xref: /qemu/target/arm/tcg/mve_helper.c (revision d54deb2a)

/*
 * M-profile MVE Operations
 *
 * Copyright (c) 2021 Linaro, Ltd.
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
 */

#include "qemu/osdep.h"
#include "cpu.h"
#include "internals.h"
#include "vec_internal.h"
#include "exec/helper-proto.h"
#include "exec/cpu_ldst.h"
#include "exec/exec-all.h"
#include "tcg/tcg.h"
#include "fpu/softfloat.h"
#include "crypto/clmul.h"

static uint16_t mve_eci_mask(CPUARMState *env)
{
    /*
     * Return the mask of which elements in the MVE vector correspond
     * to beats being executed. The mask has 1 bits for executed lanes
     * and 0 bits where ECI says this beat was already executed.
     */
    int eci;

    if ((env->condexec_bits & 0xf) != 0) {
        return 0xffff;
    }

    eci = env->condexec_bits >> 4;
    switch (eci) {
    case ECI_NONE:
        return 0xffff;
    case ECI_A0:
        return 0xfff0;
    case ECI_A0A1:
        return 0xff00;
    case ECI_A0A1A2:
    case ECI_A0A1A2B0:
        return 0xf000;
    default:
        g_assert_not_reached();
    }
}

static uint16_t mve_element_mask(CPUARMState *env)
{
    /*
     * Return the mask of which elements in the MVE vector should be
     * updated. This is a combination of multiple things:
     *  (1) by default, we update every lane in the vector
     *  (2) VPT predication stores its state in the VPR register;
     *  (3) low-overhead-branch tail predication will mask out part
     *      the vector on the final iteration of the loop
     *  (4) if EPSR.ECI is set then we must execute only some beats
     *      of the insn
     * We combine all these into a 16-bit result with the same semantics
     * as VPR.P0: 0 to mask the lane, 1 if it is active.
     * 8-bit vector ops will look at all bits of the result;
     * 16-bit ops will look at bits 0, 2, 4, ...;
     * 32-bit ops will look at bits 0, 4, 8 and 12.
     * Compare pseudocode GetCurInstrBeat(), though that only returns
     * the 4-bit slice of the mask corresponding to a single beat.
     */
    uint16_t mask = FIELD_EX32(env->v7m.vpr, V7M_VPR, P0);

    if (!(env->v7m.vpr & R_V7M_VPR_MASK01_MASK)) {
        mask |= 0xff;
    }
    if (!(env->v7m.vpr & R_V7M_VPR_MASK23_MASK)) {
        mask |= 0xff00;
    }

    if (env->v7m.ltpsize < 4 &&
        env->regs[14] <= (1 << (4 - env->v7m.ltpsize))) {
        /*
         * Tail predication active, and this is the last loop iteration.
         * The element size is (1 << ltpsize), and we only want to process
         * loopcount elements, so we want to retain the least significant
         * (loopcount * esize) predicate bits and zero out bits above that.
         */
        int masklen = env->regs[14] << env->v7m.ltpsize;
        assert(masklen <= 16);
        uint16_t ltpmask = masklen ? MAKE_64BIT_MASK(0, masklen) : 0;
        mask &= ltpmask;
    }

    /*
     * ECI bits indicate which beats are already executed;
     * we handle this by effectively predicating them out.
     */
    mask &= mve_eci_mask(env);
    return mask;
}

static void mve_advance_vpt(CPUARMState *env)
{
    /* Advance the VPT and ECI state if necessary */
    uint32_t vpr = env->v7m.vpr;
    unsigned mask01, mask23;
    uint16_t inv_mask;
    uint16_t eci_mask = mve_eci_mask(env);

    if ((env->condexec_bits & 0xf) == 0) {
        env->condexec_bits = (env->condexec_bits == (ECI_A0A1A2B0 << 4)) ?
            (ECI_A0 << 4) : (ECI_NONE << 4);
    }

    if (!(vpr & (R_V7M_VPR_MASK01_MASK | R_V7M_VPR_MASK23_MASK))) {
        /* VPT not enabled, nothing to do */
        return;
    }

    /* Invert P0 bits if needed, but only for beats we actually executed */
    mask01 = FIELD_EX32(vpr, V7M_VPR, MASK01);
    mask23 = FIELD_EX32(vpr, V7M_VPR, MASK23);
    /* Start by assuming we invert all bits corresponding to executed beats */
    inv_mask = eci_mask;
    if (mask01 <= 8) {
        /* MASK01 says don't invert low half of P0 */
        inv_mask &= ~0xff;
    }
    if (mask23 <= 8) {
        /* MASK23 says don't invert high half of P0 */
        inv_mask &= ~0xff00;
    }
    vpr ^= inv_mask;
    /* Only update MASK01 if beat 1 executed */
    if (eci_mask & 0xf0) {
        vpr = FIELD_DP32(vpr, V7M_VPR, MASK01, mask01 << 1);
    }
    /* Beat 3 always executes, so update MASK23 */
    vpr = FIELD_DP32(vpr, V7M_VPR, MASK23, mask23 << 1);
    env->v7m.vpr = vpr;
}

/* For loads, predicated lanes are zeroed instead of keeping their old values */
#define DO_VLDR(OP, MSIZE, LDTYPE, ESIZE, TYPE)                         \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, uint32_t addr)    \
    {                                                                   \
        TYPE *d = vd;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        unsigned b, e;                                                  \
        /*                                                              \
         * R_SXTM allows the dest reg to become UNKNOWN for abandoned   \
         * beats so we don't care if we update part of the dest and     \
         * then take an exception.                                      \
         */                                                             \
        for (b = 0, e = 0; b < 16; b += ESIZE, e++) {                   \
            if (eci_mask & (1 << b)) {                                  \
                d[H##ESIZE(e)] = (mask & (1 << b)) ?                    \
                    cpu_##LDTYPE##_data_ra(env, addr, GETPC()) : 0;     \
            }                                                           \
            addr += MSIZE;                                              \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_VSTR(OP, MSIZE, STTYPE, ESIZE, TYPE)                         \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, uint32_t addr)    \
    {                                                                   \
        TYPE *d = vd;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned b, e;                                                  \
        for (b = 0, e = 0; b < 16; b += ESIZE, e++) {                   \
            if (mask & (1 << b)) {                                      \
                cpu_##STTYPE##_data_ra(env, addr, d[H##ESIZE(e)], GETPC()); \
            }                                                           \
            addr += MSIZE;                                              \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_VLDR(vldrb, 1, ldub, 1, uint8_t)
DO_VLDR(vldrh, 2, lduw, 2, uint16_t)
DO_VLDR(vldrw, 4, ldl, 4, uint32_t)

DO_VSTR(vstrb, 1, stb, 1, uint8_t)
DO_VSTR(vstrh, 2, stw, 2, uint16_t)
DO_VSTR(vstrw, 4, stl, 4, uint32_t)

DO_VLDR(vldrb_sh, 1, ldsb, 2, int16_t)
DO_VLDR(vldrb_sw, 1, ldsb, 4, int32_t)
DO_VLDR(vldrb_uh, 1, ldub, 2, uint16_t)
DO_VLDR(vldrb_uw, 1, ldub, 4, uint32_t)
DO_VLDR(vldrh_sw, 2, ldsw, 4, int32_t)
DO_VLDR(vldrh_uw, 2, lduw, 4, uint32_t)

DO_VSTR(vstrb_h, 1, stb, 2, int16_t)
DO_VSTR(vstrb_w, 1, stb, 4, int32_t)
DO_VSTR(vstrh_w, 2, stw, 4, int32_t)

#undef DO_VLDR
#undef DO_VSTR

/*
 * Gather loads/scatter stores. Here each element of Qm specifies
 * an offset to use from the base register Rm. In the _os_ versions
 * that offset is scaled by the element size.
 * For loads, predicated lanes are zeroed instead of retaining
 * their previous values.
 */
#define DO_VLDR_SG(OP, LDTYPE, ESIZE, TYPE, OFFTYPE, ADDRFN, WB)        \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm,         \
                          uint32_t base)                                \
    {                                                                   \
        TYPE *d = vd;                                                   \
        OFFTYPE *m = vm;                                                \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        unsigned e;                                                     \
        uint32_t addr;                                                  \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE, eci_mask >>= ESIZE) { \
            if (!(eci_mask & 1)) {                                      \
                continue;                                               \
            }                                                           \
            addr = ADDRFN(base, m[H##ESIZE(e)]);                        \
            d[H##ESIZE(e)] = (mask & 1) ?                               \
                cpu_##LDTYPE##_data_ra(env, addr, GETPC()) : 0;         \
            if (WB) {                                                   \
                m[H##ESIZE(e)] = addr;                                  \
            }                                                           \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

/* We know here TYPE is unsigned so always the same as the offset type */
#define DO_VSTR_SG(OP, STTYPE, ESIZE, TYPE, ADDRFN, WB)                 \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm,         \
                          uint32_t base)                                \
    {                                                                   \
        TYPE *d = vd;                                                   \
        TYPE *m = vm;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        unsigned e;                                                     \
        uint32_t addr;                                                  \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE, eci_mask >>= ESIZE) { \
            if (!(eci_mask & 1)) {                                      \
                continue;                                               \
            }                                                           \
            addr = ADDRFN(base, m[H##ESIZE(e)]);                        \
            if (mask & 1) {                                             \
                cpu_##STTYPE##_data_ra(env, addr, d[H##ESIZE(e)], GETPC()); \
            }                                                           \
            if (WB) {                                                   \
                m[H##ESIZE(e)] = addr;                                  \
            }                                                           \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

/*
 * 64-bit accesses are slightly different: they are done as two 32-bit
 * accesses, controlled by the predicate mask for the relevant beat,
 * and with a single 32-bit offset in the first of the two Qm elements.
 * Note that for QEMU our IMPDEF AIRCR.ENDIANNESS is always 0 (little).
 * Address writeback happens on the odd beats and updates the address
 * stored in the even-beat element.
 */
#define DO_VLDR64_SG(OP, ADDRFN, WB)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm,         \
                          uint32_t base)                                \
    {                                                                   \
        uint32_t *d = vd;                                               \
        uint32_t *m = vm;                                               \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        unsigned e;                                                     \
        uint32_t addr;                                                  \
        for (e = 0; e < 16 / 4; e++, mask >>= 4, eci_mask >>= 4) {      \
            if (!(eci_mask & 1)) {                                      \
                continue;                                               \
            }                                                           \
            addr = ADDRFN(base, m[H4(e & ~1)]);                         \
            addr += 4 * (e & 1);                                        \
            d[H4(e)] = (mask & 1) ? cpu_ldl_data_ra(env, addr, GETPC()) : 0; \
            if (WB && (e & 1)) {                                        \
                m[H4(e & ~1)] = addr - 4;                               \
            }                                                           \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_VSTR64_SG(OP, ADDRFN, WB)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm,         \
                          uint32_t base)                                \
    {                                                                   \
        uint32_t *d = vd;                                               \
        uint32_t *m = vm;                                               \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        unsigned e;                                                     \
        uint32_t addr;                                                  \
        for (e = 0; e < 16 / 4; e++, mask >>= 4, eci_mask >>= 4) {      \
            if (!(eci_mask & 1)) {                                      \
                continue;                                               \
            }                                                           \
            addr = ADDRFN(base, m[H4(e & ~1)]);                         \
            addr += 4 * (e & 1);                                        \
            if (mask & 1) {                                             \
                cpu_stl_data_ra(env, addr, d[H4(e)], GETPC());          \
            }                                                           \
            if (WB && (e & 1)) {                                        \
                m[H4(e & ~1)] = addr - 4;                               \
            }                                                           \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define ADDR_ADD(BASE, OFFSET) ((BASE) + (OFFSET))
#define ADDR_ADD_OSH(BASE, OFFSET) ((BASE) + ((OFFSET) << 1))
#define ADDR_ADD_OSW(BASE, OFFSET) ((BASE) + ((OFFSET) << 2))
#define ADDR_ADD_OSD(BASE, OFFSET) ((BASE) + ((OFFSET) << 3))

DO_VLDR_SG(vldrb_sg_sh, ldsb, 2, int16_t, uint16_t, ADDR_ADD, false)
DO_VLDR_SG(vldrb_sg_sw, ldsb, 4, int32_t, uint32_t, ADDR_ADD, false)
DO_VLDR_SG(vldrh_sg_sw, ldsw, 4, int32_t, uint32_t, ADDR_ADD, false)

DO_VLDR_SG(vldrb_sg_ub, ldub, 1, uint8_t, uint8_t, ADDR_ADD, false)
DO_VLDR_SG(vldrb_sg_uh, ldub, 2, uint16_t, uint16_t, ADDR_ADD, false)
DO_VLDR_SG(vldrb_sg_uw, ldub, 4, uint32_t, uint32_t, ADDR_ADD, false)
DO_VLDR_SG(vldrh_sg_uh, lduw, 2, uint16_t, uint16_t, ADDR_ADD, false)
DO_VLDR_SG(vldrh_sg_uw, lduw, 4, uint32_t, uint32_t, ADDR_ADD, false)
DO_VLDR_SG(vldrw_sg_uw, ldl, 4, uint32_t, uint32_t, ADDR_ADD, false)
DO_VLDR64_SG(vldrd_sg_ud, ADDR_ADD, false)

DO_VLDR_SG(vldrh_sg_os_sw, ldsw, 4, int32_t, uint32_t, ADDR_ADD_OSH, false)
DO_VLDR_SG(vldrh_sg_os_uh, lduw, 2, uint16_t, uint16_t, ADDR_ADD_OSH, false)
DO_VLDR_SG(vldrh_sg_os_uw, lduw, 4, uint32_t, uint32_t, ADDR_ADD_OSH, false)
DO_VLDR_SG(vldrw_sg_os_uw, ldl, 4, uint32_t, uint32_t, ADDR_ADD_OSW, false)
DO_VLDR64_SG(vldrd_sg_os_ud, ADDR_ADD_OSD, false)

DO_VSTR_SG(vstrb_sg_ub, stb, 1, uint8_t, ADDR_ADD, false)
DO_VSTR_SG(vstrb_sg_uh, stb, 2, uint16_t, ADDR_ADD, false)
DO_VSTR_SG(vstrb_sg_uw, stb, 4, uint32_t, ADDR_ADD, false)
DO_VSTR_SG(vstrh_sg_uh, stw, 2, uint16_t, ADDR_ADD, false)
DO_VSTR_SG(vstrh_sg_uw, stw, 4, uint32_t, ADDR_ADD, false)
DO_VSTR_SG(vstrw_sg_uw, stl, 4, uint32_t, ADDR_ADD, false)
DO_VSTR64_SG(vstrd_sg_ud, ADDR_ADD, false)

DO_VSTR_SG(vstrh_sg_os_uh, stw, 2, uint16_t, ADDR_ADD_OSH, false)
DO_VSTR_SG(vstrh_sg_os_uw, stw, 4, uint32_t, ADDR_ADD_OSH, false)
DO_VSTR_SG(vstrw_sg_os_uw, stl, 4, uint32_t, ADDR_ADD_OSW, false)
DO_VSTR64_SG(vstrd_sg_os_ud, ADDR_ADD_OSD, false)

DO_VLDR_SG(vldrw_sg_wb_uw, ldl, 4, uint32_t, uint32_t, ADDR_ADD, true)
DO_VLDR64_SG(vldrd_sg_wb_ud, ADDR_ADD, true)
DO_VSTR_SG(vstrw_sg_wb_uw, stl, 4, uint32_t, ADDR_ADD, true)
DO_VSTR64_SG(vstrd_sg_wb_ud, ADDR_ADD, true)

/*
 * Deinterleaving loads/interleaving stores.
 *
 * For these helpers we are passed the index of the first Qreg
 * (VLD2/VST2 will also access Qn+1, VLD4/VST4 access Qn .. Qn+3)
 * and the value of the base address register Rn.
 * The helpers are specialized for pattern and element size, so
 * for instance vld42h is VLD4 with pattern 2, element size MO_16.
 *
 * These insns are beatwise but not predicated, so we must honour ECI,
 * but need not look at mve_element_mask().
 *
 * The pseudocode implements these insns with multiple memory accesses
 * of the element size, but rules R_VVVG and R_FXDM permit us to make
 * one 32-bit memory access per beat.
 */
#define DO_VLD4B(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat, e;                                                    \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 4;                                \
            data = cpu_ldl_le_data_ra(env, addr, GETPC());              \
            for (e = 0; e < 4; e++, data >>= 8) {                       \
                uint8_t *qd = (uint8_t *)aa32_vfp_qreg(env, qnidx + e); \
                qd[H1(off[beat])] = data;                               \
            }                                                           \
        }                                                               \
    }

#define DO_VLD4H(OP, O1, O2)                                            \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O1, O2, O2 };               \
        uint32_t addr, data;                                            \
        int y; /* y counts 0 2 0 2 */                                   \
        uint16_t *qd;                                                   \
        for (beat = 0, y = 0; beat < 4; beat++, mask >>= 4, y ^= 2) {   \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 8 + (beat & 1) * 4;               \
            data = cpu_ldl_le_data_ra(env, addr, GETPC());              \
            qd = (uint16_t *)aa32_vfp_qreg(env, qnidx + y);             \
            qd[H2(off[beat])] = data;                                   \
            data >>= 16;                                                \
            qd = (uint16_t *)aa32_vfp_qreg(env, qnidx + y + 1);         \
            qd[H2(off[beat])] = data;                                   \
        }                                                               \
    }

