcg_const_i32(31), l1); + /* Clear dst if shift operands were to large. */ + tcg_gen_movi_tl(d, 0); + tcg_gen_brcond_tl(TCG_COND_LT, b, tcg_const_i32(0x80000000), l1); + tcg_gen_movi_tl(d, 0xffffffff); + gen_set_label(l1); +} + static void gen_goto_tb(DisasContext *dc, int n, target_ulong dest) { TranslationBlock *tb; tb = dc->tb; if ((tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK)) { -#if 0 - /* XXX: this code is not finished */ - tcg_gen_goto_tb(n); -#endif - tcg_gen_exit_tb(0); + tcg_gen_goto_tb(n); + tcg_gen_movi_tl(cpu_T[0], dest); + gen_op_movl_pc_T0(); + tcg_gen_exit_tb((long)tb + n); } else { - tcg_gen_exit_tb(0); + gen_op_movl_pc_T0(); + tcg_gen_exit_tb(0); } } @@ -273,10 +262,22 @@ static int sign_extend(unsigned int val, unsigned int width) return sval; } +static inline void cris_clear_x_flag(DisasContext *dc) +{ + TCGv ccs; + + ccs = tcg_temp_new(TCG_TYPE_TL); + + t_gen_mov_TN_preg(ccs, PR_CCS); + tcg_gen_andi_i32(ccs, ccs, ~X_FLAG); + t_gen_mov_preg_TN(PR_CCS, ccs); + dc->flagx_live = 1; + dc->flags_x = 0; +} + static void cris_evaluate_flags(DisasContext *dc) { if (!dc->flags_live) { - switch (dc->cc_op) { case CC_OP_MCP: @@ -373,7 +374,7 @@ static void crisv32_alu_op(DisasContext *dc, int op, int rd, int size) switch (op) { case CC_OP_ADD: - gen_op_addl_T0_T1(); + tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]); /* Extended arithmetics. */ if (!dc->flagx_live) gen_op_addxl_T0_C(); @@ -381,18 +382,18 @@ static void crisv32_alu_op(DisasContext *dc, int op, int rd, int size) gen_op_addxl_T0_C(); break; case CC_OP_ADDC: - gen_op_addl_T0_T1(); + tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]); gen_op_addl_T0_C(); break; case CC_OP_MCP: - gen_op_addl_T0_T1(); + tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]); gen_op_addl_T0_R(); break; case CC_OP_SUB: - gen_op_negl_T1_T1(); - gen_op_addl_T0_T1(); + tcg_gen_sub_tl(cpu_T[1], tcg_const_i32(0), cpu_T[1]); + tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]); /* CRIS flag evaluation needs ~src. */ - gen_op_negl_T1_T1(); + tcg_gen_sub_tl(cpu_T[1], tcg_const_i32(0), cpu_T[1]); gen_op_not_T1_T1(); /* Extended arithmetics. */ @@ -402,28 +403,30 @@ static void crisv32_alu_op(DisasContext *dc, int op, int rd, int size) gen_op_subxl_T0_C(); break; case CC_OP_MOVE: - gen_op_movl_T0_T1(); + tcg_gen_mov_tl(cpu_T[0], cpu_T[1]); break; case CC_OP_OR: - gen_op_orl_T0_T1(); + tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_T[1]); break; case CC_OP_AND: - gen_op_andl_T0_T1(); + tcg_gen_and_tl(cpu_T[0], cpu_T[0], cpu_T[1]); break; case CC_OP_XOR: - gen_op_xorl_T0_T1(); + tcg_gen_xor_tl(cpu_T[0], cpu_T[0], cpu_T[1]); break; case CC_OP_LSL: - gen_op_lsll_T0_T1(); + t_gen_lsl(cpu_T[0], cpu_T[0], cpu_T[1]); break; case CC_OP_LSR: - gen_op_lsrl_T0_T1(); + t_gen_lsr(cpu_T[0], cpu_T[0], cpu_T[1]); break; case CC_OP_ASR: - gen_op_asrl_T0_T1(); + t_gen_asr(cpu_T[0], cpu_T[0], cpu_T[1]); break; case CC_OP_NEG: - gen_op_negl_T0_T1(); + /* TCG-FIXME: this is not optimal. Many archs have + fast neg insns. */ + tcg_gen_sub_tl(cpu_T[0], tcg_const_i32(0), cpu_T[1]); /* Extended arithmetics. */ gen_op_subxl_T0_C(); break; @@ -447,10 +450,11 @@ static void crisv32_alu_op(DisasContext *dc, int op, int rd, int size) gen_op_bound_T0_T1(); break; case CC_OP_CMP: - gen_op_negl_T1_T1(); - gen_op_addl_T0_T1(); + tcg_gen_sub_tl(cpu_T[1], tcg_const_i32(0), cpu_T[1]); + tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]); + /* CRIS flag evaluation needs ~src. */ + tcg_gen_sub_tl(cpu_T[1], tcg_const_i32(0), cpu_T[1]); /* CRIS flag