iwmmxt_helper.c

来自「xen虚拟机源代码安装包」· C语言 代码 · 共 683 行 · 第 1/2 页

/*
 * iwMMXt micro operations for XScale.
 *
 * Copyright (c) 2007 OpenedHand, Ltd.
 * Written by Andrzej Zaborowski <[e-mail]>
 * Copyright (c) 2008 CodeSourcery
 *
 * 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 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, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */

#include <stdlib.h>
#include <stdio.h>

#include "cpu.h"
#include "exec-all.h"
#include "helpers.h"

/* iwMMXt macros extracted from GNU gdb.  */

/* Set the SIMD wCASF flags for 8, 16, 32 or 64-bit operations.  */
#define SIMD8_SET( v, n, b)	((v != 0) << ((((b) + 1) * 4) + (n)))
#define SIMD16_SET(v, n, h)	((v != 0) << ((((h) + 1) * 8) + (n)))
#define SIMD32_SET(v, n, w)	((v != 0) << ((((w) + 1) * 16) + (n)))
#define SIMD64_SET(v, n)	((v != 0) << (32 + (n)))
/* Flags to pass as "n" above.  */
#define SIMD_NBIT	-1
#define SIMD_ZBIT	-2
#define SIMD_CBIT	-3
#define SIMD_VBIT	-4
/* Various status bit macros.  */
#define NBIT8(x)	((x) & 0x80)
#define NBIT16(x)	((x) & 0x8000)
#define NBIT32(x)	((x) & 0x80000000)
#define NBIT64(x)	((x) & 0x8000000000000000ULL)
#define ZBIT8(x)	(((x) & 0xff) == 0)
#define ZBIT16(x)	(((x) & 0xffff) == 0)
#define ZBIT32(x)	(((x) & 0xffffffff) == 0)
#define ZBIT64(x)	(x == 0)
/* Sign extension macros.  */
#define EXTEND8H(a)	((uint16_t) (int8_t) (a))
#define EXTEND8(a)	((uint32_t) (int8_t) (a))
#define EXTEND16(a)	((uint32_t) (int16_t) (a))
#define EXTEND16S(a)	((int32_t) (int16_t) (a))
#define EXTEND32(a)	((uint64_t) (int32_t) (a))

uint64_t HELPER(iwmmxt_maddsq)(uint64_t a, uint64_t b)
{
    a = ((
            EXTEND16S((a >> 0) & 0xffff) * EXTEND16S((b >> 0) & 0xffff) +
            EXTEND16S((a >> 16) & 0xffff) * EXTEND16S((b >> 16) & 0xffff)
        ) & 0xffffffff) | ((uint64_t) (
            EXTEND16S((a >> 32) & 0xffff) * EXTEND16S((b >> 32) & 0xffff) +
            EXTEND16S((a >> 48) & 0xffff) * EXTEND16S((b >> 48) & 0xffff)
        ) << 32);
    return a;
}

uint64_t HELPER(iwmmxt_madduq)(uint64_t a, uint64_t b)
{
    a = ((
            ((a >> 0) & 0xffff) * ((b >> 0) & 0xffff) +
            ((a >> 16) & 0xffff) * ((b >> 16) & 0xffff)
        ) & 0xffffffff) | ((
            ((a >> 32) & 0xffff) * ((b >> 32) & 0xffff) +
            ((a >> 48) & 0xffff) * ((b >> 48) & 0xffff)
        ) << 32);
    return a;
}

uint64_t HELPER(iwmmxt_sadb)(uint64_t a, uint64_t b)
{
#define abs(x) (((x) >= 0) ? x : -x)
#define SADB(SHR) abs((int) ((a >> SHR) & 0xff) - (int) ((b >> SHR) & 0xff))
    return
        SADB(0) + SADB(8) + SADB(16) + SADB(24) +
        SADB(32) + SADB(40) + SADB(48) + SADB(56);
#undef SADB
}

