/*[0,pi/4)に簡約*/
uint32_t fsin(uint32_t f){
if(fcmp(f,0) == 0)
return fneg(fsin(f - (1 << 31)));
if(fcmp(f,0x40c90fda) == 2)
return fsin(fadd(f,0xc0c90fda));
if(fcmp(f,MYPI) == 0){
if(fcmp(f,MYPI2) == 0){
if(fcmp(f,MYPI4) == 0)
return kernel_sin(f);
else
return kernel_cos(fadd(MYPI2,fneg(f)));
}else{
if(fcmp(f,fadd(MYPI4,MYPI2)) == 0)
kernel_cos(fadd(f,fneg(MYPI2)));
else
kernel_sin(fadd(MYPI,fneg(f)));
}
}else{
f = fadd(f,fneg(MYPI));
if(fcmp(f,MYPI2) == 0){
if(fcmp(f,MYPI4) == 0)
return fneg(kernel_sin(f));
else
return fneg(kernel_cos(fadd(MYPI2,fneg(f))));
}else{
if(fcmp(f,fadd(MYPI4,MYPI2)) == 0)
fneg(kernel_cos(fadd(f,fneg(MYPI2))));
else
fneg(kernel_sin(fadd(MYPI,fneg(f))));
}
}
}
uint32_t kernel_cos(uint32_t f){
uint32_t x2 = fmul(f,f);
uint32_t x3 = fmul(f,x2);
uint32_t x4 = fmul(x2,x2);
uint32_t x6 = fmul(x3,x3);
return fadd(fadd(0x3f800000,fmul(0xbf000000,x2)), fadd(fmul(0x3d2aa789,x4),fmul(0xbab38106,x6)));
}
uint32_t kernel_sin(uint32_t f){
uint32_t x2 = fmul(f,f);
uint32_t x3 = fmul(f,x2);
uint32_t x4 = fmul(x2,x2);
uint32_t x5 = fmul(x3,x2);
uint32_t x7 = fmul(x3,x4);
return fadd(fadd(f,fmul(0xb94d64b6,x7)), fadd(fmul(0xbe2aaaac,x3),fmul(0x3c088666,x5)));
}
void fcom(const temp_real * src1, const temp_real * src2)
{
temp_real a;
a = *src1;
a.exponent ^= 0x8000;
fadd(&a,src2,&a);
ftst(&a);
}
void draw_crusher(BITMAP *bmp)
{
static fixed angle = 0;
int x, y, x1, y1, x2, y2;
x = crusher.x+crusher.w/2-get_bitmap(BMP_CRUSHER)->w/2;
y = crusher.y+crusher.h/2-get_bitmap(BMP_CRUSHER)->h/2+8;
/* crusher body */
draw_sprite(bmp, get_bitmap(BMP_CRUSHER), x, y);
/* crusher motor */
draw_sprite(bmp, get_bitmap(BMP_CRUSHER_MOTOR),
x1 = x-get_bitmap(BMP_CRUSHER_MOTOR)->w+1+rand()%3-1,
y1 = y+rand()%3-1);
/* crusher pulley */
rotate_sprite(bmp, get_bitmap(BMP_CRUSHER_PULLEY),
x2 = x+17-get_bitmap(BMP_CRUSHER_PULLEY)->w/2+rand()%3-1,
y2 = y+13-get_bitmap(BMP_CRUSHER_PULLEY)->h/2+rand()%3-1,
angle);
/* strap */
do_line(bmp, x1+12, y1+8,
x2+get_bitmap(BMP_CRUSHER_PULLEY)->w/2-1, y2,
makecol(0, 0, 0), strap_proc);
do_line(bmp, x1+12, y1+18,
x2+get_bitmap(BMP_CRUSHER_PULLEY)->w/2-1,
y2+get_bitmap(BMP_CRUSHER_PULLEY)->h-1,
makecol(0, 0, 0), strap_proc);
angle = fadd(angle, itofix(16));
if (angle > itofix(256))
angle = fsub(angle, itofix(256));
/* scorer */
draw_sprite(bmp, get_bitmap(BMP_SCORER),
x+get_bitmap(BMP_CRUSHER)->w,
y+get_bitmap(BMP_CRUSHER)->h-get_bitmap(BMP_SCORER)->h);
/* alarm */
if (alarm_time_blue >= 0) {
masked_blit(get_bitmap(BMP_SCORER_ALARM), bmp, 0, 0,
x+get_bitmap(BMP_CRUSHER)->w+63,
y+get_bitmap(BMP_CRUSHER)->h-get_bitmap(BMP_SCORER)->h+27, 8, 9);
}
if (alarm_time_red >= 0) {
masked_blit(get_bitmap(BMP_SCORER_ALARM), bmp,
get_bitmap(BMP_SCORER_ALARM)->w-8, 0,
x+get_bitmap(BMP_CRUSHER)->w+75,
y+get_bitmap(BMP_CRUSHER)->h-get_bitmap(BMP_SCORER)->h+27, 8, 9);
}
}
void shmem_int8_add_f(FORTRAN_POINTER_T target, FORTRAN_POINTER_T value, MPI_Fint *pe)
{