#define DO_VLD4W(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        uint32_t *qd;                                                   \
        int y;                                                          \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 4;                                \
            data = cpu_ldl_le_data_ra(env, addr, GETPC());              \
            y = (beat + (O1 & 2)) & 3;                                  \
            qd = (uint32_t *)aa32_vfp_qreg(env, qnidx + y);             \
            qd[H4(off[beat] >> 2)] = data;                              \
        }                                                               \
    }

DO_VLD4B(vld40b, 0, 1, 10, 11)
DO_VLD4B(vld41b, 2, 3, 12, 13)
DO_VLD4B(vld42b, 4, 5, 14, 15)
DO_VLD4B(vld43b, 6, 7, 8, 9)

DO_VLD4H(vld40h, 0, 5)
DO_VLD4H(vld41h, 1, 6)
DO_VLD4H(vld42h, 2, 7)
DO_VLD4H(vld43h, 3, 4)

DO_VLD4W(vld40w, 0, 1, 10, 11)
DO_VLD4W(vld41w, 2, 3, 12, 13)
DO_VLD4W(vld42w, 4, 5, 14, 15)
DO_VLD4W(vld43w, 6, 7, 8, 9)

#define DO_VLD2B(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat, e;                                                    \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        uint8_t *qd;                                                    \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 2;                                \
            data = cpu_ldl_le_data_ra(env, addr, GETPC());              \
            for (e = 0; e < 4; e++, data >>= 8) {                       \
                qd = (uint8_t *)aa32_vfp_qreg(env, qnidx + (e & 1));    \
                qd[H1(off[beat] + (e >> 1))] = data;                    \
            }                                                           \
        }                                                               \
    }

#define DO_VLD2H(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        int e;                                                          \
        uint16_t *qd;                                                   \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 4;                                \
            data = cpu_ldl_le_data_ra(env, addr, GETPC());              \
            for (e = 0; e < 2; e++, data >>= 16) {                      \
                qd = (uint16_t *)aa32_vfp_qreg(env, qnidx + e);         \
                qd[H2(off[beat])] = data;                               \
            }                                                           \
        }                                                               \
    }

#define DO_VLD2W(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        uint32_t *qd;                                                   \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat];                                    \
            data = cpu_ldl_le_data_ra(env, addr, GETPC());              \
            qd = (uint32_t *)aa32_vfp_qreg(env, qnidx + (beat & 1));    \
            qd[H4(off[beat] >> 3)] = data;                              \
        }                                                               \
    }

DO_VLD2B(vld20b, 0, 2, 12, 14)
DO_VLD2B(vld21b, 4, 6, 8, 10)

DO_VLD2H(vld20h, 0, 1, 6, 7)
DO_VLD2H(vld21h, 2, 3, 4, 5)

DO_VLD2W(vld20w, 0, 4, 24, 28)
DO_VLD2W(vld21w, 8, 12, 16, 20)

#define DO_VST4B(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat, e;                                                    \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 4;                                \
            data = 0;                                                   \
            for (e = 3; e >= 0; e--) {                                  \
                uint8_t *qd = (uint8_t *)aa32_vfp_qreg(env, qnidx + e); \
                data = (data << 8) | qd[H1(off[beat])];                 \
            }                                                           \
            cpu_stl_le_data_ra(env, addr, data, GETPC());               \
        }                                                               \
    }

#define DO_VST4H(OP, O1, O2)                                            \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O1, O2, O2 };               \
        uint32_t addr, data;                                            \
        int y; /* y counts 0 2 0 2 */                                   \
        uint16_t *qd;                                                   \
        for (beat = 0, y = 0; beat < 4; beat++, mask >>= 4, y ^= 2) {   \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 8 + (beat & 1) * 4;               \
            qd = (uint16_t *)aa32_vfp_qreg(env, qnidx + y);             \
            data = qd[H2(off[beat])];                                   \
            qd = (uint16_t *)aa32_vfp_qreg(env, qnidx + y + 1);         \
            data |= qd[H2(off[beat])] << 16;                            \
            cpu_stl_le_data_ra(env, addr, data, GETPC());               \
        }                                                               \
    }

#define DO_VST4W(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        uint32_t *qd;                                                   \
        int y;                                                          \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 4;                                \
            y = (beat + (O1 & 2)) & 3;                                  \
            qd = (uint32_t *)aa32_vfp_qreg(env, qnidx + y);             \
            data = qd[H4(off[beat] >> 2)];                              \
            cpu_stl_le_data_ra(env, addr, data, GETPC());               \
        }                                                               \
    }

DO_VST4B(vst40b, 0, 1, 10, 11)
DO_VST4B(vst41b, 2, 3, 12, 13)
DO_VST4B(vst42b, 4, 5, 14, 15)
DO_VST4B(vst43b, 6, 7, 8, 9)

DO_VST4H(vst40h, 0, 5)
DO_VST4H(vst41h, 1, 6)
DO_VST4H(vst42h, 2, 7)
DO_VST4H(vst43h, 3, 4)

DO_VST4W(vst40w, 0, 1, 10, 11)
DO_VST4W(vst41w, 2, 3, 12, 13)
DO_VST4W(vst42w, 4, 5, 14, 15)
DO_VST4W(vst43w, 6, 7, 8, 9)

#define DO_VST2B(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat, e;                                                    \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        uint8_t *qd;                                                    \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 2;                                \
            data = 0;                                                   \
            for (e = 3; e >= 0; e--) {                                  \
                qd = (uint8_t *)aa32_vfp_qreg(env, qnidx + (e & 1));    \
                data = (data << 8) | qd[H1(off[beat] + (e >> 1))];      \
            }                                                           \
            cpu_stl_le_data_ra(env, addr, data, GETPC());               \
        }                                                               \
    }

#define DO_VST2H(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        int e;                                                          \
        uint16_t *qd;                                                   \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat] * 4;                                \
            data = 0;                                                   \
            for (e = 1; e >= 0; e--) {                                  \
                qd = (uint16_t *)aa32_vfp_qreg(env, qnidx + e);         \
                data = (data << 16) | qd[H2(off[beat])];                \
            }                                                           \
            cpu_stl_le_data_ra(env, addr, data, GETPC());               \
        }                                                               \
    }

#define DO_VST2W(OP, O1, O2, O3, O4)                                    \
    void HELPER(mve_##OP)(CPUARMState *env, uint32_t qnidx,             \
                          uint32_t base)                                \
    {                                                                   \
        int beat;                                                       \
        uint16_t mask = mve_eci_mask(env);                              \
        static const uint8_t off[4] = { O1, O2, O3, O4 };               \
        uint32_t addr, data;                                            \
        uint32_t *qd;                                                   \
        for (beat = 0; beat < 4; beat++, mask >>= 4) {                  \
            if ((mask & 1) == 0) {                                      \
                /* ECI says skip this beat */                           \
                continue;                                               \
            }                                                           \
            addr = base + off[beat];                                    \
            qd = (uint32_t *)aa32_vfp_qreg(env, qnidx + (beat & 1));    \
            data = qd[H4(off[beat] >> 3)];                              \
            cpu_stl_le_data_ra(env, addr, data, GETPC());               \
        }                                                               \
    }

DO_VST2B(vst20b, 0, 2, 12, 14)
DO_VST2B(vst21b, 4, 6, 8, 10)

DO_VST2H(vst20h, 0, 1, 6, 7)
DO_VST2H(vst21h, 2, 3, 4, 5)

DO_VST2W(vst20w, 0, 4, 24, 28)
DO_VST2W(vst21w, 8, 12, 16, 20)

/*
 * The mergemask(D, R, M) macro performs the operation "*D = R" but
 * storing only the bytes which correspond to 1 bits in M,
 * leaving other bytes in *D unchanged. We use _Generic
 * to select the correct implementation based on the type of D.
 */

static void mergemask_ub(uint8_t *d, uint8_t r, uint16_t mask)
{
    if (mask & 1) {
        *d = r;
    }
}

static void mergemask_sb(int8_t *d, int8_t r, uint16_t mask)
{
    mergemask_ub((uint8_t *)d, r, mask);
}

static void mergemask_uh(uint16_t *d, uint16_t r, uint16_t mask)
{
    uint16_t bmask = expand_pred_b(mask);
    *d = (*d & ~bmask) | (r & bmask);
}

static void mergemask_sh(int16_t *d, int16_t r, uint16_t mask)
{
    mergemask_uh((uint16_t *)d, r, mask);
}

static void mergemask_uw(uint32_t *d, uint32_t r, uint16_t mask)
{
    uint32_t bmask = expand_pred_b(mask);
    *d = (*d & ~bmask) | (r & bmask);
}

static void mergemask_sw(int32_t *d, int32_t r, uint16_t mask)
{
    mergemask_uw((uint32_t *)d, r, mask);
}

static void mergemask_uq(uint64_t *d, uint64_t r, uint16_t mask)
{
    uint64_t bmask = expand_pred_b(mask);
    *d = (*d & ~bmask) | (r & bmask);
}

static void mergemask_sq(int64_t *d, int64_t r, uint16_t mask)
{
    mergemask_uq((uint64_t *)d, r, mask);
}

#define mergemask(D, R, M)                      \
    _Generic(D,                                 \
             uint8_t *: mergemask_ub,           \
             int8_t *:  mergemask_sb,           \
             uint16_t *: mergemask_uh,          \
             int16_t *:  mergemask_sh,          \
             uint32_t *: mergemask_uw,          \
             int32_t *:  mergemask_sw,          \
             uint64_t *: mergemask_uq,          \
             int64_t *:  mergemask_sq)(D, R, M)

void HELPER(mve_vdup)(CPUARMState *env, void *vd, uint32_t val)
{
    /*
     * The generated code already replicated an 8 or 16 bit constant
     * into the 32-bit value, so we only need to write the 32-bit
     * value to all elements of the Qreg, allowing for predication.
     */
    uint32_t *d = vd;
    uint16_t mask = mve_element_mask(env);
    unsigned e;
    for (e = 0; e < 16 / 4; e++, mask >>= 4) {
        mergemask(&d[H4(e)], val, mask);
    }
    mve_advance_vpt(env);
}

#define DO_1OP(OP, ESIZE, TYPE, FN)                                     \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm)         \
    {                                                                   \
        TYPE *d = vd, *m = vm;                                          \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            mergemask(&d[H##ESIZE(e)], FN(m[H##ESIZE(e)]), mask);       \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_CLS_B(N)   (clrsb32(N) - 24)
#define DO_CLS_H(N)   (clrsb32(N) - 16)

DO_1OP(vclsb, 1, int8_t, DO_CLS_B)
DO_1OP(vclsh, 2, int16_t, DO_CLS_H)
DO_1OP(vclsw, 4, int32_t, clrsb32)

#define DO_CLZ_B(N)   (clz32(N) - 24)
#define DO_CLZ_H(N)   (clz32(N) - 16)

DO_1OP(vclzb, 1, uint8_t, DO_CLZ_B)
DO_1OP(vclzh, 2, uint16_t, DO_CLZ_H)
DO_1OP(vclzw, 4, uint32_t, clz32)

DO_1OP(vrev16b, 2, uint16_t, bswap16)
DO_1OP(vrev32b, 4, uint32_t, bswap32)
DO_1OP(vrev32h, 4, uint32_t, hswap32)
DO_1OP(vrev64b, 8, uint64_t, bswap64)
DO_1OP(vrev64h, 8, uint64_t, hswap64)
DO_1OP(vrev64w, 8, uint64_t, wswap64)

#define DO_NOT(N) (~(N))

DO_1OP(vmvn, 8, uint64_t, DO_NOT)

#define DO_ABS(N) ((N) < 0 ? -(N) : (N))
#define DO_FABSH(N)  ((N) & dup_const(MO_16, 0x7fff))
#define DO_FABSS(N)  ((N) & dup_const(MO_32, 0x7fffffff))

DO_1OP(vabsb, 1, int8_t, DO_ABS)
DO_1OP(vabsh, 2, int16_t, DO_ABS)
DO_1OP(vabsw, 4, int32_t, DO_ABS)

/* We can do these 64 bits at a time */
DO_1OP(vfabsh, 8, uint64_t, DO_FABSH)
DO_1OP(vfabss, 8, uint64_t, DO_FABSS)

#define DO_NEG(N)    (-(N))
#define DO_FNEGH(N) ((N) ^ dup_const(MO_16, 0x8000))
#define DO_FNEGS(N) ((N) ^ dup_const(MO_32, 0x80000000))

DO_1OP(vnegb, 1, int8_t, DO_NEG)
DO_1OP(vnegh, 2, int16_t, DO_NEG)
DO_1OP(vnegw, 4, int32_t, DO_NEG)

/* We can do these 64 bits at a time */
DO_1OP(vfnegh, 8, uint64_t, DO_FNEGH)
DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS)

/*
 * 1 operand immediates: Vda is destination and possibly also one source.
 * All these insns work at 64-bit widths.
 */
#define DO_1OP_IMM(OP, FN)                                              \
    void HELPER(mve_##OP)(CPUARMState *env, void *vda, uint64_t imm)    \
    {                                                                   \
        uint64_t *da = vda;                                             \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        for (e = 0; e < 16 / 8; e++, mask >>= 8) {                      \
            mergemask(&da[H8(e)], FN(da[H8(e)], imm), mask);            \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_MOVI(N, I) (I)
#define DO_ANDI(N, I) ((N) & (I))
#define DO_ORRI(N, I) ((N) | (I))

DO_1OP_IMM(vmovi, DO_MOVI)
DO_1OP_IMM(vandi, DO_ANDI)
DO_1OP_IMM(vorri, DO_ORRI)

#define DO_2OP(OP, ESIZE, TYPE, FN)                                     \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vn, void *vm)           \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            mergemask(&d[H##ESIZE(e)],                                  \
                      FN(n[H##ESIZE(e)], m[H##ESIZE(e)]), mask);        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

/* provide unsigned 2-op helpers for all sizes */
#define DO_2OP_U(OP, FN)                        \
    DO_2OP(OP##b, 1, uint8_t, FN)               \
    DO_2OP(OP##h, 2, uint16_t, FN)              \
    DO_2OP(OP##w, 4, uint32_t, FN)

/* provide signed 2-op helpers for all sizes */
#define DO_2OP_S(OP, FN)                        \
    DO_2OP(OP##b, 1, int8_t, FN)                \
    DO_2OP(OP##h, 2, int16_t, FN)               \
    DO_2OP(OP##w, 4, int32_t, FN)

/*
 * "Long" operations where two half-sized inputs (taken from either the
 * top or the bottom of the input vector) produce a double-width result.
 * Here ESIZE, TYPE are for the input, and LESIZE, LTYPE for the output.
 */
#define DO_2OP_L(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN)               \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, void *vm) \
    {                                                                   \
        LTYPE *d = vd;                                                  \
        TYPE *n = vn, *m = vm;                                          \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned le;                                                    \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) {         \
            LTYPE r = FN((LTYPE)n[H##ESIZE(le * 2 + TOP)],              \
                         m[H##ESIZE(le * 2 + TOP)]);                    \
            mergemask(&d[H##LESIZE(le)], r, mask);                      \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_2OP_SAT(OP, ESIZE, TYPE, FN)                                 \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, void *vm) \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        bool qc = false;                                                \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            bool sat = false;                                           \
            TYPE r_ = FN(n[H##ESIZE(e)], m[H##ESIZE(e)], &sat);         \
            mergemask(&d[H##ESIZE(e)], r_, mask);                       \
            qc |= sat & mask & 1;                                       \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

/* provide unsigned 2-op helpers for all sizes */
#define DO_2OP_SAT_U(OP, FN)                    \
    DO_2OP_SAT(OP##b, 1, uint8_t, FN)           \
    DO_2OP_SAT(OP##h, 2, uint16_t, FN)          \
    DO_2OP_SAT(OP##w, 4, uint32_t, FN)

/* provide signed 2-op helpers for all sizes */
#define DO_2OP_SAT_S(OP, FN)                    \
    DO_2OP_SAT(OP##b, 1, int8_t, FN)            \
    DO_2OP_SAT(OP##h, 2, int16_t, FN)           \
    DO_2OP_SAT(OP##w, 4, int32_t, FN)

#define DO_AND(N, M)  ((N) & (M))
#define DO_BIC(N, M)  ((N) & ~(M))
#define DO_ORR(N, M)  ((N) | (M))
#define DO_ORN(N, M)  ((N) | ~(M))
#define DO_EOR(N, M)  ((N) ^ (M))

DO_2OP(vand, 8, uint64_t, DO_AND)
DO_2OP(vbic, 8, uint64_t, DO_BIC)
DO_2OP(vorr, 8, uint64_t, DO_ORR)
DO_2OP(vorn, 8, uint64_t, DO_ORN)
DO_2OP(veor, 8, uint64_t, DO_EOR)

#define DO_ADD(N, M) ((N) + (M))
#define DO_SUB(N, M) ((N) - (M))
#define DO_MUL(N, M) ((N) * (M))

DO_2OP_U(vadd, DO_ADD)
DO_2OP_U(vsub, DO_SUB)
DO_2OP_U(vmul, DO_MUL)

DO_2OP_L(vmullbsb, 0, 1, int8_t, 2, int16_t, DO_MUL)
DO_2OP_L(vmullbsh, 0, 2, int16_t, 4, int32_t, DO_MUL)
DO_2OP_L(vmullbsw, 0, 4, int32_t, 8, int64_t, DO_MUL)
DO_2OP_L(vmullbub, 0, 1, uint8_t, 2, uint16_t, DO_MUL)
DO_2OP_L(vmullbuh, 0, 2, uint16_t, 4, uint32_t, DO_MUL)
DO_2OP_L(vmullbuw, 0, 4, uint32_t, 8, uint64_t, DO_MUL)