evaluation needs ~src. */ - gen_op_negl_T1_T1(); gen_op_not_T1_T1(); /* Extended arithmetics. */ @@ -467,23 +471,24 @@ static void crisv32_alu_op(DisasContext *dc, int op, int rd, int size) gen_op_update_cc_src_T1(); if (size == 1) - gen_op_andl_T0_im(0xff); + tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0xff); else if (size == 2) - gen_op_andl_T0_im(0xffff); + tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0xffff); + /* Writeback. */ if (writeback) { if (size == 4) - gen_movl_reg_T0[rd](); + t_gen_mov_reg_TN(rd, cpu_T[0]); else { - gen_op_movl_T1_T0(); - gen_movl_T0_reg[rd](); + tcg_gen_mov_tl(cpu_T[1], cpu_T[0]); + t_gen_mov_TN_reg(cpu_T[0], rd); if (size == 1) - gen_op_andl_T0_im(~0xff); + tcg_gen_andi_tl(cpu_T[0], cpu_T[0], ~0xff); else - gen_op_andl_T0_im(~0xffff); - gen_op_orl_T0_T1(); - gen_movl_reg_T0[rd](); - gen_op_movl_T0_T1(); + tcg_gen_andi_tl(cpu_T[0], cpu_T[0], ~0xffff); + tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_T[1]); + t_gen_mov_reg_TN(rd, cpu_T[0]); + tcg_gen_mov_tl(cpu_T[0], cpu_T[1]); } } if (dc->update_cc) @@ -674,53 +679,21 @@ void gen_store_T0_T1 (DisasContext *dc, unsigned int size) gen_op_stl_T0_T1(dc); } -/* sign extend T1 according to size. */ -static void gen_sext_T1_T0(int size) -{ - if (size == 1) - gen_op_extb_T1_T0(); - else if (size == 2) - gen_op_extw_T1_T0(); -} - -static void gen_sext_T1_T1(int size) -{ - if (size == 1) - gen_op_extb_T1_T1(); - else if (size == 2) - gen_op_extw_T1_T1(); -} - -static void gen_sext_T0_T0(int size) -{ - if (size == 1) - gen_op_extb_T0_T0(); - else if (size == 2) - gen_op_extw_T0_T0(); -} - -static void gen_zext_T0_T0(int size) -{ - if (size == 1) - gen_op_zextb_T0_T0(); - else if (size == 2) - gen_op_zextw_T0_T0(); -} - -static void gen_zext_T1_T0(int size) +static inline void t_gen_sext(TCGv d, TCGv s, int size) { if (size == 1) - gen_op_zextb_T1_T0(); + tcg_gen_ext8s_i32(d, s); else if (size == 2) - gen_op_zextw_T1_T0(); + tcg_gen_ext16s_i32(d, s); } -static void gen_zext_T1_T1(int size) +static inline void t_gen_zext(TCGv d, TCGv s, int size) { + /* TCG-FIXME: this is not optimal. Many archs have fast zext insns. */ if (size == 1) - gen_op_zextb_T1_T1(); + tcg_gen_andi_i32(d, s, 0xff); else if (size == 2) - gen_op_zextw_T1_T1(); + tcg_gen_andi_i32(d, s, 0xffff); } #if DISAS_CRIS @@ -758,21 +731,20 @@ static void do_postinc (DisasContext *dc, int size) { if (!dc->postinc) return; - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); gen_op_addl_T0_im(size); - gen_movl_reg_T0[dc->op1](); + t_gen_mov_reg_TN(dc->op1, cpu_T[0]); } static void dec_prep_move_r(DisasContext *dc, int rs, int rd, int size, int s_ext) { - gen_movl_T0_reg[rs](); - gen_op_movl_T1_T0(); + t_gen_mov_TN_reg(cpu_T[1], rs); if (s_ext) - gen_sext_T1_T1(size); + t_gen_sext(cpu_T[1], cpu_T[1], size); else - gen_zext_T1_T1(size); + t_gen_zext(cpu_T[1], cpu_T[1], size); } /* Prepare T0 and T1 for a register alu operation. @@ -783,11 +755,11 @@ static void dec_prep_alu_r(DisasContext *dc, int rs, int rd, { dec_prep_move_r(dc, rs, rd, size, s_ext); - gen_movl_T0_reg[rd](); + t_gen_mov_TN_reg(cpu_T[0], rd); if (s_ext) - gen_sext_T0_T0(size); + t_gen_sext(cpu_T[0], cpu_T[0], size); else - gen_zext_T0_T0(size); + t_gen_zext(cpu_T[0], cpu_T[0], size); } /* Prepare T0 and T1 for a memory + alu operation. @@ -823,20 +795,20 @@ static int dec_prep_alu_m(DisasContext *dc, int s_ext, int memsize) } DIS(fprintf (logfile, "imm=%x rd=%d sext=%d ms=%d\n", imm, rd, s_ext, memsize)); - gen_op_movl_T1_im (imm); + tcg_gen_movi_tl(cpu_T[1], imm); dc->postinc = 0; } else { - gen_movl_T0_reg[rs](); + t_gen_mov_TN_reg(cpu_T[0], rs); gen_load_T0_T0(dc, memsize, 0); - gen_op_movl_T1_T0(); + tcg_gen_mov_tl(cpu_T[1], cpu_T[0]); if (s_ext) - gen_sext_T1_T1(memsize); + t_gen_sext(cpu_T[1], cpu_T[1], memsize); else - gen_zext_T1_T1(memsize); + t_gen_zext(cpu_T[1], cpu_T[1], memsize); } /* put dest in T0. */ - gen_movl_T0_reg[rd](); + t_gen_mov_TN_reg(cpu_T[0], rd); return insn_len; } @@ -882,8 +854,8 @@ static unsigned int dec_addoq(DisasContext *dc) DIS(fprintf (logfile, "addoq %d, $r%u\n", imm, dc->op2)); cris_cc_mask(dc, 0); /* Fetch register operand, */ - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(imm); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + tcg_gen_movi_tl(cpu_T[1], imm); crisv32_alu_op(dc, CC_OP_ADD, R_ACR, 4); return 2; } @@ -895,8 +867,8 @@ static unsigned int dec_addq(DisasContext *dc) cris_cc_mask(dc, CC_MASK_NZVC); /* Fetch register operand, */ - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(dc->op1); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + tcg_gen_movi_tl(cpu_T[1], dc->op1); crisv32_alu_op(dc, CC_OP_ADD, dc->op2, 4); return 2; } @@ -908,10 +880,9 @@ static unsigned int dec_moveq(DisasContext *dc) imm = sign_extend(dc->op1, 5); DIS(fprintf (logfile, "moveq %d, $r%u\n", imm, dc->op2)); - cris_cc_mask(dc, 0); - gen_op_movl_T1_im(imm); - crisv32_alu_op(dc, CC_OP_MOVE, dc->op2, 4); - + t_gen_mov_reg_TN(dc->op2, tcg_const_i32(imm)); + if (!dc->flagx_live || dc->flags_x) + cris_clear_x_flag(dc); return 2; } static unsigned int dec_subq(DisasContext *dc) @@ -922,8 +893,8 @@ static unsigned int dec_subq(DisasContext *dc) cris_cc_mask(dc, CC_MASK_NZVC); /* Fetch register operand, */ - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(dc->op1); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], dc->op1); crisv32_alu_op(dc, CC_OP_SUB, dc->op2, 4); return 2; } @@ -935,8 +906,8 @@ static unsigned int dec_cmpq(DisasContext *dc) DIS(fprintf (logfile, "cmpq %d, $r%d\n", imm, dc->op2)); cris_cc_mask(dc, CC_MASK_NZVC); - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(imm); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], imm); crisv32_alu_op(dc, CC_OP_CMP, dc->op2, 4); return 2; } @@ -948,8 +919,8 @@ static unsigned int dec_andq(DisasContext *dc) DIS(fprintf (logfile, "andq %d, $r%d\n", imm, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(imm); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], imm); crisv32_alu_op(dc, CC_OP_AND, dc->op2, 4); return 2; } @@ -960,8 +931,8 @@ static unsigned int dec_orq(DisasContext *dc) imm = sign_extend(dc->op1, 5); DIS(fprintf (logfile, "orq %d, $r%d\n", imm, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(imm); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], imm); crisv32_alu_op(dc, CC_OP_OR, dc->op2, 4); return 2; } @@ -970,8 +941,8 @@ static unsigned int dec_btstq(DisasContext *dc) dc->op1 = EXTRACT_FIELD(dc->ir, 0, 4); DIS(fprintf (logfile, "btstq %u, $r%d\n", dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(dc->op1); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], dc->op1); crisv32_alu_op(dc, CC_OP_BTST, dc->op2, 4); cris_update_cc_op(dc, CC_OP_FLAGS); @@ -984,8 +955,8 @@ static unsigned int dec_asrq(DisasContext *dc) dc->op1 = EXTRACT_FIELD(dc->ir, 0, 4); DIS(fprintf (logfile, "asrq %u, $r%d\n", dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(dc->op1); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], dc->op1); crisv32_alu_op(dc, CC_OP_ASR, dc->op2, 4); return 2; } @@ -995,8 +966,8 @@ static unsigned int dec_lslq(DisasContext *dc) DIS(fprintf (logfile, "lslq %u, $r%d\n", dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(dc->op1); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], dc->op1); crisv32_alu_op(dc, CC_OP_LSL, dc->op2, 4); return 2; } @@ -1006,8 +977,8 @@ static unsigned int dec_lsrq(DisasContext *dc) DIS(fprintf (logfile, "lsrq %u, $r%d\n", dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_im(dc->op1); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + t_gen_mov_TN_im(cpu_T[1], dc->op1); crisv32_alu_op(dc, CC_OP_LSR, dc->op2, 4); return 2; } @@ -1075,7 +1046,7 @@ static unsigned int dec_lsl_r(DisasContext *dc) memsize_char(size), dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); dec_prep_alu_r(dc, dc->op1, dc->op2, size, 0); - gen_op_andl_T1_im(63); + tcg_gen_andi_tl(cpu_T[1], cpu_T[1], 63); crisv32_alu_op(dc, CC_OP_LSL, dc->op2, size); return 2; } @@ -1088,7 +1059,7 @@ static unsigned int dec_lsr_r(DisasContext *dc) memsize_char(size), dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); dec_prep_alu_r(dc, dc->op1, dc->op2, size, 0); - gen_op_andl_T1_im(63); + tcg_gen_andi_tl(cpu_T[1], cpu_T[1], 63); crisv32_alu_op(dc, CC_OP_LSR, dc->op2, size); return 2; } @@ -1101,7 +1072,7 @@ static unsigned int dec_asr_r(DisasContext *dc) memsize_char(size), dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); dec_prep_alu_r(dc, dc->op1, dc->op2, size, 1); - gen_op_andl_T1_im(63); + tcg_gen_andi_tl(cpu_T[1], cpu_T[1], 63); crisv32_alu_op(dc, CC_OP_ASR, dc->op2, size); return 2; } @@ -1114,7 +1085,7 @@ static unsigned int dec_muls_r(DisasContext *dc) memsize_char(size), dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZV); dec_prep_alu_r(dc, dc->op1, dc->op2, size, 1); - gen_sext_T0_T0(size); + t_gen_sext(cpu_T[0], cpu_T[0], size); crisv32_alu_op(dc, CC_OP_MULS, dc->op2, 4); return 2; } @@ -1127,7 +1098,7 @@ static unsigned int dec_mulu_r(DisasContext *dc) memsize_char(size), dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZV); dec_prep_alu_r(dc, dc->op1, dc->op2, size, 0); - gen_zext_T0_T0(size); + t_gen_zext(cpu_T[0], cpu_T[0], size); crisv32_alu_op(dc, CC_OP_MULU, dc->op2, 4); return 2; } @@ -1137,9 +1108,8 @@ static unsigned int dec_dstep_r(DisasContext *dc) { DIS(fprintf (logfile, "dstep $r%u, $r%u\n", dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op1](); - gen_op_movl_T1_T0(); - gen_movl_T0_reg[dc->op2](); + t_gen_mov_TN_reg(cpu_T[1], dc->op1); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); crisv32_alu_op(dc, CC_OP_DSTEP, dc->op2, 4); return 2; } @@ -1165,7 +1135,7 @@ static unsigned int dec_bound_r(DisasContext *dc) /* TODO: needs optmimization. */ dec_prep_alu_r(dc, dc->op1, dc->op2, size, 0); /* rd should be 4. */ - gen_movl_T0_reg[dc->op2](); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); crisv32_alu_op(dc, CC_OP_BOUND, dc->op2, 4); return 2; } @@ -1220,9 +1190,8 @@ static unsigned int dec_mcp_r(DisasContext *dc) dc->op2, dc->op1)); cris_evaluate_flags(dc); cris_cc_mask(dc, CC_MASK_RNZV); - gen_movl_T0_preg[dc->op2](); - gen_op_movl_T1_T0(); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); + t_gen_mov_TN_preg(cpu_T[1], dc->op2); crisv32_alu_op(dc, CC_OP_MCP, dc->op1, 4); return 2; } @@ -1250,7 +1219,7 @@ static unsigned int dec_swap_r(DisasContext *dc) swapmode_name(dc->op2, modename), dc->op1)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); if (dc->op2 & 8) gen_op_not_T0_T0(); if (dc->op2 & 4) @@ -1281,9 +1250,9 @@ static unsigned int dec_addi_r(DisasContext *dc) memsize_char(memsize_zz(dc)), dc->op2, dc->op1)); cris_cc_mask(dc, 