ompi_fortran_integer8_t out_value = 0;
oshmem_op_t* op = oshmem_op_sum_fint8;
MCA_ATOMIC_CALL(fadd(FPTR_2_VOID_PTR(target),
(void *)&out_value,
FPTR_2_VOID_PTR(value),
sizeof(out_value),
OMPI_FINT_2_INT(*pe),
op));
}
ompi_fortran_real4_t shmem_real4_fetch_f(FORTRAN_POINTER_T target, MPI_Fint *pe)
{
ompi_fortran_real4_t out_value = 0;
ompi_fortran_real4_t value = 0;
MCA_ATOMIC_CALL(fadd(oshmem_ctx_default, FPTR_2_VOID_PTR(target),
(void *)&out_value,
value,
sizeof(out_value),
OMPI_FINT_2_INT(*pe)));
return out_value;
}
void shmem_int4_inc_f(FORTRAN_POINTER_T target, MPI_Fint *pe)
{
ompi_fortran_integer4_t out_value = 0;
ompi_fortran_integer4_t value = 1;
oshmem_op_t* op = oshmem_op_sum_fint4;
MCA_ATOMIC_CALL(fadd(FPTR_2_VOID_PTR(target),
(void *)&out_value,
(const void*)&value,
sizeof(out_value),
OMPI_FINT_2_INT(*pe),
op));
}
int
main(int argc, const char *argv[])
{
int i;
int *mem = malloc(1100000);
capture_stdio();
for (i = 0;; ++i) {
int32_t t0[2], t1[2];
char *msg;
int n;
gp_get_usertime(t0);
switch (i) {
case 0:
iadd(0, n = 10000000, &msg);
break;
case 1:
imul(1, n = 1000000, &msg);
break;
case 2:
idiv(1, n = 1000000, &msg);
break;
case 3:
fadd(3.14, n = 10000000, &msg);
break;
case 4:
fmul(1.0000001, n = 10000000, &msg);
break;
case 5:
fdiv(1.0000001, n = 1000000, &msg);
break;
case 6:
fconv(12345, n = 10000000, &msg);
break;
case 7:
mfast(mem, n = 10000000, &msg);
break;
case 8:
mslow(mem, n = 1000000, &msg);
break;
default:
free(mem);
exit(0);
}
gp_get_usertime(t1);
fprintf(stdout, "Time for %9d %s = %g ms\n", n, msg,
(t1[0] - t0[0]) * 1000.0 + (t1[1] - t0[1]) / 1000000.0);
fflush(stdout);
}
}
ompi_fortran_real4_t shmem_real4_fetch_f(FORTRAN_POINTER_T target, MPI_Fint *pe)
{
ompi_fortran_real4_t out_value = 0;
oshmem_op_t* op = oshmem_op_sum_freal4;
int value = 0;
MCA_ATOMIC_CALL(fadd(FPTR_2_VOID_PTR(target),
(void *)&out_value,
(const void *)&value,
sizeof(out_value),
OMPI_FINT_2_INT(*pe),
op));
return out_value;
}
int main(int argc, char* argv[])
{
printf("easy test!\n");
int t;
int a = 4;
int b = 5;
int c = add(a, b);
printf("%d + %d = %d\n", a, b, c);
float x = 5.5;
float y = .6;
float z = fadd(x, y);
printf("%f + %f = %f\n", x, y, z);
return 0;
}
void fasinh(_MIPD_ flash x,flash y)
{ /* calculate y=asinh(x) */
#ifdef MR_OS_THREADS
miracl *mr_mip=get_mip();
#endif
copy(x,y);
if (mr_mip->ERNUM || size(y)==0) return;
MR_IN(66)
fmul(_MIPP_ y,y,mr_mip->w11);
fincr(_MIPP_ mr_mip->w11,1,1,mr_mip->w11);
froot(_MIPP_ mr_mip->w11,2,mr_mip->w11);
fadd(_MIPP_ y,mr_mip->w11,y);
flog(_MIPP_ y,y);
MR_OUT
}
void facosh(_MIPD_ flash x,flash y)
{ /* calculate y=acosh(x) */
#ifndef MR_GENERIC_MT
miracl *mr_mip=get_mip();
#endif
copy(x,y);
if (mr_mip->ERNUM) return;
MR_IN(62)
fmul(_MIPP_ y,y,mr_mip->w11);
fincr(_MIPP_ mr_mip->w11,(-1),1,mr_mip->w11);
froot(_MIPP_ mr_mip->w11,2,mr_mip->w11);
fadd(_MIPP_ y,mr_mip->w11,y);
flog(_MIPP_ y,y);
MR_OUT
}
/** Adds two matrices.
