Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180510' into staging

target-arm queue:
 * hw/arm/iotkit.c: fix minor memory leak
 * softfloat: fix wrong-exception-flags bug for multiply-add corner case
 * arm: isolate and clean up DTB generation
 * implement Arm v8.1-Atomics extension
 * Fix some bugs and missing instructions in the v8.2-FP16 extension

# gpg: Signature made Thu 10 May 2018 18:44:34 BST
# gpg:                using RSA key 3C2525ED14360CDE
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>"
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>"
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>"
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20180510: (21 commits)
  target/arm: Clear SVE high bits for FMOV
  target/arm: Fix float16 to/from int16
  target/arm: Implement vector shifted FCVT for fp16
  target/arm: Implement vector shifted SCVF/UCVF for fp16
  target/arm: Enable ARM_FEATURE_V8_ATOMICS for user-only
  target/arm: Implement CAS and CASP
  target/arm: Fill in disas_ldst_atomic
  target/arm: Introduce ARM_FEATURE_V8_ATOMICS and initial decode
  target/riscv: Use new atomic min/max expanders
  tcg: Use GEN_ATOMIC_HELPER_FN for opposite endian atomic add
  tcg: Introduce atomic helpers for integer min/max
  target/xtensa: Use new min/max expanders
  target/arm: Use new min/max expanders
  tcg: Introduce helpers for integer min/max
  atomic.h: Work around gcc spurious "unused value" warning
  make sure that we aren't overwriting mc->get_hotplug_handler by accident
  arm/boot: split load_dtb() from arm_load_kernel()
  platform-bus-device: use device plug callback instead of machine_done notifier
  pc: simplify MachineClass::get_hotplug_handler handling
  softfloat: Handle default NaN mode after pickNaNMulAdd, not before
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

# Conflicts:
#	target/riscv/translate.c

9 files changed, 514 insertions, 188 deletions

diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 44e6b77..3b086be 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -1449,6 +1449,7 @@ enum arm_features {
     ARM_FEATURE_V8_SHA3, /* implements SHA3 part of v8 Crypto Extensions */
     ARM_FEATURE_V8_SM3, /* implements SM3 part of v8 Crypto Extensions */
     ARM_FEATURE_V8_SM4, /* implements SM4 part of v8 Crypto Extensions */
+    ARM_FEATURE_V8_ATOMICS, /* ARMv8.1-Atomics feature */
     ARM_FEATURE_V8_RDM, /* implements v8.1 simd round multiply */
     ARM_FEATURE_V8_FP16, /* implements v8.2 half-precision float */
     ARM_FEATURE_V8_FCMA, /* has complex number part of v8.3 extensions.  */
diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c
index 991d764..c50dcd4 100644
--- a/target/arm/cpu64.c
+++ b/target/arm/cpu64.c
@@ -248,6 +248,7 @@ static void aarch64_max_initfn(Object *obj)
         set_feature(&cpu->env, ARM_FEATURE_V8_SM4);
         set_feature(&cpu->env, ARM_FEATURE_V8_PMULL);
         set_feature(&cpu->env, ARM_FEATURE_CRC);
+        set_feature(&cpu->env, ARM_FEATURE_V8_ATOMICS);
         set_feature(&cpu->env, ARM_FEATURE_V8_RDM);
         set_feature(&cpu->env, ARM_FEATURE_V8_FP16);
         set_feature(&cpu->env, ARM_FEATURE_V8_FCMA);
diff --git a/target/arm/helper-a64.c b/target/arm/helper-a64.c
index afb25ad..549ed35 100644
--- a/target/arm/helper-a64.c
+++ b/target/arm/helper-a64.c
@@ -636,6 +636,49 @@ uint64_t HELPER(paired_cmpxchg64_be_parallel)(CPUARMState *env, uint64_t addr,
     return do_paired_cmpxchg64_be(env, addr, new_lo, new_hi, true, GETPC());
 }
 
