Make improvements to locking

This change makes pthread_mutex_lock() as fast as _spinlock() by
default. Thread instability issues on NetBSD have been resolved.
Improvements made to gdtoa thread code. Crash reporting will now
synchronize between threads in a slightly better way.
This commit is contained in:
Justine Tunney 2022-06-19 01:13:03 -07:00
parent 25041b8026
commit d5312b60f7
60 changed files with 890 additions and 629 deletions

View file

@ -30,6 +30,7 @@
*/
#include "libc/assert.h"
#include "libc/calls/calls.h"
#include "libc/macros.internal.h"
#include "libc/runtime/runtime.h"
#include "third_party/gdtoa/gdtoa.internal.h"
@ -62,69 +63,80 @@ __attribute__((__constructor__)) static void
__gdtoa_Binit(void)
{
atexit(__gdtoa_Bclear);
TI0.P5s = __gdtoa_i2b(625);
TI0.P5s->next = 0;
}
static ThInfo *
__gdtoa_get_TI(void)
{
return &TI0;
}
Bigint *
__gdtoa_Balloc(int k)
__gdtoa_Balloc(int k, ThInfo **PTI)
{
int x;
Bigint *rv;
assert(k >= 0);
__gdtoa_lock();
if (k <= Kmax && (rv = TI0.Freelist[k]) != 0) {
TI0.Freelist[k] = rv->next;
unsigned int len;
ThInfo *TI;
if (!(TI = *PTI))
*PTI = TI = __gdtoa_get_TI();
if (TI == &TI0)
__gdtoa_lock();
if (k <= Kmax && (rv = TI->Freelist[k]) != 0) {
TI->Freelist[k] = rv->next;
} else {
x = 1 << k;
rv = malloc(sizeof(Bigint) + (x-1)*sizeof(ULong));
rv->k = k;
rv->maxwds = x;
}
rv->sign = 0;
rv->wds = 0;
__gdtoa_unlock();
if (TI == &TI0)
__gdtoa_unlock();
rv->sign = rv->wds = 0;
return rv;
}
void
__gdtoa_Bfree(Bigint *v)
__gdtoa_Bfree(Bigint *v, ThInfo **PTI)
{
ThInfo *TI;
if (v) {
if (v->k > Kmax) {
free(v);
} else {
assert(v->k >= 0);
__gdtoa_lock();
v->next = TI0.Freelist[v->k];
TI0.Freelist[v->k] = v;
__gdtoa_unlock();
if (v->k > Kmax)
free((void *)v);
else {
if (!(TI = *PTI))
*PTI = TI = __gdtoa_get_TI();
if (TI == &TI0)
__gdtoa_lock();
v->next = TI->Freelist[v->k];
TI->Freelist[v->k] = v;
if (TI == &TI0)
__gdtoa_unlock();
}
}
}
Bigint * /* multiply by m and add a */
__gdtoa_multadd(Bigint *b, int m, int a)
Bigint *
__gdtoa_multadd(Bigint *b, int m, int a, ThInfo **PTI)
{
int i, wds;
ULong *x;
ULLong carry, y;
uint64_t carry, y;
Bigint *b1;
wds = b->wds;
x = b->x;
i = 0;
carry = a;
do {
y = *x * (ULLong)m + carry;
y = *x * (uint64_t)m + carry;
carry = y >> 32;
*x++ = y & 0xffffffffUL;
}
while(++i < wds);
} while (++i < wds);
if (carry) {
if (wds >= b->maxwds) {
b1 = __gdtoa_Balloc(b->k+1);
Bcopy(b1, b);
__gdtoa_Bfree(b);
b1 = __gdtoa_Balloc(b->k + 1, PTI);