DO_2OP_L(vmulltsb, 1, 1, int8_t, 2, int16_t, DO_MUL)
DO_2OP_L(vmulltsh, 1, 2, int16_t, 4, int32_t, DO_MUL)
DO_2OP_L(vmulltsw, 1, 4, int32_t, 8, int64_t, DO_MUL)
DO_2OP_L(vmulltub, 1, 1, uint8_t, 2, uint16_t, DO_MUL)
DO_2OP_L(vmulltuh, 1, 2, uint16_t, 4, uint32_t, DO_MUL)
DO_2OP_L(vmulltuw, 1, 4, uint32_t, 8, uint64_t, DO_MUL)

/*
 * Polynomial multiply. We can always do this generating 64 bits
 * of the result at a time, so we don't need to use DO_2OP_L.
 */
DO_2OP(vmullpbh, 8, uint64_t, clmul_8x4_even)
DO_2OP(vmullpth, 8, uint64_t, clmul_8x4_odd)
DO_2OP(vmullpbw, 8, uint64_t, clmul_16x2_even)
DO_2OP(vmullptw, 8, uint64_t, clmul_16x2_odd)

/*
 * Because the computation type is at least twice as large as required,
 * these work for both signed and unsigned source types.
 */
static inline uint8_t do_mulh_b(int32_t n, int32_t m)
{
    return (n * m) >> 8;
}

static inline uint16_t do_mulh_h(int32_t n, int32_t m)
{
    return (n * m) >> 16;
}

static inline uint32_t do_mulh_w(int64_t n, int64_t m)
{
    return (n * m) >> 32;
}

static inline uint8_t do_rmulh_b(int32_t n, int32_t m)
{
    return (n * m + (1U << 7)) >> 8;
}

static inline uint16_t do_rmulh_h(int32_t n, int32_t m)
{
    return (n * m + (1U << 15)) >> 16;
}

static inline uint32_t do_rmulh_w(int64_t n, int64_t m)
{
    return (n * m + (1U << 31)) >> 32;
}

DO_2OP(vmulhsb, 1, int8_t, do_mulh_b)
DO_2OP(vmulhsh, 2, int16_t, do_mulh_h)
DO_2OP(vmulhsw, 4, int32_t, do_mulh_w)
DO_2OP(vmulhub, 1, uint8_t, do_mulh_b)
DO_2OP(vmulhuh, 2, uint16_t, do_mulh_h)
DO_2OP(vmulhuw, 4, uint32_t, do_mulh_w)

DO_2OP(vrmulhsb, 1, int8_t, do_rmulh_b)
DO_2OP(vrmulhsh, 2, int16_t, do_rmulh_h)
DO_2OP(vrmulhsw, 4, int32_t, do_rmulh_w)
DO_2OP(vrmulhub, 1, uint8_t, do_rmulh_b)
DO_2OP(vrmulhuh, 2, uint16_t, do_rmulh_h)
DO_2OP(vrmulhuw, 4, uint32_t, do_rmulh_w)

#define DO_MAX(N, M)  ((N) >= (M) ? (N) : (M))
#define DO_MIN(N, M)  ((N) >= (M) ? (M) : (N))

DO_2OP_S(vmaxs, DO_MAX)
DO_2OP_U(vmaxu, DO_MAX)
DO_2OP_S(vmins, DO_MIN)
DO_2OP_U(vminu, DO_MIN)

#define DO_ABD(N, M)  ((N) >= (M) ? (N) - (M) : (M) - (N))

DO_2OP_S(vabds, DO_ABD)
DO_2OP_U(vabdu, DO_ABD)

static inline uint32_t do_vhadd_u(uint32_t n, uint32_t m)
{
    return ((uint64_t)n + m) >> 1;
}

static inline int32_t do_vhadd_s(int32_t n, int32_t m)
{
    return ((int64_t)n + m) >> 1;
}

static inline uint32_t do_vhsub_u(uint32_t n, uint32_t m)
{
    return ((uint64_t)n - m) >> 1;
}

static inline int32_t do_vhsub_s(int32_t n, int32_t m)
{
    return ((int64_t)n - m) >> 1;
}

DO_2OP_S(vhadds, do_vhadd_s)
DO_2OP_U(vhaddu, do_vhadd_u)
DO_2OP_S(vhsubs, do_vhsub_s)
DO_2OP_U(vhsubu, do_vhsub_u)

#define DO_VSHLS(N, M) do_sqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, false, NULL)
#define DO_VSHLU(N, M) do_uqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, false, NULL)
#define DO_VRSHLS(N, M) do_sqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, true, NULL)
#define DO_VRSHLU(N, M) do_uqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, true, NULL)

DO_2OP_S(vshls, DO_VSHLS)
DO_2OP_U(vshlu, DO_VSHLU)
DO_2OP_S(vrshls, DO_VRSHLS)
DO_2OP_U(vrshlu, DO_VRSHLU)

#define DO_RHADD_S(N, M) (((int64_t)(N) + (M) + 1) >> 1)
#define DO_RHADD_U(N, M) (((uint64_t)(N) + (M) + 1) >> 1)

DO_2OP_S(vrhadds, DO_RHADD_S)
DO_2OP_U(vrhaddu, DO_RHADD_U)

static void do_vadc(CPUARMState *env, uint32_t *d, uint32_t *n, uint32_t *m,
                    uint32_t inv, uint32_t carry_in, bool update_flags)
{
    uint16_t mask = mve_element_mask(env);
    unsigned e;

    /* If any additions trigger, we will update flags. */
    if (mask & 0x1111) {
        update_flags = true;
    }

    for (e = 0; e < 16 / 4; e++, mask >>= 4) {
        uint64_t r = carry_in;
        r += n[H4(e)];
        r += m[H4(e)] ^ inv;
        if (mask & 1) {
            carry_in = r >> 32;
        }
        mergemask(&d[H4(e)], r, mask);
    }

    if (update_flags) {
        /* Store C, clear NZV. */
        env->vfp.xregs[ARM_VFP_FPSCR] &= ~FPCR_NZCV_MASK;
        env->vfp.xregs[ARM_VFP_FPSCR] |= carry_in * FPCR_C;
    }
    mve_advance_vpt(env);
}

void HELPER(mve_vadc)(CPUARMState *env, void *vd, void *vn, void *vm)
{
    bool carry_in = env->vfp.xregs[ARM_VFP_FPSCR] & FPCR_C;
    do_vadc(env, vd, vn, vm, 0, carry_in, false);
}

void HELPER(mve_vsbc)(CPUARMState *env, void *vd, void *vn, void *vm)
{
    bool carry_in = env->vfp.xregs[ARM_VFP_FPSCR] & FPCR_C;
    do_vadc(env, vd, vn, vm, -1, carry_in, false);
}


void HELPER(mve_vadci)(CPUARMState *env, void *vd, void *vn, void *vm)
{
    do_vadc(env, vd, vn, vm, 0, 0, true);
}

void HELPER(mve_vsbci)(CPUARMState *env, void *vd, void *vn, void *vm)
{
    do_vadc(env, vd, vn, vm, -1, 1, true);
}

#define DO_VCADD(OP, ESIZE, TYPE, FN0, FN1)                             \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, void *vm) \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        TYPE r[16 / ESIZE];                                             \
        /* Calculate all results first to avoid overwriting inputs */   \
        for (e = 0; e < 16 / ESIZE; e++) {                              \
            if (!(e & 1)) {                                             \
                r[e] = FN0(n[H##ESIZE(e)], m[H##ESIZE(e + 1)]);         \
            } else {                                                    \
                r[e] = FN1(n[H##ESIZE(e)], m[H##ESIZE(e - 1)]);         \
            }                                                           \
        }                                                               \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            mergemask(&d[H##ESIZE(e)], r[e], mask);                     \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_VCADD_ALL(OP, FN0, FN1)              \
    DO_VCADD(OP##b, 1, int8_t, FN0, FN1)        \
    DO_VCADD(OP##h, 2, int16_t, FN0, FN1)       \
    DO_VCADD(OP##w, 4, int32_t, FN0, FN1)

DO_VCADD_ALL(vcadd90, DO_SUB, DO_ADD)
DO_VCADD_ALL(vcadd270, DO_ADD, DO_SUB)
DO_VCADD_ALL(vhcadd90, do_vhsub_s, do_vhadd_s)
DO_VCADD_ALL(vhcadd270, do_vhadd_s, do_vhsub_s)

static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s)
{
    if (val > max) {
        *s = true;
        return max;
    } else if (val < min) {
        *s = true;
        return min;
    }
    return val;
}

#define DO_SQADD_B(n, m, s) do_sat_bhw((int64_t)n + m, INT8_MIN, INT8_MAX, s)
#define DO_SQADD_H(n, m, s) do_sat_bhw((int64_t)n + m, INT16_MIN, INT16_MAX, s)
#define DO_SQADD_W(n, m, s) do_sat_bhw((int64_t)n + m, INT32_MIN, INT32_MAX, s)

#define DO_UQADD_B(n, m, s) do_sat_bhw((int64_t)n + m, 0, UINT8_MAX, s)
#define DO_UQADD_H(n, m, s) do_sat_bhw((int64_t)n + m, 0, UINT16_MAX, s)
#define DO_UQADD_W(n, m, s) do_sat_bhw((int64_t)n + m, 0, UINT32_MAX, s)

#define DO_SQSUB_B(n, m, s) do_sat_bhw((int64_t)n - m, INT8_MIN, INT8_MAX, s)
#define DO_SQSUB_H(n, m, s) do_sat_bhw((int64_t)n - m, INT16_MIN, INT16_MAX, s)
#define DO_SQSUB_W(n, m, s) do_sat_bhw((int64_t)n - m, INT32_MIN, INT32_MAX, s)

#define DO_UQSUB_B(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT8_MAX, s)
#define DO_UQSUB_H(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT16_MAX, s)
#define DO_UQSUB_W(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT32_MAX, s)

/*
 * For QDMULH and QRDMULH we simplify "double and shift by esize" into
 * "shift by esize-1", adjusting the QRDMULH rounding constant to match.
 */
#define DO_QDMULH_B(n, m, s) do_sat_bhw(((int64_t)n * m) >> 7, \
                                        INT8_MIN, INT8_MAX, s)
#define DO_QDMULH_H(n, m, s) do_sat_bhw(((int64_t)n * m) >> 15, \
                                        INT16_MIN, INT16_MAX, s)
#define DO_QDMULH_W(n, m, s) do_sat_bhw(((int64_t)n * m) >> 31, \
                                        INT32_MIN, INT32_MAX, s)

#define DO_QRDMULH_B(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 6)) >> 7, \
                                         INT8_MIN, INT8_MAX, s)
#define DO_QRDMULH_H(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 14)) >> 15, \
                                         INT16_MIN, INT16_MAX, s)
#define DO_QRDMULH_W(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 30)) >> 31, \
                                         INT32_MIN, INT32_MAX, s)

DO_2OP_SAT(vqdmulhb, 1, int8_t, DO_QDMULH_B)
DO_2OP_SAT(vqdmulhh, 2, int16_t, DO_QDMULH_H)
DO_2OP_SAT(vqdmulhw, 4, int32_t, DO_QDMULH_W)

DO_2OP_SAT(vqrdmulhb, 1, int8_t, DO_QRDMULH_B)
DO_2OP_SAT(vqrdmulhh, 2, int16_t, DO_QRDMULH_H)
DO_2OP_SAT(vqrdmulhw, 4, int32_t, DO_QRDMULH_W)

DO_2OP_SAT(vqaddub, 1, uint8_t, DO_UQADD_B)
DO_2OP_SAT(vqadduh, 2, uint16_t, DO_UQADD_H)
DO_2OP_SAT(vqadduw, 4, uint32_t, DO_UQADD_W)
DO_2OP_SAT(vqaddsb, 1, int8_t, DO_SQADD_B)
DO_2OP_SAT(vqaddsh, 2, int16_t, DO_SQADD_H)
DO_2OP_SAT(vqaddsw, 4, int32_t, DO_SQADD_W)

DO_2OP_SAT(vqsubub, 1, uint8_t, DO_UQSUB_B)
DO_2OP_SAT(vqsubuh, 2, uint16_t, DO_UQSUB_H)
DO_2OP_SAT(vqsubuw, 4, uint32_t, DO_UQSUB_W)
DO_2OP_SAT(vqsubsb, 1, int8_t, DO_SQSUB_B)
DO_2OP_SAT(vqsubsh, 2, int16_t, DO_SQSUB_H)
DO_2OP_SAT(vqsubsw, 4, int32_t, DO_SQSUB_W)

/*
 * This wrapper fixes up the impedance mismatch between do_sqrshl_bhs()
 * and friends wanting a uint32_t* sat and our needing a bool*.
 */
#define WRAP_QRSHL_HELPER(FN, N, M, ROUND, satp)                        \
    ({                                                                  \
        uint32_t su32 = 0;                                              \
        typeof(N) qrshl_ret = FN(N, (int8_t)(M), sizeof(N) * 8, ROUND, &su32); \
        if (su32) {                                                     \
            *satp = true;                                               \
        }                                                               \
        qrshl_ret;                                                      \
    })

#define DO_SQSHL_OP(N, M, satp) \
    WRAP_QRSHL_HELPER(do_sqrshl_bhs, N, M, false, satp)
#define DO_UQSHL_OP(N, M, satp) \
    WRAP_QRSHL_HELPER(do_uqrshl_bhs, N, M, false, satp)
#define DO_SQRSHL_OP(N, M, satp) \
    WRAP_QRSHL_HELPER(do_sqrshl_bhs, N, M, true, satp)
#define DO_UQRSHL_OP(N, M, satp) \
    WRAP_QRSHL_HELPER(do_uqrshl_bhs, N, M, true, satp)
#define DO_SUQSHL_OP(N, M, satp) \
    WRAP_QRSHL_HELPER(do_suqrshl_bhs, N, M, false, satp)

DO_2OP_SAT_S(vqshls, DO_SQSHL_OP)
DO_2OP_SAT_U(vqshlu, DO_UQSHL_OP)
DO_2OP_SAT_S(vqrshls, DO_SQRSHL_OP)
DO_2OP_SAT_U(vqrshlu, DO_UQRSHL_OP)

/*
 * Multiply add dual returning high half
 * The 'FN' here takes four inputs A, B, C, D, a 0/1 indicator of
 * whether to add the rounding constant, and the pointer to the
 * saturation flag, and should do "(A * B + C * D) * 2 + rounding constant",
 * saturate to twice the input size and return the high half; or
 * (A * B - C * D) etc for VQDMLSDH.
 */
#define DO_VQDMLADH_OP(OP, ESIZE, TYPE, XCHG, ROUND, FN)                \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                void *vm)                               \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        bool qc = false;                                                \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            bool sat = false;                                           \
            if ((e & 1) == XCHG) {                                      \
                TYPE vqdmladh_ret = FN(n[H##ESIZE(e)],                  \
                            m[H##ESIZE(e - XCHG)],                      \
                            n[H##ESIZE(e + (1 - 2 * XCHG))],            \
                            m[H##ESIZE(e + (1 - XCHG))],                \
                            ROUND, &sat);                               \
                mergemask(&d[H##ESIZE(e)], vqdmladh_ret, mask);         \
                qc |= sat & mask & 1;                                   \
            }                                                           \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

static int8_t do_vqdmladh_b(int8_t a, int8_t b, int8_t c, int8_t d,
                            int round, bool *sat)
{
    int64_t r = ((int64_t)a * b + (int64_t)c * d) * 2 + (round << 7);
    return do_sat_bhw(r, INT16_MIN, INT16_MAX, sat) >> 8;
}

static int16_t do_vqdmladh_h(int16_t a, int16_t b, int16_t c, int16_t d,
                             int round, bool *sat)
{
    int64_t r = ((int64_t)a * b + (int64_t)c * d) * 2 + (round << 15);
    return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat) >> 16;
}

static int32_t do_vqdmladh_w(int32_t a, int32_t b, int32_t c, int32_t d,
                             int round, bool *sat)
{
    int64_t m1 = (int64_t)a * b;
    int64_t m2 = (int64_t)c * d;
    int64_t r;
    /*
     * Architecturally we should do the entire add, double, round
     * and then check for saturation. We do three saturating adds,
     * but we need to be careful about the order. If the first
     * m1 + m2 saturates then it's impossible for the *2+rc to
     * bring it back into the non-saturated range. However, if
     * m1 + m2 is negative then it's possible that doing the doubling
     * would take the intermediate result below INT64_MAX and the
     * addition of the rounding constant then brings it back in range.
     * So we add half the rounding constant before doubling rather
     * than adding the rounding constant after the doubling.
     */
    if (sadd64_overflow(m1, m2, &r) ||
        sadd64_overflow(r, (round << 30), &r) ||
        sadd64_overflow(r, r, &r)) {
        *sat = true;
        return r < 0 ? INT32_MAX : INT32_MIN;
    }
    return r >> 32;
}

static int8_t do_vqdmlsdh_b(int8_t a, int8_t b, int8_t c, int8_t d,
                            int round, bool *sat)
{
    int64_t r = ((int64_t)a * b - (int64_t)c * d) * 2 + (round << 7);
    return do_sat_bhw(r, INT16_MIN, INT16_MAX, sat) >> 8;
}

static int16_t do_vqdmlsdh_h(int16_t a, int16_t b, int16_t c, int16_t d,
                             int round, bool *sat)
{
    int64_t r = ((int64_t)a * b - (int64_t)c * d) * 2 + (round << 15);
    return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat) >> 16;
}

static int32_t do_vqdmlsdh_w(int32_t a, int32_t b, int32_t c, int32_t d,
                             int round, bool *sat)
{
    int64_t m1 = (int64_t)a * b;
    int64_t m2 = (int64_t)c * d;
    int64_t r;
    /* The same ordering issue as in do_vqdmladh_w applies here too */
    if (ssub64_overflow(m1, m2, &r) ||
        sadd64_overflow(r, (round << 30), &r) ||
        sadd64_overflow(r, r, &r)) {
        *sat = true;
        return r < 0 ? INT32_MAX : INT32_MIN;
    }
    return r >> 32;
}