+/* Writes back the old data into Rs.  */
+void HELPER(casp_le_parallel)(CPUARMState *env, uint32_t rs, uint64_t addr,
+                              uint64_t new_lo, uint64_t new_hi)
+{
+    uintptr_t ra = GETPC();
+#ifndef CONFIG_ATOMIC128
+    cpu_loop_exit_atomic(ENV_GET_CPU(env), ra);
+#else
+    Int128 oldv, cmpv, newv;
+
+    cmpv = int128_make128(env->xregs[rs], env->xregs[rs + 1]);
+    newv = int128_make128(new_lo, new_hi);
+
+    int mem_idx = cpu_mmu_index(env, false);
+    TCGMemOpIdx oi = make_memop_idx(MO_LEQ | MO_ALIGN_16, mem_idx);
+    oldv = helper_atomic_cmpxchgo_le_mmu(env, addr, cmpv, newv, oi, ra);
+
+    env->xregs[rs] = int128_getlo(oldv);
+    env->xregs[rs + 1] = int128_gethi(oldv);
+#endif
+}
+
+void HELPER(casp_be_parallel)(CPUARMState *env, uint32_t rs, uint64_t addr,
+                              uint64_t new_hi, uint64_t new_lo)
+{
+    uintptr_t ra = GETPC();
+#ifndef CONFIG_ATOMIC128
+    cpu_loop_exit_atomic(ENV_GET_CPU(env), ra);
+#else
+    Int128 oldv, cmpv, newv;
+
+    cmpv = int128_make128(env->xregs[rs + 1], env->xregs[rs]);
+    newv = int128_make128(new_lo, new_hi);
+
+    int mem_idx = cpu_mmu_index(env, false);
+    TCGMemOpIdx oi = make_memop_idx(MO_LEQ | MO_ALIGN_16, mem_idx);
+    oldv = helper_atomic_cmpxchgo_be_mmu(env, addr, cmpv, newv, oi, ra);
+
+    env->xregs[rs + 1] = int128_getlo(oldv);
+    env->xregs[rs] = int128_gethi(oldv);
+#endif
+}
+
 /*
  * AdvSIMD half-precision
  */
diff --git a/target/arm/helper-a64.h b/target/arm/helper-a64.h
index ef4ddfe..b8028ac 100644
--- a/target/arm/helper-a64.h
+++ b/target/arm/helper-a64.h
@@ -51,6 +51,8 @@ DEF_HELPER_FLAGS_4(paired_cmpxchg64_le_parallel, TCG_CALL_NO_WG,
 DEF_HELPER_FLAGS_4(paired_cmpxchg64_be, TCG_CALL_NO_WG, i64, env, i64, i64, i64)
 DEF_HELPER_FLAGS_4(paired_cmpxchg64_be_parallel, TCG_CALL_NO_WG,
                    i64, env, i64, i64, i64)
+DEF_HELPER_5(casp_le_parallel, void, env, i32, i64, i64, i64)
+DEF_HELPER_5(casp_be_parallel, void, env, i32, i64, i64, i64)
 DEF_HELPER_FLAGS_3(advsimd_maxh, TCG_CALL_NO_RWG, f16, f16, f16, ptr)
 DEF_HELPER_FLAGS_3(advsimd_minh, TCG_CALL_NO_RWG, f16, f16, f16, ptr)
 DEF_HELPER_FLAGS_3(advsimd_maxnumh, TCG_CALL_NO_RWG, f16, f16, f16, ptr)
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 0fef5d4..817f9d8 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -11420,11 +11420,60 @@ VFP_CONV_FIX_A64(sq, s, 32, 64, int64)
 VFP_CONV_FIX(uh, s, 32, 32, uint16)
 VFP_CONV_FIX(ul, s, 32, 32, uint32)
 VFP_CONV_FIX_A64(uq, s, 32, 64, uint64)
-VFP_CONV_FIX_A64(sl, h, 16, 32, int32)
-VFP_CONV_FIX_A64(ul, h, 16, 32, uint32)
+
 #undef VFP_CONV_FIX
 #undef VFP_CONV_FIX_FLOAT
 #undef VFP_CONV_FLOAT_FIX_ROUND
+#undef VFP_CONV_FIX_A64
+
+/* Conversion to/from f16 can overflow to infinity before/after scaling.
+ * Therefore we convert to f64 (which does not round), scale,
+ * and then convert f64 to f16 (which may round).
+ */
+
+static float16 do_postscale_fp16(float64 f, int shift, float_status *fpst)
+{
+    return float64_to_float16(float64_scalbn(f, -shift, fpst), true, fpst);
+}
+
+float16 HELPER(vfp_sltoh)(uint32_t x, uint32_t shift, void *fpst)
+{
+    return do_postscale_fp16(int32_to_float64(x, fpst), shift, fpst);
+}
+
+float16 HELPER(vfp_ultoh)(uint32_t x, uint32_t shift, void *fpst)
+{
+    return do_postscale_fp16(uint32_to_float64(x, fpst), shift, fpst);
+}
+
+static float64 do_prescale_fp16(float16 f, int shift, float_status *fpst)
+{
+    if (unlikely(float16_is_any_nan(f))) {
+        float_raise(float_flag_invalid, fpst);
+        return 0;
+    } else {
+        int old_exc_flags = get_float_exception_flags(fpst);
+        float64 ret;
+
+        ret = float16_to_float64(f, true, fpst);
+        ret = float64_scalbn(ret, shift, fpst);
+        old_exc_flags |= get_float_exception_flags(fpst)
+            & float_flag_input_denormal;
+        set_float_exception_flags(old_exc_flags, fpst);
+
+        return ret;
+    }
+}
+
+uint32_t HELPER(vfp_toshh)(float16 x, uint32_t shift, void *fpst)
+{
+    return float64_to_int16(do_prescale_fp16(x, shift, fpst), fpst);
+}
+
+uint32_t HELPER(vfp_touhh)(float16 x, uint32_t shift, void *fpst)
+{
+    return float64_to_uint16(do_prescale_fp16(x, shift, fpst), fpst);
+}
 
 /* Set the current fp rounding mode and return the old one.
  * The argument is a softfloat float_round_ value.
diff --git a/target/arm/helper.h b/target/arm/helper.h
index 34e8cc8..1969b37 100644
--- a/target/arm/helper.h
+++ b/target/arm/helper.h
@@ -149,8 +149,8 @@ DEF_HELPER_3(vfp_toshd_round_to_zero, i64, f64, i32, ptr)
 DEF_HELPER_3(vfp_tosld_round_to_zero, i64, f64, i32, ptr)
 DEF_HELPER_3(vfp_touhd_round_to_zero, i64, f64, i32, ptr)
 DEF_HELPER_3(vfp_tould_round_to_zero, i64, f64, i32, ptr)
-DEF_HELPER_3(vfp_toulh, i32, f16, i32, ptr)
-DEF_HELPER_3(vfp_toslh, i32, f16, i32, ptr)
+DEF_HELPER_3(vfp_touhh, i32, f16, i32, ptr)
+DEF_HELPER_3(vfp_toshh, i32, f16, i32, ptr)
 DEF_HELPER_3(vfp_toshs, i32, f32, i32, ptr)
 DEF_HELPER_3(vfp_tosls, i32, f32, i32, ptr)
 DEF_HELPER_3(vfp_tosqs, i64, f32, i32, ptr)
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 1e7c150..4d1b220 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -84,6 +84,7 @@ typedef void NeonGenOneOpFn(TCGv_i64, TCGv_i64);
 typedef void CryptoTwoOpFn(TCGv_ptr, TCGv_ptr);
 typedef void CryptoThreeOpIntFn(TCGv_ptr, TCGv_ptr, TCGv_i32);
 typedef void CryptoThreeOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr);
+typedef void AtomicThreeOpFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGArg, TCGMemOp);
 
 /* Note that the gvec expanders operate on offsets + sizes.  */
 typedef void GVecGen2Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t);
@@ -2113,6 +2114,103 @@ static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
     tcg_gen_movi_i64(cpu_exclusive_addr, -1);
 }
 