uint64_t HELPER(iwmmxt_sadw)(uint64_t a, uint64_t b)
{
#define SADW(SHR) \
    abs((int) ((a >> SHR) & 0xffff) - (int) ((b >> SHR) & 0xffff))
    return SADW(0) + SADW(16) + SADW(32) + SADW(48);
#undef SADW
}

uint64_t HELPER(iwmmxt_mulslw)(uint64_t a, uint64_t b)
{
#define MULS(SHR) ((uint64_t) ((( \
        EXTEND16S((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff) \
    ) >> 0) & 0xffff) << SHR)
    return MULS(0) | MULS(16) | MULS(32) | MULS(48);
#undef MULS
}

uint64_t HELPER(iwmmxt_mulshw)(uint64_t a, uint64_t b)
{
#define MULS(SHR) ((uint64_t) ((( \
        EXTEND16S((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff) \
    ) >> 16) & 0xffff) << SHR)
    return MULS(0) | MULS(16) | MULS(32) | MULS(48);
#undef MULS
}

uint64_t HELPER(iwmmxt_mululw)(uint64_t a, uint64_t b)
{
#define MULU(SHR) ((uint64_t) ((( \
        ((a >> SHR) & 0xffff) * ((b >> SHR) & 0xffff) \
    ) >> 0) & 0xffff) << SHR)
    return MULU(0) | MULU(16) | MULU(32) | MULU(48);
#undef MULU
}

uint64_t HELPER(iwmmxt_muluhw)(uint64_t a, uint64_t b)
{
#define MULU(SHR) ((uint64_t) ((( \
        ((a >> SHR) & 0xffff) * ((b >> SHR) & 0xffff) \
    ) >> 16) & 0xffff) << SHR)
    return MULU(0) | MULU(16) | MULU(32) | MULU(48);
#undef MULU
}

uint64_t HELPER(iwmmxt_macsw)(uint64_t a, uint64_t b)
{
#define MACS(SHR) ( \
        EXTEND16((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff))
    return (int64_t) (MACS(0) + MACS(16) + MACS(32) + MACS(48));
#undef MACS
}

uint64_t HELPER(iwmmxt_macuw)(uint64_t a, uint64_t b)
{
#define MACU(SHR) ( \
        (uint32_t) ((a >> SHR) & 0xffff) * \
        (uint32_t) ((b >> SHR) & 0xffff))
    return MACU(0) + MACU(16) + MACU(32) + MACU(48);
#undef MACU
}