DO_VQDMLADH_OP(vqdmladhb, 1, int8_t, 0, 0, do_vqdmladh_b)
DO_VQDMLADH_OP(vqdmladhh, 2, int16_t, 0, 0, do_vqdmladh_h)
DO_VQDMLADH_OP(vqdmladhw, 4, int32_t, 0, 0, do_vqdmladh_w)
DO_VQDMLADH_OP(vqdmladhxb, 1, int8_t, 1, 0, do_vqdmladh_b)
DO_VQDMLADH_OP(vqdmladhxh, 2, int16_t, 1, 0, do_vqdmladh_h)
DO_VQDMLADH_OP(vqdmladhxw, 4, int32_t, 1, 0, do_vqdmladh_w)

DO_VQDMLADH_OP(vqrdmladhb, 1, int8_t, 0, 1, do_vqdmladh_b)
DO_VQDMLADH_OP(vqrdmladhh, 2, int16_t, 0, 1, do_vqdmladh_h)
DO_VQDMLADH_OP(vqrdmladhw, 4, int32_t, 0, 1, do_vqdmladh_w)
DO_VQDMLADH_OP(vqrdmladhxb, 1, int8_t, 1, 1, do_vqdmladh_b)
DO_VQDMLADH_OP(vqrdmladhxh, 2, int16_t, 1, 1, do_vqdmladh_h)
DO_VQDMLADH_OP(vqrdmladhxw, 4, int32_t, 1, 1, do_vqdmladh_w)

DO_VQDMLADH_OP(vqdmlsdhb, 1, int8_t, 0, 0, do_vqdmlsdh_b)
DO_VQDMLADH_OP(vqdmlsdhh, 2, int16_t, 0, 0, do_vqdmlsdh_h)
DO_VQDMLADH_OP(vqdmlsdhw, 4, int32_t, 0, 0, do_vqdmlsdh_w)
DO_VQDMLADH_OP(vqdmlsdhxb, 1, int8_t, 1, 0, do_vqdmlsdh_b)
DO_VQDMLADH_OP(vqdmlsdhxh, 2, int16_t, 1, 0, do_vqdmlsdh_h)
DO_VQDMLADH_OP(vqdmlsdhxw, 4, int32_t, 1, 0, do_vqdmlsdh_w)

DO_VQDMLADH_OP(vqrdmlsdhb, 1, int8_t, 0, 1, do_vqdmlsdh_b)
DO_VQDMLADH_OP(vqrdmlsdhh, 2, int16_t, 0, 1, do_vqdmlsdh_h)
DO_VQDMLADH_OP(vqrdmlsdhw, 4, int32_t, 0, 1, do_vqdmlsdh_w)
DO_VQDMLADH_OP(vqrdmlsdhxb, 1, int8_t, 1, 1, do_vqdmlsdh_b)
DO_VQDMLADH_OP(vqrdmlsdhxh, 2, int16_t, 1, 1, do_vqdmlsdh_h)
DO_VQDMLADH_OP(vqrdmlsdhxw, 4, int32_t, 1, 1, do_vqdmlsdh_w)

#define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN)                              \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                uint32_t rm)                            \
    {                                                                   \
        TYPE *d = vd, *n = vn;                                          \
        TYPE m = rm;                                                    \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            mergemask(&d[H##ESIZE(e)], FN(n[H##ESIZE(e)], m), mask);    \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_2OP_SAT_SCALAR(OP, ESIZE, TYPE, FN)                          \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                uint32_t rm)                            \
    {                                                                   \
        TYPE *d = vd, *n = vn;                                          \
        TYPE m = rm;                                                    \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        bool qc = false;                                                \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            bool sat = false;                                           \
            mergemask(&d[H##ESIZE(e)], FN(n[H##ESIZE(e)], m, &sat),     \
                      mask);                                            \
            qc |= sat & mask & 1;                                       \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

/* "accumulating" version where FN takes d as well as n and m */
#define DO_2OP_ACC_SCALAR(OP, ESIZE, TYPE, FN)                          \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                uint32_t rm)                            \
    {                                                                   \
        TYPE *d = vd, *n = vn;                                          \
        TYPE m = rm;                                                    \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            mergemask(&d[H##ESIZE(e)],                                  \
                      FN(d[H##ESIZE(e)], n[H##ESIZE(e)], m), mask);     \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_2OP_SAT_ACC_SCALAR(OP, ESIZE, TYPE, FN)                      \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                uint32_t rm)                            \
    {                                                                   \
        TYPE *d = vd, *n = vn;                                          \
        TYPE m = rm;                                                    \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        bool qc = false;                                                \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            bool sat = false;                                           \
            mergemask(&d[H##ESIZE(e)],                                  \
                      FN(d[H##ESIZE(e)], n[H##ESIZE(e)], m, &sat),      \
                      mask);                                            \
            qc |= sat & mask & 1;                                       \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

/* provide unsigned 2-op scalar helpers for all sizes */
#define DO_2OP_SCALAR_U(OP, FN)                 \
    DO_2OP_SCALAR(OP##b, 1, uint8_t, FN)        \
    DO_2OP_SCALAR(OP##h, 2, uint16_t, FN)       \
    DO_2OP_SCALAR(OP##w, 4, uint32_t, FN)
#define DO_2OP_SCALAR_S(OP, FN)                 \
    DO_2OP_SCALAR(OP##b, 1, int8_t, FN)         \
    DO_2OP_SCALAR(OP##h, 2, int16_t, FN)        \
    DO_2OP_SCALAR(OP##w, 4, int32_t, FN)

#define DO_2OP_ACC_SCALAR_U(OP, FN)             \
    DO_2OP_ACC_SCALAR(OP##b, 1, uint8_t, FN)    \
    DO_2OP_ACC_SCALAR(OP##h, 2, uint16_t, FN)   \
    DO_2OP_ACC_SCALAR(OP##w, 4, uint32_t, FN)

DO_2OP_SCALAR_U(vadd_scalar, DO_ADD)
DO_2OP_SCALAR_U(vsub_scalar, DO_SUB)
DO_2OP_SCALAR_U(vmul_scalar, DO_MUL)
DO_2OP_SCALAR_S(vhadds_scalar, do_vhadd_s)
DO_2OP_SCALAR_U(vhaddu_scalar, do_vhadd_u)
DO_2OP_SCALAR_S(vhsubs_scalar, do_vhsub_s)
DO_2OP_SCALAR_U(vhsubu_scalar, do_vhsub_u)

DO_2OP_SAT_SCALAR(vqaddu_scalarb, 1, uint8_t, DO_UQADD_B)
DO_2OP_SAT_SCALAR(vqaddu_scalarh, 2, uint16_t, DO_UQADD_H)
DO_2OP_SAT_SCALAR(vqaddu_scalarw, 4, uint32_t, DO_UQADD_W)
DO_2OP_SAT_SCALAR(vqadds_scalarb, 1, int8_t, DO_SQADD_B)
DO_2OP_SAT_SCALAR(vqadds_scalarh, 2, int16_t, DO_SQADD_H)
DO_2OP_SAT_SCALAR(vqadds_scalarw, 4, int32_t, DO_SQADD_W)

DO_2OP_SAT_SCALAR(vqsubu_scalarb, 1, uint8_t, DO_UQSUB_B)
DO_2OP_SAT_SCALAR(vqsubu_scalarh, 2, uint16_t, DO_UQSUB_H)
DO_2OP_SAT_SCALAR(vqsubu_scalarw, 4, uint32_t, DO_UQSUB_W)
DO_2OP_SAT_SCALAR(vqsubs_scalarb, 1, int8_t, DO_SQSUB_B)
DO_2OP_SAT_SCALAR(vqsubs_scalarh, 2, int16_t, DO_SQSUB_H)
DO_2OP_SAT_SCALAR(vqsubs_scalarw, 4, int32_t, DO_SQSUB_W)

DO_2OP_SAT_SCALAR(vqdmulh_scalarb, 1, int8_t, DO_QDMULH_B)
DO_2OP_SAT_SCALAR(vqdmulh_scalarh, 2, int16_t, DO_QDMULH_H)
DO_2OP_SAT_SCALAR(vqdmulh_scalarw, 4, int32_t, DO_QDMULH_W)
DO_2OP_SAT_SCALAR(vqrdmulh_scalarb, 1, int8_t, DO_QRDMULH_B)
DO_2OP_SAT_SCALAR(vqrdmulh_scalarh, 2, int16_t, DO_QRDMULH_H)
DO_2OP_SAT_SCALAR(vqrdmulh_scalarw, 4, int32_t, DO_QRDMULH_W)

static int8_t do_vqdmlah_b(int8_t a, int8_t b, int8_t c, int round, bool *sat)
{
    int64_t r = (int64_t)a * b * 2 + ((int64_t)c << 8) + (round << 7);
    return do_sat_bhw(r, INT16_MIN, INT16_MAX, sat) >> 8;
}

static int16_t do_vqdmlah_h(int16_t a, int16_t b, int16_t c,
                           int round, bool *sat)
{
    int64_t r = (int64_t)a * b * 2 + ((int64_t)c << 16) + (round << 15);
    return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat) >> 16;
}

static int32_t do_vqdmlah_w(int32_t a, int32_t b, int32_t c,
                            int round, bool *sat)
{
    /*
     * Architecturally we should do the entire add, double, round
     * and then check for saturation. We do three saturating adds,
     * but we need to be careful about the order. If the first
     * m1 + m2 saturates then it's impossible for the *2+rc to
     * bring it back into the non-saturated range. However, if
     * m1 + m2 is negative then it's possible that doing the doubling
     * would take the intermediate result below INT64_MAX and the
     * addition of the rounding constant then brings it back in range.
     * So we add half the rounding constant and half the "c << esize"
     * before doubling rather than adding the rounding constant after
     * the doubling.
     */
    int64_t m1 = (int64_t)a * b;
    int64_t m2 = (int64_t)c << 31;
    int64_t r;
    if (sadd64_overflow(m1, m2, &r) ||
        sadd64_overflow(r, (round << 30), &r) ||
        sadd64_overflow(r, r, &r)) {
        *sat = true;
        return r < 0 ? INT32_MAX : INT32_MIN;
    }
    return r >> 32;
}

/*
 * The *MLAH insns are vector * scalar + vector;
 * the *MLASH insns are vector * vector + scalar
 */
#define DO_VQDMLAH_B(D, N, M, S) do_vqdmlah_b(N, M, D, 0, S)
#define DO_VQDMLAH_H(D, N, M, S) do_vqdmlah_h(N, M, D, 0, S)
#define DO_VQDMLAH_W(D, N, M, S) do_vqdmlah_w(N, M, D, 0, S)
#define DO_VQRDMLAH_B(D, N, M, S) do_vqdmlah_b(N, M, D, 1, S)
#define DO_VQRDMLAH_H(D, N, M, S) do_vqdmlah_h(N, M, D, 1, S)
#define DO_VQRDMLAH_W(D, N, M, S) do_vqdmlah_w(N, M, D, 1, S)

#define DO_VQDMLASH_B(D, N, M, S) do_vqdmlah_b(N, D, M, 0, S)
#define DO_VQDMLASH_H(D, N, M, S) do_vqdmlah_h(N, D, M, 0, S)
#define DO_VQDMLASH_W(D, N, M, S) do_vqdmlah_w(N, D, M, 0, S)
#define DO_VQRDMLASH_B(D, N, M, S) do_vqdmlah_b(N, D, M, 1, S)
#define DO_VQRDMLASH_H(D, N, M, S) do_vqdmlah_h(N, D, M, 1, S)
#define DO_VQRDMLASH_W(D, N, M, S) do_vqdmlah_w(N, D, M, 1, S)

DO_2OP_SAT_ACC_SCALAR(vqdmlahb, 1, int8_t, DO_VQDMLAH_B)
DO_2OP_SAT_ACC_SCALAR(vqdmlahh, 2, int16_t, DO_VQDMLAH_H)
DO_2OP_SAT_ACC_SCALAR(vqdmlahw, 4, int32_t, DO_VQDMLAH_W)
DO_2OP_SAT_ACC_SCALAR(vqrdmlahb, 1, int8_t, DO_VQRDMLAH_B)
DO_2OP_SAT_ACC_SCALAR(vqrdmlahh, 2, int16_t, DO_VQRDMLAH_H)
DO_2OP_SAT_ACC_SCALAR(vqrdmlahw, 4, int32_t, DO_VQRDMLAH_W)

DO_2OP_SAT_ACC_SCALAR(vqdmlashb, 1, int8_t, DO_VQDMLASH_B)
DO_2OP_SAT_ACC_SCALAR(vqdmlashh, 2, int16_t, DO_VQDMLASH_H)
DO_2OP_SAT_ACC_SCALAR(vqdmlashw, 4, int32_t, DO_VQDMLASH_W)
DO_2OP_SAT_ACC_SCALAR(vqrdmlashb, 1, int8_t, DO_VQRDMLASH_B)
DO_2OP_SAT_ACC_SCALAR(vqrdmlashh, 2, int16_t, DO_VQRDMLASH_H)
DO_2OP_SAT_ACC_SCALAR(vqrdmlashw, 4, int32_t, DO_VQRDMLASH_W)

/* Vector by scalar plus vector */
#define DO_VMLA(D, N, M) ((N) * (M) + (D))

DO_2OP_ACC_SCALAR_U(vmla, DO_VMLA)

/* Vector by vector plus scalar */
#define DO_VMLAS(D, N, M) ((N) * (D) + (M))

DO_2OP_ACC_SCALAR_U(vmlas, DO_VMLAS)

/*
 * Long saturating scalar ops. As with DO_2OP_L, TYPE and H are for the
 * input (smaller) type and LESIZE, LTYPE, LH for the output (long) type.
 * SATMASK specifies which bits of the predicate mask matter for determining
 * whether to propagate a saturation indication into FPSCR.QC -- for
 * the 16x16->32 case we must check only the bit corresponding to the T or B
 * half that we used, but for the 32x32->64 case we propagate if the mask
 * bit is set for either half.
 */
#define DO_2OP_SAT_SCALAR_L(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN, SATMASK) \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                uint32_t rm)                            \
    {                                                                   \
        LTYPE *d = vd;                                                  \
        TYPE *n = vn;                                                   \
        TYPE m = rm;                                                    \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned le;                                                    \
        bool qc = false;                                                \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) {         \
            bool sat = false;                                           \
            LTYPE r = FN((LTYPE)n[H##ESIZE(le * 2 + TOP)], m, &sat);    \
            mergemask(&d[H##LESIZE(le)], r, mask);                      \
            qc |= sat && (mask & SATMASK);                              \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

static inline int32_t do_qdmullh(int16_t n, int16_t m, bool *sat)
{
    int64_t r = ((int64_t)n * m) * 2;
    return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat);
}

static inline int64_t do_qdmullw(int32_t n, int32_t m, bool *sat)
{
    /* The multiply can't overflow, but the doubling might */
    int64_t r = (int64_t)n * m;
    if (r > INT64_MAX / 2) {
        *sat = true;
        return INT64_MAX;
    } else if (r < INT64_MIN / 2) {
        *sat = true;
        return INT64_MIN;
    } else {
        return r * 2;
    }
}

#define SATMASK16B 1
#define SATMASK16T (1 << 2)
#define SATMASK32 ((1 << 4) | 1)

DO_2OP_SAT_SCALAR_L(vqdmullb_scalarh, 0, 2, int16_t, 4, int32_t, \
                    do_qdmullh, SATMASK16B)
DO_2OP_SAT_SCALAR_L(vqdmullb_scalarw, 0, 4, int32_t, 8, int64_t, \
                    do_qdmullw, SATMASK32)
DO_2OP_SAT_SCALAR_L(vqdmullt_scalarh, 1, 2, int16_t, 4, int32_t, \
                    do_qdmullh, SATMASK16T)
DO_2OP_SAT_SCALAR_L(vqdmullt_scalarw, 1, 4, int32_t, 8, int64_t, \
                    do_qdmullw, SATMASK32)

/*
 * Long saturating ops
 */
#define DO_2OP_SAT_L(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN, SATMASK)  \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                void *vm)                               \
    {                                                                   \
        LTYPE *d = vd;                                                  \
        TYPE *n = vn, *m = vm;                                          \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned le;                                                    \
        bool qc = false;                                                \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) {         \
            bool sat = false;                                           \
            LTYPE op1 = n[H##ESIZE(le * 2 + TOP)];                      \
            LTYPE op2 = m[H##ESIZE(le * 2 + TOP)];                      \
            mergemask(&d[H##LESIZE(le)], FN(op1, op2, &sat), mask);     \
            qc |= sat && (mask & SATMASK);                              \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_2OP_SAT_L(vqdmullbh, 0, 2, int16_t, 4, int32_t, do_qdmullh, SATMASK16B)
DO_2OP_SAT_L(vqdmullbw, 0, 4, int32_t, 8, int64_t, do_qdmullw, SATMASK32)
DO_2OP_SAT_L(vqdmullth, 1, 2, int16_t, 4, int32_t, do_qdmullh, SATMASK16T)
DO_2OP_SAT_L(vqdmulltw, 1, 4, int32_t, 8, int64_t, do_qdmullw, SATMASK32)

static inline uint32_t do_vbrsrb(uint32_t n, uint32_t m)
{
    m &= 0xff;
    if (m == 0) {
        return 0;
    }
    n = revbit8(n);
    if (m < 8) {
        n >>= 8 - m;
    }
    return n;
}

static inline uint32_t do_vbrsrh(uint32_t n, uint32_t m)
{
    m &= 0xff;
    if (m == 0) {
        return 0;
    }
    n = revbit16(n);
    if (m < 16) {
        n >>= 16 - m;
    }
    return n;
}

static inline uint32_t do_vbrsrw(uint32_t n, uint32_t m)
{
    m &= 0xff;
    if (m == 0) {
        return 0;
    }
    n = revbit32(n);
    if (m < 32) {
        n >>= 32 - m;
    }
    return n;
}

