softfloat: Move round_to_int to softfloat-parts.c.inc

At the same time, convert to pointers, split out
parts$N_round_to_int_normal, define a macro for
parts_round_to_int using QEMU_GENERIC.

This necessarily meant some rearrangement to the
rount_to_{,u}int_and_pack routines, so go ahead and
convert to parts_round_to_int_normal, which in turn
allows cleaning up of the raised exception handling.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This commit is contained in:
Richard Henderson 2020-11-14 12:53:12 -08:00
parent 9882ccaff9
commit afc34931eb
2 changed files with 263 additions and 328 deletions

View file

@ -594,3 +594,160 @@ static FloatPartsN *partsN(div)(FloatPartsN *a, FloatPartsN *b,
a->cls = float_class_inf;
return a;
}
/*
* Rounds the floating-point value `a' to an integer, and returns the
* result as a floating-point value. The operation is performed
* according to the IEC/IEEE Standard for Binary Floating-Point
* Arithmetic.
*
* parts_round_to_int_normal is an internal helper function for
* normal numbers only, returning true for inexact but not directly
* raising float_flag_inexact.
*/
static bool partsN(round_to_int_normal)(FloatPartsN *a, FloatRoundMode rmode,
int scale, int frac_size)
{
uint64_t frac_lsb, frac_lsbm1, rnd_even_mask, rnd_mask, inc;
int shift_adj;
scale = MIN(MAX(scale, -0x10000), 0x10000);
a->exp += scale;
if (a->exp < 0) {
bool one;
/* All fractional */
switch (rmode) {
case float_round_nearest_even:
one = false;
if (a->exp == -1) {
FloatPartsN tmp;
/* Shift left one, discarding DECOMPOSED_IMPLICIT_BIT */
frac_add(&tmp, a, a);
/* Anything remaining means frac > 0.5. */
one = !frac_eqz(&tmp);
}
break;
case float_round_ties_away:
one = a->exp == -1;
break;
case float_round_to_zero:
one = false;
break;
case float_round_up:
one = !a->sign;
break;
case float_round_down:
one = a->sign;
break;
case float_round_to_odd:
one = true;
break;
default:
g_assert_not_reached();
}
frac_clear(a);
a->exp = 0;
if (one) {
a->frac_hi = DECOMPOSED_IMPLICIT_BIT;
} else {
a->cls = float_class_zero;
}
return true;
}
if (a->exp >= frac_size) {
/* All integral */
return false;
}
if (N > 64 && a->exp < N - 64) {
/*
* Rounding is not in the low word -- shift lsb to bit 2,
* which leaves room for sticky and rounding bit.
*/
shift_adj = (N - 1) - (a->exp + 2);
frac_shrjam(a, shift_adj);
frac_lsb = 1 << 2;
} else {
shift_adj = 0;
frac_lsb = DECOMPOSED_IMPLICIT_BIT >> (a->exp & 63);
}
frac_lsbm1 = frac_lsb >> 1;
rnd_mask = frac_lsb - 1;
rnd_even_mask = rnd_mask | frac_lsb;
if (!(a->frac_lo & rnd_mask)) {
/* Fractional bits already clear, undo the shift above. */
frac_shl(a, shift_adj);
return false;
}
switch (rmode) {
case float_round_nearest_even:
inc = ((a->frac_lo & rnd_even_mask) != frac_lsbm1 ? frac_lsbm1 : 0);
break;
case float_round_ties_away:
inc = frac_lsbm1;
break;
case float_round_to_zero:
inc = 0;
break;
case float_round_up:
inc = a->sign ? 0 : rnd_mask;
break;
case float_round_down:
inc = a->sign ? rnd_mask : 0;
break;
case float_round_to_odd:
inc = a->frac_lo & frac_lsb ? 0 : rnd_mask;
break;
default:
g_assert_not_reached();
}
if (shift_adj == 0) {
if (frac_addi(a, a, inc)) {
frac_shr(a, 1);
a->frac_hi |= DECOMPOSED_IMPLICIT_BIT;
a->exp++;
}
a->frac_lo &= ~rnd_mask;
} else {
frac_addi(a, a, inc);
a->frac_lo &= ~rnd_mask;
/* Be careful shifting back, not to overflow */
frac_shl(a, shift_adj - 1);
if (a->frac_hi & DECOMPOSED_IMPLICIT_BIT) {
a->exp++;
} else {
frac_add(a, a, a);
}
}
return true;
}
static void partsN(round_to_int)(FloatPartsN *a, FloatRoundMode rmode,
int scale, float_status *s,
const FloatFmt *fmt)
{
switch (a->cls) {
case float_class_qnan:
case float_class_snan:
parts_return_nan(a, s);
break;
case float_class_zero:
case float_class_inf:
break;
case float_class_normal:
if (parts_round_to_int_normal(a, rmode, scale, fmt->frac_size)) {
float_raise(float_flag_inexact, s);
}
break;
default:
g_assert_not_reached();
}
}

View file

@ -811,6 +811,24 @@ static FloatParts128 *parts128_div(FloatParts128 *a, FloatParts128 *b,
#define parts_div(A, B, S) \
PARTS_GENERIC_64_128(div, A)(A, B, S)
static bool parts64_round_to_int_normal(FloatParts64 *a, FloatRoundMode rm,
int scale, int frac_size);
static bool parts128_round_to_int_normal(FloatParts128 *a, FloatRoundMode r,
int scale, int frac_size);
#define parts_round_to_int_normal(A, R, C, F) \
PARTS_GENERIC_64_128(round_to_int_normal, A)(A, R, C, F)
static void parts64_round_to_int(FloatParts64 *a, FloatRoundMode rm,
int scale, float_status *s,
const FloatFmt *fmt);
static void parts128_round_to_int(FloatParts128 *a, FloatRoundMode r,
int scale, float_status *s,
const FloatFmt *fmt);
#define parts_round_to_int(A, R, C, S, F) \
PARTS_GENERIC_64_128(round_to_int, A)(A, R, C, S, F)
/*
* Helper functions for softfloat-parts.c.inc, per-size operations.
*/
@ -2285,153 +2303,52 @@ float128 float64_to_float128(float64 a, float_status *s)
}
/*
* Rounds the floating-point value `a' to an integer, and returns the
* result as a floating-point value. The operation is performed
* according to the IEC/IEEE Standard for Binary Floating-Point
* Arithmetic.
* Round to integral value
*/
static FloatParts64 round_to_int(FloatParts64 a, FloatRoundMode rmode,
int scale, float_status *s)
{
switch (a.cls) {
case float_class_qnan:
case float_class_snan:
parts_return_nan(&a, s);
break;
case float_class_zero:
case float_class_inf:
/* already "integral" */
break;
case float_class_normal:
scale = MIN(MAX(scale, -0x10000), 0x10000);
a.exp += scale;
if (a.exp >= DECOMPOSED_BINARY_POINT) {
/* already integral */
break;
}
if (a.exp < 0) {
bool one;
/* all fractional */
float_raise(float_flag_inexact, s);
switch (rmode) {
case float_round_nearest_even:
one = a.exp == -1 && a.frac > DECOMPOSED_IMPLICIT_BIT;
break;
case float_round_ties_away:
one = a.exp == -1 && a.frac >= DECOMPOSED_IMPLICIT_BIT;
break;
case float_round_to_zero:
one = false;
break;
case float_round_up:
one = !a.sign;
break;
case float_round_down:
one = a.sign;
break;
case float_round_to_odd:
one = true;