+static void gen_compare_and_swap(DisasContext *s, int rs, int rt,
+                                 int rn, int size)
+{
+    TCGv_i64 tcg_rs = cpu_reg(s, rs);
+    TCGv_i64 tcg_rt = cpu_reg(s, rt);
+    int memidx = get_mem_index(s);
+    TCGv_i64 addr = cpu_reg_sp(s, rn);
+
+    if (rn == 31) {
+        gen_check_sp_alignment(s);
+    }
+    tcg_gen_atomic_cmpxchg_i64(tcg_rs, addr, tcg_rs, tcg_rt, memidx,
+                               size | MO_ALIGN | s->be_data);
+}
+
+static void gen_compare_and_swap_pair(DisasContext *s, int rs, int rt,
+                                      int rn, int size)
+{
+    TCGv_i64 s1 = cpu_reg(s, rs);
+    TCGv_i64 s2 = cpu_reg(s, rs + 1);
+    TCGv_i64 t1 = cpu_reg(s, rt);
+    TCGv_i64 t2 = cpu_reg(s, rt + 1);
+    TCGv_i64 addr = cpu_reg_sp(s, rn);
+    int memidx = get_mem_index(s);
+
+    if (rn == 31) {
+        gen_check_sp_alignment(s);
+    }
+
+    if (size == 2) {
+        TCGv_i64 cmp = tcg_temp_new_i64();
+        TCGv_i64 val = tcg_temp_new_i64();
+
+        if (s->be_data == MO_LE) {
+            tcg_gen_concat32_i64(val, t1, t2);
+            tcg_gen_concat32_i64(cmp, s1, s2);
+        } else {
+            tcg_gen_concat32_i64(val, t2, t1);
+            tcg_gen_concat32_i64(cmp, s2, s1);
+        }
+
+        tcg_gen_atomic_cmpxchg_i64(cmp, addr, cmp, val, memidx,
+                                   MO_64 | MO_ALIGN | s->be_data);
+        tcg_temp_free_i64(val);
+
+        if (s->be_data == MO_LE) {
+            tcg_gen_extr32_i64(s1, s2, cmp);
+        } else {
+            tcg_gen_extr32_i64(s2, s1, cmp);
+        }
+        tcg_temp_free_i64(cmp);
+    } else if (tb_cflags(s->base.tb) & CF_PARALLEL) {
+        TCGv_i32 tcg_rs = tcg_const_i32(rs);
+
+        if (s->be_data == MO_LE) {
+            gen_helper_casp_le_parallel(cpu_env, tcg_rs, addr, t1, t2);
+        } else {
+            gen_helper_casp_be_parallel(cpu_env, tcg_rs, addr, t1, t2);
+        }
+        tcg_temp_free_i32(tcg_rs);
+    } else {
+        TCGv_i64 d1 = tcg_temp_new_i64();
+        TCGv_i64 d2 = tcg_temp_new_i64();
+        TCGv_i64 a2 = tcg_temp_new_i64();
+        TCGv_i64 c1 = tcg_temp_new_i64();
+        TCGv_i64 c2 = tcg_temp_new_i64();
+        TCGv_i64 zero = tcg_const_i64(0);
+
+        /* Load the two words, in memory order.  */
+        tcg_gen_qemu_ld_i64(d1, addr, memidx,
+                            MO_64 | MO_ALIGN_16 | s->be_data);
+        tcg_gen_addi_i64(a2, addr, 8);
+        tcg_gen_qemu_ld_i64(d2, addr, memidx, MO_64 | s->be_data);
+
+        /* Compare the two words, also in memory order.  */
+        tcg_gen_setcond_i64(TCG_COND_EQ, c1, d1, s1);
+        tcg_gen_setcond_i64(TCG_COND_EQ, c2, d2, s2);
+        tcg_gen_and_i64(c2, c2, c1);
+
+        /* If compare equal, write back new data, else write back old data.  */
+        tcg_gen_movcond_i64(TCG_COND_NE, c1, c2, zero, t1, d1);
+        tcg_gen_movcond_i64(TCG_COND_NE, c2, c2, zero, t2, d2);
+        tcg_gen_qemu_st_i64(c1, addr, memidx, MO_64 | s->be_data);
+        tcg_gen_qemu_st_i64(c2, a2, memidx, MO_64 | s->be_data);
+        tcg_temp_free_i64(a2);
+        tcg_temp_free_i64(c1);
+        tcg_temp_free_i64(c2);
+        tcg_temp_free_i64(zero);
+
+        /* Write back the data from memory to Rs.  */
+        tcg_gen_mov_i64(s1, d1);
+        tcg_gen_mov_i64(s2, d2);
+        tcg_temp_free_i64(d1);
+        tcg_temp_free_i64(d2);
+    }
+}
+
 /* Update the Sixty-Four bit (SF) registersize. This logic is derived
  * from the ARMv8 specs for LDR (Shared decode for all encodings).
  */
@@ -2147,62 +2245,114 @@ static void disas_ldst_excl(DisasContext *s, uint32_t insn)
     int rt = extract32(insn, 0, 5);
     int rn = extract32(insn, 5, 5);
     int rt2 = extract32(insn, 10, 5);
-    int is_lasr = extract32(insn, 15, 1);
     int rs = extract32(insn, 16, 5);
-    int is_pair = extract32(insn, 21, 1);
-    int is_store = !extract32(insn, 22, 1);
-    int is_excl = !extract32(insn, 23, 1);
+    int is_lasr = extract32(insn, 15, 1);
+    int o2_L_o1_o0 = extract32(insn, 21, 3) * 2 | is_lasr;
     int size = extract32(insn, 30, 2);
     TCGv_i64 tcg_addr;
 