#define NZBIT8(x, i) \
    SIMD8_SET(NBIT8((x) & 0xff), SIMD_NBIT, i) | \
    SIMD8_SET(ZBIT8((x) & 0xff), SIMD_ZBIT, i)
#define NZBIT16(x, i) \
    SIMD16_SET(NBIT16((x) & 0xffff), SIMD_NBIT, i) | \
    SIMD16_SET(ZBIT16((x) & 0xffff), SIMD_ZBIT, i)
#define NZBIT32(x, i) \
    SIMD32_SET(NBIT32((x) & 0xffffffff), SIMD_NBIT, i) | \
    SIMD32_SET(ZBIT32((x) & 0xffffffff), SIMD_ZBIT, i)
#define NZBIT64(x) \
    SIMD64_SET(NBIT64(x), SIMD_NBIT) | \
    SIMD64_SET(ZBIT64(x), SIMD_ZBIT)
#define IWMMXT_OP_UNPACK(S, SH0, SH1, SH2, SH3)			\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, b)))(CPUState *env, \
                                                 uint64_t a, uint64_t b) \
{								\
    a =							        \
        (((a >> SH0) & 0xff) << 0) | (((b >> SH0) & 0xff) << 8) |	\
        (((a >> SH1) & 0xff) << 16) | (((b >> SH1) & 0xff) << 24) |	\
        (((a >> SH2) & 0xff) << 32) | (((b >> SH2) & 0xff) << 40) |	\
        (((a >> SH3) & 0xff) << 48) | (((b >> SH3) & 0xff) << 56);	\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |		        \
        NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |		\
        NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |		\
        NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);		\
    return a;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, w)))(CPUState *env, \
                                        uint64_t a, uint64_t b) \
{								\
    a =							        \
        (((a >> SH0) & 0xffff) << 0) |				\
        (((b >> SH0) & 0xffff) << 16) |			        \
        (((a >> SH2) & 0xffff) << 32) |			        \
        (((b >> SH2) & 0xffff) << 48);				\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT8(a >> 0, 0) | NZBIT8(a >> 16, 1) |		\
        NZBIT8(a >> 32, 2) | NZBIT8(a >> 48, 3);		\
    return a;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, l)))(CPUState *env, \
                                        uint64_t a, uint64_t b) \
{								\
    a =							        \
        (((a >> SH0) & 0xffffffff) << 0) |			\
        (((b >> SH0) & 0xffffffff) << 32);			\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);		\
    return a;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, ub)))(CPUState *env, \
                                                  uint64_t x)   \
{								\
    x =							        \
        (((x >> SH0) & 0xff) << 0) |				\
        (((x >> SH1) & 0xff) << 16) |				\
        (((x >> SH2) & 0xff) << 32) |				\
        (((x >> SH3) & 0xff) << 48);				\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |		\
        NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);		\
    return x;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, uw)))(CPUState *env, \
                                                  uint64_t x)   \
{								\
    x =							        \
        (((x >> SH0) & 0xffff) << 0) |				\
        (((x >> SH2) & 0xffff) << 32);				\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);		\
    return x;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, ul)))(CPUState *env, \
                                                  uint64_t x)   \
{								\
    x = (((x >> SH0) & 0xffffffff) << 0);			\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x >> 0);	\
    return x;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sb)))(CPUState *env, \
                                                  uint64_t x)   \
{								\
    x =							        \
        ((uint64_t) EXTEND8H((x >> SH0) & 0xff) << 0) |	        \
        ((uint64_t) EXTEND8H((x >> SH1) & 0xff) << 16) |	\
        ((uint64_t) EXTEND8H((x >> SH2) & 0xff) << 32) |	\
        ((uint64_t) EXTEND8H((x >> SH3) & 0xff) << 48);	        \
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |		\
        NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);		\
    return x;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sw)))(CPUState *env, \
                                                  uint64_t x)   \
{								\
    x =							        \
        ((uint64_t) EXTEND16((x >> SH0) & 0xffff) << 0) |	\
        ((uint64_t) EXTEND16((x >> SH2) & 0xffff) << 32);	\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);		\
    return x;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sl)))(CPUState *env, \
                                                  uint64_t x)   \
{								\
    x = EXTEND32((x >> SH0) & 0xffffffff);			\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x >> 0);	\
    return x;                                                   \
}
IWMMXT_OP_UNPACK(l, 0, 8, 16, 24)
IWMMXT_OP_UNPACK(h, 32, 40, 48, 56)