break;
default:
g_assert_not_reached();
}
if (one) {
a.frac = DECOMPOSED_IMPLICIT_BIT;
a.exp = 0;
} else {
a.cls = float_class_zero;
}
} else {
uint64_t frac_lsb = DECOMPOSED_IMPLICIT_BIT >> a.exp;
uint64_t frac_lsbm1 = frac_lsb >> 1;
uint64_t rnd_even_mask = (frac_lsb - 1) | frac_lsb;
uint64_t rnd_mask = rnd_even_mask >> 1;
uint64_t inc;
switch (rmode) {
case float_round_nearest_even:
inc = ((a.frac & rnd_even_mask) != frac_lsbm1 ? frac_lsbm1 : 0);
break;
case float_round_ties_away:
inc = frac_lsbm1;
break;
case float_round_to_zero:
inc = 0;
break;
case float_round_up:
inc = a.sign ? 0 : rnd_mask;
break;
case float_round_down:
inc = a.sign ? rnd_mask : 0;
break;
case float_round_to_odd:
inc = a.frac & frac_lsb ? 0 : rnd_mask;
break;
default:
g_assert_not_reached();
}
if (a.frac & rnd_mask) {
float_raise(float_flag_inexact, s);
if (uadd64_overflow(a.frac, inc, &a.frac)) {
a.frac >>= 1;
a.frac |= DECOMPOSED_IMPLICIT_BIT;
a.exp++;
}
a.frac &= ~rnd_mask;
}
}
break;
default:
g_assert_not_reached();
}
return a;
}
float16 float16_round_to_int(float16 a, float_status *s)
{
FloatParts64 pa, pr;
FloatParts64 p;
float16_unpack_canonical(&pa, a, s);
pr = round_to_int(pa, s->float_rounding_mode, 0, s);
return float16_round_pack_canonical(&pr, s);
float16_unpack_canonical(&p, a, s);
parts_round_to_int(&p, s->float_rounding_mode, 0, s, &float16_params);
return float16_round_pack_canonical(&p, s);
}
float32 float32_round_to_int(float32 a, float_status *s)
{
FloatParts64 pa, pr;
FloatParts64 p;
float32_unpack_canonical(&pa, a, s);
pr = round_to_int(pa, s->float_rounding_mode, 0, s);
return float32_round_pack_canonical(&pr, s);
float32_unpack_canonical(&p, a, s);
parts_round_to_int(&p, s->float_rounding_mode, 0, s, &float32_params);
return float32_round_pack_canonical(&p, s);
}
float64 float64_round_to_int(float64 a, float_status *s)
{
FloatParts64 pa, pr;
FloatParts64 p;
float64_unpack_canonical(&pa, a, s);
pr = round_to_int(pa, s->float_rounding_mode, 0, s);
return float64_round_pack_canonical(&pr, s);
float64_unpack_canonical(&p, a, s);
parts_round_to_int(&p, s->float_rounding_mode, 0, s, &float64_params);
return float64_round_pack_canonical(&p, s);
}
/*
* Rounds the bfloat16 value `a' to an integer, and returns the
* result as a bfloat16 value.
*/
bfloat16 bfloat16_round_to_int(bfloat16 a, float_status *s)
{
FloatParts64 pa, pr;
FloatParts64 p;
bfloat16_unpack_canonical(&pa, a, s);
pr = round_to_int(pa, s->float_rounding_mode, 0, s);
return bfloat16_round_pack_canonical(&pr, s);
bfloat16_unpack_canonical(&p, a, s);
parts_round_to_int(&p, s->float_rounding_mode, 0, s, &bfloat16_params);
return bfloat16_round_pack_canonical(&p, s);
}
float128 float128_round_to_int(float128 a, float_status *s)
{
FloatParts128 p;
float128_unpack_canonical(&p, a, s);
parts_round_to_int(&p, s->float_rounding_mode, 0, s, &float128_params);
return float128_round_pack_canonical(&p, s);
}
/*
@ -2445,48 +2362,58 @@ bfloat16 bfloat16_round_to_int(bfloat16 a, float_status *s)
* is returned.