-    if ((!is_excl && !is_pair && !is_lasr) ||
-        (!is_excl && is_pair) ||
-        (is_pair && size < 2)) {
-        unallocated_encoding(s);
+    switch (o2_L_o1_o0) {
+    case 0x0: /* STXR */
+    case 0x1: /* STLXR */
+        if (rn == 31) {
+            gen_check_sp_alignment(s);
+        }
+        if (is_lasr) {
+            tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
+        }
+        tcg_addr = read_cpu_reg_sp(s, rn, 1);
+        gen_store_exclusive(s, rs, rt, rt2, tcg_addr, size, false);
         return;
-    }
 
-    if (rn == 31) {
-        gen_check_sp_alignment(s);
-    }
-    tcg_addr = read_cpu_reg_sp(s, rn, 1);
+    case 0x4: /* LDXR */
+    case 0x5: /* LDAXR */
+        if (rn == 31) {
+            gen_check_sp_alignment(s);
+        }
+        tcg_addr = read_cpu_reg_sp(s, rn, 1);
+        s->is_ldex = true;
+        gen_load_exclusive(s, rt, rt2, tcg_addr, size, false);
+        if (is_lasr) {
+            tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
+        }
+        return;
 
-    /* Note that since TCG is single threaded load-acquire/store-release
-     * semantics require no extra if (is_lasr) { ... } handling.
-     */
+    case 0x9: /* STLR */
+        /* Generate ISS for non-exclusive accesses including LASR.  */
+        if (rn == 31) {
+            gen_check_sp_alignment(s);
+        }
+        tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
+        tcg_addr = read_cpu_reg_sp(s, rn, 1);
+        do_gpr_st(s, cpu_reg(s, rt), tcg_addr, size, true, rt,
+                  disas_ldst_compute_iss_sf(size, false, 0), is_lasr);
+        return;
 
-    if (is_excl) {
-        if (!is_store) {
-            s->is_ldex = true;
-            gen_load_exclusive(s, rt, rt2, tcg_addr, size, is_pair);
-            if (is_lasr) {
-                tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
+    case 0xd: /* LDAR */
+        /* Generate ISS for non-exclusive accesses including LASR.  */
+        if (rn == 31) {
+            gen_check_sp_alignment(s);
+        }
+        tcg_addr = read_cpu_reg_sp(s, rn, 1);
+        do_gpr_ld(s, cpu_reg(s, rt), tcg_addr, size, false, false, true, rt,
+                  disas_ldst_compute_iss_sf(size, false, 0), is_lasr);
+        tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
+        return;
+
+    case 0x2: case 0x3: /* CASP / STXP */
+        if (size & 2) { /* STXP / STLXP */
+            if (rn == 31) {
+                gen_check_sp_alignment(s);
             }
-        } else {
             if (is_lasr) {
                 tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
             }
-            gen_store_exclusive(s, rs, rt, rt2, tcg_addr, size, is_pair);
+            tcg_addr = read_cpu_reg_sp(s, rn, 1);
+            gen_store_exclusive(s, rs, rt, rt2, tcg_addr, size, true);
+            return;
         }
-    } else {
-        TCGv_i64 tcg_rt = cpu_reg(s, rt);
-        bool iss_sf = disas_ldst_compute_iss_sf(size, false, 0);
+        if (rt2 == 31
+            && ((rt | rs) & 1) == 0
+            && arm_dc_feature(s, ARM_FEATURE_V8_ATOMICS)) {
+            /* CASP / CASPL */
+            gen_compare_and_swap_pair(s, rs, rt, rn, size | 2);
+            return;
+        }
+        break;
 
-        /* Generate ISS for non-exclusive accesses including LASR.  */
-        if (is_store) {
-            if (is_lasr) {
-                tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
+    case 0x6: case 0x7: /* CASPA / LDXP */
+        if (size & 2) { /* LDXP / LDAXP */
+            if (rn == 31) {
+                gen_check_sp_alignment(s);
             }
-            do_gpr_st(s, tcg_rt, tcg_addr, size,
-                      true, rt, iss_sf, is_lasr);
-        } else {
-            do_gpr_ld(s, tcg_rt, tcg_addr, size, false, false,
-                      true, rt, iss_sf, is_lasr);
+            tcg_addr = read_cpu_reg_sp(s, rn, 1);
+            s->is_ldex = true;
+            gen_load_exclusive(s, rt, rt2, tcg_addr, size, true);
             if (is_lasr) {
                 tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
             }
+            return;
+        }
+        if (rt2 == 31
+            && ((rt | rs) & 1) == 0
+            && arm_dc_feature(s, ARM_FEATURE_V8_ATOMICS)) {
+            /* CASPA / CASPAL */
+            gen_compare_and_swap_pair(s, rs, rt, rn, size | 2);
+            return;
         }
+        break;
+
+    case 0xa: /* CAS */
+    case 0xb: /* CASL */
+    case 0xe: /* CASA */
+    case 0xf: /* CASAL */
+        if (rt2 == 31 && arm_dc_feature(s, ARM_FEATURE_V8_ATOMICS)) {
+            gen_compare_and_swap(s, rs, rt, rn, size);
+            return;
+        }
+        break;
     }
+    unallocated_encoding(s);
 }
 
 /*
@@ -2715,6 +2865,88 @@ static void disas_ldst_reg_unsigned_imm(DisasContext *s, uint32_t insn,
     }
 }
 
