diff options
author | Richard Henderson <rth@twiddle.net> | 2012-10-16 19:32:14 +1000 |
---|---|---|
committer | Blue Swirl <blauwirbel@gmail.com> | 2012-10-20 07:56:06 +0000 |
commit | 9d1d4e342cb154a0f801eabd9758e90599893073 (patch) | |
tree | 53e5d8c475e8c775ce64bca619d00f954cb6acd0 /target-sparc | |
parent | 97ea285917d473c8c8ecb627a637d76082cd6584 (diff) | |
download | qemu-9d1d4e342cb154a0f801eabd9758e90599893073.zip qemu-9d1d4e342cb154a0f801eabd9758e90599893073.tar.gz qemu-9d1d4e342cb154a0f801eabd9758e90599893073.tar.bz2 |
target-sparc: Use gen_load_gpr in get_src[12]
This means we can avoid the incoming temporary, though the cleanup
of the existing temporaries is not performed in this patch.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Diffstat (limited to 'target-sparc')
-rw-r--r-- | target-sparc/translate.c | 75 |
1 files changed, 29 insertions, 46 deletions
diff --git a/target-sparc/translate.c b/target-sparc/translate.c index 0ec3d48..760cfd6 100644 --- a/target-sparc/translate.c +++ b/target-sparc/translate.c @@ -2248,40 +2248,23 @@ static inline void gen_ldstub_asi(TCGv dst, TCGv addr, int insn) } #endif -static inline TCGv get_src1(unsigned int insn, TCGv def) +static TCGv get_src1(DisasContext *dc, unsigned int insn) { - TCGv r_rs1 = def; - unsigned int rs1; - - rs1 = GET_FIELD(insn, 13, 17); - if (rs1 == 0) { - tcg_gen_movi_tl(def, 0); - } else if (rs1 < 8) { - r_rs1 = cpu_gregs[rs1]; - } else { - tcg_gen_ld_tl(def, cpu_regwptr, (rs1 - 8) * sizeof(target_ulong)); - } - return r_rs1; + unsigned int rs1 = GET_FIELD(insn, 13, 17); + return gen_load_gpr(dc, rs1); } -static inline TCGv get_src2(unsigned int insn, TCGv def) +static TCGv get_src2(DisasContext *dc, unsigned int insn) { - TCGv r_rs2 = def; - if (IS_IMM) { /* immediate */ target_long simm = GET_FIELDs(insn, 19, 31); - tcg_gen_movi_tl(def, simm); - } else { /* register */ + TCGv t = get_temp_tl(dc); + tcg_gen_movi_tl(t, simm); + return t; + } else { /* register */ unsigned int rs2 = GET_FIELD(insn, 27, 31); - if (rs2 == 0) { - tcg_gen_movi_tl(def, 0); - } else if (rs2 < 8) { - r_rs2 = cpu_gregs[rs2]; - } else { - tcg_gen_ld_tl(def, cpu_regwptr, (rs2 - 8) * sizeof(target_ulong)); - } + return gen_load_gpr(dc, rs2); } - return r_rs2; } #ifdef TARGET_SPARC64 @@ -2560,7 +2543,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) (GET_FIELD_SP(insn, 20, 21) << 14); target = sign_extend(target, 16); target <<= 2; - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); do_branch_reg(dc, target, insn, cpu_src1); goto jmp_insn; } @@ -3187,7 +3170,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) do { \ DisasCompare cmp; \ cond = GET_FIELD_SP(insn, 14, 17); \ - cpu_src1 = get_src1(insn, cpu_src1); \ + cpu_src1 = get_src1(dc, insn); \ gen_compare_reg(&cmp, cond, cpu_src1); \ gen_fmov##sz(dc, &cmp, rd, rs2); \ free_compare(&cmp); \ @@ -3344,7 +3327,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) } } } else { - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); if (IS_IMM) { /* immediate */ simm = GET_FIELDs(insn, 19, 31); tcg_gen_ori_tl(dst, cpu_src1, simm); @@ -3363,7 +3346,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) } #ifdef TARGET_SPARC64 } else if (xop == 0x25) { /* sll, V9 sllx */ - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); if (IS_IMM) { /* immediate */ simm = GET_FIELDs(insn, 20, 31); if (insn & (1 << 12)) { @@ -3383,7 +3366,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) } gen_store_gpr(dc, rd, cpu_dst); } else if (xop == 0x26) { /* srl, V9 srlx */ - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); if (IS_IMM) { /* immediate */ simm = GET_FIELDs(insn, 20, 31); if (insn & (1 << 12)) { @@ -3406,7 +3389,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) } gen_store_gpr(dc, rd, cpu_dst); } else if (xop == 0x27) { /* sra, V9 srax */ - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); if (IS_IMM) { /* immediate */ simm = GET_FIELDs(insn, 20, 31); if (insn & (1 << 12)) { @@ -3431,8 +3414,8 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) #endif } else if (xop < 0x36) { if (xop < 0x20) { - cpu_src1 = get_src1(insn, cpu_src1); - cpu_src2 = get_src2(insn, cpu_src2); + cpu_src1 = get_src1(dc, insn); + cpu_src2 = get_src2(dc, insn); switch (xop & ~0x10) { case 0x0: /* add */ if (xop & 0x10) { @@ -3563,8 +3546,8 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) } gen_store_gpr(dc, rd, cpu_dst); } else { - cpu_src1 = get_src1(insn, cpu_src1); - cpu_src2 = get_src2(insn, cpu_src2); + cpu_src1 = get_src1(dc, insn); + cpu_src2 = get_src2(dc, insn); switch (xop) { case 0x20: /* taddcc */ gen_op_add_cc(cpu_dst, cpu_src1, cpu_src2); @@ -4153,14 +4136,14 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) break; case 0x010: /* VIS I array8 */ CHECK_FPU_FEATURE(dc, VIS1); - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = gen_load_gpr(dc, rs1); cpu_src2 = gen_load_gpr(dc, rs2); gen_helper_array8(cpu_dst, cpu_src1, cpu_src2); gen_store_gpr(dc, rd, cpu_dst); break; case 0x012: /* VIS I array16 */ CHECK_FPU_FEATURE(dc, VIS1); - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = gen_load_gpr(dc, rs1); cpu_src2 = gen_load_gpr(dc, rs2); gen_helper_array8(cpu_dst, cpu_src1, cpu_src2); tcg_gen_shli_i64(cpu_dst, cpu_dst, 1); @@ -4168,7 +4151,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) break; case 0x014: /* VIS I array32 */ CHECK_FPU_FEATURE(dc, VIS1); - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = gen_load_gpr(dc, rs1); cpu_src2 = gen_load_gpr(dc, rs2); gen_helper_array8(cpu_dst, cpu_src1, cpu_src2); tcg_gen_shli_i64(cpu_dst, cpu_dst, 2); @@ -4176,22 +4159,22 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) break; case 0x018: /* VIS I alignaddr */ CHECK_FPU_FEATURE(dc, VIS1); - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = gen_load_gpr(dc, rs1); cpu_src2 = gen_load_gpr(dc, rs2); gen_alignaddr(cpu_dst, cpu_src1, cpu_src2, 0); gen_store_gpr(dc, rd, cpu_dst); break; case 0x01a: /* VIS I alignaddrl */ CHECK_FPU_FEATURE(dc, VIS1); - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = gen_load_gpr(dc, rs1); cpu_src2 = gen_load_gpr(dc, rs2); gen_alignaddr(cpu_dst, cpu_src1, cpu_src2, 1); gen_store_gpr(dc, rd, cpu_dst); break; case 0x019: /* VIS II bmask */ CHECK_FPU_FEATURE(dc, VIS2); - cpu_src1 = get_src1(insn, cpu_src1); - cpu_src2 = get_src2(insn, cpu_src2); + cpu_src1 = gen_load_gpr(dc, rs1); + cpu_src2 = gen_load_gpr(dc, rs2); tcg_gen_add_tl(cpu_dst, cpu_src1, cpu_src2); tcg_gen_deposit_tl(cpu_gsr, cpu_gsr, cpu_dst, 32, 32); gen_store_gpr(dc, rd, cpu_dst); @@ -4511,7 +4494,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) TCGv_i32 r_const; save_state(dc); - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); if (IS_IMM) { /* immediate */ simm = GET_FIELDs(insn, 19, 31); tcg_gen_addi_tl(cpu_dst, cpu_src1, simm); @@ -4534,7 +4517,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) goto jmp_insn; #endif } else { - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); if (IS_IMM) { /* immediate */ simm = GET_FIELDs(insn, 19, 31); tcg_gen_addi_tl(cpu_dst, cpu_src1, simm); @@ -4632,7 +4615,7 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn) { unsigned int xop = GET_FIELD(insn, 7, 12); - cpu_src1 = get_src1(insn, cpu_src1); + cpu_src1 = get_src1(dc, insn); if (xop == 0x3c || xop == 0x3e) { // V9 casa/casxa rs2 = GET_FIELD(insn, 27, 31); cpu_src2 = gen_load_gpr(dc, rs2); |