memcpy(&b1->sign, &b->sign, b->wds * sizeof(ULong) + 2 * sizeof(int));
__gdtoa_Bfree(b, PTI);
b = b1;
}
b->x[wds++] = carry;
@ -134,23 +146,23 @@ __gdtoa_multadd(Bigint *b, int m, int a)
}
Bigint *
__gdtoa_i2b(int i)
__gdtoa_i2b(int i, ThInfo **PTI)
{
Bigint *b;
b = __gdtoa_Balloc(1);
b = __gdtoa_Balloc(1, PTI);
b->x[0] = i;
b->wds = 1;
return b;
}
Bigint *
__gdtoa_mult(Bigint *a, Bigint *b)
__gdtoa_mult(Bigint *a, Bigint *b, ThInfo **PTI)
{
Bigint *c;
int k, wa, wb, wc;
ULong *x, *xa, *xae, *xb, *xbe, *xc, *xc0;
ULong y;
ULLong carry, z;
uint64_t carry, z;
if (a->wds < b->wds) {
c = a;
a = b;
@ -162,96 +174,115 @@ __gdtoa_mult(Bigint *a, Bigint *b)
wc = wa + wb;
if (wc > a->maxwds)
k++;
c = __gdtoa_Balloc(k);
for(x = c->x, xa = x + wc; x < xa; x++)
c = __gdtoa_Balloc(k, PTI);
for (x = c->x, xa = x + wc; x < xa; x++)
*x = 0;
xa = a->x;
xae = xa + wa;
xb = b->x;
xbe = xb + wb;
xc0 = c->x;
for(; xb < xbe; xc0++) {
if ( (y = *xb++) !=0) {
for (; xb < xbe; xc0++) {
if ((y = *xb++) != 0) {
x = xa;
xc = xc0;
carry = 0;
do {
z = *x++ * (ULLong)y + *xc + carry;
z = *x++ * (uint64_t)y + *xc + carry;
carry = z >> 32;
*xc++ = z & 0xffffffffUL;
}
while(x < xae);
} while (x < xae);
*xc = carry;
}
}
for(xc0 = c->x, xc = xc0 + wc; wc > 0 && !*--xc; --wc) ;
for (xc0 = c->x, xc = xc0 + wc; wc > 0 && !*--xc; --wc)
;
c->wds = wc;
return c;
}
Bigint *
__gdtoa_pow5mult(Bigint *b, int k)
__gdtoa_pow5mult(Bigint *b, int k, ThInfo **PTI)
{
int i;
Bigint *b1, *p5, *p51;
static const int p05[3] = { 5, 25, 125 };
if ((i = k & 3))
b = __gdtoa_multadd(b, p05[i-1], 0);
ThInfo *TI;
int i;
static int p05[3] = {5, 25, 125};
if ((i = k & 3) != 0)
b = __gdtoa_multadd(b, p05[i - 1], 0, PTI);
if (!(k >>= 2))
return b;
__gdtoa_lock();
p5 = TI0.P5s;
for(;;) {
if (!(TI = *PTI))
*PTI = TI = __gdtoa_get_TI();
if ((p5 = TI->P5s) == 0) {
if (!(TI = *PTI))
*PTI = TI = __gdtoa_get_TI();
if (TI == &TI0)
__gdtoa_lock1();
if (!(p5 = TI->P5s)) {
p5 = TI->P5s = __gdtoa_i2b(625, PTI);
p5->next = 0;
}
if (TI == &TI0)
__gdtoa_unlock1();
}
for (;;) {
if (k & 1) {
b1 = __gdtoa_mult(b, p5);
__gdtoa_Bfree(b);
b1 = __gdtoa_mult(b, p5, PTI);
__gdtoa_Bfree(b, PTI);
b = b1;
}
if (!(k >>= 1))
break;
if ((p51 = p5->next) == 0) {
p51 = p5->next = __gdtoa_mult(p5,p5);
p51->next = 0;
if (!TI && !(TI = *PTI))
*PTI = TI = __gdtoa_get_TI();
if (TI == &TI0)
__gdtoa_lock1();
if (!(p51 = p5->next)) {
p51 = p5->next = __gdtoa_mult(p5, p5, PTI);
p51->next = 0;
}