* Returns materr(2) if the two matrices can't
* be added, materr(1) if the initialization
* didn't go right.
*/
matrix MatrixAdd(matrix m1, matrix m2)
{
if (!isValidOp(m1,m2,'a')) {
return materr(2);
}
matrix mx = initialize(m1.rows, m1.columns);
if (mx.mx == NULL) {
return materr(1);
}
for(int row = 0; row < mx.rows; row++)
for(int col = 0; col < mx.columns; col++)
mx.mx[row][col] = fadd(m1.mx[row][col], m2.mx[row][col]);
return mx;
}
/** Multiplies m1 by m2.
* Returns materr(2) if the two matrices can't
* be multiplied, materr(1) if the initialization
* didn't go right.
*/
matrix MatrixMul(matrix m1, matrix m2)
{
if (!isValidOp(m1,m2,'m')) {
return materr(2);
}
matrix mx = initialize(m1.rows, m2.columns);
if (mx.mx == NULL) {
return materr(1);
}
for(int row = 0; row < mx.rows; row++) {
for(int col = 0; col < mx.columns; col++) {
for(int i = 0; i < m1.columns; i++) {
mx.mx[row][col] = fadd(mx.mx[row][col], fmul(m1.mx[row][i], m2.mx[i][col]));
}
}
}
return mx;
}
char *
all_tests (void)
{
fesetround (FE_TOWARDZERO);
srand((unsigned)time(NULL));
union uint32_f a, b, c;
for (int i = 0; i < 2000000 ; i++)
{
char aa[33],bb[33],cc[33];
a.i = (uint32_t)( (rand () << 2) + rand ());
b.i = (uint32_t)( (rand () << 2) + rand ());
if (fpclassify (a.f) != FP_NORMAL || fpclassify (b.f) != FP_NORMAL)
continue;
tests_run++;
c.i = fadd (a.i, b.i);
if (fpclassify (c.f) != FP_NORMAL || fpclassify (a.f*b.f) != FP_NORMAL)
continue;
for (int t = 0; t < 32;++t)
{
aa[31 - t] = a.i & (1 << t) ? '1' : '0';
bb[31 - t] = b.i & (1 << t) ? '1' : '0';
cc[31 - t] = c.i & (1 << t) ? '1' : '0';
}
aa[32] = '\0';
bb[32] = '\0';
cc[32] = '\0';
// 非正規化数とかはやらない
if (isnormal (a.f) && isnormal (b.f))
{
if (!isnan (c.f))
printf ("%s\t%s\t%s\n",aa,bb,cc);
}
}
return NULL;
}
static void do_emu(struct info * info)
{
unsigned short code;
temp_real tmp;
char * address;
if (I387.cwd & I387.swd & 0x3f)
I387.swd |= 0x8000;
else
I387.swd &= 0x7fff;
ORIG_EIP = EIP;
/* 0x0007 means user code space */
if (CS != 0x000F) {
printk("math_emulate: %04x:%08x\n\r",CS,EIP);
panic("Math emulation needed in kernel");
}
code = get_fs_word((unsigned short *) EIP);
bswapw(code);
code &= 0x7ff;
I387.fip = EIP;
*(unsigned short *) &I387.fcs = CS;
*(1+(unsigned short *) &I387.fcs) = code;
EIP += 2;
switch (code) {
case 0x1d0: /* fnop */
return;
case 0x1d1: case 0x1d2: case 0x1d3:
case 0x1d4: case 0x1d5: case 0x1d6: case 0x1d7:
math_abort(info,1<<(SIGILL-1));
case 0x1e0:
ST(0).exponent ^= 0x8000;
return;
case 0x1e1:
ST(0).exponent &= 0x7fff;
return;
case 0x1e2: case 0x1e3:
math_abort(info,1<<(SIGILL-1));
case 0x1e4:
ftst(PST(0));
return;
case 0x1e5:
printk("fxam not implemented\n\r");
math_abort(info,1<<(SIGILL-1));
case 0x1e6: case 0x1e7:
math_abort(info,1<<(SIGILL-1));
case 0x1e8:
fpush();
ST(0) = CONST1;
return;
case 0x1e9:
fpush();
ST(0) = CONSTL2T;
return;
case 0x1ea:
fpush();
ST(0) = CONSTL2E;
return;
case 0x1eb:
fpush();
ST(0) = CONSTPI;
return;
case 0x1ec:
fpush();
ST(0) = CONSTLG2;
return;
case 0x1ed:
fpush();
ST(0) = CONSTLN2;
return;
case 0x1ee:
fpush();
ST(0) = CONSTZ;
return;
case 0x1ef:
math_abort(info,1<<(SIGILL-1));
case 0x1f0: case 0x1f1: case 0x1f2: case 0x1f3:
case 0x1f4: case 0x1f5: case 0x1f6: case 0x1f7:
case 0x1f8: case 0x1f9: case 0x1fa: case 0x1fb:
case 0x1fc: case 0x1fd: case 0x1fe: case 0x1ff:
printk("%04x fxxx not implemented\n\r",code + 0xc800);
math_abort(info,1<<(SIGILL-1));
case 0x2e9:
fucom(PST(1),PST(0));
fpop(); fpop();
return;
case 0x3d0: case 0x3d1:
return;
case 0x3e2:
I387.swd &= 0x7f00;
return;
case 0x3e3:
I387.cwd = 0x037f;
I387.swd = 0x0000;
I387.twd = 0x0000;
return;
case 0x3e4:
return;
case 0x6d9:
fcom(PST(1),PST(0));
fpop(); fpop();
return;
case 0x7e0:
*(short *) &EAX = I387.swd;
return;
}
switch (code >> 3) {
case 0x18:
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 0x19:
fmul(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 0x1a:
fcom(PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 0x1b:
fcom(PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
fpop();
return;
case 0x1c:
real_to_real(&ST(code & 7),&tmp);
tmp.exponent ^= 0x8000;
fadd(PST(0),&tmp,&tmp);
real_to_real(&tmp,&ST(0));
return;
case 0x1d:
ST(0).exponent ^= 0x8000;
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 0x1e:
fdiv(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 0x1f:
fdiv(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 0x38:
fpush();
ST(0) = ST((code & 7)+1);
return;
case 0x39:
fxchg(&ST(0),&ST(code & 7));
return;
case 0x3b:
ST(code & 7) = ST(0);
fpop();
return;
case 0x98:
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return;
case 0x99:
fmul(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return;
case 0x9a:
fcom(PST(code & 7),PST(0));
return;
case 0x9b:
fcom(PST(code & 7),PST(0));
fpop();
return;
case 0x9c:
ST(code & 7).exponent ^= 0x8000;
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return;
case 0x9d:
real_to_real(&ST(0),&tmp);
tmp.exponent ^= 0x8000;
fadd(PST(code & 7),&tmp,&tmp);
real_to_real(&tmp,&ST(code & 7));
return;
case 0x9e:
fdiv(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return;
case 0x9f:
fdiv(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
return;
case 0xb8:
printk("ffree not implemented\n\r");
math_abort(info,1<<(SIGILL-1));
case 0xb9:
fxchg(&ST(0),&ST(code & 7));
return;
case 0xba:
ST(code & 7) = ST(0);
return;
case 0xbb:
ST(code & 7) = ST(0);
fpop();
return;
case 0xbc:
fucom(PST(code & 7),PST(0));
return;
case 0xbd:
fucom(PST(code & 7),PST(0));
fpop();
return;
case 0xd8:
fadd(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return;
case 0xd9:
fmul(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return;
case 0xda:
fcom(PST(code & 7),PST(0));
fpop();
return;
case 0xdc:
ST(code & 7).exponent ^= 0x8000;
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return;
case 0xdd:
real_to_real(&ST(0),&tmp);
tmp.exponent ^= 0x8000;
fadd(PST(code & 7),&tmp,&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return;
case 0xde:
fdiv(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return;
case 0xdf:
fdiv(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return;
case 0xf8:
printk("ffree not implemented\n\r");
math_abort(info,1<<(SIGILL-1));
fpop();
return;
case 0xf9:
fxchg(&ST(0),&ST(code & 7));
return;
case 0xfa:
case 0xfb:
ST(code & 7) = ST(0);
fpop();
return;
}
switch ((code>>3) & 0xe7) {
case 0x22:
put_short_real(PST(0),info,code);
return;
case 0x23:
put_short_real(PST(0),info,code);
fpop();
return;
case 0x24:
address = ea(info,code);
for (code = 0 ; code < 7 ; code++) {
((long *) & I387)[code] =
get_fs_long((unsigned long *) address);
address += 4;
}
return;
case 0x25:
address = ea(info,code);
*(unsigned short *) &I387.cwd =
get_fs_word((unsigned short *) address);
return;
case 0x26:
address = ea(info,code);
verify_area(address,28);
for (code = 0 ; code < 7 ; code++) {
put_fs_long( ((long *) & I387)[code],
(unsigned long *) address);
address += 4;
}
return;
case 0x27:
address = ea(info,code);
verify_area(address,2);
put_fs_word(I387.cwd,(short *) address);
return;
case 0x62:
put_long_int(PST(0),info,code);
return;
case 0x63:
put_long_int(PST(0),info,code);
fpop();
return;
case 0x65:
fpush();
get_temp_real(&tmp,info,code);
real_to_real(&tmp,&ST(0));
return;
case 0x67:
put_temp_real(PST(0),info,code);
fpop();
return;
case 0xa2:
put_long_real(PST(0),info,code);
return;
case 0xa3:
put_long_real(PST(0),info,code);
fpop();
return;
case 0xa4:
address = ea(info,code);
for (code = 0 ; code < 27 ; code++) {
((long *) & I387)[code] =
get_fs_long((unsigned long *) address);
address += 4;
}
return;
case 0xa6:
address = ea(info,code);
verify_area(address,108);
for (code = 0 ; code < 27 ; code++) {
put_fs_long( ((long *) & I387)[code],
(unsigned long *) address);
address += 4;
}
I387.cwd = 0x037f;
I387.swd = 0x0000;
I387.twd = 0x0000;
return;
case 0xa7:
address = ea(info,code);
verify_area(address,2);
put_fs_word(I387.swd,(short *) address);
return;
case 0xe2:
put_short_int(PST(0),info,code);
return;
case 0xe3:
put_short_int(PST(0),info,code);
fpop();
return;
case 0xe4:
fpush();
get_BCD(&tmp,info,code);
real_to_real(&tmp,&ST(0));
return;
case 0xe5:
fpush();
get_longlong_int(&tmp,info,code);
real_to_real(&tmp,&ST(0));
return;
case 0xe6:
put_BCD(PST(0),info,code);
fpop();
return;
case 0xe7:
put_longlong_int(PST(0),info,code);
fpop();
return;
}
switch (code >> 9) {
case 0:
get_short_real(&tmp,info,code);
break;
case 1:
get_long_int(&tmp,info,code);
break;
case 2:
get_long_real(&tmp,info,code);
break;
case 4:
get_short_int(&tmp,info,code);
}
switch ((code>>3) & 0x27) {
case 0:
fadd(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 1:
fmul(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 2:
fcom(&tmp,PST(0));
return;
case 3:
fcom(&tmp,PST(0));
fpop();
return;
case 4:
tmp.exponent ^= 0x8000;
fadd(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 5:
ST(0).exponent ^= 0x8000;
fadd(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return;
case 6:
fdiv(PST(0),&tmp,&tmp);
real_to_real(&tmp,&ST(0));
return;
case 7:
fdiv(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return;
}
if ((code & 0x138) == 0x100) {
fpush();
real_to_real(&tmp,&ST(0));
return;
}
printk("Unknown math-insns: %04x:%08x %04x\n\r",CS,EIP,code);
math_abort(info,1<<(SIGFPE-1));
}
BOOL gauss(flash A[][50],flash b[],int n)
{ /* solve Ax=b using Gaussian elimination *
* solution x returned in b */
int i,j,k,m;
BOOL ok;
flash w,s;
w=mirvar(0);
s=mirvar(0);
ok=TRUE;
for (i=0;i<n;i++)
copy(b[i],A[i][n]);
for (i=0;i<n;i++)
{ /* Gaussian elimination */
m=i;
for (j=i+1;j<n;j++)
{
absol(A[j][i],w);
absol(A[m][i],s);
if (fcomp(w,s)>0) m=j;
}
if (m!=i) for (k=i;k<=n;k++)
{
copy(A[i][k],w);
copy(A[m][k],A[i][k]);
copy(w,A[m][k]);
}
if (size(A[i][i])==0)
{
ok=FALSE;
break;
}
for (j=i+1;j<n;j++)
{
fdiv(A[j][i],A[i][i],s);
for (k=n;k>=i;k--)
{
fmul(s,A[i][k],w);
fsub(A[j][k],w,A[j][k]);
}
}
}
if (ok) for (j=n-1;j>=0;j--)
{ /* Backward substitution */
zero(s);
for (k=j+1;k<n;k++)