*/
static int64_t round_to_int_and_pack(FloatParts64 in, FloatRoundMode rmode,
static int64_t round_to_int_and_pack(FloatParts64 p, FloatRoundMode rmode,
int scale, int64_t min, int64_t max,
float_status *s)
{
int flags = 0;
uint64_t r;
int orig_flags = get_float_exception_flags(s);
FloatParts64 p = round_to_int(in, rmode, scale, s);
switch (p.cls) {
case float_class_snan:
case float_class_qnan:
s->float_exception_flags = orig_flags | float_flag_invalid;
return max;
flags = float_flag_invalid;
r = max;
break;
case float_class_inf:
s->float_exception_flags = orig_flags | float_flag_invalid;
return p.sign ? min : max;
flags = float_flag_invalid;
r = p.sign ? min : max;
break;
case float_class_zero:
return 0;
case float_class_normal:
/* TODO: 62 = N - 2, frac_size for rounding */
if (parts_round_to_int_normal(&p, rmode, scale, 62)) {
flags = float_flag_inexact;
}
if (p.exp <= DECOMPOSED_BINARY_POINT) {
r = p.frac >> (DECOMPOSED_BINARY_POINT - p.exp);
} else {
r = UINT64_MAX;
}
if (p.sign) {
if (r <= -(uint64_t) min) {
return -r;
if (r <= -(uint64_t)min) {
r = -r;
} else {
s->float_exception_flags = orig_flags | float_flag_invalid;
return min;
}
} else {
if (r <= max) {
return r;
} else {
s->float_exception_flags = orig_flags | float_flag_invalid;
return max;
flags = float_flag_invalid;
r = min;
}
} else if (r > max) {
flags = float_flag_invalid;
r = max;
}
break;
default:
g_assert_not_reached();
}
float_raise(flags, s);
return r;
}
int8_t float16_to_int8_scalbn(float16 a, FloatRoundMode rmode, int scale,
@ -2749,49 +2676,59 @@ int64_t bfloat16_to_int64_round_to_zero(bfloat16 a, float_status *s)
* flag.
*/
static uint64_t round_to_uint_and_pack(FloatParts64 in, FloatRoundMode rmode,
static uint64_t round_to_uint_and_pack(FloatParts64 p, FloatRoundMode rmode,
int scale, uint64_t max,
float_status *s)
{
int orig_flags = get_float_exception_flags(s);
FloatParts64 p = round_to_int(in, rmode, scale, s);
int flags = 0;
uint64_t r;
switch (p.cls) {
case float_class_snan:
case float_class_qnan:
s->float_exception_flags = orig_flags | float_flag_invalid;
return max;
flags = float_flag_invalid;
r = max;
break;
case float_class_inf:
s->float_exception_flags = orig_flags | float_flag_invalid;
return p.sign ? 0 : max;
flags = float_flag_invalid;
r = p.sign ? 0 : max;
break;
case float_class_zero:
return 0;
case float_class_normal:
/* TODO: 62 = N - 2, frac_size for rounding */
if (parts_round_to_int_normal(&p, rmode, scale, 62)) {
flags = float_flag_inexact;
if (p.cls == float_class_zero) {
r = 0;
break;
}
}
if (p.sign) {
s->float_exception_flags = orig_flags | float_flag_invalid;
return 0;
}
if (p.exp <= DECOMPOSED_BINARY_POINT) {
r = p.frac >> (DECOMPOSED_BINARY_POINT - p.exp);
flags = float_flag_invalid;
r = 0;
} else if (p.exp > DECOMPOSED_BINARY_POINT) {
flags = float_flag_invalid;
r = max;
} else {
s->float_exception_flags = orig_flags | float_flag_invalid;
return max;
r = p.frac >> (DECOMPOSED_BINARY_POINT - p.exp);
if (r > max) {
flags = float_flag_invalid;
r = max;
}
}
break;
/* For uint64 this will never trip, but if p.exp is too large
* to shift a decomposed fraction we shall have exited via the
* 3rd leg above.
*/
if (r > max) {
s->float_exception_flags = orig_flags | float_flag_invalid;
return max;
}
return r;
default:
g_assert_not_reached();
}
float_raise(flags, s);
return r;
}
uint8_t float16_to_uint8_scalbn(float16 a, FloatRoundMode rmode, int scale,
@ -6956,165 +6893,6 @@ floatx80 float128_to_floatx80(float128 a, float_status *status)
}
/*----------------------------------------------------------------------------
| Rounds the quadruple-precision floating-point value `a' to an integer, and
| returns the result as a quadruple-precision floating-point value. The
| operation is performed according to the IEC/IEEE Standard for Binary
| Floating-Point Arithmetic.