0); dec_prep_alu_r(dc, dc->op1, dc->op2, 4, 0); - gen_op_lsll_T0_im(dc->zzsize); - gen_op_addl_T0_T1(); - gen_movl_reg_T0[dc->op1](); + t_gen_lsl(cpu_T[0], cpu_T[0], tcg_const_i32(dc->zzsize)); + tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]); + t_gen_mov_reg_TN(dc->op1, cpu_T[0]); return 2; } @@ -1293,9 +1262,10 @@ static unsigned int dec_addi_acr(DisasContext *dc) memsize_char(memsize_zz(dc)), dc->op2, dc->op1)); cris_cc_mask(dc, 0); dec_prep_alu_r(dc, dc->op1, dc->op2, 4, 0); - gen_op_lsll_T0_im(dc->zzsize); - gen_op_addl_T0_T1(); - gen_movl_reg_T0[R_ACR](); + t_gen_lsl(cpu_T[0], cpu_T[0], tcg_const_i32(dc->zzsize)); + + tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]); + t_gen_mov_reg_TN(R_ACR, cpu_T[0]); return 2; } @@ -1358,9 +1328,9 @@ static unsigned int dec_movs_r(DisasContext *dc) dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZ); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); /* Size can only be qi or hi. */ - gen_sext_T1_T0(size); + t_gen_sext(cpu_T[1], cpu_T[0], size); crisv32_alu_op(dc, CC_OP_MOVE, dc->op2, 4); return 2; } @@ -1374,13 +1344,14 @@ static unsigned int dec_addu_r(DisasContext *dc) dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZVC); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[1], dc->op1); /* Size can only be qi or hi. */ - gen_zext_T1_T0(size); - gen_movl_T0_reg[dc->op2](); + t_gen_zext(cpu_T[1], cpu_T[1], size); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); crisv32_alu_op(dc, CC_OP_ADD, dc->op2, 4); return 2; } + /* Sign extension. From size to dword. */ static unsigned int dec_adds_r(DisasContext *dc) { @@ -1390,10 +1361,11 @@ static unsigned int dec_adds_r(DisasContext *dc) dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZVC); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[1], dc->op1); /* Size can only be qi or hi. */ - gen_sext_T1_T0(size); - gen_movl_T0_reg[dc->op2](); + t_gen_sext(cpu_T[1], cpu_T[1], size); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); + crisv32_alu_op(dc, CC_OP_ADD, dc->op2, 4); return 2; } @@ -1407,10 +1379,10 @@ static unsigned int dec_subu_r(DisasContext *dc) dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZVC); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[1], dc->op1); /* Size can only be qi or hi. */ - gen_zext_T1_T0(size); - gen_movl_T0_reg[dc->op2](); + t_gen_zext(cpu_T[1], cpu_T[1], size); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); crisv32_alu_op(dc, CC_OP_SUB, dc->op2, 4); return 2; } @@ -1424,10 +1396,10 @@ static unsigned int dec_subs_r(DisasContext *dc) dc->op1, dc->op2)); cris_cc_mask(dc, CC_MASK_NZVC); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[1], dc->op1); /* Size can only be qi or hi. */ - gen_sext_T1_T0(size); - gen_movl_T0_reg[dc->op2](); + t_gen_sext(cpu_T[1], cpu_T[1], size); + t_gen_mov_TN_reg(cpu_T[0], dc->op2); crisv32_alu_op(dc, CC_OP_SUB, dc->op2, 4); return 2; } @@ -1469,29 +1441,34 @@ static unsigned int dec_move_rs(DisasContext *dc) { DIS(fprintf (logfile, "move $r%u, $s%u\n", dc->op1, dc->op2)); cris_cc_mask(dc, 0); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); gen_op_movl_sreg_T0(dc->op2); - if (dc->op2 == 5) /* srs is checked at runtime. */ +#if !defined(CONFIG_USER_ONLY) + if (dc->op2 == 6) + gen_op_movl_tlb_hi_T0(); + else if (dc->op2 == 5) { /* srs is checked at runtime. */ + tcg_gen_helper_0_1(helper_tlb_update, cpu_T[0]); gen_op_movl_tlb_lo_T0(); + } +#endif return 2; } static unsigned int dec_move_sr(DisasContext *dc) { - DIS(fprintf (logfile, "move $s%u, $r%u\n", dc->op1, dc->op2)); + DIS(fprintf (logfile, "move $s%u, $r%u\n", dc->op2, dc->op1)); cris_cc_mask(dc, 0); - gen_op_movl_T0_sreg(dc->op1); - gen_op_movl_T1_T0(); - crisv32_alu_op(dc, CC_OP_MOVE, dc->op2, 4); + gen_op_movl_T0_sreg(dc->op2); + tcg_gen_mov_tl(cpu_T[1], cpu_T[0]); + crisv32_alu_op(dc, CC_OP_MOVE, dc->op1, 4); return 2; } static unsigned int dec_move_rp(DisasContext *dc) { DIS(fprintf (logfile, "move $r%u, $p%u\n", dc->op1, dc->op2)); cris_cc_mask(dc, 0); - gen_movl_T0_reg[dc->op1](); - gen_op_movl_T1_T0(); - gen_movl_preg_T0[dc->op2](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); + t_gen_mov_preg_TN(dc->op2, cpu_T[0]); return 2; } static unsigned int dec_move_pr(DisasContext *dc) @@ -1501,11 +1478,9 @@ static unsigned int dec_move_pr(DisasContext *dc) /* Support register 0 is hardwired to zero. Treat it specially. */ if (dc->op2 == 0) - gen_op_movl_T1_im(0); - else { - gen_movl_T0_preg[dc->op2](); - gen_op_movl_T1_T0(); - } + tcg_gen_movi_tl(cpu_T[1], 0); + else + t_gen_mov_TN_preg(cpu_T[1], dc->op2); crisv32_alu_op(dc, CC_OP_MOVE, dc->op1, preg_sizes[dc->op2]); return 2; } @@ -1688,8 +1663,8 @@ static unsigned int dec_test_m(DisasContext *dc) cris_cc_mask(dc, CC_MASK_NZ); gen_op_clrf(3); insn_len = dec_prep_alu_m(dc, 0, memsize); - gen_op_swp_T0_T1(); - gen_op_movl_T1_im(0); + tcg_gen_mov_tl(cpu_T[0], cpu_T[1]); + tcg_gen_movi_tl(cpu_T[1], 0); crisv32_alu_op(dc, CC_OP_CMP, dc->op2, memsize_zz(dc)); do_postinc(dc, memsize); return insn_len; @@ -1819,8 +1794,7 @@ static unsigned int dec_move_mp(DisasContext *dc) cris_cc_mask(dc, 0); insn_len = dec_prep_alu_m(dc, 0, memsize); - gen_op_movl_T0_T1(); - gen_movl_preg_T0[dc->op2](); + t_gen_mov_preg_TN(dc->op2, cpu_T[1]); do_postinc(dc, memsize); return insn_len; @@ -1838,21 +1812,13 @@ static unsigned int dec_move_pm(DisasContext *dc) cris_cc_mask(dc, 0); /* prepare store. Address in T0, value in T1. */ - /* Support register 0 is hardwired to zero. - Treat it specially. */ - if (dc->op2 == 0) - gen_op_movl_T1_im(0); - else - { - gen_movl_T0_preg[dc->op2](); - gen_op_movl_T1_T0(); - } - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_preg(cpu_T[1], dc->op2); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); gen_store_T0_T1(dc, memsize); if (dc->postinc) { - gen_op_addl_T0_im(memsize); - gen_movl_reg_T0[dc->op1](); + tcg_gen_addi_tl(cpu_T[0], cpu_T[0], memsize); + t_gen_mov_reg_TN(dc->op1, cpu_T[0]); } return 2; } @@ -1865,21 +1831,18 @@ static unsigned int dec_movem_mr(DisasContext *dc) dc->postinc ? "+]" : "]", dc->op2)); cris_cc_mask(dc, 0); - /* fetch the address into T1. */ - gen_movl_T0_reg[dc->op1](); - gen_op_movl_T1_T0(); + /* fetch the address into T0 and T1. */ + t_gen_mov_TN_reg(cpu_T[1], dc->op1); for (i = 0; i <= dc->op2; i++) { /* Perform the load onto regnum i. Always dword wide. */ + tcg_gen_mov_tl(cpu_T[0], cpu_T[1]); gen_load_T0_T0(dc, 4, 0); - gen_movl_reg_T0[i](); - /* Update the address. */ - gen_op_addl_T1_im(4); - gen_op_movl_T0_T1(); - } - if (dc->postinc) { - /* writeback the updated pointer value. */ - gen_movl_reg_T0[dc->op1](); + t_gen_mov_reg_TN(i, cpu_T[0]); + tcg_gen_addi_tl(cpu_T[1], cpu_T[1], 4); } + /* writeback the updated pointer value. */ + if (dc->postinc) + t_gen_mov_reg_TN(dc->op1, cpu_T[1]); return 2; } @@ -1893,22 +1856,20 @@ static unsigned int dec_movem_rm(DisasContext *dc) cris_cc_mask(dc, 0); for (i = 0; i <= dc->op2; i++) { /* Fetch register i into T1. */ - gen_movl_T0_reg[i](); - gen_op_movl_T1_T0(); - + t_gen_mov_TN_reg(cpu_T[1], i); /* Fetch the address into T0. */ - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); /* Displace it. */ - gen_op_addl_T0_im(i * 4); - + tcg_gen_addi_tl(cpu_T[0], cpu_T[0], i * 4); /* Perform the store. */ gen_store_T0_T1(dc, 4); } if (dc->postinc) { - /* Update the address. */ - gen_op_addl_T0_im(4); + /* T0 should point to the last written addr, advance one more + step. */ + tcg_gen_addi_tl(cpu_T[0], cpu_T[0], 4); /* writeback the updated pointer value. */ - gen_movl_reg_T0[dc->op1](); + t_gen_mov_reg_TN(dc->op1, cpu_T[0]); } return 2; } @@ -1924,25 +1885,23 @@ static unsigned int dec_move_rm(DisasContext *dc) cris_cc_mask(dc, 0); /* prepare store. */ - gen_movl_T0_reg[dc->op2](); - gen_op_movl_T1_T0(); - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); + t_gen_mov_TN_reg(cpu_T[1], dc->op2); gen_store_T0_T1(dc, memsize); if (dc->postinc) { - gen_op_addl_T0_im(memsize); - gen_movl_reg_T0[dc->op1](); + tcg_gen_addi_tl(cpu_T[0], cpu_T[0], memsize); + t_gen_mov_reg_TN(dc->op1, cpu_T[0]); } return 2; } - static unsigned int dec_lapcq(DisasContext *dc) { DIS(fprintf (logfile, "lapcq %x, $r%u\n", dc->pc + dc->op1*2, dc->op2)); cris_cc_mask(dc, 0); - gen_op_movl_T1_im(dc->pc + dc->op1*2); + tcg_gen_movi_tl(cpu_T[1], dc->pc + dc->op1 * 2); crisv32_alu_op(dc, CC_OP_MOVE, dc->op2, 4); return 2; } @@ -1951,16 +1910,14 @@ static unsigned int dec_lapc_im(DisasContext *dc) { unsigned int rd; int32_t imm; - int insn_len = 6; rd = dc->op2; cris_cc_mask(dc, 0); imm = ldl_code(dc->pc + 2); DIS(fprintf (logfile, "lapc 0x%x, $r%u\n", imm + dc->pc, dc->op2)); - gen_op_movl_T0_im (dc->pc + imm); - gen_movl_reg_T0[rd] (); - return insn_len; + t_gen_mov_reg_TN(rd, tcg_const_i32(dc->pc + imm)); + return 6; } /* Jump to special reg. */ @@ -1969,7 +1926,7 @@ static unsigned int dec_jump_p(DisasContext *dc) DIS(fprintf (logfile, "jump $p%u\n", dc->op2)); cris_cc_mask(dc, 0); /* Store the return address in Pd. */ - gen_movl_T0_preg[dc->op2](); + t_gen_mov_TN_preg(cpu_T[0], dc->op2); gen_op_movl_btarget_T0(); cris_prepare_dyn_jmp(dc); return 2; @@ -1981,10 +1938,10 @@ static unsigned int dec_jas_r(DisasContext *dc) DIS(fprintf (logfile, "jas $r%u, $p%u\n", dc->op1, dc->op2)); cris_cc_mask(dc, 0); /* Stor the return address in Pd. */ - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); gen_op_movl_btarget_T0(); - gen_op_movl_T0_im(dc->pc + 4); - gen_movl_preg_T0[dc->op2](); + tcg_gen_movi_tl(cpu_T[0], dc->pc + 4); + t_gen_mov_preg_TN(dc->op2, cpu_T[0]); cris_prepare_dyn_jmp(dc); return 2; } @@ -1998,10 +1955,10 @@ static unsigned int dec_jas_im(DisasContext *dc) DIS(fprintf (logfile, "jas 0x%x\n", imm)); cris_cc_mask(dc, 0); /* Stor the return address in Pd. */ - gen_op_movl_T0_im(imm); + tcg_gen_movi_tl(cpu_T[0], imm); gen_op_movl_btarget_T0(); - gen_op_movl_T0_im(dc->pc + 8); - gen_movl_preg_T0[dc->op2](); + tcg_gen_movi_tl(cpu_T[0], dc->pc + 8); + t_gen_mov_preg_TN(dc->op2, cpu_T[0]); cris_prepare_dyn_jmp(dc); return 6; } @@ -2015,10 +1972,10 @@ static unsigned int dec_jasc_im(DisasContext *dc) DIS(fprintf (logfile, "jasc 0x%x\n", imm)); cris_cc_mask(dc, 0); /* Stor the return address in Pd. */ - gen_op_movl_T0_im(imm); + tcg_gen_movi_tl(cpu_T[0], imm); gen_op_movl_btarget_T0(); - gen_op_movl_T0_im(dc->pc + 8 + 4); - gen_movl_preg_T0[dc->op2](); + tcg_gen_movi_tl(cpu_T[0], dc->pc + 8 + 4); + t_gen_mov_preg_TN(dc->op2, cpu_T[0]); cris_prepare_dyn_jmp(dc); return 6; } @@ -2028,10 +1985,10 @@ static unsigned int dec_jasc_r(DisasContext *dc) DIS(fprintf (logfile, "jasc_r $r%u, $p%u\n", dc->op1, dc->op2)); cris_cc_mask(dc, 0); /* Stor the return address in Pd. */ - gen_movl_T0_reg[dc->op1](); + t_gen_mov_TN_reg(cpu_T[0], dc->op1); gen_op_movl_btarget_T0(); - gen_op_movl_T0_im(dc->pc + 4 + 4); - gen_movl_preg_T0[dc->op2](); + tcg_gen_movi_tl(cpu_T[0], dc->pc + 4 + 4); + t_gen_mov_preg_TN(dc->op2, cpu_T[0]); cris_prepare_dyn_jmp(dc); return 2; } @@ -2064,10 +2021,10 @@ static unsigned int dec_bas_im(DisasContext *dc) DIS(fprintf (logfile, "bas 0x%x, $p%u\n", dc->pc + simm, dc->op2)); cris_cc_mask(dc, 0); /* Stor the return address in Pd. */ - gen_op_movl_T0_im(dc->pc + simm); + tcg_gen_movi_tl(cpu_T[0], dc->pc + simm); gen_op_movl_btarget_T0(); - gen_op_movl_T0_im(dc->pc + 8); - gen_movl_preg_T0[dc->op2](); + tcg_gen_movi_tl(cpu_T[0], dc->pc + 8); + t_gen_mov_preg_TN(dc->op2, cpu_T[0]); cris_prepare_dyn_jmp(dc); return 6; } @@ -2080,10 +2037,10 @@ static unsigned int dec_basc_im(DisasContext *dc) DIS(fprintf (logfile, "basc 0x%x, $p%u\n", dc->pc + simm, dc->op2)); cris_cc_mask(dc, 0); /* Stor the return address in Pd. */ - gen_op_movl_T0_im(dc->pc + simm); + tcg_gen_movi_tl(cpu_T[0], dc->pc + simm); gen_op_movl_btarget_T0(); - gen_op_movl_T0_im(dc->pc + 12); - gen_movl_preg_T0[dc->op2](); + tcg_gen_movi_tl(cpu_T[0], dc->pc + 12); + t_gen_mov_preg_TN(dc->op2, cpu_T[0]); cris_prepare_dyn_jmp(dc); return 6; } @@ -2096,7 +2053,7 @@ static unsigned int dec_rfe_etc(DisasContext *dc) cris_cc_mask(dc, 0); if (dc->op2 == 15) /* ignore halt. */ - goto done; + return 2; switch (dc->op2 & 7) { case 2: @@ -2110,7 +2067,7 @@ static unsigned int dec_rfe_etc(DisasContext *dc) break; case 6: /* break. */ - gen_op_movl_T0_im(dc->pc); + tcg_gen_movi_tl(cpu_T[0], dc->pc); gen_op_movl_pc_T0(); /* Breaks start at 16 in the exception vector. */ gen_op_break_im(dc->op1 + 16); @@ -2122,7 +2079,6 @@ static unsigned int dec_rfe_etc(DisasContext *dc) break; } - done: return 2; } @@ -2296,7 +2252,7 @@ static void check_breakpoint(CPUState *env, DisasContext *dc) for(j = 0; j < env->nb_breakpoints; j++) { if (env->breakpoints[j] == dc->pc) { cris_evaluate_flags (dc); - gen_op_movl_T0_im((long)dc->pc); + tcg_gen_movi_tl(cpu_T[0], dc->pc); gen_op_movl_pc_T0(); gen_op_debug(); dc->is_jmp = DISAS_UPDATE; @@ -2305,7 +2261,6 @@ static void check_breakpoint(CPUState *env, DisasContext *dc) } } - /* generate intermediate code for basic block 'tb'. */ struct DisasContext ctx; static int @@ -2356,11 +2311,7 @@ gen_intermediate_code_internal(CPUState *env, TranslationBlock *tb, if (!dc->flagx_live || (dc->flagx_live && !(dc->cc_op == CC_OP_FLAGS && dc->flags_x))) { - gen_movl_T0_preg[PR_CCS](); - gen_op_andl_T0_im(~X_FLAG); - gen_movl_preg_T0[PR_CCS](); - dc->flagx_live = 1; - dc->flags_x = 0; + cris_clear_x_flag(dc); } /* Check for delayed branches here. If we do it before @@ -2488,6 +2439,10 @@ void cpu_dump_state (CPUState *env, FILE *f, } +static void tcg_macro_func(TCGContext *s, int macro_id, const int *dead_args) +{ +} + CPUCRISState *cpu_cris_init (const char *cpu_model) { CPUCRISState *env; @@ -2496,6 +2451,19 @@ CPUCRISState *cpu_cris_init (const char *cpu_model) if (!env) return NULL; cpu_exec_init(env); + + tcg_set_macro_func(&tcg_ctx, tcg_macro_func); + cpu_env = tcg_global_reg_new(TCG_TYPE_PTR, TCG_AREG0, "env"); +#if TARGET_LONG_BITS > HOST_LONG_BITS + cpu_T[0] = tcg_global_mem_new(TCG_TYPE_TL, + TCG_AREG0, offsetof(CPUState, t0), "T0"); + cpu_T[1] = tcg_global_mem_new(TCG_TYPE_TL, + TCG_AREG0, offsetof(CPUState, t1), "T1"); +#else + cpu_T[0] = tcg_global_reg_new(TCG_TYPE_TL, TCG_AREG1, "T0"); + cpu_T[1] = tcg_global_reg_new(TCG_TYPE_TL, TCG_AREG2, "T1"); +#endif + cpu_reset(env); return env; } |