{
fmul(A[j][k],b[k],w);
fadd(s,w,s);
}
fsub(A[j][n],s,w);
if (size(A[j][j])==0)
{
ok=FALSE;
break;
}
fdiv(w,A[j][j],b[j]);
}
mirkill(s);
mirkill(w);
return ok;
}
uint32_t fsub(uint32_t a, uint32_t b) {
return fadd(a, fneg(b));
}
int encode_op(char *opcode, char *op_data)
{
int rd,rs,rt,imm,funct,shaft,target;
char tmp[256];
const char *fi = "%s %d";
const char *fg = "%s %%g%d";
const char *ff = "%s %%f%d";
const char *fl = "%s %s";
const char *fgi = "%s %%g%d, %d";
const char *fgl = "%s %%g%d, %s";
const char *fgg = "%s %%g%d, %%g%d";
const char *fggl = "%s %%g%d, %%g%d, %s";
const char *fggi = "%s %%g%d, %%g%d, %d";
const char *fggg = "%s %%g%d, %%g%d, %%g%d";
const char *fff = "%s %%f%d, %%f%d";
const char *fgf = "%s %%g%d, %%f%d";
const char *ffg = "%s %%f%d, %%g%d";
const char *fffl = "%s %%f%d, %%f%d, %s";
const char *ffff = "%s %%f%d, %%f%d, %%f%d";
const char *ffgi = "%s %%f%d, %%g%d, %d";
const char *ffgg = "%s %%f%d, %%g%d, %%g%d";
char lname[256];
shaft = funct = target = 0;
if(strcmp(opcode, "mvhi") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return mvhi(rs,0,imm);
}
if(strcmp(opcode, "mvlo") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return mvlo(rs,0,imm);
}
if(strcmp(opcode, "add") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return add(rs,rt,rd,0);
}
if(strcmp(opcode, "nor") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return nor(rs,rt,rd,0);
}
if(strcmp(opcode, "sub") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return sub(rs,rt,rd,0);
}
if(strcmp(opcode, "mul") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return mul(rs,rt,rd,0);
}
if(strcmp(opcode, "addi") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return addi(rs,rt,imm);
}
if(strcmp(opcode, "subi") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return subi(rs,rt,imm);
}
if(strcmp(opcode, "muli") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return muli(rs,rt,imm);
}
if(strcmp(opcode, "input") == 0){
if(sscanf(op_data, fg, tmp, &rd) == 2)
return input(0,0,rd,0);
}
if(strcmp(opcode, "inputw") == 0){
if(sscanf(op_data, fg, tmp, &rd) == 2)
return inputw(0,0,rd,0);
}
if(strcmp(opcode, "inputf") == 0){
if(sscanf(op_data, ff, tmp, &rd) == 2)
return inputf(0,0,rd,0);
}
if(strcmp(opcode, "output") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return output(rs,0,0,0);
}
if(strcmp(opcode, "outputw") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return outputw(rs,0,0,0);
}
if(strcmp(opcode, "outputf") == 0){
if(sscanf(op_data, ff, tmp, &rs) == 2)
return outputf(rs,0,0,0);
}
if(strcmp(opcode, "and") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return _and(rs,rt,rd,0);
}
if(strcmp(opcode, "or") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return _or(rs,rt,rd,0);
}
if(strcmp(opcode, "sll") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return sll(rs,rt,rd,0);
}
if(strcmp(opcode, "srl") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return srl(rs,rt,rd,0);
}
if(strcmp(opcode, "slli") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return slli(rs,rt,imm);
}
if(strcmp(opcode, "srli") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return srli(rs,rt,imm);
}
if(strcmp(opcode, "b") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return b(rs,0,0,0);
}
if(strcmp(opcode, "jmp") == 0){
