On Thu, Jul 26, 2018 at 11:40:20AM +1000, David Gibson wrote:
> On Tue, Jul 24, 2018 at 12:13:04PM +0000, Yasmin Beatriz wrote:
> > After solving a corner case in bcdsub, this patch simplifies the logic
> > of both bcdadd/sub instructions by removing some unnecessary local flags.
> >
> > Signed-off-by: Yasmin Beatriz <yasm...@linux.ibm.com>
> > ---
> > target/ppc/int_helper.c | 33 +++++++++------------------------
> > 1 file changed, 9 insertions(+), 24 deletions(-)
> >
> > diff --git a/target/ppc/int_helper.c b/target/ppc/int_helper.c
> > index fa18e6e..b8ac4bb 100644
> > --- a/target/ppc/int_helper.c
> > +++ b/target/ppc/int_helper.c
> > @@ -2671,16 +2671,14 @@ static int bcd_cmp_mag(ppc_avr_t *a, ppc_avr_t *b)
> > return 0;
> > }
> >
> > -static int bcd_add_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int
> > *invalid,
> > +static void bcd_add_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int
> > *invalid,
> > int *overflow)
> > {
> > int carry = 0;
> > int i;
> > - int is_zero = 1;
> > for (i = 1; i <= 31; i++) {
> > uint8_t digit = bcd_get_digit(a, i, invalid) +
> > bcd_get_digit(b, i, invalid) + carry;
> > - is_zero &= (digit == 0);
> > if (digit > 9) {
> > carry = 1;
> > digit -= 10;
> > @@ -2689,26 +2687,20 @@ static int bcd_add_mag(ppc_avr_t *t, ppc_avr_t *a,
> > ppc_avr_t *b, int *invalid,
> > }
> >
> > bcd_put_digit(t, digit, i);
> > -
> > - if (unlikely(*invalid)) {
> > - return -1;
> > - }
> > }
> >
> > *overflow = carry;
> > - return is_zero;
> > }
> >
> > -static int bcd_sub_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int
> > *invalid,
> > +static void bcd_sub_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int
> > *invalid,
> > int *overflow)
> > {
> > int carry = 0;
> > int i;
> > - int is_zero = 1;
> > +
> > for (i = 1; i <= 31; i++) {
> > uint8_t digit = bcd_get_digit(a, i, invalid) -
> > bcd_get_digit(b, i, invalid) + carry;
> > - is_zero &= (digit == 0);
> > if (digit & 0x80) {
> > carry = -1;
> > digit += 10;
> > @@ -2717,14 +2709,9 @@ static int bcd_sub_mag(ppc_avr_t *t, ppc_avr_t *a,
> > ppc_avr_t *b, int *invalid,
> > }
> >
> > bcd_put_digit(t, digit, i);
> > -
> > - if (unlikely(*invalid)) {
> > - return -1;
> > - }
> > }
> >
> > *overflow = carry;
> > - return is_zero;
> > }
> >
> > uint32_t helper_bcdadd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t
> > ps)
> > @@ -2734,25 +2721,25 @@ uint32_t helper_bcdadd(ppc_avr_t *r, ppc_avr_t *a,
> > ppc_avr_t *b, uint32_t ps)
> > int sgnb = bcd_get_sgn(b);
> > int invalid = (sgna == 0) || (sgnb == 0);
> > int overflow = 0;
> > - int zero = 0;
> > uint32_t cr = 0;
> > ppc_avr_t result = { .u64 = { 0, 0 } };
> >
> > if (!invalid) {
> > if (sgna == sgnb) {
> > result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps);
> > - zero = bcd_add_mag(&result, a, b, &invalid, &overflow);
> > - cr = (sgna > 0) ? CRF_GT : CRF_LT;
> > + bcd_add_mag(&result, a, b, &invalid, &overflow);
> > + cr = bcd_cmp_zero(&result);
> > } else if (bcd_cmp_mag(a, b) > 0) {
> > result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps);
> > - zero = bcd_sub_mag(&result, a, b, &invalid, &overflow);
> > + bcd_sub_mag(&result, a, b, &invalid, &overflow);
> > cr = (sgna > 0) ? CRF_GT : CRF_LT;
> > } else if (bcd_cmp_mag(a, b) == 0) {
> > result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(0, ps);
> > - zero = bcd_sub_mag(&result, b, a, &invalid, &overflow);
> > + bcd_sub_mag(&result, b, a, &invalid, &overflow);
>
> I don't think you actually need the sub here, since you know the
> result is going to be zero.
Right. Will fix this in v2.
> Although.. in all of the different-sign cases aren't we effectively
> doing the subtraction twice - once in bcd_cmp_mag() then again in
> bcd_sub_mag()?
Actually no, bcd_cmp_mag() compares the magnitude between 'a' and 'b'
starting from the most significant digit and returns as soon as it finds
a difference between the two of them. It helps to decide whether we're
going to perform a - b or b - a.
Anyway, I think the following code can make it easier to understand:
int magnitude;
if (sgna == sgnb) {
// same thing
} else {
magnitude = bdc_cmp_mag(a, b);
if (magnitude > 0) {
// do a - b
} else if (magnitude < 0) {
// do b - a
} else {
// 0
}
}
> > + cr = CRF_EQ;
> > } else {
> > result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgnb, ps);
> > - zero = bcd_sub_mag(&result, b, a, &invalid, &overflow);
> > + bcd_sub_mag(&result, b, a, &invalid, &overflow);
> > cr = (sgnb > 0) ? CRF_GT : CRF_LT;
> > }
> > }
> > @@ -2762,8 +2749,6 @@ uint32_t helper_bcdadd(ppc_avr_t *r, ppc_avr_t *a,
> > ppc_avr_t *b, uint32_t ps)
> > cr = CRF_SO;
> > } else if (overflow) {
> > cr |= CRF_SO;
> > - } else if (zero) {
> > - cr = CRF_EQ;
> > }
> >
> > *r = result;
>
> --
> David Gibson | I'll have my music baroque, and my code
> david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_
> _other_
> | _way_ _around_!
> http://www.ozlabs.org/~dgibson