+/* Atomic memory operations
+ *
+ *  31  30      27  26    24    22  21   16   15    12    10    5     0
+ * +------+-------+---+-----+-----+---+----+----+-----+-----+----+-----+
+ * | size | 1 1 1 | V | 0 0 | A R | 1 | Rs | o3 | opc | 0 0 | Rn |  Rt |
+ * +------+-------+---+-----+-----+--------+----+-----+-----+----+-----+
+ *
+ * Rt: the result register
+ * Rn: base address or SP
+ * Rs: the source register for the operation
+ * V: vector flag (always 0 as of v8.3)
+ * A: acquire flag
+ * R: release flag
+ */
+static void disas_ldst_atomic(DisasContext *s, uint32_t insn,
+                              int size, int rt, bool is_vector)
+{
+    int rs = extract32(insn, 16, 5);
+    int rn = extract32(insn, 5, 5);
+    int o3_opc = extract32(insn, 12, 4);
+    int feature = ARM_FEATURE_V8_ATOMICS;
+    TCGv_i64 tcg_rn, tcg_rs;
+    AtomicThreeOpFn *fn;
+
+    if (is_vector) {
+        unallocated_encoding(s);
+        return;
+    }
+    switch (o3_opc) {
+    case 000: /* LDADD */
+        fn = tcg_gen_atomic_fetch_add_i64;
+        break;
+    case 001: /* LDCLR */
+        fn = tcg_gen_atomic_fetch_and_i64;
+        break;
+    case 002: /* LDEOR */
+        fn = tcg_gen_atomic_fetch_xor_i64;
+        break;
+    case 003: /* LDSET */
+        fn = tcg_gen_atomic_fetch_or_i64;
+        break;
+    case 004: /* LDSMAX */
+        fn = tcg_gen_atomic_fetch_smax_i64;
+        break;
+    case 005: /* LDSMIN */
+        fn = tcg_gen_atomic_fetch_smin_i64;
+        break;
+    case 006: /* LDUMAX */
+        fn = tcg_gen_atomic_fetch_umax_i64;
+        break;
+    case 007: /* LDUMIN */
+        fn = tcg_gen_atomic_fetch_umin_i64;
+        break;
+    case 010: /* SWP */
+        fn = tcg_gen_atomic_xchg_i64;
+        break;
+    default:
+        unallocated_encoding(s);
+        return;
+    }
+    if (!arm_dc_feature(s, feature)) {
+        unallocated_encoding(s);
+        return;
+    }
+
+    if (rn == 31) {
+        gen_check_sp_alignment(s);
+    }
+    tcg_rn = cpu_reg_sp(s, rn);
+    tcg_rs = read_cpu_reg(s, rs, true);
+
+    if (o3_opc == 1) { /* LDCLR */
+        tcg_gen_not_i64(tcg_rs, tcg_rs);
+    }
+
+    /* The tcg atomic primitives are all full barriers.  Therefore we
+     * can ignore the Acquire and Release bits of this instruction.
+     */
+    fn(cpu_reg(s, rt), tcg_rn, tcg_rs, get_mem_index(s),
+       s->be_data | size | MO_ALIGN);
+}
+
 /* Load/store register (all forms) */
 static void disas_ldst_reg(DisasContext *s, uint32_t insn)
 {
@@ -2725,23 +2957,28 @@ static void disas_ldst_reg(DisasContext *s, uint32_t insn)
 
     switch (extract32(insn, 24, 2)) {
     case 0:
-        if (extract32(insn, 21, 1) == 1 && extract32(insn, 10, 2) == 2) {
-            disas_ldst_reg_roffset(s, insn, opc, size, rt, is_vector);
-        } else {
+        if (extract32(insn, 21, 1) == 0) {
             /* Load/store register (unscaled immediate)
              * Load/store immediate pre/post-indexed
              * Load/store register unprivileged
              */
             disas_ldst_reg_imm9(s, insn, opc, size, rt, is_vector);
+            return;
+        }
+        switch (extract32(insn, 10, 2)) {
+        case 0:
+            disas_ldst_atomic(s, insn, size, rt, is_vector);
+            return;
+        case 2:
+            disas_ldst_reg_roffset(s, insn, opc, size, rt, is_vector);
+            return;
         }
         break;
     case 1:
         disas_ldst_reg_unsigned_imm(s, insn, opc, size, rt, is_vector);
-        break;
-    default:
-        unallocated_encoding(s);
-        break;
+        return;
     }
+    unallocated_encoding(s);
 }
 
 /* AdvSIMD load/store multiple structures
@@ -5444,31 +5681,24 @@ static void handle_fmov(DisasContext *s, int rd, int rn, int type, bool itof)
 
     if (itof) {
         TCGv_i64 tcg_rn = cpu_reg(s, rn);
+        TCGv_i64 tmp;
 
         switch (type) {
         case 0:
-        {
             /* 32 bit */
-            TCGv_i64 tmp = tcg_temp_new_i64();
+            tmp = tcg_temp_new_i64();
             tcg_gen_ext32u_i64(tmp, tcg_rn);
-            tcg_gen_st_i64(tmp, cpu_env, fp_reg_offset(s, rd, MO_64));
-            tcg_gen_movi_i64(tmp, 0);
-            tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(s, rd));
+            write_fp_dreg(s, rd, tmp);
             tcg_temp_free_i64(tmp);
             break;
-        }
         case 1:
-        {
             /* 64 bit */
-            TCGv_i64 tmp = tcg_const_i64(0);
-            tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_offset(s, rd, MO_64));
-            tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(s, rd));
-            tcg_temp_free_i64(tmp);
+            write_fp_dreg(s, rd, tcg_rn);
             break;
-        }
         case 2:
             /* 64 bit to top half. */
             tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_hi_offset(s, rd));
+            clear_vec_high(s, true, rd);
             break;
         }
     } else {
@@ -6021,15 +6251,18 @@ static void disas_simd_across_lanes(DisasContext *s, uint32_t insn)
                 tcg_gen_add_i64(tcg_res, tcg_res, tcg_elt);
                 break;
             case 0x0a: /* SMAXV / UMAXV */
-                tcg_gen_movcond_i64(is_u ? TCG_COND_GEU : TCG_COND_GE,
-                                    tcg_res,
-                                    tcg_res, tcg_elt, tcg_res, tcg_elt);
+                if (is_u) {
+                    tcg_gen_umax_i64(tcg_res, tcg_res, tcg_elt);
+                } else {
+                    tcg_gen_smax_i64(tcg_res, tcg_res, tcg_elt);
+                }
                 break;
             case 0x1a: /* SMINV / UMINV */
-                tcg_gen_movcond_i64(is_u ? TCG_COND_LEU : TCG_COND_LE,
-                                    tcg_res,
-                                    tcg_res, tcg_elt, tcg_res, tcg_elt);
-                break;
+                if (is_u) {
+                    tcg_gen_umin_i64(tcg_res, tcg_res, tcg_elt);
+                } else {
+                    tcg_gen_smin_i64(tcg_res, tcg_res, tcg_elt);
+                }
                 break;
             default:
                 g_assert_not_reached();
@@ -7165,13 +7398,26 @@ static void handle_simd_shift_intfp_conv(DisasContext *s, bool is_scalar,
                                          int immh, int immb, int opcode,
                                          int rn, int rd)
 {
-    bool is_double = extract32(immh, 3, 1);
-    int size = is_double ? MO_64 : MO_32;
-    int elements;
+    int size, elements, fracbits;
     int immhb = immh << 3 | immb;
-    int fracbits = (is_double ? 128 : 64) - immhb;
 
