@@ -9,6 +9,8 @@ DEF_HELPER_4(divull, void, env, int, int, i32)
DEF_HELPER_4(divsll, void, env, int, int, s32)
DEF_HELPER_2(set_sr, void, env, i32)
DEF_HELPER_3(movec, void, env, i32, i32)
+DEF_HELPER_4(cas2w, void, env, i32, i32, i32)
+DEF_HELPER_4(cas2l, void, env, i32, i32, i32)
DEF_HELPER_2(f64_to_i32, f32, env, f64)
DEF_HELPER_2(f64_to_f32, f32, env, f64)
@@ -359,3 +359,136 @@ void HELPER(divsll)(CPUM68KState *env, int numr, int regr, int32_t den)
env->dregs[regr] = rem;
env->dregs[numr] = quot;
}
+
+void HELPER(cas2w)(CPUM68KState *env, uint32_t regs, uint32_t a1, uint32_t a2)
+{
+ uint32_t Dc1 = extract32(regs, 9, 3);
+ uint32_t Dc2 = extract32(regs, 6, 3);
+ uint32_t Du1 = extract32(regs, 3, 3);
+ uint32_t Du2 = extract32(regs, 0, 3);
+ int16_t c1 = env->dregs[Dc1];
+ int16_t c2 = env->dregs[Dc2];
+ int16_t u1 = env->dregs[Du1];
+ int16_t u2 = env->dregs[Du2];
+ int16_t l1, l2;
+ uintptr_t ra = GETPC();
+
+ if (parallel_cpus) {
+ /* Tell the main loop we need to serialize this insn. */
+ cpu_loop_exit_atomic(ENV_GET_CPU(env), ra);
+ } else {
+ /* We're executing in a serial context -- no need to be atomic. */
+#ifdef CONFIG_USER_ONLY
+ int16_t *ha1 = g2h(a1);
+ int16_t *ha2 = g2h(a2);
+ l1 = ldsw_be_p(ha1);
+ l2 = ldsw_be_p(ha2);
+ if (l1 == c1 && l2 == c2) {
+ stw_be_p(ha1, u1);
+ stw_be_p(ha2, u2);
+ }
+#else
+ int mmu_idx = cpu_mmu_index(env, 0);
+ TCGMemOpIdx oi = make_memop_idx(MO_BEUW, mmu_idx);
+ l1 = helper_be_ldsw_mmu(env, a1, oi, ra);
+ l2 = helper_be_ldsw_mmu(env, a2, oi, ra);
+ if (l1 == c1 && l2 == c2) {
+ helper_be_stw_mmu(env, a1, u1, oi, ra);
+ helper_be_stw_mmu(env, a2, u2, oi, ra);
+ }
+#endif
+ }
+
+ if (c1 != l1) {
+ env->cc_n = l1;
+ env->cc_v = c1;
+ } else {
+ env->cc_n = l2;
+ env->cc_v = c2;
+ }
+ env->cc_op = CC_OP_CMPL;
+ env->dregs[Dc1] = deposit32(env->dregs[Dc1], 0, 16, l1);
+ env->dregs[Dc2] = deposit32(env->dregs[Dc2], 0, 16, l2);
+}
+
+void HELPER(cas2l)(CPUM68KState *env, uint32_t regs, uint32_t a1, uint32_t a2)
+{
+ uint32_t Dc1 = extract32(regs, 9, 3);
+ uint32_t Dc2 = extract32(regs, 6, 3);
+ uint32_t Du1 = extract32(regs, 3, 3);
+ uint32_t Du2 = extract32(regs, 0, 3);
+ uint32_t c1 = env->dregs[Dc1];
+ uint32_t c2 = env->dregs[Dc2];
+ uint32_t u1 = env->dregs[Du1];
+ uint32_t u2 = env->dregs[Du2];
+ uint32_t l1, l2;
+ uint64_t c, u, l;
+ uintptr_t ra = GETPC();
+#ifndef CONFIG_USER_ONLY
+ int mmu_idx = cpu_mmu_index(env, 0);
+ TCGMemOpIdx oi;
+#endif
+
+ if (parallel_cpus) {