DO_2OP_SCALAR(vbrsrb, 1, uint8_t, do_vbrsrb)
DO_2OP_SCALAR(vbrsrh, 2, uint16_t, do_vbrsrh)
DO_2OP_SCALAR(vbrsrw, 4, uint32_t, do_vbrsrw)

/*
 * Multiply add long dual accumulate ops.
 */
#define DO_LDAV(OP, ESIZE, TYPE, XCHG, EVENACC, ODDACC)                 \
    uint64_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vn,         \
                                    void *vm, uint64_t a)               \
    {                                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        TYPE *n = vn, *m = vm;                                          \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if (mask & 1) {                                             \
                if (e & 1) {                                            \
                    a ODDACC                                            \
                        (int64_t)n[H##ESIZE(e - 1 * XCHG)] * m[H##ESIZE(e)]; \
                } else {                                                \
                    a EVENACC                                           \
                        (int64_t)n[H##ESIZE(e + 1 * XCHG)] * m[H##ESIZE(e)]; \
                }                                                       \
            }                                                           \
        }                                                               \
        mve_advance_vpt(env);                                           \
        return a;                                                       \
    }

DO_LDAV(vmlaldavsh, 2, int16_t, false, +=, +=)
DO_LDAV(vmlaldavxsh, 2, int16_t, true, +=, +=)
DO_LDAV(vmlaldavsw, 4, int32_t, false, +=, +=)
DO_LDAV(vmlaldavxsw, 4, int32_t, true, +=, +=)

DO_LDAV(vmlaldavuh, 2, uint16_t, false, +=, +=)
DO_LDAV(vmlaldavuw, 4, uint32_t, false, +=, +=)

DO_LDAV(vmlsldavsh, 2, int16_t, false, +=, -=)
DO_LDAV(vmlsldavxsh, 2, int16_t, true, +=, -=)
DO_LDAV(vmlsldavsw, 4, int32_t, false, +=, -=)
DO_LDAV(vmlsldavxsw, 4, int32_t, true, +=, -=)

/*
 * Multiply add dual accumulate ops
 */
#define DO_DAV(OP, ESIZE, TYPE, XCHG, EVENACC, ODDACC) \
    uint32_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vn,         \
                                    void *vm, uint32_t a)               \
    {                                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        TYPE *n = vn, *m = vm;                                          \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if (mask & 1) {                                             \
                if (e & 1) {                                            \
                    a ODDACC                                            \
                        n[H##ESIZE(e - 1 * XCHG)] * m[H##ESIZE(e)];     \
                } else {                                                \
                    a EVENACC                                           \
                        n[H##ESIZE(e + 1 * XCHG)] * m[H##ESIZE(e)];     \
                }                                                       \
            }                                                           \
        }                                                               \
        mve_advance_vpt(env);                                           \
        return a;                                                       \
    }

#define DO_DAV_S(INSN, XCHG, EVENACC, ODDACC)           \
    DO_DAV(INSN##b, 1, int8_t, XCHG, EVENACC, ODDACC)   \
    DO_DAV(INSN##h, 2, int16_t, XCHG, EVENACC, ODDACC)  \
    DO_DAV(INSN##w, 4, int32_t, XCHG, EVENACC, ODDACC)

#define DO_DAV_U(INSN, XCHG, EVENACC, ODDACC)           \
    DO_DAV(INSN##b, 1, uint8_t, XCHG, EVENACC, ODDACC)  \
    DO_DAV(INSN##h, 2, uint16_t, XCHG, EVENACC, ODDACC) \
    DO_DAV(INSN##w, 4, uint32_t, XCHG, EVENACC, ODDACC)

DO_DAV_S(vmladavs, false, +=, +=)
DO_DAV_U(vmladavu, false, +=, +=)
DO_DAV_S(vmlsdav, false, +=, -=)
DO_DAV_S(vmladavsx, true, +=, +=)
DO_DAV_S(vmlsdavx, true, +=, -=)

/*
 * Rounding multiply add long dual accumulate high. In the pseudocode
 * this is implemented with a 72-bit internal accumulator value of which
 * the top 64 bits are returned. We optimize this to avoid having to
 * use 128-bit arithmetic -- we can do this because the 74-bit accumulator
 * is squashed back into 64-bits after each beat.
 */
#define DO_LDAVH(OP, TYPE, LTYPE, XCHG, SUB)                            \
    uint64_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vn,         \
                                    void *vm, uint64_t a)               \
    {                                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        TYPE *n = vn, *m = vm;                                          \
        for (e = 0; e < 16 / 4; e++, mask >>= 4) {                      \
            if (mask & 1) {                                             \
                LTYPE mul;                                              \
                if (e & 1) {                                            \
                    mul = (LTYPE)n[H4(e - 1 * XCHG)] * m[H4(e)];        \
                    if (SUB) {                                          \
                        mul = -mul;                                     \
                    }                                                   \
                } else {                                                \
                    mul = (LTYPE)n[H4(e + 1 * XCHG)] * m[H4(e)];        \
                }                                                       \
                mul = (mul >> 8) + ((mul >> 7) & 1);                    \
                a += mul;                                               \
            }                                                           \
        }                                                               \
        mve_advance_vpt(env);                                           \
        return a;                                                       \
    }

DO_LDAVH(vrmlaldavhsw, int32_t, int64_t, false, false)
DO_LDAVH(vrmlaldavhxsw, int32_t, int64_t, true, false)

DO_LDAVH(vrmlaldavhuw, uint32_t, uint64_t, false, false)

DO_LDAVH(vrmlsldavhsw, int32_t, int64_t, false, true)
DO_LDAVH(vrmlsldavhxsw, int32_t, int64_t, true, true)

/* Vector add across vector */
#define DO_VADDV(OP, ESIZE, TYPE)                               \
    uint32_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vm, \
                                    uint32_t ra)                \
    {                                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        TYPE *m = vm;                                           \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            if (mask & 1) {                                     \
                ra += m[H##ESIZE(e)];                           \
            }                                                   \
        }                                                       \
        mve_advance_vpt(env);                                   \
        return ra;                                              \
    }                                                           \

DO_VADDV(vaddvsb, 1, int8_t)
DO_VADDV(vaddvsh, 2, int16_t)
DO_VADDV(vaddvsw, 4, int32_t)
DO_VADDV(vaddvub, 1, uint8_t)
DO_VADDV(vaddvuh, 2, uint16_t)
DO_VADDV(vaddvuw, 4, uint32_t)

/*
 * Vector max/min across vector. Unlike VADDV, we must
 * read ra as the element size, not its full width.
 * We work with int64_t internally for simplicity.
 */
#define DO_VMAXMINV(OP, ESIZE, TYPE, RATYPE, FN)                \
    uint32_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vm, \
                                    uint32_t ra_in)             \
    {                                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        TYPE *m = vm;                                           \
        int64_t ra = (RATYPE)ra_in;                             \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            if (mask & 1) {                                     \
                ra = FN(ra, m[H##ESIZE(e)]);                    \
            }                                                   \
        }                                                       \
        mve_advance_vpt(env);                                   \
        return ra;                                              \
    }                                                           \

#define DO_VMAXMINV_U(INSN, FN)                         \
    DO_VMAXMINV(INSN##b, 1, uint8_t, uint8_t, FN)       \
    DO_VMAXMINV(INSN##h, 2, uint16_t, uint16_t, FN)     \
    DO_VMAXMINV(INSN##w, 4, uint32_t, uint32_t, FN)
#define DO_VMAXMINV_S(INSN, FN)                         \
    DO_VMAXMINV(INSN##b, 1, int8_t, int8_t, FN)         \
    DO_VMAXMINV(INSN##h, 2, int16_t, int16_t, FN)       \
    DO_VMAXMINV(INSN##w, 4, int32_t, int32_t, FN)

/*
 * Helpers for max and min of absolute values across vector:
 * note that we only take the absolute value of 'm', not 'n'
 */
static int64_t do_maxa(int64_t n, int64_t m)
{
    if (m < 0) {
        m = -m;
    }
    return MAX(n, m);
}

static int64_t do_mina(int64_t n, int64_t m)
{
    if (m < 0) {
        m = -m;
    }
    return MIN(n, m);
}

DO_VMAXMINV_S(vmaxvs, DO_MAX)
DO_VMAXMINV_U(vmaxvu, DO_MAX)
DO_VMAXMINV_S(vminvs, DO_MIN)
DO_VMAXMINV_U(vminvu, DO_MIN)
/*
 * VMAXAV, VMINAV treat the general purpose input as unsigned
 * and the vector elements as signed.
 */
DO_VMAXMINV(vmaxavb, 1, int8_t, uint8_t, do_maxa)
DO_VMAXMINV(vmaxavh, 2, int16_t, uint16_t, do_maxa)
DO_VMAXMINV(vmaxavw, 4, int32_t, uint32_t, do_maxa)
DO_VMAXMINV(vminavb, 1, int8_t, uint8_t, do_mina)
DO_VMAXMINV(vminavh, 2, int16_t, uint16_t, do_mina)
DO_VMAXMINV(vminavw, 4, int32_t, uint32_t, do_mina)

#define DO_VABAV(OP, ESIZE, TYPE)                               \
    uint32_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vn, \
                                    void *vm, uint32_t ra)      \
    {                                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        TYPE *m = vm, *n = vn;                                  \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            if (mask & 1) {                                     \
                int64_t n0 = n[H##ESIZE(e)];                    \
                int64_t m0 = m[H##ESIZE(e)];                    \
                uint32_t r = n0 >= m0 ? (n0 - m0) : (m0 - n0);  \
                ra += r;                                        \
            }                                                   \
        }                                                       \
        mve_advance_vpt(env);                                   \
        return ra;                                              \
    }

DO_VABAV(vabavsb, 1, int8_t)
DO_VABAV(vabavsh, 2, int16_t)
DO_VABAV(vabavsw, 4, int32_t)
DO_VABAV(vabavub, 1, uint8_t)
DO_VABAV(vabavuh, 2, uint16_t)
DO_VABAV(vabavuw, 4, uint32_t)

#define DO_VADDLV(OP, TYPE, LTYPE)                              \
    uint64_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vm, \
                                    uint64_t ra)                \
    {                                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        TYPE *m = vm;                                           \
        for (e = 0; e < 16 / 4; e++, mask >>= 4) {              \
            if (mask & 1) {                                     \
                ra += (LTYPE)m[H4(e)];                          \
            }                                                   \
        }                                                       \
        mve_advance_vpt(env);                                   \
        return ra;                                              \
    }                                                           \

DO_VADDLV(vaddlv_s, int32_t, int64_t)
DO_VADDLV(vaddlv_u, uint32_t, uint64_t)

/* Shifts by immediate */
#define DO_2SHIFT(OP, ESIZE, TYPE, FN)                          \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd,     \
                                void *vm, uint32_t shift)       \
    {                                                           \
        TYPE *d = vd, *m = vm;                                  \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            mergemask(&d[H##ESIZE(e)],                          \
                      FN(m[H##ESIZE(e)], shift), mask);         \
        }                                                       \
        mve_advance_vpt(env);                                   \
    }

#define DO_2SHIFT_SAT(OP, ESIZE, TYPE, FN)                      \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd,     \
                                void *vm, uint32_t shift)       \
    {                                                           \
        TYPE *d = vd, *m = vm;                                  \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        bool qc = false;                                        \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            bool sat = false;                                   \
            mergemask(&d[H##ESIZE(e)],                          \
                      FN(m[H##ESIZE(e)], shift, &sat), mask);   \
            qc |= sat & mask & 1;                               \
        }                                                       \
        if (qc) {                                               \
            env->vfp.qc[0] = qc;                                \
        }                                                       \
        mve_advance_vpt(env);                                   \
    }

/* provide unsigned 2-op shift helpers for all sizes */
#define DO_2SHIFT_U(OP, FN)                     \
    DO_2SHIFT(OP##b, 1, uint8_t, FN)            \
    DO_2SHIFT(OP##h, 2, uint16_t, FN)           \
    DO_2SHIFT(OP##w, 4, uint32_t, FN)
#define DO_2SHIFT_S(OP, FN)                     \
    DO_2SHIFT(OP##b, 1, int8_t, FN)             \
    DO_2SHIFT(OP##h, 2, int16_t, FN)            \
    DO_2SHIFT(OP##w, 4, int32_t, FN)

#define DO_2SHIFT_SAT_U(OP, FN)                 \
    DO_2SHIFT_SAT(OP##b, 1, uint8_t, FN)        \
    DO_2SHIFT_SAT(OP##h, 2, uint16_t, FN)       \
    DO_2SHIFT_SAT(OP##w, 4, uint32_t, FN)
#define DO_2SHIFT_SAT_S(OP, FN)                 \
    DO_2SHIFT_SAT(OP##b, 1, int8_t, FN)         \
    DO_2SHIFT_SAT(OP##h, 2, int16_t, FN)        \
    DO_2SHIFT_SAT(OP##w, 4, int32_t, FN)

DO_2SHIFT_U(vshli_u, DO_VSHLU)
DO_2SHIFT_S(vshli_s, DO_VSHLS)
DO_2SHIFT_SAT_U(vqshli_u, DO_UQSHL_OP)
DO_2SHIFT_SAT_S(vqshli_s, DO_SQSHL_OP)
DO_2SHIFT_SAT_S(vqshlui_s, DO_SUQSHL_OP)
DO_2SHIFT_U(vrshli_u, DO_VRSHLU)
DO_2SHIFT_S(vrshli_s, DO_VRSHLS)
DO_2SHIFT_SAT_U(vqrshli_u, DO_UQRSHL_OP)
DO_2SHIFT_SAT_S(vqrshli_s, DO_SQRSHL_OP)

/* Shift-and-insert; we always work with 64 bits at a time */
#define DO_2SHIFT_INSERT(OP, ESIZE, SHIFTFN, MASKFN)                    \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd,             \
                                void *vm, uint32_t shift)               \
    {                                                                   \
        uint64_t *d = vd, *m = vm;                                      \
        uint16_t mask;                                                  \
        uint64_t shiftmask;                                             \
        unsigned e;                                                     \
        if (shift == ESIZE * 8) {                                       \
            /*                                                          \
             * Only VSRI can shift by <dt>; it should mean "don't       \
             * update the destination". The generic logic can't handle  \
             * this because it would try to shift by an out-of-range    \
             * amount, so special case it here.                         \
             */                                                         \
            goto done;                                                  \
        }                                                               \
        assert(shift < ESIZE * 8);                                      \
        mask = mve_element_mask(env);                                   \
        /* ESIZE / 2 gives the MO_* value if ESIZE is in [1,2,4] */     \
        shiftmask = dup_const(ESIZE / 2, MASKFN(ESIZE * 8, shift));     \
        for (e = 0; e < 16 / 8; e++, mask >>= 8) {                      \
            uint64_t r = (SHIFTFN(m[H8(e)], shift) & shiftmask) |       \
                (d[H8(e)] & ~shiftmask);                                \
            mergemask(&d[H8(e)], r, mask);                              \
        }                                                               \
done:                                                                   \
        mve_advance_vpt(env);                                           \
    }

#define DO_SHL(N, SHIFT) ((N) << (SHIFT))
#define DO_SHR(N, SHIFT) ((N) >> (SHIFT))
#define SHL_MASK(EBITS, SHIFT) MAKE_64BIT_MASK((SHIFT), (EBITS) - (SHIFT))
#define SHR_MASK(EBITS, SHIFT) MAKE_64BIT_MASK(0, (EBITS) - (SHIFT))

DO_2SHIFT_INSERT(vsrib, 1, DO_SHR, SHR_MASK)
DO_2SHIFT_INSERT(vsrih, 2, DO_SHR, SHR_MASK)
DO_2SHIFT_INSERT(vsriw, 4, DO_SHR, SHR_MASK)
DO_2SHIFT_INSERT(vslib, 1, DO_SHL, SHL_MASK)
DO_2SHIFT_INSERT(vslih, 2, DO_SHL, SHL_MASK)
DO_2SHIFT_INSERT(vsliw, 4, DO_SHL, SHL_MASK)

/*
 * Long shifts taking half-sized inputs from top or bottom of the input
 * vector and producing a double-width result. ESIZE, TYPE are for
 * the input, and LESIZE, LTYPE for the output.
 * Unlike the normal shift helpers, we do not handle negative shift counts,
 * because the long shift is strictly left-only.
 */
#define DO_VSHLL(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE)                   \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd,             \
                                void *vm, uint32_t shift)               \
    {                                                                   \
        LTYPE *d = vd;                                                  \
        TYPE *m = vm;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned le;                                                    \
        assert(shift <= 16);                                            \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) {         \
            LTYPE r = (LTYPE)m[H##ESIZE(le * 2 + TOP)] << shift;        \
            mergemask(&d[H##LESIZE(le)], r, mask);                      \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_VSHLL_ALL(OP, TOP)                                \
    DO_VSHLL(OP##sb, TOP, 1, int8_t, 2, int16_t)             \
    DO_VSHLL(OP##ub, TOP, 1, uint8_t, 2, uint16_t)           \
    DO_VSHLL(OP##sh, TOP, 2, int16_t, 4, int32_t)            \
    DO_VSHLL(OP##uh, TOP, 2, uint16_t, 4, uint32_t)          \