-    if (!extract32(immh, 2, 2)) {
+    if (immh & 8) {
+        size = MO_64;
+        if (!is_scalar && !is_q) {
+            unallocated_encoding(s);
+            return;
+        }
+    } else if (immh & 4) {
+        size = MO_32;
+    } else if (immh & 2) {
+        size = MO_16;
+        if (!arm_dc_feature(s, ARM_FEATURE_V8_FP16)) {
+            unallocated_encoding(s);
+            return;
+        }
+    } else {
+        /* immh == 0 would be a failure of the decode logic */
+        g_assert(immh == 1);
         unallocated_encoding(s);
         return;
     }
@@ -7179,20 +7425,14 @@ static void handle_simd_shift_intfp_conv(DisasContext *s, bool is_scalar,
     if (is_scalar) {
         elements = 1;
     } else {
-        elements = is_double ? 2 : is_q ? 4 : 2;
-        if (is_double && !is_q) {
-            unallocated_encoding(s);
-            return;
-        }
+        elements = (8 << is_q) >> size;
     }
+    fracbits = (16 << size) - immhb;
 
     if (!fp_access_check(s)) {
         return;
     }
 
-    /* immh == 0 would be a failure of the decode logic */
-    g_assert(immh);
-
     handle_simd_intfp_conv(s, rd, rn, elements, !is_u, fracbits, size);
 }
 
@@ -7201,19 +7441,28 @@ static void handle_simd_shift_fpint_conv(DisasContext *s, bool is_scalar,
                                          bool is_q, bool is_u,
                                          int immh, int immb, int rn, int rd)
 {
-    bool is_double = extract32(immh, 3, 1);
     int immhb = immh << 3 | immb;
-    int fracbits = (is_double ? 128 : 64) - immhb;
-    int pass;
+    int pass, size, fracbits;
     TCGv_ptr tcg_fpstatus;
     TCGv_i32 tcg_rmode, tcg_shift;
 
-    if (!extract32(immh, 2, 2)) {
-        unallocated_encoding(s);
-        return;
-    }
-
-    if (!is_scalar && !is_q && is_double) {
+    if (immh & 0x8) {
+        size = MO_64;
+        if (!is_scalar && !is_q) {
+            unallocated_encoding(s);
+            return;
+        }
+    } else if (immh & 0x4) {
+        size = MO_32;
+    } else if (immh & 0x2) {
+        size = MO_16;
+        if (!arm_dc_feature(s, ARM_FEATURE_V8_FP16)) {
+            unallocated_encoding(s);
+            return;
+        }
+    } else {
+        /* Should have split out AdvSIMD modified immediate earlier.  */
+        assert(immh == 1);
         unallocated_encoding(s);
         return;
     }
@@ -7225,11 +7474,12 @@ static void handle_simd_shift_fpint_conv(DisasContext *s, bool is_scalar,
     assert(!(is_scalar && is_q));
 
     tcg_rmode = tcg_const_i32(arm_rmode_to_sf(FPROUNDING_ZERO));
-    tcg_fpstatus = get_fpstatus_ptr(false);
+    tcg_fpstatus = get_fpstatus_ptr(size == MO_16);
     gen_helper_set_rmode(tcg_rmode, tcg_rmode, tcg_fpstatus);
+    fracbits = (16 << size) - immhb;
     tcg_shift = tcg_const_i32(fracbits);
 
-    if (is_double) {
+    if (size == MO_64) {
         int maxpass = is_scalar ? 1 : 2;
 
         for (pass = 0; pass < maxpass; pass++) {
@@ -7246,20 +7496,37 @@ static void handle_simd_shift_fpint_conv(DisasContext *s, bool is_scalar,
         }
         clear_vec_high(s, is_q, rd);
     } else {
-        int maxpass = is_scalar ? 1 : is_q ? 4 : 2;
-        for (pass = 0; pass < maxpass; pass++) {
-            TCGv_i32 tcg_op = tcg_temp_new_i32();
+        void (*fn)(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_ptr);
+        int maxpass = is_scalar ? 1 : ((8 << is_q) >> size);
 
-            read_vec_element_i32(s, tcg_op, rn, pass, MO_32);
+        switch (size) {
+        case MO_16:
             if (is_u) {
-                gen_helper_vfp_touls(tcg_op, tcg_op, tcg_shift, tcg_fpstatus);
+                fn = gen_helper_vfp_touhh;
             } else {
-                gen_helper_vfp_tosls(tcg_op, tcg_op, tcg_shift, tcg_fpstatus);
+                fn = gen_helper_vfp_toshh;
             }
+            break;
+        case MO_32:
+            if (is_u) {
+                fn = gen_helper_vfp_touls;
+            } else {
+                fn = gen_helper_vfp_tosls;
+            }
+            break;
+        default:
+            g_assert_not_reached();
+        }
+
+        for (pass = 0; pass < maxpass; pass++) {
+            TCGv_i32 tcg_op = tcg_temp_new_i32();
+
+            read_vec_element_i32(s, tcg_op, rn, pass, size);
+            fn(tcg_op, tcg_op, tcg_shift, tcg_fpstatus);
             if (is_scalar) {
                 write_fp_sreg(s, rd, tcg_op);
             } else {
-                write_vec_element_i32(s, tcg_op, rd, pass, MO_32);
+                write_vec_element_i32(s, tcg_op, rd, pass, size);
             }
             tcg_temp_free_i32(tcg_op);
         }
@@ -9927,27 +10194,6 @@ static void disas_simd_3same_logic(DisasContext *s, uint32_t insn)
     }
 }
 