+ /* We're executing in a parallel context -- must be atomic. */
+ if ((a1 & 7) == 0 && a2 == a1 + 4) {
+ c = deposit64(c2, 32, 32, c1);
+ u = deposit64(u2, 32, 32, u1);
+#ifdef CONFIG_USER_ONLY
+ uint64_t *ha1 = g2h(a1);
+ l = atomic_cmpxchg__nocheck(ha1, c, u);
+#else
+ oi = make_memop_idx(MO_BEQ, mmu_idx);
+ l = helper_atomic_cmpxchgq_be_mmu(env, a1, c, u, oi, ra);
+#endif
+ l1 = l >> 32;
+ l2 = l;
+ } else if ((a2 & 7) == 0 && a1 == a2 + 4) {
+ c = deposit64(c1, 32, 32, c2);
+ u = deposit64(u1, 32, 32, u2);
+#ifdef CONFIG_USER_ONLY
+ uint64_t *ha1 = g2h(a1);
+ l = atomic_cmpxchg__nocheck(ha1, c, u);
+#else
+ oi = make_memop_idx(MO_BEQ, mmu_idx);
+ l = helper_atomic_cmpxchgq_be_mmu(env, a1, c, u, oi, ra);
+#endif
+ l2 = l >> 32;
+ l1 = l;
+ } else {
+ /* Tell the main loop we need to serialize this insn. */
+ cpu_loop_exit_atomic(ENV_GET_CPU(env), ra);
+ }
+ } else {
+#ifdef CONFIG_USER_ONLY
+ uint32_t *ha1 = g2h(a1);
+ uint32_t *ha2 = g2h(a2);
+ l1 = ldl_be_p(ha1);
+ l2 = ldl_be_p(ha2);
+ if (l1 == c1 && l2 == c2) {
+ stl_be_p(ha1, u1);
+ stl_be_p(ha2, u2);
+ }
+#else
+ /* We're executing in a serial context -- no need to be atomic. */
+ oi = make_memop_idx(MO_BEUL, mmu_idx);
+ l1 = helper_be_ldul_mmu(env, a1, oi, ra);
+ l2 = helper_be_ldul_mmu(env, a2, oi, ra);
+ if (l1 == c1 && l2 == c2) {
+ helper_be_stl_mmu(env, a1, u1, oi, ra);
+ helper_be_stl_mmu(env, a2, u2, oi, ra);
+ }
+#endif
+ }
+
+ if (c1 != l1) {
+ env->cc_n = l1;
+ env->cc_v = c1;
+ } else {
+ env->cc_n = l2;
+ env->cc_v = c2;
+ }
+ env->cc_op = CC_OP_CMPL;
+ env->dregs[Dc1] = l1;
+ env->dregs[Dc2] = l2;
+}
@@ -1880,6 +1880,146 @@ DISAS_INSN(arith_im)
tcg_temp_free(dest);
}
+DISAS_INSN(cas)
+{
+ int opsize;
+ TCGv addr;
+ uint16_t ext;
+ TCGv load;
+ TCGv cmp;
+ TCGMemOp opc;
+
+ switch ((insn >> 9) & 3) {
+ case 1:
+ opsize = OS_BYTE;
+ opc = MO_UB;
+ break;
+ case 2:
+ opsize = OS_WORD;
+ opc = MO_TEUW;
+ break;
+ case 3:
+ opsize = OS_LONG;
+ opc = MO_TEUL;
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ opc |= MO_ALIGN;
+
+ ext = read_im16(env, s);
+
+ /* cas Dc,Du,<EA> */
+
+ addr = gen_lea(env, s, insn, opsize);
+ if (IS_NULL_QREG(addr)) {
+ gen_addr_fault(s);
+ return;
+ }
+
+ cmp = gen_extend(DREG(ext, 0), opsize, 0);
+
+ /* if <EA> == Dc then
+ * <EA> = Du
+ * Dc = <EA> (because <EA> == Dc)
+ * else
+ * Dc = <EA>
+ */
+
+ load = tcg_temp_new();
+ tcg_gen_atomic_cmpxchg_i32(load, addr, cmp, DREG(ext, 6),
+ IS_USER(s), opc);
+ gen_partset_reg(opsize, DREG(ext, 0), load);
