// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * High speed xor_block operation for RAID4/5 utilizing the
 * ldd/std SPARC instructions.
 *
 * Copyright (C) 1999 Jakub Jelinek (jj@ultra.linux.cz)
 */
#include "xor_impl.h"
#include "xor_arch.h"

static void
sparc_2(unsigned long bytes, unsigned long * __restrict p1,
        const unsigned long * __restrict p2)
{
        int lines = bytes / (sizeof (long)) / 8;

        do {
                __asm__ __volatile__(
                  "ldd [%0 + 0x00], %%g2\n\t"
                  "ldd [%0 + 0x08], %%g4\n\t"
                  "ldd [%0 + 0x10], %%o0\n\t"
                  "ldd [%0 + 0x18], %%o2\n\t"
                  "ldd [%1 + 0x00], %%o4\n\t"
                  "ldd [%1 + 0x08], %%l0\n\t"
                  "ldd [%1 + 0x10], %%l2\n\t"
                  "ldd [%1 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "std %%g2, [%0 + 0x00]\n\t"
                  "std %%g4, [%0 + 0x08]\n\t"
                  "std %%o0, [%0 + 0x10]\n\t"
                  "std %%o2, [%0 + 0x18]\n"
                :
                : "r" (p1), "r" (p2)
                : "g2", "g3", "g4", "g5",
                  "o0", "o1", "o2", "o3", "o4", "o5",
                  "l0", "l1", "l2", "l3", "l4", "l5");
                p1 += 8;
                p2 += 8;
        } while (--lines > 0);
}

static void
sparc_3(unsigned long bytes, unsigned long * __restrict p1,
        const unsigned long * __restrict p2,
        const unsigned long * __restrict p3)
{
        int lines = bytes / (sizeof (long)) / 8;

        do {
                __asm__ __volatile__(
                  "ldd [%0 + 0x00], %%g2\n\t"
                  "ldd [%0 + 0x08], %%g4\n\t"
                  "ldd [%0 + 0x10], %%o0\n\t"
                  "ldd [%0 + 0x18], %%o2\n\t"
                  "ldd [%1 + 0x00], %%o4\n\t"
                  "ldd [%1 + 0x08], %%l0\n\t"
                  "ldd [%1 + 0x10], %%l2\n\t"
                  "ldd [%1 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "ldd [%2 + 0x00], %%o4\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "ldd [%2 + 0x08], %%l0\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "ldd [%2 + 0x10], %%l2\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "ldd [%2 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "std %%g2, [%0 + 0x00]\n\t"
                  "std %%g4, [%0 + 0x08]\n\t"
                  "std %%o0, [%0 + 0x10]\n\t"
                  "std %%o2, [%0 + 0x18]\n"
                :
                : "r" (p1), "r" (p2), "r" (p3)
                : "g2", "g3", "g4", "g5",
                  "o0", "o1", "o2", "o3", "o4", "o5",
                  "l0", "l1", "l2", "l3", "l4", "l5");
                p1 += 8;
                p2 += 8;
                p3 += 8;
        } while (--lines > 0);
}

static void
sparc_4(unsigned long bytes, unsigned long * __restrict p1,
        const unsigned long * __restrict p2,
        const unsigned long * __restrict p3,
        const unsigned long * __restrict p4)
{
        int lines = bytes / (sizeof (long)) / 8;

        do {
                __asm__ __volatile__(
                  "ldd [%0 + 0x00], %%g2\n\t"
                  "ldd [%0 + 0x08], %%g4\n\t"
                  "ldd [%0 + 0x10], %%o0\n\t"
                  "ldd [%0 + 0x18], %%o2\n\t"
                  "ldd [%1 + 0x00], %%o4\n\t"
                  "ldd [%1 + 0x08], %%l0\n\t"
                  "ldd [%1 + 0x10], %%l2\n\t"
                  "ldd [%1 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "ldd [%2 + 0x00], %%o4\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "ldd [%2 + 0x08], %%l0\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "ldd [%2 + 0x10], %%l2\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "ldd [%2 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "ldd [%3 + 0x00], %%o4\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "ldd [%3 + 0x08], %%l0\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "ldd [%3 + 0x10], %%l2\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "ldd [%3 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "std %%g2, [%0 + 0x00]\n\t"
                  "std %%g4, [%0 + 0x08]\n\t"
                  "std %%o0, [%0 + 0x10]\n\t"
                  "std %%o2, [%0 + 0x18]\n"
                :
                : "r" (p1), "r" (p2), "r" (p3), "r" (p4)
                : "g2", "g3", "g4", "g5",
                  "o0", "o1", "o2", "o3", "o4", "o5",
                  "l0", "l1", "l2", "l3", "l4", "l5");
                p1 += 8;
                p2 += 8;
                p3 += 8;
                p4 += 8;
        } while (--lines > 0);
}

static void
sparc_5(unsigned long bytes, unsigned long * __restrict p1,
        const unsigned long * __restrict p2,
        const unsigned long * __restrict p3,
        const unsigned long * __restrict p4,
        const unsigned long * __restrict p5)
{
        int lines = bytes / (sizeof (long)) / 8;

        do {
                __asm__ __volatile__(
                  "ldd [%0 + 0x00], %%g2\n\t"
                  "ldd [%0 + 0x08], %%g4\n\t"
                  "ldd [%0 + 0x10], %%o0\n\t"
                  "ldd [%0 + 0x18], %%o2\n\t"
                  "ldd [%1 + 0x00], %%o4\n\t"
                  "ldd [%1 + 0x08], %%l0\n\t"
                  "ldd [%1 + 0x10], %%l2\n\t"
                  "ldd [%1 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "ldd [%2 + 0x00], %%o4\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "ldd [%2 + 0x08], %%l0\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "ldd [%2 + 0x10], %%l2\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "ldd [%2 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "ldd [%3 + 0x00], %%o4\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "ldd [%3 + 0x08], %%l0\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "ldd [%3 + 0x10], %%l2\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "ldd [%3 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "ldd [%4 + 0x00], %%o4\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "ldd [%4 + 0x08], %%l0\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "ldd [%4 + 0x10], %%l2\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "ldd [%4 + 0x18], %%l4\n\t"
                  "xor %%g2, %%o4, %%g2\n\t"
                  "xor %%g3, %%o5, %%g3\n\t"
                  "xor %%g4, %%l0, %%g4\n\t"
                  "xor %%g5, %%l1, %%g5\n\t"
                  "xor %%o0, %%l2, %%o0\n\t"
                  "xor %%o1, %%l3, %%o1\n\t"
                  "xor %%o2, %%l4, %%o2\n\t"
                  "xor %%o3, %%l5, %%o3\n\t"
                  "std %%g2, [%0 + 0x00]\n\t"
                  "std %%g4, [%0 + 0x08]\n\t"
                  "std %%o0, [%0 + 0x10]\n\t"
                  "std %%o2, [%0 + 0x18]\n"
                :
                : "r" (p1), "r" (p2), "r" (p3), "r" (p4), "r" (p5)
                : "g2", "g3", "g4", "g5",
                  "o0", "o1", "o2", "o3", "o4", "o5",
                  "l0", "l1", "l2", "l3", "l4", "l5");
                p1 += 8;
                p2 += 8;
                p3 += 8;
                p4 += 8;
                p5 += 8;
        } while (--lines > 0);
}

DO_XOR_BLOCKS(sparc32, sparc_2, sparc_3, sparc_4, sparc_5);

struct xor_block_template xor_block_SPARC = {
        .name           = "SPARC",
        .xor_gen        = xor_gen_sparc32,
};