-/* Helper functions for 32 bit comparisons */
-static void gen_max_s32(TCGv_i32 res, TCGv_i32 op1, TCGv_i32 op2)
-{
-    tcg_gen_movcond_i32(TCG_COND_GE, res, op1, op2, op1, op2);
-}
-
-static void gen_max_u32(TCGv_i32 res, TCGv_i32 op1, TCGv_i32 op2)
-{
-    tcg_gen_movcond_i32(TCG_COND_GEU, res, op1, op2, op1, op2);
-}
-
-static void gen_min_s32(TCGv_i32 res, TCGv_i32 op1, TCGv_i32 op2)
-{
-    tcg_gen_movcond_i32(TCG_COND_LE, res, op1, op2, op1, op2);
-}
-
-static void gen_min_u32(TCGv_i32 res, TCGv_i32 op1, TCGv_i32 op2)
-{
-    tcg_gen_movcond_i32(TCG_COND_LEU, res, op1, op2, op1, op2);
-}
-
 /* Pairwise op subgroup of C3.6.16.
  *
  * This is called directly or via the handle_3same_float for float pairwise
@@ -10047,7 +10293,7 @@ static void handle_simd_3same_pair(DisasContext *s, int is_q, int u, int opcode,
                 static NeonGenTwoOpFn * const fns[3][2] = {
                     { gen_helper_neon_pmax_s8, gen_helper_neon_pmax_u8 },
                     { gen_helper_neon_pmax_s16, gen_helper_neon_pmax_u16 },
-                    { gen_max_s32, gen_max_u32 },
+                    { tcg_gen_smax_i32, tcg_gen_umax_i32 },
                 };
                 genfn = fns[size][u];
                 break;
@@ -10057,7 +10303,7 @@ static void handle_simd_3same_pair(DisasContext *s, int is_q, int u, int opcode,
                 static NeonGenTwoOpFn * const fns[3][2] = {
                     { gen_helper_neon_pmin_s8, gen_helper_neon_pmin_u8 },
                     { gen_helper_neon_pmin_s16, gen_helper_neon_pmin_u16 },
-                    { gen_min_s32, gen_min_u32 },
+                    { tcg_gen_smin_i32, tcg_gen_umin_i32 },
                 };
                 genfn = fns[size][u];
                 break;
@@ -10512,7 +10758,7 @@ static void disas_simd_3same_int(DisasContext *s, uint32_t insn)
                 static NeonGenTwoOpFn * const fns[3][2] = {
                     { gen_helper_neon_max_s8, gen_helper_neon_max_u8 },
                     { gen_helper_neon_max_s16, gen_helper_neon_max_u16 },
-                    { gen_max_s32, gen_max_u32 },
+                    { tcg_gen_smax_i32, tcg_gen_umax_i32 },
                 };
                 genfn = fns[size][u];
                 break;
@@ -10523,7 +10769,7 @@ static void disas_simd_3same_int(DisasContext *s, uint32_t insn)
                 static NeonGenTwoOpFn * const fns[3][2] = {
                     { gen_helper_neon_min_s8, gen_helper_neon_min_u8 },
                     { gen_helper_neon_min_s16, gen_helper_neon_min_u16 },
-                    { gen_min_s32, gen_min_u32 },
+                    { tcg_gen_smin_i32, tcg_gen_umin_i32 },
                 };
                 genfn = fns[size][u];
                 break;
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index 1788668..ee2bbc5 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -716,7 +716,6 @@ static void gen_atomic(DisasContext *ctx, uint32_t opc,
     TCGv src1, src2, dat;
     TCGLabel *l1, *l2;
     TCGMemOp mop;
-    TCGCond cond;
     bool aq, rl;
 
