private static ref obj.Prog zerorange(ref gc.Progs pp, ref obj.Prog p, long off, long cnt, ref uint r0)
{
if (cnt == 0L)
{
return(p);
}
if (r0 == 0L.Value)
{
p = pp.Appendpp(p, arm.AMOVW, obj.TYPE_CONST, 0L, 0L, obj.TYPE_REG, arm.REG_R0, 0L);
r0.Value = 1L;
}
if (cnt < int64(4L * gc.Widthptr))
{
{
var i = int64(0L);
while (i < cnt)
{
p = pp.Appendpp(p, arm.AMOVW, obj.TYPE_REG, arm.REG_R0, 0L, obj.TYPE_MEM, arm.REGSP, 4L + off + i);
i += int64(gc.Widthptr);
}
}
}
else if (!gc.Nacl && (cnt <= int64(128L * gc.Widthptr)))
{
p = pp.Appendpp(p, arm.AADD, obj.TYPE_CONST, 0L, 4L + off, obj.TYPE_REG, arm.REG_R1, 0L);
p.Reg = arm.REGSP;
p = pp.Appendpp(p, obj.ADUFFZERO, obj.TYPE_NONE, 0L, 0L, obj.TYPE_MEM, 0L, 0L);
p.To.Name = obj.NAME_EXTERN;
p.To.Sym = gc.Duffzero;
p.To.Offset = 4L * (128L - cnt / int64(gc.Widthptr));
}
else
{
p = pp.Appendpp(p, arm.AADD, obj.TYPE_CONST, 0L, 4L + off, obj.TYPE_REG, arm.REG_R1, 0L);
p.Reg = arm.REGSP;
p = pp.Appendpp(p, arm.AADD, obj.TYPE_CONST, 0L, cnt, obj.TYPE_REG, arm.REG_R2, 0L);
p.Reg = arm.REG_R1;
p = pp.Appendpp(p, arm.AMOVW, obj.TYPE_REG, arm.REG_R0, 0L, obj.TYPE_MEM, arm.REG_R1, 4L);
var p1 = p;
p.Scond |= arm.C_PBIT;
p = pp.Appendpp(p, arm.ACMP, obj.TYPE_REG, arm.REG_R1, 0L, obj.TYPE_NONE, 0L, 0L);
p.Reg = arm.REG_R2;
p = pp.Appendpp(p, arm.ABNE, obj.TYPE_NONE, 0L, 0L, obj.TYPE_BRANCH, 0L, 0L);
gc.Patch(p, p1);
}
return(p);
}
private static ref obj.Prog zerorange(ref gc.Progs pp, ref obj.Prog p, long off, long cnt, ref uint state)
{
const long ax = 1L << (int)(iota);
const var x0 = 0;
if (cnt == 0L)
{
return(p);
}
if (cnt % int64(gc.Widthreg) != 0L)
{
// should only happen with nacl
if (cnt % int64(gc.Widthptr) != 0L)
{
gc.Fatalf("zerorange count not a multiple of widthptr %d", cnt);
}
if (state & ax == 0L.Value)
{
p = pp.Appendpp(p, x86.AMOVQ, obj.TYPE_CONST, 0L, 0L, obj.TYPE_REG, x86.REG_AX, 0L);
state.Value |= ax;
}
p = pp.Appendpp(p, x86.AMOVL, obj.TYPE_REG, x86.REG_AX, 0L, obj.TYPE_MEM, x86.REG_SP, off);
off += int64(gc.Widthptr);
cnt -= int64(gc.Widthptr);
}
if (cnt == 8L)
{
if (state & ax == 0L.Value)
{
p = pp.Appendpp(p, x86.AMOVQ, obj.TYPE_CONST, 0L, 0L, obj.TYPE_REG, x86.REG_AX, 0L);
state.Value |= ax;
}
p = pp.Appendpp(p, x86.AMOVQ, obj.TYPE_REG, x86.REG_AX, 0L, obj.TYPE_MEM, x86.REG_SP, off);
}
else if (!isPlan9 && cnt <= int64(8L * gc.Widthreg))
{
if (state & x0 == 0L.Value)
{
p = pp.Appendpp(p, x86.AXORPS, obj.TYPE_REG, x86.REG_X0, 0L, obj.TYPE_REG, x86.REG_X0, 0L);
state.Value |= x0;
}
for (var i = int64(0L); i < cnt / 16L; i++)
{
p = pp.Appendpp(p, x86.AMOVUPS, obj.TYPE_REG, x86.REG_X0, 0L, obj.TYPE_MEM, x86.REG_SP, off + i * 16L);
}
if (cnt % 16L != 0L)
{
p = pp.Appendpp(p, x86.AMOVUPS, obj.TYPE_REG, x86.REG_X0, 0L, obj.TYPE_MEM, x86.REG_SP, off + cnt - int64(16L));
}
}
else if (!gc.Nacl && !isPlan9 && (cnt <= int64(128L * gc.Widthreg)))
{
if (state & x0 == 0L.Value)
{
p = pp.Appendpp(p, x86.AXORPS, obj.TYPE_REG, x86.REG_X0, 0L, obj.TYPE_REG, x86.REG_X0, 0L);
state.Value |= x0;
}
p = pp.Appendpp(p, leaptr, obj.TYPE_MEM, x86.REG_SP, off + dzDI(cnt), obj.TYPE_REG, x86.REG_DI, 0L);
p = pp.Appendpp(p, obj.ADUFFZERO, obj.TYPE_NONE, 0L, 0L, obj.TYPE_ADDR, 0L, dzOff(cnt));
p.To.Sym = gc.Duffzero;
if (cnt % 16L != 0L)
{
p = pp.Appendpp(p, x86.AMOVUPS, obj.TYPE_REG, x86.REG_X0, 0L, obj.TYPE_MEM, x86.REG_DI, -int64(8L));
}
}
else
{
if (state & ax == 0L.Value)
{
p = pp.Appendpp(p, x86.AMOVQ, obj.TYPE_CONST, 0L, 0L, obj.TYPE_REG, x86.REG_AX, 0L);
state.Value |= ax;
}
p = pp.Appendpp(p, x86.AMOVQ, obj.TYPE_CONST, 0L, cnt / int64(gc.Widthreg), obj.TYPE_REG, x86.REG_CX, 0L);
p = pp.Appendpp(p, leaptr, obj.TYPE_MEM, x86.REG_SP, off, obj.TYPE_REG, x86.REG_DI, 0L);
p = pp.Appendpp(p, x86.AREP, obj.TYPE_NONE, 0L, 0L, obj.TYPE_NONE, 0L, 0L);
p = pp.Appendpp(p, x86.ASTOSQ, obj.TYPE_NONE, 0L, 0L, obj.TYPE_NONE, 0L, 0L);
}
return(p);
}