DO_VSHLL_ALL(vshllb, false)
DO_VSHLL_ALL(vshllt, true)

/*
 * Narrowing right shifts, taking a double sized input, shifting it
 * and putting the result in either the top or bottom half of the output.
 * ESIZE, TYPE are the output, and LESIZE, LTYPE the input.
 */
#define DO_VSHRN(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN)       \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd,     \
                                void *vm, uint32_t shift)       \
    {                                                           \
        LTYPE *m = vm;                                          \
        TYPE *d = vd;                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned le;                                            \
        mask >>= ESIZE * TOP;                                   \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) { \
            TYPE r = FN(m[H##LESIZE(le)], shift);               \
            mergemask(&d[H##ESIZE(le * 2 + TOP)], r, mask);     \
        }                                                       \
        mve_advance_vpt(env);                                   \
    }

#define DO_VSHRN_ALL(OP, FN)                                    \
    DO_VSHRN(OP##bb, false, 1, uint8_t, 2, uint16_t, FN)        \
    DO_VSHRN(OP##bh, false, 2, uint16_t, 4, uint32_t, FN)       \
    DO_VSHRN(OP##tb, true, 1, uint8_t, 2, uint16_t, FN)         \
    DO_VSHRN(OP##th, true, 2, uint16_t, 4, uint32_t, FN)

static inline uint64_t do_urshr(uint64_t x, unsigned sh)
{
    if (likely(sh < 64)) {
        return (x >> sh) + ((x >> (sh - 1)) & 1);
    } else if (sh == 64) {
        return x >> 63;
    } else {
        return 0;
    }
}

static inline int64_t do_srshr(int64_t x, unsigned sh)
{
    if (likely(sh < 64)) {
        return (x >> sh) + ((x >> (sh - 1)) & 1);
    } else {
        /* Rounding the sign bit always produces 0. */
        return 0;
    }
}

DO_VSHRN_ALL(vshrn, DO_SHR)
DO_VSHRN_ALL(vrshrn, do_urshr)

static inline int32_t do_sat_bhs(int64_t val, int64_t min, int64_t max,
                                 bool *satp)
{
    if (val > max) {
        *satp = true;
        return max;
    } else if (val < min) {
        *satp = true;
        return min;
    } else {
        return val;
    }
}

/* Saturating narrowing right shifts */
#define DO_VSHRN_SAT(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN)   \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd,     \
                                void *vm, uint32_t shift)       \
    {                                                           \
        LTYPE *m = vm;                                          \
        TYPE *d = vd;                                           \
        uint16_t mask = mve_element_mask(env);                  \
        bool qc = false;                                        \
        unsigned le;                                            \
        mask >>= ESIZE * TOP;                                   \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) { \
            bool sat = false;                                   \
            TYPE r = FN(m[H##LESIZE(le)], shift, &sat);         \
            mergemask(&d[H##ESIZE(le * 2 + TOP)], r, mask);     \
            qc |= sat & mask & 1;                               \
        }                                                       \
        if (qc) {                                               \
            env->vfp.qc[0] = qc;                                \
        }                                                       \
        mve_advance_vpt(env);                                   \
    }

#define DO_VSHRN_SAT_UB(BOP, TOP, FN)                           \
    DO_VSHRN_SAT(BOP, false, 1, uint8_t, 2, uint16_t, FN)       \
    DO_VSHRN_SAT(TOP, true, 1, uint8_t, 2, uint16_t, FN)

#define DO_VSHRN_SAT_UH(BOP, TOP, FN)                           \
    DO_VSHRN_SAT(BOP, false, 2, uint16_t, 4, uint32_t, FN)      \
    DO_VSHRN_SAT(TOP, true, 2, uint16_t, 4, uint32_t, FN)

#define DO_VSHRN_SAT_SB(BOP, TOP, FN)                           \
    DO_VSHRN_SAT(BOP, false, 1, int8_t, 2, int16_t, FN)         \
    DO_VSHRN_SAT(TOP, true, 1, int8_t, 2, int16_t, FN)

#define DO_VSHRN_SAT_SH(BOP, TOP, FN)                           \
    DO_VSHRN_SAT(BOP, false, 2, int16_t, 4, int32_t, FN)        \
    DO_VSHRN_SAT(TOP, true, 2, int16_t, 4, int32_t, FN)

#define DO_SHRN_SB(N, M, SATP)                                  \
    do_sat_bhs((int64_t)(N) >> (M), INT8_MIN, INT8_MAX, SATP)
#define DO_SHRN_UB(N, M, SATP)                                  \
    do_sat_bhs((uint64_t)(N) >> (M), 0, UINT8_MAX, SATP)
#define DO_SHRUN_B(N, M, SATP)                                  \
    do_sat_bhs((int64_t)(N) >> (M), 0, UINT8_MAX, SATP)

#define DO_SHRN_SH(N, M, SATP)                                  \
    do_sat_bhs((int64_t)(N) >> (M), INT16_MIN, INT16_MAX, SATP)
#define DO_SHRN_UH(N, M, SATP)                                  \
    do_sat_bhs((uint64_t)(N) >> (M), 0, UINT16_MAX, SATP)
#define DO_SHRUN_H(N, M, SATP)                                  \
    do_sat_bhs((int64_t)(N) >> (M), 0, UINT16_MAX, SATP)

#define DO_RSHRN_SB(N, M, SATP)                                 \
    do_sat_bhs(do_srshr(N, M), INT8_MIN, INT8_MAX, SATP)
#define DO_RSHRN_UB(N, M, SATP)                                 \
    do_sat_bhs(do_urshr(N, M), 0, UINT8_MAX, SATP)
#define DO_RSHRUN_B(N, M, SATP)                                 \
    do_sat_bhs(do_srshr(N, M), 0, UINT8_MAX, SATP)

#define DO_RSHRN_SH(N, M, SATP)                                 \
    do_sat_bhs(do_srshr(N, M), INT16_MIN, INT16_MAX, SATP)
#define DO_RSHRN_UH(N, M, SATP)                                 \
    do_sat_bhs(do_urshr(N, M), 0, UINT16_MAX, SATP)
#define DO_RSHRUN_H(N, M, SATP)                                 \
    do_sat_bhs(do_srshr(N, M), 0, UINT16_MAX, SATP)

DO_VSHRN_SAT_SB(vqshrnb_sb, vqshrnt_sb, DO_SHRN_SB)
DO_VSHRN_SAT_SH(vqshrnb_sh, vqshrnt_sh, DO_SHRN_SH)
DO_VSHRN_SAT_UB(vqshrnb_ub, vqshrnt_ub, DO_SHRN_UB)
DO_VSHRN_SAT_UH(vqshrnb_uh, vqshrnt_uh, DO_SHRN_UH)
DO_VSHRN_SAT_SB(vqshrunbb, vqshruntb, DO_SHRUN_B)
DO_VSHRN_SAT_SH(vqshrunbh, vqshrunth, DO_SHRUN_H)

DO_VSHRN_SAT_SB(vqrshrnb_sb, vqrshrnt_sb, DO_RSHRN_SB)
DO_VSHRN_SAT_SH(vqrshrnb_sh, vqrshrnt_sh, DO_RSHRN_SH)
DO_VSHRN_SAT_UB(vqrshrnb_ub, vqrshrnt_ub, DO_RSHRN_UB)
DO_VSHRN_SAT_UH(vqrshrnb_uh, vqrshrnt_uh, DO_RSHRN_UH)
DO_VSHRN_SAT_SB(vqrshrunbb, vqrshruntb, DO_RSHRUN_B)
DO_VSHRN_SAT_SH(vqrshrunbh, vqrshrunth, DO_RSHRUN_H)

#define DO_VMOVN(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE)                   \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm)         \
    {                                                                   \
        LTYPE *m = vm;                                                  \
        TYPE *d = vd;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned le;                                                    \
        mask >>= ESIZE * TOP;                                           \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) {         \
            mergemask(&d[H##ESIZE(le * 2 + TOP)],                       \
                      m[H##LESIZE(le)], mask);                          \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_VMOVN(vmovnbb, false, 1, uint8_t, 2, uint16_t)
DO_VMOVN(vmovnbh, false, 2, uint16_t, 4, uint32_t)
DO_VMOVN(vmovntb, true, 1, uint8_t, 2, uint16_t)
DO_VMOVN(vmovnth, true, 2, uint16_t, 4, uint32_t)

#define DO_VMOVN_SAT(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN)           \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm)         \
    {                                                                   \
        LTYPE *m = vm;                                                  \
        TYPE *d = vd;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        bool qc = false;                                                \
        unsigned le;                                                    \
        mask >>= ESIZE * TOP;                                           \
        for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) {         \
            bool sat = false;                                           \
            TYPE r = FN(m[H##LESIZE(le)], &sat);                        \
            mergemask(&d[H##ESIZE(le * 2 + TOP)], r, mask);             \
            qc |= sat & mask & 1;                                       \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_VMOVN_SAT_UB(BOP, TOP, FN)                           \
    DO_VMOVN_SAT(BOP, false, 1, uint8_t, 2, uint16_t, FN)       \
    DO_VMOVN_SAT(TOP, true, 1, uint8_t, 2, uint16_t, FN)

#define DO_VMOVN_SAT_UH(BOP, TOP, FN)                           \
    DO_VMOVN_SAT(BOP, false, 2, uint16_t, 4, uint32_t, FN)      \
    DO_VMOVN_SAT(TOP, true, 2, uint16_t, 4, uint32_t, FN)

#define DO_VMOVN_SAT_SB(BOP, TOP, FN)                           \
    DO_VMOVN_SAT(BOP, false, 1, int8_t, 2, int16_t, FN)         \
    DO_VMOVN_SAT(TOP, true, 1, int8_t, 2, int16_t, FN)

#define DO_VMOVN_SAT_SH(BOP, TOP, FN)                           \
    DO_VMOVN_SAT(BOP, false, 2, int16_t, 4, int32_t, FN)        \
    DO_VMOVN_SAT(TOP, true, 2, int16_t, 4, int32_t, FN)

#define DO_VQMOVN_SB(N, SATP)                           \
    do_sat_bhs((int64_t)(N), INT8_MIN, INT8_MAX, SATP)
#define DO_VQMOVN_UB(N, SATP)                           \
    do_sat_bhs((uint64_t)(N), 0, UINT8_MAX, SATP)
#define DO_VQMOVUN_B(N, SATP)                           \
    do_sat_bhs((int64_t)(N), 0, UINT8_MAX, SATP)

#define DO_VQMOVN_SH(N, SATP)                           \
    do_sat_bhs((int64_t)(N), INT16_MIN, INT16_MAX, SATP)
#define DO_VQMOVN_UH(N, SATP)                           \
    do_sat_bhs((uint64_t)(N), 0, UINT16_MAX, SATP)
#define DO_VQMOVUN_H(N, SATP)                           \
    do_sat_bhs((int64_t)(N), 0, UINT16_MAX, SATP)

DO_VMOVN_SAT_SB(vqmovnbsb, vqmovntsb, DO_VQMOVN_SB)
DO_VMOVN_SAT_SH(vqmovnbsh, vqmovntsh, DO_VQMOVN_SH)
DO_VMOVN_SAT_UB(vqmovnbub, vqmovntub, DO_VQMOVN_UB)
DO_VMOVN_SAT_UH(vqmovnbuh, vqmovntuh, DO_VQMOVN_UH)
DO_VMOVN_SAT_SB(vqmovunbb, vqmovuntb, DO_VQMOVUN_B)
DO_VMOVN_SAT_SH(vqmovunbh, vqmovunth, DO_VQMOVUN_H)

uint32_t HELPER(mve_vshlc)(CPUARMState *env, void *vd, uint32_t rdm,
                           uint32_t shift)
{
    uint32_t *d = vd;
    uint16_t mask = mve_element_mask(env);
    unsigned e;
    uint32_t r;

    /*
     * For each 32-bit element, we shift it left, bringing in the
     * low 'shift' bits of rdm at the bottom. Bits shifted out at
     * the top become the new rdm, if the predicate mask permits.
     * The final rdm value is returned to update the register.
     * shift == 0 here means "shift by 32 bits".
     */
    if (shift == 0) {
        for (e = 0; e < 16 / 4; e++, mask >>= 4) {
            r = rdm;
            if (mask & 1) {
                rdm = d[H4(e)];
            }
            mergemask(&d[H4(e)], r, mask);
        }
    } else {
        uint32_t shiftmask = MAKE_64BIT_MASK(0, shift);

        for (e = 0; e < 16 / 4; e++, mask >>= 4) {
            r = (d[H4(e)] << shift) | (rdm & shiftmask);
            if (mask & 1) {
                rdm = d[H4(e)] >> (32 - shift);
            }
            mergemask(&d[H4(e)], r, mask);
        }
    }
    mve_advance_vpt(env);
    return rdm;
}

uint64_t HELPER(mve_sshrl)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_sqrshl_d(n, -(int8_t)shift, false, NULL);
}

uint64_t HELPER(mve_ushll)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_uqrshl_d(n, (int8_t)shift, false, NULL);
}

uint64_t HELPER(mve_sqshll)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_sqrshl_d(n, (int8_t)shift, false, &env->QF);
}

uint64_t HELPER(mve_uqshll)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_uqrshl_d(n, (int8_t)shift, false, &env->QF);
}

uint64_t HELPER(mve_sqrshrl)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_sqrshl_d(n, -(int8_t)shift, true, &env->QF);
}

uint64_t HELPER(mve_uqrshll)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_uqrshl_d(n, (int8_t)shift, true, &env->QF);
}

/* Operate on 64-bit values, but saturate at 48 bits */
static inline int64_t do_sqrshl48_d(int64_t src, int64_t shift,
                                    bool round, uint32_t *sat)
{
    int64_t val, extval;

    if (shift <= -48) {
        /* Rounding the sign bit always produces 0. */
        if (round) {
            return 0;
        }
        return src >> 63;
    } else if (shift < 0) {
        if (round) {
            src >>= -shift - 1;
            val = (src >> 1) + (src & 1);
        } else {
            val = src >> -shift;
        }
        extval = sextract64(val, 0, 48);
        if (!sat || val == extval) {
            return extval;
        }
    } else if (shift < 48) {
        extval = sextract64(src << shift, 0, 48);
        if (!sat || src == (extval >> shift)) {
            return extval;
        }
    } else if (!sat || src == 0) {
        return 0;
    }

    *sat = 1;
    return src >= 0 ? MAKE_64BIT_MASK(0, 47) : MAKE_64BIT_MASK(47, 17);
}

/* Operate on 64-bit values, but saturate at 48 bits */
static inline uint64_t do_uqrshl48_d(uint64_t src, int64_t shift,
                                     bool round, uint32_t *sat)
{
    uint64_t val, extval;

    if (shift <= -(48 + round)) {
        return 0;
    } else if (shift < 0) {
        if (round) {
            val = src >> (-shift - 1);
            val = (val >> 1) + (val & 1);
        } else {
            val = src >> -shift;
        }
        extval = extract64(val, 0, 48);
        if (!sat || val == extval) {
            return extval;
        }
    } else if (shift < 48) {
        extval = extract64(src << shift, 0, 48);
        if (!sat || src == (extval >> shift)) {
            return extval;
        }
    } else if (!sat || src == 0) {
        return 0;
    }

    *sat = 1;
    return MAKE_64BIT_MASK(0, 48);
}

uint64_t HELPER(mve_sqrshrl48)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_sqrshl48_d(n, -(int8_t)shift, true, &env->QF);
}

uint64_t HELPER(mve_uqrshll48)(CPUARMState *env, uint64_t n, uint32_t shift)
{
    return do_uqrshl48_d(n, (int8_t)shift, true, &env->QF);
}

uint32_t HELPER(mve_uqshl)(CPUARMState *env, uint32_t n, uint32_t shift)
{
    return do_uqrshl_bhs(n, (int8_t)shift, 32, false, &env->QF);
}

uint32_t HELPER(mve_sqshl)(CPUARMState *env, uint32_t n, uint32_t shift)
{
    return do_sqrshl_bhs(n, (int8_t)shift, 32, false, &env->QF);
}

uint32_t HELPER(mve_uqrshl)(CPUARMState *env, uint32_t n, uint32_t shift)
{
    return do_uqrshl_bhs(n, (int8_t)shift, 32, true, &env->QF);
}

uint32_t HELPER(mve_sqrshr)(CPUARMState *env, uint32_t n, uint32_t shift)
{
    return do_sqrshl_bhs(n, -(int8_t)shift, 32, true, &env->QF);
}

#define DO_VIDUP(OP, ESIZE, TYPE, FN)                           \
    uint32_t HELPER(mve_##OP)(CPUARMState *env, void *vd,       \
                           uint32_t offset, uint32_t imm)       \
    {                                                           \
        TYPE *d = vd;                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            mergemask(&d[H##ESIZE(e)], offset, mask);           \
            offset = FN(offset, imm);                           \
        }                                                       \
        mve_advance_vpt(env);                                   \
        return offset;                                          \
    }

#define DO_VIWDUP(OP, ESIZE, TYPE, FN)                          \
    uint32_t HELPER(mve_##OP)(CPUARMState *env, void *vd,       \
                              uint32_t offset, uint32_t wrap,   \
                              uint32_t imm)                     \
    {                                                           \
        TYPE *d = vd;                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            mergemask(&d[H##ESIZE(e)], offset, mask);           \
            offset = FN(offset, wrap, imm);                     \
        }                                                       \
        mve_advance_vpt(env);                                   \
        return offset;                                          \
    }