+
+ gen_update_cc_cmp(s, load, cmp, opsize);
+ tcg_temp_free(load);
+}
+
+DISAS_INSN(cas2w)
+{
+ uint16_t ext1, ext2;
+ TCGv addr1, addr2;
+ TCGv regs;
+
+ /* cas2 Dc1:Dc2,Du1:Du2,(Rn1):(Rn2) */
+
+ ext1 = read_im16(env, s);
+
+ if (ext1 & 0x8000) {
+ /* Address Register */
+ addr1 = AREG(ext1, 12);
+ } else {
+ /* Data Register */
+ addr1 = DREG(ext1, 12);
+ }
+
+ ext2 = read_im16(env, s);
+ if (ext2 & 0x8000) {
+ /* Address Register */
+ addr2 = AREG(ext2, 12);
+ } else {
+ /* Data Register */
+ addr2 = DREG(ext2, 12);
+ }
+
+ /* if (R1) == Dc1 && (R2) == Dc2 then
+ * (R1) = Du1
+ * (R2) = Du2
+ * else
+ * Dc1 = (R1)
+ * Dc2 = (R2)
+ */
+
+ regs = tcg_const_i32(REG(ext2, 6) |
+ (REG(ext1, 6) << 3) |
+ (REG(ext2, 0) << 6) |
+ (REG(ext1, 0) << 9));
+ gen_helper_cas2w(cpu_env, regs, addr1, addr2);
+ tcg_temp_free(regs);
+}
+
+DISAS_INSN(cas2l)
+{
+ uint16_t ext1, ext2;
+ TCGv addr1, addr2, regs;
+
+ /* cas2 Dc1:Dc2,Du1:Du2,(Rn1):(Rn2) */
+
+ ext1 = read_im16(env, s);
+
+ if (ext1 & 0x8000) {
+ /* Address Register */
+ addr1 = AREG(ext1, 12);
+ } else {
+ /* Data Register */
+ addr1 = DREG(ext1, 12);
+ }
+
+ ext2 = read_im16(env, s);
+ if (ext2 & 0x8000) {
+ /* Address Register */
+ addr2 = AREG(ext2, 12);
+ } else {
+ /* Data Register */
+ addr2 = DREG(ext2, 12);
+ }
+
+ /* if (R1) == Dc1 && (R2) == Dc2 then
+ * (R1) = Du1
+ * (R2) = Du2
+ * else
+ * Dc1 = (R1)
+ * Dc2 = (R2)
+ */
+
+ regs = tcg_const_i32(REG(ext2, 6) |
+ (REG(ext1, 6) << 3) |
+ (REG(ext2, 0) << 6) |
+ (REG(ext1, 0) << 9));
+ gen_helper_cas2w(cpu_env, regs, addr1, addr2);
+ tcg_temp_free(regs);
+}
+
DISAS_INSN(byterev)
{
TCGv reg;
@@ -3885,6 +4025,9 @@ void register_m68k_insns (CPUM68KState *env)
INSN(arith_im, 0680, fff8, CF_ISA_A);
INSN(arith_im, 0c00, ff38, CF_ISA_A);
INSN(arith_im, 0c00, ff00, M68000);
+ INSN(cas, 08c0, f9c0, CAS);
+ INSN(cas2w, 0cfc, ffff, CAS);
+ INSN(cas2l, 0efc, ffff, CAS);
BASE(bitop_im, 0800, ffc0);
BASE(bitop_im, 0840, ffc0);
BASE(bitop_im, 0880, ffc0);
Implement CAS using cmpxchg. Implement CAS2 using helper and either cmpxchg when the 32bit addresses are consecutive, or with parallel_cpus+cpu_loop_exit_atomic() otherwise. Suggested-by: Richard Henderson <rth@twiddle.net> Signed-off-by: Laurent Vivier <laurent@vivier.eu> --- target-m68k/helper.h | 2 + target-m68k/op_helper.c | 133 ++++++++++++++++++++++++++++++++++++++++++++ target-m68k/translate.c | 143 ++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 278 insertions(+)