if (TI == &TI0)
__gdtoa_unlock1();
}
p5 = p51;
}
__gdtoa_unlock();
return b;
}
Bigint *
__gdtoa_lshift(Bigint *b, int k)
__gdtoa_lshift(Bigint *b, int k, ThInfo **PTI)
{
int i, k1, n, n1;
Bigint *b1;
ULong *x, *x1, *xe, z;
n = k >> kshift;
n = k >> 5;
k1 = b->k;
n1 = n + b->wds + 1;
for(i = b->maxwds; n1 > i; i <<= 1)
for (i = b->maxwds; n1 > i; i <<= 1)
k1++;
b1 = __gdtoa_Balloc(k1);
b1 = __gdtoa_Balloc(k1, PTI);
x1 = b1->x;
for(i = 0; i < n; i++)
for (i = 0; i < n; i++)
*x1++ = 0;
x = b->x;
xe = x + b->wds;
if (k &= kmask) {
if (k &= 31) {
k1 = 32 - k;
z = 0;
do {
*x1++ = *x << k | z;
z = *x++ >> k1;
}
while(x < xe);
if ((*x1 = z) !=0)
} while (x < xe);
if ((*x1 = z) != 0)
++n1;
}
else do
*x1++ = *x++;
while(x < xe);
} else
do
*x1++ = *x++;
while (x < xe);
b1->wds = n1 - 1;
__gdtoa_Bfree(b);
__gdtoa_Bfree(b, PTI);
return b1;
}
@ -262,19 +293,13 @@ __gdtoa_cmp(Bigint *a, Bigint *b)
int i, j;
i = a->wds;
j = b->wds;
#ifdef DEBUG
if (i > 1 && !a->x[i-1])
Bug("__gdtoa_cmp called with a->x[a->wds-1] == 0");
if (j > 1 && !b->x[j-1])
Bug("__gdtoa_cmp called with b->x[b->wds-1] == 0");
#endif
if (i -= j)
return i;
xa0 = a->x;
xa = xa0 + j;
xb0 = b->x;
xb = xb0 + j;
for(;;) {
for (;;) {
if (*--xa != *--xb)
return *xa < *xb ? -1 : 1;
if (xa <= xa0)
@ -284,15 +309,15 @@ __gdtoa_cmp(Bigint *a, Bigint *b)
}
Bigint *
__gdtoa_diff(Bigint *a, Bigint *b)
__gdtoa_diff(Bigint *a, Bigint *b, ThInfo **PTI)
{
Bigint *c;
int i, wa, wb;
ULong *xa, *xae, *xb, *xbe, *xc;
ULLong borrow, y;
i = __gdtoa_cmp(a,b);
uint64_t borrow, y;
i = __gdtoa_cmp(a, b);
if (!i) {
c = __gdtoa_Balloc(0);
c = __gdtoa_Balloc(0, PTI);
c->wds = 1;
c->x[0] = 0;
return c;
@ -302,10 +327,9 @@ __gdtoa_diff(Bigint *a, Bigint *b)
a = b;
b = c;
i = 1;
}
else
} else
i = 0;
c = __gdtoa_Balloc(a->k);
c = __gdtoa_Balloc(a->k, PTI);
c->sign = i;
wa = a->wds;
xa = a->x;
@ -316,17 +340,16 @@ __gdtoa_diff(Bigint *a, Bigint *b)
xc = c->x;
borrow = 0;
do {
y = (ULLong)*xa++ - *xb++ - borrow;
y = (uint64_t)*xa++ - *xb++ - borrow;
borrow = y >> 32 & 1UL;
*xc++ = y & 0xffffffffUL;
}
while(xb < xbe);
while(xa < xae) {
} while (xb < xbe);
while (xa < xae) {
y = *xa++ - borrow;
borrow = y >> 32 & 1UL;
*xc++ = y & 0xffffffffUL;
}
while(!*--xc)
while (!*--xc)
wa--;
c->wds = wa;
return c;
@ -341,33 +364,29 @@ __gdtoa_b2d(Bigint *a, int *e)
xa0 = a->x;
xa = xa0 + a->wds;
y = *--xa;
#ifdef DEBUG
if (!y) Bug("zero y in __gdtoa_b2d");
#endif
k = hi0bits(y);
k = hi0bits((ULong)(y));