#define DO_VIDUP_ALL(OP, FN)                    \
    DO_VIDUP(OP##b, 1, int8_t, FN)              \
    DO_VIDUP(OP##h, 2, int16_t, FN)             \
    DO_VIDUP(OP##w, 4, int32_t, FN)

#define DO_VIWDUP_ALL(OP, FN)                   \
    DO_VIWDUP(OP##b, 1, int8_t, FN)             \
    DO_VIWDUP(OP##h, 2, int16_t, FN)            \
    DO_VIWDUP(OP##w, 4, int32_t, FN)

static uint32_t do_add_wrap(uint32_t offset, uint32_t wrap, uint32_t imm)
{
    offset += imm;
    if (offset == wrap) {
        offset = 0;
    }
    return offset;
}

static uint32_t do_sub_wrap(uint32_t offset, uint32_t wrap, uint32_t imm)
{
    if (offset == 0) {
        offset = wrap;
    }
    offset -= imm;
    return offset;
}

DO_VIDUP_ALL(vidup, DO_ADD)
DO_VIWDUP_ALL(viwdup, do_add_wrap)
DO_VIWDUP_ALL(vdwdup, do_sub_wrap)

/*
 * Vector comparison.
 * P0 bits for non-executed beats (where eci_mask is 0) are unchanged.
 * P0 bits for predicated lanes in executed beats (where mask is 0) are 0.
 * P0 bits otherwise are updated with the results of the comparisons.
 * We must also keep unchanged the MASK fields at the top of v7m.vpr.
 */
#define DO_VCMP(OP, ESIZE, TYPE, FN)                                    \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vn, void *vm)   \
    {                                                                   \
        TYPE *n = vn, *m = vm;                                          \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        uint16_t beatpred = 0;                                          \
        uint16_t emask = MAKE_64BIT_MASK(0, ESIZE);                     \
        unsigned e;                                                     \
        for (e = 0; e < 16 / ESIZE; e++) {                              \
            bool r = FN(n[H##ESIZE(e)], m[H##ESIZE(e)]);                \
            /* Comparison sets 0/1 bits for each byte in the element */ \
            beatpred |= r * emask;                                      \
            emask <<= ESIZE;                                            \
        }                                                               \
        beatpred &= mask;                                               \
        env->v7m.vpr = (env->v7m.vpr & ~(uint32_t)eci_mask) |           \
            (beatpred & eci_mask);                                      \
        mve_advance_vpt(env);                                           \
    }

#define DO_VCMP_SCALAR(OP, ESIZE, TYPE, FN)                             \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vn,             \
                                uint32_t rm)                            \
    {                                                                   \
        TYPE *n = vn;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        uint16_t beatpred = 0;                                          \
        uint16_t emask = MAKE_64BIT_MASK(0, ESIZE);                     \
        unsigned e;                                                     \
        for (e = 0; e < 16 / ESIZE; e++) {                              \
            bool r = FN(n[H##ESIZE(e)], (TYPE)rm);                      \
            /* Comparison sets 0/1 bits for each byte in the element */ \
            beatpred |= r * emask;                                      \
            emask <<= ESIZE;                                            \
        }                                                               \
        beatpred &= mask;                                               \
        env->v7m.vpr = (env->v7m.vpr & ~(uint32_t)eci_mask) |           \
            (beatpred & eci_mask);                                      \
        mve_advance_vpt(env);                                           \
    }

#define DO_VCMP_S(OP, FN)                               \
    DO_VCMP(OP##b, 1, int8_t, FN)                       \
    DO_VCMP(OP##h, 2, int16_t, FN)                      \
    DO_VCMP(OP##w, 4, int32_t, FN)                      \
    DO_VCMP_SCALAR(OP##_scalarb, 1, int8_t, FN)         \
    DO_VCMP_SCALAR(OP##_scalarh, 2, int16_t, FN)        \
    DO_VCMP_SCALAR(OP##_scalarw, 4, int32_t, FN)

#define DO_VCMP_U(OP, FN)                               \
    DO_VCMP(OP##b, 1, uint8_t, FN)                      \
    DO_VCMP(OP##h, 2, uint16_t, FN)                     \
    DO_VCMP(OP##w, 4, uint32_t, FN)                     \
    DO_VCMP_SCALAR(OP##_scalarb, 1, uint8_t, FN)        \
    DO_VCMP_SCALAR(OP##_scalarh, 2, uint16_t, FN)       \
    DO_VCMP_SCALAR(OP##_scalarw, 4, uint32_t, FN)

#define DO_EQ(N, M) ((N) == (M))
#define DO_NE(N, M) ((N) != (M))
#define DO_EQ(N, M) ((N) == (M))
#define DO_EQ(N, M) ((N) == (M))
#define DO_GE(N, M) ((N) >= (M))
#define DO_LT(N, M) ((N) < (M))
#define DO_GT(N, M) ((N) > (M))
#define DO_LE(N, M) ((N) <= (M))

DO_VCMP_U(vcmpeq, DO_EQ)
DO_VCMP_U(vcmpne, DO_NE)
DO_VCMP_U(vcmpcs, DO_GE)
DO_VCMP_U(vcmphi, DO_GT)
DO_VCMP_S(vcmpge, DO_GE)
DO_VCMP_S(vcmplt, DO_LT)
DO_VCMP_S(vcmpgt, DO_GT)
DO_VCMP_S(vcmple, DO_LE)

void HELPER(mve_vpsel)(CPUARMState *env, void *vd, void *vn, void *vm)
{
    /*
     * Qd[n] = VPR.P0[n] ? Qn[n] : Qm[n]
     * but note that whether bytes are written to Qd is still subject
     * to (all forms of) predication in the usual way.
     */
    uint64_t *d = vd, *n = vn, *m = vm;
    uint16_t mask = mve_element_mask(env);
    uint16_t p0 = FIELD_EX32(env->v7m.vpr, V7M_VPR, P0);
    unsigned e;
    for (e = 0; e < 16 / 8; e++, mask >>= 8, p0 >>= 8) {
        uint64_t r = m[H8(e)];
        mergemask(&r, n[H8(e)], p0);
        mergemask(&d[H8(e)], r, mask);
    }
    mve_advance_vpt(env);
}

void HELPER(mve_vpnot)(CPUARMState *env)
{
    /*
     * P0 bits for unexecuted beats (where eci_mask is 0) are unchanged.
     * P0 bits for predicated lanes in executed bits (where mask is 0) are 0.
     * P0 bits otherwise are inverted.
     * (This is the same logic as VCMP.)
     * This insn is itself subject to predication and to beat-wise execution,
     * and after it executes VPT state advances in the usual way.
     */
    uint16_t mask = mve_element_mask(env);
    uint16_t eci_mask = mve_eci_mask(env);
    uint16_t beatpred = ~env->v7m.vpr & mask;
    env->v7m.vpr = (env->v7m.vpr & ~(uint32_t)eci_mask) | (beatpred & eci_mask);
    mve_advance_vpt(env);
}

/*
 * VCTP: P0 unexecuted bits unchanged, predicated bits zeroed,
 * otherwise set according to value of Rn. The calculation of
 * newmask here works in the same way as the calculation of the
 * ltpmask in mve_element_mask(), but we have pre-calculated
 * the masklen in the generated code.
 */
void HELPER(mve_vctp)(CPUARMState *env, uint32_t masklen)
{
    uint16_t mask = mve_element_mask(env);
    uint16_t eci_mask = mve_eci_mask(env);
    uint16_t newmask;

    assert(masklen <= 16);
    newmask = masklen ? MAKE_64BIT_MASK(0, masklen) : 0;
    newmask &= mask;
    env->v7m.vpr = (env->v7m.vpr & ~(uint32_t)eci_mask) | (newmask & eci_mask);
    mve_advance_vpt(env);
}

#define DO_1OP_SAT(OP, ESIZE, TYPE, FN)                                 \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm)         \
    {                                                                   \
        TYPE *d = vd, *m = vm;                                          \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        bool qc = false;                                                \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            bool sat = false;                                           \
            mergemask(&d[H##ESIZE(e)], FN(m[H##ESIZE(e)], &sat), mask); \
            qc |= sat & mask & 1;                                       \
        }                                                               \
        if (qc) {                                                       \
            env->vfp.qc[0] = qc;                                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_VQABS_B(N, SATP) \
    do_sat_bhs(DO_ABS((int64_t)N), INT8_MIN, INT8_MAX, SATP)
#define DO_VQABS_H(N, SATP) \
    do_sat_bhs(DO_ABS((int64_t)N), INT16_MIN, INT16_MAX, SATP)
#define DO_VQABS_W(N, SATP) \
    do_sat_bhs(DO_ABS((int64_t)N), INT32_MIN, INT32_MAX, SATP)

#define DO_VQNEG_B(N, SATP) do_sat_bhs(-(int64_t)N, INT8_MIN, INT8_MAX, SATP)
#define DO_VQNEG_H(N, SATP) do_sat_bhs(-(int64_t)N, INT16_MIN, INT16_MAX, SATP)
#define DO_VQNEG_W(N, SATP) do_sat_bhs(-(int64_t)N, INT32_MIN, INT32_MAX, SATP)

DO_1OP_SAT(vqabsb, 1, int8_t, DO_VQABS_B)
DO_1OP_SAT(vqabsh, 2, int16_t, DO_VQABS_H)
DO_1OP_SAT(vqabsw, 4, int32_t, DO_VQABS_W)

DO_1OP_SAT(vqnegb, 1, int8_t, DO_VQNEG_B)
DO_1OP_SAT(vqnegh, 2, int16_t, DO_VQNEG_H)
DO_1OP_SAT(vqnegw, 4, int32_t, DO_VQNEG_W)

/*
 * VMAXA, VMINA: vd is unsigned; vm is signed, and we take its
 * absolute value; we then do an unsigned comparison.
 */
#define DO_VMAXMINA(OP, ESIZE, STYPE, UTYPE, FN)                        \
    void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm)         \
    {                                                                   \
        UTYPE *d = vd;                                                  \
        STYPE *m = vm;                                                  \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            UTYPE r = DO_ABS(m[H##ESIZE(e)]);                           \
            r = FN(d[H##ESIZE(e)], r);                                  \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_VMAXMINA(vmaxab, 1, int8_t, uint8_t, DO_MAX)
DO_VMAXMINA(vmaxah, 2, int16_t, uint16_t, DO_MAX)
DO_VMAXMINA(vmaxaw, 4, int32_t, uint32_t, DO_MAX)
DO_VMAXMINA(vminab, 1, int8_t, uint8_t, DO_MIN)
DO_VMAXMINA(vminah, 2, int16_t, uint16_t, DO_MIN)
DO_VMAXMINA(vminaw, 4, int32_t, uint32_t, DO_MIN)

/*
 * 2-operand floating point. Note that if an element is partially
 * predicated we must do the FP operation to update the non-predicated
 * bytes, but we must be careful to avoid updating the FP exception
 * state unless byte 0 of the element was unpredicated.
 */
#define DO_2OP_FP(OP, ESIZE, TYPE, FN)                                  \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vn, void *vm)           \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        TYPE r;                                                         \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE)) == 0) {              \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & 1)) {                                          \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(n[H##ESIZE(e)], m[H##ESIZE(e)], fpst);               \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_2OP_FP_ALL(OP, FN)                  \
    DO_2OP_FP(OP##h, 2, float16, float16_##FN) \
    DO_2OP_FP(OP##s, 4, float32, float32_##FN)

DO_2OP_FP_ALL(vfadd, add)
DO_2OP_FP_ALL(vfsub, sub)
DO_2OP_FP_ALL(vfmul, mul)

static inline float16 float16_abd(float16 a, float16 b, float_status *s)
{
    return float16_abs(float16_sub(a, b, s));
}

static inline float32 float32_abd(float32 a, float32 b, float_status *s)
{
    return float32_abs(float32_sub(a, b, s));
}

DO_2OP_FP_ALL(vfabd, abd)
DO_2OP_FP_ALL(vmaxnm, maxnum)
DO_2OP_FP_ALL(vminnm, minnum)

static inline float16 float16_maxnuma(float16 a, float16 b, float_status *s)
{
    return float16_maxnum(float16_abs(a), float16_abs(b), s);
}

static inline float32 float32_maxnuma(float32 a, float32 b, float_status *s)
{
    return float32_maxnum(float32_abs(a), float32_abs(b), s);
}

static inline float16 float16_minnuma(float16 a, float16 b, float_status *s)
{
    return float16_minnum(float16_abs(a), float16_abs(b), s);
}

static inline float32 float32_minnuma(float32 a, float32 b, float_status *s)
{
    return float32_minnum(float32_abs(a), float32_abs(b), s);
}

DO_2OP_FP_ALL(vmaxnma, maxnuma)
DO_2OP_FP_ALL(vminnma, minnuma)

#define DO_VCADD_FP(OP, ESIZE, TYPE, FN0, FN1)                          \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vn, void *vm)           \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        TYPE r[16 / ESIZE];                                             \
        uint16_t tm, mask = mve_element_mask(env);                      \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        /* Calculate all results first to avoid overwriting inputs */   \
        for (e = 0, tm = mask; e < 16 / ESIZE; e++, tm >>= ESIZE) {     \
            if ((tm & MAKE_64BIT_MASK(0, ESIZE)) == 0) {                \
                r[e] = 0;                                               \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(tm & 1)) {                                            \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            if (!(e & 1)) {                                             \
                r[e] = FN0(n[H##ESIZE(e)], m[H##ESIZE(e + 1)], fpst);   \
            } else {                                                    \
                r[e] = FN1(n[H##ESIZE(e)], m[H##ESIZE(e - 1)], fpst);   \
            }                                                           \
        }                                                               \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            mergemask(&d[H##ESIZE(e)], r[e], mask);                     \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_VCADD_FP(vfcadd90h, 2, float16, float16_sub, float16_add)
DO_VCADD_FP(vfcadd90s, 4, float32, float32_sub, float32_add)
DO_VCADD_FP(vfcadd270h, 2, float16, float16_add, float16_sub)
DO_VCADD_FP(vfcadd270s, 4, float32, float32_add, float32_sub)

#define DO_VFMA(OP, ESIZE, TYPE, CHS)                                   \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vn, void *vm)           \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        TYPE r;                                                         \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE)) == 0) {              \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & 1)) {                                          \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = n[H##ESIZE(e)];                                         \
            if (CHS) {                                                  \
                r = TYPE##_chs(r);                                      \
            }                                                           \
            r = TYPE##_muladd(r, m[H##ESIZE(e)], d[H##ESIZE(e)],        \
                              0, fpst);                                 \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_VFMA(vfmah, 2, float16, false)
DO_VFMA(vfmas, 4, float32, false)
DO_VFMA(vfmsh, 2, float16, true)
DO_VFMA(vfmss, 4, float32, true)

#define DO_VCMLA(OP, ESIZE, TYPE, ROT, FN)                              \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vn, void *vm)           \
    {                                                                   \
        TYPE *d = vd, *n = vn, *m = vm;                                 \
        TYPE r0, r1, e1, e2, e3, e4;                                    \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst0, *fpst1;                                    \
        float_status scratch_fpst;                                      \
        /* We loop through pairs of elements at a time */               \
        for (e = 0; e < 16 / ESIZE; e += 2, mask >>= ESIZE * 2) {       \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE * 2)) == 0) {          \
                continue;                                               \
            }                                                           \
            fpst0 = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :   \
                &env->vfp.standard_fp_status;                           \
            fpst1 = fpst0;                                              \
            if (!(mask & 1)) {                                          \
                scratch_fpst = *fpst0;                                  \
                fpst0 = &scratch_fpst;                                  \
            }                                                           \
            if (!(mask & (1 << ESIZE))) {                               \
                scratch_fpst = *fpst1;                                  \
                fpst1 = &scratch_fpst;                                  \
            }                                                           \
            switch (ROT) {                                              \
            case 0:                                                     \
                e1 = m[H##ESIZE(e)];                                    \
                e2 = n[H##ESIZE(e)];                                    \
                e3 = m[H##ESIZE(e + 1)];                                \
                e4 = n[H##ESIZE(e)];                                    \
                break;                                                  \
            case 1:                                                     \
                e1 = TYPE##_chs(m[H##ESIZE(e + 1)]);                    \
                e2 = n[H##ESIZE(e + 1)];                                \
                e3 = m[H##ESIZE(e)];                                    \
                e4 = n[H##ESIZE(e + 1)];                                \
                break;                                                  \
            case 2:                                                     \
                e1 = TYPE##_chs(m[H##ESIZE(e)]);                        \
                e2 = n[H##ESIZE(e)];                                    \
                e3 = TYPE##_chs(m[H##ESIZE(e + 1)]);                    \
                e4 = n[H##ESIZE(e)];                                    \
                break;                                                  \
            case 3:                                                     \
                e1 = m[H##ESIZE(e + 1)];                                \
                e2 = n[H##ESIZE(e + 1)];                                \
                e3 = TYPE##_chs(m[H##ESIZE(e)]);                        \
                e4 = n[H##ESIZE(e + 1)];                                \
                break;                                                  \
            default:                                                    \
                g_assert_not_reached();                                 \
            }                                                           \
            r0 = FN(e2, e1, d[H##ESIZE(e)], fpst0);                     \
            r1 = FN(e4, e3, d[H##ESIZE(e + 1)], fpst1);                 \
            mergemask(&d[H##ESIZE(e)], r0, mask);                       \
            mergemask(&d[H##ESIZE(e + 1)], r1, mask >> ESIZE);          \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_VCMULH(N, M, D, S) float16_mul(N, M, S)
#define DO_VCMULS(N, M, D, S) float32_mul(N, M, S)

#define DO_VCMLAH(N, M, D, S) float16_muladd(N, M, D, 0, S)
#define DO_VCMLAS(N, M, D, S) float32_muladd(N, M, D, 0, S)