if(sscanf(op_data, fl, tmp, lname) == 2) {
strcpy(label_name[label_cnt],lname);
return jmp(label_cnt++);
}
}
if(strcmp(opcode, "jeq") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jeq(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "jne") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jne(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "jlt") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jlt(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "jle") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jle(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "call") == 0){
if(sscanf(op_data, fl, tmp, lname) == 2) {
strcpy(label_name[label_cnt],lname);
return call(label_cnt++);
}
}
if(strcmp(opcode, "callR") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return callr(rs,0,0,0);
}
if(strcmp(opcode, "return") == 0){
return _return(0);
}
if(strcmp(opcode, "ld") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return ld(rs,rt,rd,0);
}
if(strcmp(opcode, "ldi") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return ldi(rs,rt,imm);
}
if(strcmp(opcode, "ldlr") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return ldlr(rs,0,imm);
}
if(strcmp(opcode, "fld") == 0){
if(sscanf(op_data, ffgg, tmp, &rd, &rs,&rt) == 4)
return fld(rs,rt,rd,0);
}
if(strcmp(opcode, "st") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return st(rs,rt,rd,0);
}
if(strcmp(opcode, "sti") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return sti(rs,rt,imm);
}
if(strcmp(opcode, "stlr") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return stlr(rs,0,imm);
}
if(strcmp(opcode, "fst") == 0){
if(sscanf(op_data, ffgg, tmp, &rd, &rs,&rt) == 4)
return fst(rs,rt,rd,0);
}
if(strcmp(opcode, "fadd") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fadd(rs,rt,rd,0);
}
if(strcmp(opcode, "fsub") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fsub(rs,rt,rd,0);
}
if(strcmp(opcode, "fmul") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fmul(rs,rt,rd,0);
}
if(strcmp(opcode, "fdiv") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fdiv(rs,rt,rd,0);
}
if(strcmp(opcode, "fsqrt") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return fsqrt(rs,0,rd,0);
}
if(strcmp(opcode, "fabs") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return _fabs(rs,0,rd,0);
}
if(strcmp(opcode, "fmov") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return fmov(rs,0,rd,0);
}
if(strcmp(opcode, "fneg") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return fneg(rs,0,rd,0);
}
if(strcmp(opcode, "fldi") == 0){
if(sscanf(op_data, ffgi, tmp, &rt, &rs, &imm) == 4)
return fldi(rs,rt,imm);
}
if(strcmp(opcode, "fsti") == 0){
if(sscanf(op_data, ffgi, tmp, &rt, &rs, &imm) == 4)
return fsti(rs,rt,imm);
}
if(strcmp(opcode, "fjeq") == 0){
if(sscanf(op_data, fffl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return fjeq(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "fjlt") == 0){
if(sscanf(op_data, fffl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return fjlt(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "halt") == 0){
return halt(0,0,0,0);
}
if(strcmp(opcode, "setL") == 0){
if(sscanf(op_data, fgl, tmp, &rd, lname) == 3) {
strcpy(label_name[label_cnt],lname);
return setl(0,rd,label_cnt++);
}
}
if(strcmp(opcode, "padd") == 0){
if(sscanf(op_data, fgi, tmp, &rt, &imm) == 3) {