#define IWMMXT_OP_CMP(SUFF, Tb, Tw, Tl, O)			\
uint64_t HELPER(glue(iwmmxt_, glue(SUFF, b)))(CPUState *env,    \
                                        uint64_t a, uint64_t b) \
{								\
    a =							        \
        CMP(0, Tb, O, 0xff) | CMP(8, Tb, O, 0xff) |		\
        CMP(16, Tb, O, 0xff) | CMP(24, Tb, O, 0xff) |		\
        CMP(32, Tb, O, 0xff) | CMP(40, Tb, O, 0xff) |		\
        CMP(48, Tb, O, 0xff) | CMP(56, Tb, O, 0xff);		\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |		        \
        NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |		\
        NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |		\
        NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);		\
    return a;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_, glue(SUFF, w)))(CPUState *env,    \
                                        uint64_t a, uint64_t b) \
{								\
    a = CMP(0, Tw, O, 0xffff) | CMP(16, Tw, O, 0xffff) |	\
        CMP(32, Tw, O, 0xffff) | CMP(48, Tw, O, 0xffff);	\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT16(a >> 0, 0) | NZBIT16(a >> 16, 1) |		\
        NZBIT16(a >> 32, 2) | NZBIT16(a >> 48, 3);		\
    return a;                                                   \
}								\
uint64_t HELPER(glue(iwmmxt_, glue(SUFF, l)))(CPUState *env,    \
                                        uint64_t a, uint64_t b) \
{								\
    a = CMP(0, Tl, O, 0xffffffff) |				\
        CMP(32, Tl, O, 0xffffffff);				\
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =			\
        NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);		\
    return a;                                                   \
}
#define CMP(SHR, TYPE, OPER, MASK) ((((TYPE) ((a >> SHR) & MASK) OPER \
            (TYPE) ((b >> SHR) & MASK)) ? (uint64_t) MASK : 0) << SHR)
IWMMXT_OP_CMP(cmpeq, uint8_t, uint16_t, uint32_t, ==)
IWMMXT_OP_CMP(cmpgts, int8_t, int16_t, int32_t, >)
IWMMXT_OP_CMP(cmpgtu, uint8_t, uint16_t, uint32_t, >)
#undef CMP
#define CMP(SHR, TYPE, OPER, MASK) ((((TYPE) ((a >> SHR) & MASK) OPER \
            (TYPE) ((b >> SHR) & MASK)) ? a : b) & ((uint64_t) MASK << SHR))
IWMMXT_OP_CMP(mins, int8_t, int16_t, int32_t, <)
IWMMXT_OP_CMP(minu, uint8_t, uint16_t, uint32_t, <)
IWMMXT_OP_CMP(maxs, int8_t, int16_t, int32_t, >)
IWMMXT_OP_CMP(maxu, uint8_t, uint16_t, uint32_t, >)
#undef CMP
#define CMP(SHR, TYPE, OPER, MASK) ((uint64_t) (((TYPE) ((a >> SHR) & MASK) \
            OPER (TYPE) ((b >> SHR) & MASK)) & MASK) << SHR)
IWMMXT_OP_CMP(subn, uint8_t, uint16_t, uint32_t, -)
IWMMXT_OP_CMP(addn, uint8_t, uint16_t, uint32_t, +)
#undef CMP
/* TODO Signed- and Unsigned-Saturation */
#define CMP(SHR, TYPE, OPER, MASK) ((uint64_t) (((TYPE) ((a >> SHR) & MASK) \
            OPER (TYPE) ((b >> SHR) & MASK)) & MASK) << SHR)
IWMMXT_OP_CMP(subu, uint8_t, uint16_t, uint32_t, -)
IWMMXT_OP_CMP(addu, uint8_t, uint16_t, uint32_t, +)
IWMMXT_OP_CMP(subs, int8_t, int16_t, int32_t, -)
IWMMXT_OP_CMP(adds, int8_t, int16_t, int32_t, +)
#undef CMP
#undef IWMMXT_OP_CMP

#define AVGB(SHR) ((( \
        ((a >> SHR) & 0xff) + ((b >> SHR) & 0xff) + round) >> 1) << SHR)
#define IWMMXT_OP_AVGB(r)                                                 \
uint64_t HELPER(iwmmxt_avgb##r)(CPUState *env, uint64_t a, uint64_t b)    \
{                                                                         \
    const int round = r;                                                  \
    a = AVGB(0) | AVGB(8) | AVGB(16) | AVGB(24) |                         \
        AVGB(32) | AVGB(40) | AVGB(48) | AVGB(56);                        \
    env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                                 \
        SIMD8_SET(ZBIT8((a >> 0) & 0xff), SIMD_ZBIT, 0) |                 \
        SIMD8_SET(ZBIT8((a >> 8) & 0xff), SIMD_ZBIT, 1) |                 \
        SIMD8_SET(ZBIT8((a >> 16) & 0xff), SIMD_ZBIT, 2) |                \
        SIMD8_SET(ZBIT8((a >> 24) & 0xff), SIMD_ZBIT, 3) |                \