DO_VCMLA(vcmul0h, 2, float16, 0, DO_VCMULH)
DO_VCMLA(vcmul0s, 4, float32, 0, DO_VCMULS)
DO_VCMLA(vcmul90h, 2, float16, 1, DO_VCMULH)
DO_VCMLA(vcmul90s, 4, float32, 1, DO_VCMULS)
DO_VCMLA(vcmul180h, 2, float16, 2, DO_VCMULH)
DO_VCMLA(vcmul180s, 4, float32, 2, DO_VCMULS)
DO_VCMLA(vcmul270h, 2, float16, 3, DO_VCMULH)
DO_VCMLA(vcmul270s, 4, float32, 3, DO_VCMULS)

DO_VCMLA(vcmla0h, 2, float16, 0, DO_VCMLAH)
DO_VCMLA(vcmla0s, 4, float32, 0, DO_VCMLAS)
DO_VCMLA(vcmla90h, 2, float16, 1, DO_VCMLAH)
DO_VCMLA(vcmla90s, 4, float32, 1, DO_VCMLAS)
DO_VCMLA(vcmla180h, 2, float16, 2, DO_VCMLAH)
DO_VCMLA(vcmla180s, 4, float32, 2, DO_VCMLAS)
DO_VCMLA(vcmla270h, 2, float16, 3, DO_VCMLAH)
DO_VCMLA(vcmla270s, 4, float32, 3, DO_VCMLAS)

#define DO_2OP_FP_SCALAR(OP, ESIZE, TYPE, FN)                           \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vn, uint32_t rm)        \
    {                                                                   \
        TYPE *d = vd, *n = vn;                                          \
        TYPE r, m = rm;                                                 \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE)) == 0) {              \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & 1)) {                                          \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(n[H##ESIZE(e)], m, fpst);                            \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

#define DO_2OP_FP_SCALAR_ALL(OP, FN)                    \
    DO_2OP_FP_SCALAR(OP##h, 2, float16, float16_##FN)   \
    DO_2OP_FP_SCALAR(OP##s, 4, float32, float32_##FN)

DO_2OP_FP_SCALAR_ALL(vfadd_scalar, add)
DO_2OP_FP_SCALAR_ALL(vfsub_scalar, sub)
DO_2OP_FP_SCALAR_ALL(vfmul_scalar, mul)

#define DO_2OP_FP_ACC_SCALAR(OP, ESIZE, TYPE, FN)                       \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vn, uint32_t rm)        \
    {                                                                   \
        TYPE *d = vd, *n = vn;                                          \
        TYPE r, m = rm;                                                 \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE)) == 0) {              \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & 1)) {                                          \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(n[H##ESIZE(e)], m, d[H##ESIZE(e)], 0, fpst);         \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

/* VFMAS is vector * vector + scalar, so swap op2 and op3 */
#define DO_VFMAS_SCALARH(N, M, D, F, S) float16_muladd(N, D, M, F, S)
#define DO_VFMAS_SCALARS(N, M, D, F, S) float32_muladd(N, D, M, F, S)

/* VFMA is vector * scalar + vector */
DO_2OP_FP_ACC_SCALAR(vfma_scalarh, 2, float16, float16_muladd)
DO_2OP_FP_ACC_SCALAR(vfma_scalars, 4, float32, float32_muladd)
DO_2OP_FP_ACC_SCALAR(vfmas_scalarh, 2, float16, DO_VFMAS_SCALARH)
DO_2OP_FP_ACC_SCALAR(vfmas_scalars, 4, float32, DO_VFMAS_SCALARS)

/* Floating point max/min across vector. */
#define DO_FP_VMAXMINV(OP, ESIZE, TYPE, ABS, FN)                \
    uint32_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vm, \
                                    uint32_t ra_in)             \
    {                                                           \
        uint16_t mask = mve_element_mask(env);                  \
        unsigned e;                                             \
        TYPE *m = vm;                                           \
        TYPE ra = (TYPE)ra_in;                                  \
        float_status *fpst = (ESIZE == 2) ?                     \
            &env->vfp.standard_fp_status_f16 :                  \
            &env->vfp.standard_fp_status;                       \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {      \
            if (mask & 1) {                                     \
                TYPE v = m[H##ESIZE(e)];                        \
                if (TYPE##_is_signaling_nan(ra, fpst)) {        \
                    ra = TYPE##_silence_nan(ra, fpst);          \
                    float_raise(float_flag_invalid, fpst);      \
                }                                               \
                if (TYPE##_is_signaling_nan(v, fpst)) {         \
                    v = TYPE##_silence_nan(v, fpst);            \
                    float_raise(float_flag_invalid, fpst);      \
                }                                               \
                if (ABS) {                                      \
                    v = TYPE##_abs(v);                          \
                }                                               \
                ra = FN(ra, v, fpst);                           \
            }                                                   \
        }                                                       \
        mve_advance_vpt(env);                                   \
        return ra;                                              \
    }                                                           \

#define NOP(X) (X)

DO_FP_VMAXMINV(vmaxnmvh, 2, float16, false, float16_maxnum)
DO_FP_VMAXMINV(vmaxnmvs, 4, float32, false, float32_maxnum)
DO_FP_VMAXMINV(vminnmvh, 2, float16, false, float16_minnum)
DO_FP_VMAXMINV(vminnmvs, 4, float32, false, float32_minnum)
DO_FP_VMAXMINV(vmaxnmavh, 2, float16, true, float16_maxnum)
DO_FP_VMAXMINV(vmaxnmavs, 4, float32, true, float32_maxnum)
DO_FP_VMAXMINV(vminnmavh, 2, float16, true, float16_minnum)
DO_FP_VMAXMINV(vminnmavs, 4, float32, true, float32_minnum)

/* FP compares; note that all comparisons signal InvalidOp for QNaNs */
#define DO_VCMP_FP(OP, ESIZE, TYPE, FN)                                 \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vn, void *vm)   \
    {                                                                   \
        TYPE *n = vn, *m = vm;                                          \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        uint16_t beatpred = 0;                                          \
        uint16_t emask = MAKE_64BIT_MASK(0, ESIZE);                     \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        bool r;                                                         \
        for (e = 0; e < 16 / ESIZE; e++, emask <<= ESIZE) {             \
            if ((mask & emask) == 0) {                                  \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & (1 << (e * ESIZE)))) {                         \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(n[H##ESIZE(e)], m[H##ESIZE(e)], fpst);               \
            /* Comparison sets 0/1 bits for each byte in the element */ \
            beatpred |= r * emask;                                      \
        }                                                               \
        beatpred &= mask;                                               \
        env->v7m.vpr = (env->v7m.vpr & ~(uint32_t)eci_mask) |           \
            (beatpred & eci_mask);                                      \
        mve_advance_vpt(env);                                           \
    }

#define DO_VCMP_FP_SCALAR(OP, ESIZE, TYPE, FN)                          \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vn,             \
                                uint32_t rm)                            \
    {                                                                   \
        TYPE *n = vn;                                                   \
        uint16_t mask = mve_element_mask(env);                          \
        uint16_t eci_mask = mve_eci_mask(env);                          \
        uint16_t beatpred = 0;                                          \
        uint16_t emask = MAKE_64BIT_MASK(0, ESIZE);                     \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        bool r;                                                         \
        for (e = 0; e < 16 / ESIZE; e++, emask <<= ESIZE) {             \
            if ((mask & emask) == 0) {                                  \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & (1 << (e * ESIZE)))) {                         \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(n[H##ESIZE(e)], (TYPE)rm, fpst);                     \
            /* Comparison sets 0/1 bits for each byte in the element */ \
            beatpred |= r * emask;                                      \
        }                                                               \
        beatpred &= mask;                                               \
        env->v7m.vpr = (env->v7m.vpr & ~(uint32_t)eci_mask) |           \
            (beatpred & eci_mask);                                      \
        mve_advance_vpt(env);                                           \
    }

#define DO_VCMP_FP_BOTH(VOP, SOP, ESIZE, TYPE, FN)      \
    DO_VCMP_FP(VOP, ESIZE, TYPE, FN)                    \
    DO_VCMP_FP_SCALAR(SOP, ESIZE, TYPE, FN)

/*
 * Some care is needed here to get the correct result for the unordered case.
 * Architecturally EQ, GE and GT are defined to be false for unordered, but
 * the NE, LT and LE comparisons are defined as simple logical inverses of
 * EQ, GE and GT and so they must return true for unordered. The softfloat
 * comparison functions float*_{eq,le,lt} all return false for unordered.
 */
#define DO_GE16(X, Y, S) float16_le(Y, X, S)
#define DO_GE32(X, Y, S) float32_le(Y, X, S)
#define DO_GT16(X, Y, S) float16_lt(Y, X, S)
#define DO_GT32(X, Y, S) float32_lt(Y, X, S)

DO_VCMP_FP_BOTH(vfcmpeqh, vfcmpeq_scalarh, 2, float16, float16_eq)
DO_VCMP_FP_BOTH(vfcmpeqs, vfcmpeq_scalars, 4, float32, float32_eq)

DO_VCMP_FP_BOTH(vfcmpneh, vfcmpne_scalarh, 2, float16, !float16_eq)
DO_VCMP_FP_BOTH(vfcmpnes, vfcmpne_scalars, 4, float32, !float32_eq)

DO_VCMP_FP_BOTH(vfcmpgeh, vfcmpge_scalarh, 2, float16, DO_GE16)
DO_VCMP_FP_BOTH(vfcmpges, vfcmpge_scalars, 4, float32, DO_GE32)

DO_VCMP_FP_BOTH(vfcmplth, vfcmplt_scalarh, 2, float16, !DO_GE16)
DO_VCMP_FP_BOTH(vfcmplts, vfcmplt_scalars, 4, float32, !DO_GE32)

DO_VCMP_FP_BOTH(vfcmpgth, vfcmpgt_scalarh, 2, float16, DO_GT16)
DO_VCMP_FP_BOTH(vfcmpgts, vfcmpgt_scalars, 4, float32, DO_GT32)

DO_VCMP_FP_BOTH(vfcmpleh, vfcmple_scalarh, 2, float16, !DO_GT16)
DO_VCMP_FP_BOTH(vfcmples, vfcmple_scalars, 4, float32, !DO_GT32)

#define DO_VCVT_FIXED(OP, ESIZE, TYPE, FN)                              \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vm,   \
                                uint32_t shift)                         \
    {                                                                   \
        TYPE *d = vd, *m = vm;                                          \
        TYPE r;                                                         \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE)) == 0) {              \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & 1)) {                                          \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(m[H##ESIZE(e)], shift, fpst);                        \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_VCVT_FIXED(vcvt_sh, 2, int16_t, helper_vfp_shtoh)
DO_VCVT_FIXED(vcvt_uh, 2, uint16_t, helper_vfp_uhtoh)
DO_VCVT_FIXED(vcvt_hs, 2, int16_t, helper_vfp_toshh_round_to_zero)
DO_VCVT_FIXED(vcvt_hu, 2, uint16_t, helper_vfp_touhh_round_to_zero)
DO_VCVT_FIXED(vcvt_sf, 4, int32_t, helper_vfp_sltos)
DO_VCVT_FIXED(vcvt_uf, 4, uint32_t, helper_vfp_ultos)
DO_VCVT_FIXED(vcvt_fs, 4, int32_t, helper_vfp_tosls_round_to_zero)
DO_VCVT_FIXED(vcvt_fu, 4, uint32_t, helper_vfp_touls_round_to_zero)

/* VCVT with specified rmode */
#define DO_VCVT_RMODE(OP, ESIZE, TYPE, FN)                              \
    void HELPER(glue(mve_, OP))(CPUARMState *env,                       \
                                void *vd, void *vm, uint32_t rmode)     \
    {                                                                   \
        TYPE *d = vd, *m = vm;                                          \
        TYPE r;                                                         \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        float_status *base_fpst = (ESIZE == 2) ?                        \
            &env->vfp.standard_fp_status_f16 :                          \
            &env->vfp.standard_fp_status;                               \
        uint32_t prev_rmode = get_float_rounding_mode(base_fpst);       \
        set_float_rounding_mode(rmode, base_fpst);                      \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE)) == 0) {              \
                continue;                                               \
            }                                                           \
            fpst = base_fpst;                                           \
            if (!(mask & 1)) {                                          \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(m[H##ESIZE(e)], 0, fpst);                            \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        set_float_rounding_mode(prev_rmode, base_fpst);                 \
        mve_advance_vpt(env);                                           \
    }

DO_VCVT_RMODE(vcvt_rm_sh, 2, uint16_t, helper_vfp_toshh)
DO_VCVT_RMODE(vcvt_rm_uh, 2, uint16_t, helper_vfp_touhh)
DO_VCVT_RMODE(vcvt_rm_ss, 4, uint32_t, helper_vfp_tosls)
DO_VCVT_RMODE(vcvt_rm_us, 4, uint32_t, helper_vfp_touls)

#define DO_VRINT_RM_H(M, F, S) helper_rinth(M, S)
#define DO_VRINT_RM_S(M, F, S) helper_rints(M, S)

DO_VCVT_RMODE(vrint_rm_h, 2, uint16_t, DO_VRINT_RM_H)
DO_VCVT_RMODE(vrint_rm_s, 4, uint32_t, DO_VRINT_RM_S)

/*
 * VCVT between halfprec and singleprec. As usual for halfprec
 * conversions, FZ16 is ignored and AHP is observed.
 */
static void do_vcvt_sh(CPUARMState *env, void *vd, void *vm, int top)
{
    uint16_t *d = vd;
    uint32_t *m = vm;
    uint16_t r;
    uint16_t mask = mve_element_mask(env);
    bool ieee = !(env->vfp.xregs[ARM_VFP_FPSCR] & FPCR_AHP);
    unsigned e;
    float_status *fpst;
    float_status scratch_fpst;
    float_status *base_fpst = &env->vfp.standard_fp_status;
    bool old_fz = get_flush_to_zero(base_fpst);
    set_flush_to_zero(false, base_fpst);
    for (e = 0; e < 16 / 4; e++, mask >>= 4) {
        if ((mask & MAKE_64BIT_MASK(0, 4)) == 0) {
            continue;
        }
        fpst = base_fpst;
        if (!(mask & 1)) {
            /* We need the result but without updating flags */
            scratch_fpst = *fpst;
            fpst = &scratch_fpst;
        }
        r = float32_to_float16(m[H4(e)], ieee, fpst);
        mergemask(&d[H2(e * 2 + top)], r, mask >> (top * 2));
    }
    set_flush_to_zero(old_fz, base_fpst);
    mve_advance_vpt(env);
}

static void do_vcvt_hs(CPUARMState *env, void *vd, void *vm, int top)
{
    uint32_t *d = vd;
    uint16_t *m = vm;
    uint32_t r;
    uint16_t mask = mve_element_mask(env);
    bool ieee = !(env->vfp.xregs[ARM_VFP_FPSCR] & FPCR_AHP);
    unsigned e;
    float_status *fpst;
    float_status scratch_fpst;
    float_status *base_fpst = &env->vfp.standard_fp_status;
    bool old_fiz = get_flush_inputs_to_zero(base_fpst);
    set_flush_inputs_to_zero(false, base_fpst);
    for (e = 0; e < 16 / 4; e++, mask >>= 4) {
        if ((mask & MAKE_64BIT_MASK(0, 4)) == 0) {
            continue;
        }
        fpst = base_fpst;
        if (!(mask & (1 << (top * 2)))) {
            /* We need the result but without updating flags */
            scratch_fpst = *fpst;
            fpst = &scratch_fpst;
        }
        r = float16_to_float32(m[H2(e * 2 + top)], ieee, fpst);
        mergemask(&d[H4(e)], r, mask);
    }
    set_flush_inputs_to_zero(old_fiz, base_fpst);
    mve_advance_vpt(env);
}

void HELPER(mve_vcvtb_sh)(CPUARMState *env, void *vd, void *vm)
{
    do_vcvt_sh(env, vd, vm, 0);
}
void HELPER(mve_vcvtt_sh)(CPUARMState *env, void *vd, void *vm)
{
    do_vcvt_sh(env, vd, vm, 1);
}
void HELPER(mve_vcvtb_hs)(CPUARMState *env, void *vd, void *vm)
{
    do_vcvt_hs(env, vd, vm, 0);
}
void HELPER(mve_vcvtt_hs)(CPUARMState *env, void *vd, void *vm)
{
    do_vcvt_hs(env, vd, vm, 1);
}

#define DO_1OP_FP(OP, ESIZE, TYPE, FN)                                  \
    void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vm)   \
    {                                                                   \
        TYPE *d = vd, *m = vm;                                          \
        TYPE r;                                                         \
        uint16_t mask = mve_element_mask(env);                          \
        unsigned e;                                                     \
        float_status *fpst;                                             \
        float_status scratch_fpst;                                      \
        for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) {              \
            if ((mask & MAKE_64BIT_MASK(0, ESIZE)) == 0) {              \
                continue;                                               \
            }                                                           \
            fpst = (ESIZE == 2) ? &env->vfp.standard_fp_status_f16 :    \
                &env->vfp.standard_fp_status;                           \
            if (!(mask & 1)) {                                          \
                /* We need the result but without updating flags */     \
                scratch_fpst = *fpst;                                   \
                fpst = &scratch_fpst;                                   \
            }                                                           \
            r = FN(m[H##ESIZE(e)], fpst);                               \
            mergemask(&d[H##ESIZE(e)], r, mask);                        \
        }                                                               \
        mve_advance_vpt(env);                                           \
    }

DO_1OP_FP(vrintx_h, 2, float16, float16_round_to_int)
DO_1OP_FP(vrintx_s, 4, float32, float32_round_to_int)