*e = 32 - k;
if (k < Ebits) {
word0(&d) = Exp_1 | y >> (Ebits - k);
if (k < 11) {
(&d)->L[1] = 0x3ff00000 | y >> (11 - k);
w = xa > xa0 ? *--xa : 0;
word1(&d) = y << ((32-Ebits) + k) | w >> (Ebits - k);
(&d)->L[0] = y << ((32 - 11) + k) | w >> (11 - k);
goto ret_d;
}
z = xa > xa0 ? *--xa : 0;
if (k -= Ebits) {
word0(&d) = Exp_1 | y << k | z >> (32 - k);
if (k -= 11) {
(&d)->L[1] = 0x3ff00000 | y << k | z >> (32 - k);
y = xa > xa0 ? *--xa : 0;
word1(&d) = z << k | y >> (32 - k);
}
else {
word0(&d) = Exp_1 | y;
word1(&d) = z;
(&d)->L[0] = z << k | y >> (32 - k);
} else {
(&d)->L[1] = 0x3ff00000 | y;
(&d)->L[0] = z;
}
ret_d:
return dval(&d);
return (&d)->d;
}
Bigint *
__gdtoa_d2b(double dd, int *e, int *bits)
__gdtoa_d2b(double dd, int *e, int *bits, ThInfo **PTI)
{
Bigint *b;
U d;
@ -375,47 +394,37 @@ __gdtoa_d2b(double dd, int *e, int *bits)
int de, k;
ULong *x, y, z;
d.d = dd;
b = __gdtoa_Balloc(1);
b = __gdtoa_Balloc(1, PTI);
x = b->x;
z = word0(&d) & Frac_mask;
word0(&d) &= 0x7fffffff; /* clear sign bit, which we ignore */
if ( (de = (int)(word0(&d) >> Exp_shift)) !=0)
z |= Exp_msk1;
if ( (y = word1(&d)) !=0) {
if ( (k = lo0bits(&y)) !=0) {
z = (&d)->L[1] & 0xfffff;
(&d)->L[1] &= 0x7fffffff;
if ((de = (int)((&d)->L[1] >> 20)) != 0)
z |= 0x100000;
if ((y = (&d)->L[0]) != 0) {
if ((k = lo0bits(&y)) != 0) {
x[0] = y | z << (32 - k);
z >>= k;
}
else
} else
x[0] = y;
i = b->wds = (x[1] = z) !=0 ? 2 : 1;
}
else {
i = b->wds = (x[1] = z) != 0 ? 2 : 1;
} else {
k = lo0bits(&z);
x[0] = z;
i = b->wds = 1;
k += 32;
}
if (de) {
*e = de - Bias - (P-1) + k;
*bits = P - k;
}
else {
*e = de - Bias - (P-1) + 1 + k;
*bits = 32*i - hi0bits(x[i-1]);
*e = de - 1023 - (53 - 1) + k;
*bits = 53 - k;
} else {
*e = de - 1023 - (53 - 1) + 1 + k;
*bits = 32 * i - hi0bits((ULong)(x[i - 1]));
}
return b;
}
const double
__gdtoa_bigtens[] = { 1e16, 1e32, 1e64, 1e128, 1e256 };
const double
__gdtoa_tinytens[] = { 1e-16, 1e-32, 1e-64, 1e-128, 1e-256 };
const double
__gdtoa_tens[] = {
1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
1e10, 1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19,
1e20, 1e21, 1e22
};
const double __gdtoa_bigtens[] = {1e16, 1e32, 1e64, 1e128, 1e256};
const double __gdtoa_tinytens[] = {1e-16, 1e-32, 1e-64, 1e-128, 1e-256};
const double __gdtoa_tens[] = {1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7,
1e8, 1e9, 1e10, 1e11, 1e12, 1e13, 1e14, 1e15,
1e16, 1e17, 1e18, 1e19, 1e20, 1e21, 1e22};