*----------------------------------------------------------------------------*/
float128 float128_round_to_int(float128 a, float_status *status)
{
bool aSign;
int32_t aExp;
uint64_t lastBitMask, roundBitsMask;
float128 z;
aExp = extractFloat128Exp( a );
if ( 0x402F <= aExp ) {
if ( 0x406F <= aExp ) {
if ( ( aExp == 0x7FFF )
&& ( extractFloat128Frac0( a ) | extractFloat128Frac1( a ) )
) {
return propagateFloat128NaN(a, a, status);
}
return a;
}
lastBitMask = 1;
lastBitMask = ( lastBitMask<<( 0x406E - aExp ) )<<1;
roundBitsMask = lastBitMask - 1;
z = a;
switch (status->float_rounding_mode) {
case float_round_nearest_even:
if ( lastBitMask ) {
add128( z.high, z.low, 0, lastBitMask>>1, &z.high, &z.low );
if ( ( z.low & roundBitsMask ) == 0 ) z.low &= ~ lastBitMask;
}
else {
if ( (int64_t) z.low < 0 ) {
++z.high;
if ( (uint64_t) ( z.low<<1 ) == 0 ) z.high &= ~1;
}
}
break;
case float_round_ties_away:
if (lastBitMask) {
add128(z.high, z.low, 0, lastBitMask >> 1, &z.high, &z.low);
} else {
if ((int64_t) z.low < 0) {
++z.high;
}
}
break;
case float_round_to_zero:
break;
case float_round_up:
if (!extractFloat128Sign(z)) {
add128(z.high, z.low, 0, roundBitsMask, &z.high, &z.low);
}
break;
case float_round_down:
if (extractFloat128Sign(z)) {
add128(z.high, z.low, 0, roundBitsMask, &z.high, &z.low);
}
break;
case float_round_to_odd:
/*
* Note that if lastBitMask == 0, the last bit is the lsb
* of high, and roundBitsMask == -1.
*/
if ((lastBitMask ? z.low & lastBitMask : z.high & 1) == 0) {
add128(z.high, z.low, 0, roundBitsMask, &z.high, &z.low);
}
break;
default:
abort();
}
z.low &= ~ roundBitsMask;
}
else {
if ( aExp < 0x3FFF ) {
if ( ( ( (uint64_t) ( a.high<<1 ) ) | a.low ) == 0 ) return a;
float_raise(float_flag_inexact, status);
aSign = extractFloat128Sign( a );
switch (status->float_rounding_mode) {
case float_round_nearest_even:
if ( ( aExp == 0x3FFE )
&& ( extractFloat128Frac0( a )
| extractFloat128Frac1( a ) )
) {
return packFloat128( aSign, 0x3FFF, 0, 0 );
}
break;
case float_round_ties_away:
if (aExp == 0x3FFE) {
return packFloat128(aSign, 0x3FFF, 0, 0);
}
break;
case float_round_down:
return
aSign ? packFloat128( 1, 0x3FFF, 0, 0 )
: packFloat128( 0, 0, 0, 0 );
case float_round_up:
return
aSign ? packFloat128( 1, 0, 0, 0 )
: packFloat128( 0, 0x3FFF, 0, 0 );
case float_round_to_odd:
return packFloat128(aSign, 0x3FFF, 0, 0);
case float_round_to_zero:
break;
}
return packFloat128( aSign, 0, 0, 0 );
}
lastBitMask = 1;
lastBitMask <<= 0x402F - aExp;
roundBitsMask = lastBitMask - 1;
z.low = 0;
z.high = a.high;
switch (status->float_rounding_mode) {
case float_round_nearest_even:
z.high += lastBitMask>>1;
if ( ( ( z.high & roundBitsMask ) | a.low ) == 0 ) {
z.high &= ~ lastBitMask;
}
break;
case float_round_ties_away:
z.high += lastBitMask>>1;
break;
case float_round_to_zero:
break;
case float_round_up:
if (!extractFloat128Sign(z)) {
z.high |= ( a.low != 0 );
z.high += roundBitsMask;
}
break;
case float_round_down:
if (extractFloat128Sign(z)) {
z.high |= (a.low != 0);
z.high += roundBitsMask;
}
break;
case float_round_to_odd:
if ((z.high & lastBitMask) == 0) {
z.high |= (a.low != 0);
z.high += roundBitsMask;
}
break;
default:
abort();
}
z.high &= ~ roundBitsMask;
}
if ( ( z.low != a.low ) || ( z.high != a.high ) ) {
float_raise(float_flag_inexact, status);
}
return z;
}
/*----------------------------------------------------------------------------
| Returns the remainder of the quadruple-precision floating-point value `a'
| with respect to the corresponding value `b'. The operation is performed