     /* Extract the size of the atomic operation.  */
@@ -814,60 +813,29 @@ static void gen_atomic(DisasContext *ctx, uint32_t opc,
         tcg_gen_atomic_fetch_or_tl(src2, src1, src2, ctx->mem_idx, mop);
         gen_set_gpr(rd, src2);
         break;
-
     case OPC_RISC_AMOMIN:
-        cond = TCG_COND_LT;
-        goto do_minmax;
+        gen_get_gpr(src1, rs1);
+        gen_get_gpr(src2, rs2);
+        tcg_gen_atomic_fetch_smin_tl(src2, src1, src2, ctx->mem_idx, mop);
+        gen_set_gpr(rd, src2);
+        break;
     case OPC_RISC_AMOMAX:
-        cond = TCG_COND_GT;
-        goto do_minmax;
+        gen_get_gpr(src1, rs1);
+        gen_get_gpr(src2, rs2);
+        tcg_gen_atomic_fetch_smax_tl(src2, src1, src2, ctx->mem_idx, mop);
+        gen_set_gpr(rd, src2);
+        break;
     case OPC_RISC_AMOMINU:
-        cond = TCG_COND_LTU;
-        goto do_minmax;
+        gen_get_gpr(src1, rs1);
+        gen_get_gpr(src2, rs2);
+        tcg_gen_atomic_fetch_umin_tl(src2, src1, src2, ctx->mem_idx, mop);
+        gen_set_gpr(rd, src2);
+        break;
     case OPC_RISC_AMOMAXU:
-        cond = TCG_COND_GTU;
-        goto do_minmax;
-    do_minmax:
-        /* Handle the RL barrier.  The AQ barrier is handled along the
-           parallel path by the SC atomic cmpxchg.  On the serial path,
-           of course, barriers do not matter.  */
-        if (rl) {
-            tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
-        }
-        if (tb_cflags(ctx->base.tb) & CF_PARALLEL) {
-            l1 = gen_new_label();
-            gen_set_label(l1);
-        } else {
-            l1 = NULL;
-        }
-
         gen_get_gpr(src1, rs1);
         gen_get_gpr(src2, rs2);
-        if ((mop & MO_SSIZE) == MO_SL) {
-            /* Sign-extend the register comparison input.  */
-            tcg_gen_ext32s_tl(src2, src2);
-        }
-        dat = tcg_temp_local_new();
-        tcg_gen_qemu_ld_tl(dat, src1, ctx->mem_idx, mop);
-        tcg_gen_movcond_tl(cond, src2, dat, src2, dat, src2);
-
-        if (tb_cflags(ctx->base.tb) & CF_PARALLEL) {
-            /* Parallel context.  Make this operation atomic by verifying
-               that the memory didn't change while we computed the result.  */
-            tcg_gen_atomic_cmpxchg_tl(src2, src1, dat, src2, ctx->mem_idx, mop);
-
-            /* If the cmpxchg failed, retry. */
-            /* ??? There is an assumption here that this will eventually
-               succeed, such that we don't live-lock.  This is not unlike
-               a similar loop that the compiler would generate for e.g.
-               __atomic_fetch_and_xor, so don't worry about it.  */
-            tcg_gen_brcond_tl(TCG_COND_NE, dat, src2, l1);
-        } else {
-            /* Serial context.  Directly store the result.  */
-            tcg_gen_qemu_st_tl(src2, src1, ctx->mem_idx, mop);
-        }
-        gen_set_gpr(rd, dat);
-        tcg_temp_free(dat);
+        tcg_gen_atomic_fetch_umax_tl(src2, src1, src2, ctx->mem_idx, mop);
+        gen_set_gpr(rd, src2);
         break;
 
     default:
diff --git a/target/xtensa/translate.c b/target/xtensa/translate.c
index aad4963..ae0feb0 100644
--- a/target/xtensa/translate.c
+++ b/target/xtensa/translate.c
@@ -1526,10 +1526,8 @@ static void translate_clamps(DisasContext *dc, const uint32_t arg[],
         TCGv_i32 tmp1 = tcg_const_i32(-1u << arg[2]);
         TCGv_i32 tmp2 = tcg_const_i32((1 << arg[2]) - 1);
 
-        tcg_gen_movcond_i32(TCG_COND_GT, tmp1,
-                            cpu_R[arg[1]], tmp1, cpu_R[arg[1]], tmp1);
-        tcg_gen_movcond_i32(TCG_COND_LT, cpu_R[arg[0]],
-                            tmp1, tmp2, tmp1, tmp2);
+        tcg_gen_smax_i32(tmp1, tmp1, cpu_R[arg[1]]);
+        tcg_gen_smin_i32(cpu_R[arg[0]], tmp1, tmp2);
         tcg_temp_free(tmp1);
         tcg_temp_free(tmp2);
     }
@@ -1854,13 +1852,35 @@ static void translate_memw(DisasContext *dc, const uint32_t arg[],
     tcg_gen_mb(TCG_BAR_SC | TCG_MO_ALL);
 }
 
-static void translate_minmax(DisasContext *dc, const uint32_t arg[],
-                             const uint32_t par[])
+static void translate_smin(DisasContext *dc, const uint32_t arg[],
+                           const uint32_t par[])
 {
     if (gen_window_check3(dc, arg[0], arg[1], arg[2])) {
-        tcg_gen_movcond_i32(par[0], cpu_R[arg[0]],
-                            cpu_R[arg[1]], cpu_R[arg[2]],
-                            cpu_R[arg[1]], cpu_R[arg[2]]);
+        tcg_gen_smin_i32(cpu_R[arg[0]], cpu_R[arg[1]], cpu_R[arg[2]]);
+    }
+}
+
+static void translate_umin(DisasContext *dc, const uint32_t arg[],
+                           const uint32_t par[])
+{
+    if (gen_window_check3(dc, arg[0], arg[1], arg[2])) {
+        tcg_gen_umin_i32(cpu_R[arg[0]], cpu_R[arg[1]], cpu_R[arg[2]]);
+    }
+}
+
+static void translate_smax(DisasContext *dc, const uint32_t arg[],
+                           const uint32_t par[])
+{
+    if (gen_window_check3(dc, arg[0], arg[1], arg[2])) {
+        tcg_gen_smax_i32(cpu_R[arg[0]], cpu_R[arg[1]], cpu_R[arg[2]]);
+    }
+}
+
+static void translate_umax(DisasContext *dc, const uint32_t arg[],
+                           const uint32_t par[])
+{
+    if (gen_window_check3(dc, arg[0], arg[1], arg[2])) {
+        tcg_gen_umax_i32(cpu_R[arg[0]], cpu_R[arg[1]], cpu_R[arg[2]]);
     }
 }
 
@@ -2983,23 +3003,19 @@ static const XtensaOpcodeOps core_ops[] = {
         .par = (const uint32_t[]){TCG_COND_NE},
     }, {
         .name = "max",
-        .translate = translate_minmax,
-        .par = (const uint32_t[]){TCG_COND_GE},
+        .translate = translate_smax,
     }, {
         .name = "maxu",
-        .translate = translate_minmax,
-        .par = (const uint32_t[]){TCG_COND_GEU},
+        .translate = translate_umax,
     }, {
         .name = "memw",
         .translate = translate_memw,
     }, {
         .name = "min",
-        .translate = translate_minmax,
-        .par = (const uint32_t[]){TCG_COND_LT},
+        .translate = translate_smin,
     }, {
         .name = "minu",
-        .translate = translate_minmax,
-        .par = (const uint32_t[]){TCG_COND_LTU},
+        .translate = translate_umin,
     }, {
         .name = "mov",
         .translate = translate_mov,