return padd(0,rt,imm);
}
}
if(strcmp(opcode, "link") == 0){
if(sscanf(op_data, fi, tmp, &imm) == 2) {
return link(0,0,imm);
}
}
if(strcmp(opcode, "movlr") == 0){
return movlr(0,0,0,0);
}
if(strcmp(opcode, "btmplr") == 0){
return btmplr(0,0,0,0);
}
/*
if(strcmp(opcode, "padd") == 0){
if(sscanf(op_data, fgg, tmp, &rd, &rt) == 3) {
return padd(0,rt,d,0);
}
}
*/
return -1;
}
void ANCFBeamBE2D::EvalF2(Vector& f, double t)
{
Body2D::EvalF2(f, t);
TMStartTimer(22);
// do the computation
/*ConstVector<ANCFBeamBE2DmaxDOF> fadd(2*NS()); // temporary storage vector
ConstVector<ANCFBeamBE2DmaxDOF> delta_eps(2*NS()), delta_kappa(2*NS());
double EI = GetBeamEIy();
double EA = GetBeamEA();
for (IntegrationPointsIterator ip(integrationRuleStiffness); !ip.IsEnd(); ++ip)
{
double x = ip.Point2D().X();
double eps = GetEpsAxial(x);
double kappa = GetKappa(x);
GetDeltaEpsAxial(x, delta_eps);
GetDeltaKappa(x, delta_kappa);
fadd += ((EA*eps)*delta_eps + (EI*kappa)*delta_kappa) * (ip.Weight() * lx*0.5);
}
f -= fadd;*/
ConstVector<ANCFBeamBE2DmaxDOF> fadd(2*NS()); // temporary storage vector
ConstVector<ANCFBeamBE2DmaxDOF> delta_eps(2*NS()), delta_kappa(2*NS());
double EI = GetBeamEIy();
double EA = GetBeamEA();
for (IntegrationPointsIterator ip(integrationRuleStiffness); !ip.IsEnd(); ++ip)
{
double x = ip.Point2D().X();
//Vector2D rx = GetPosx2D(x);
//Vector2D rxx = GetPosxx2D(x);
Vector2D rx, rxx;
for (int i = 1; i <= NS(); i++)
{
//double sx = ip.GetVectorData(1)->Get(i) * 2./lx;
//double sxx = ip.GetVectorData(2)->Get(i) * 4./(lx*lx);
double sx = GetS0x(x, i) * 2./GetLx();
double sxx = GetS0xx(x, i) * 4./(GetLx()*GetLx());
double d1 = q0(2*i-1) + XG(2*i-1);
double d2 = q0(2*i ) + XG(2*i );
rx(1) += sx * d1;
rx(2) += sx * d2;
rxx(1) += sxx * d1;
rxx(2) += sxx * d2;
}
double rxn2 = rx.Norm2();
double eps = rx.Norm() - 1.0;
double kappa = (rx.Cross(rxx))/rxn2;
double d = 1./rx.Norm();
double kdux = -2.*rx(1)*kappa / rxn2 + rxx(2) / rxn2;
double kduxx = -rx(2) / rxn2;
double kdwx = -2.*rx(2)*kappa / rxn2 - rxx(1) / rxn2;
double kdwxx = rx(1) / rxn2;
for (int j = 1; j <= NS(); j++)
{
double sx = GetS0x(x, j) * 2./GetLx();
double sxx = GetS0xx(x, j) * 4./(GetLx()*GetLx());
//double sx = ip.GetVectorData(1)->Get(j) * 2./lx;
//double sxx = ip.GetVectorData(2)->Get(j) * 4./(lx*lx);
delta_eps(2*j-1) = d * rx(1) * sx;
delta_eps(2*j ) = d * rx(2) * sx;
delta_kappa(2*j-1) = (kdux * sx + kduxx * sxx);
delta_kappa(2*j ) = (kdwx * sx + kdwxx * sxx);
}
fadd += ((EA*eps)*delta_eps + (EI*kappa)*delta_kappa) * (ip.Weight() * GetLx()*0.5);
}
f -= fadd;
TMStopTimer(22);
};
STATIC
unsigned fsub(unsigned fa, unsigned fb)
{
return fadd(fa, fneg(fb));
}
float fsub(float in1, float in2)
{
return fadd(in1,(in2*-1));
}
int main() {
float_b f1,f2,f3,f4;
int i = 0;
int c1 = 0;
int c2 = 0;
int c3 = 0;
int c4 = 0;
srand(time(NULL));
while(i < 1000000) { //正規化数ランダム
f1.ieee.sign = rand()%2;
f1.ieee.exp = rand()%254+1;
f1.ieee.fraction = rand()%8388608;
f2.ieee.sign = rand()%2;
f2.ieee.exp = rand()%254+1;
f2.ieee.fraction = rand()%8388608;
f3.f = f1.f + f2.f;
f4.f = fadd(f1.f,f2.f);
if (f3.f != f4.f && (f3.ieee.exp != 0 || f4.ieee.exp != 0) && (f3.ieee.exp != 255 || f4.ieee.exp != 255)) {
printf("%f+%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c1++;
}
f3.f = f1.f * f2.f;
f4.f = fmul(f1.f,f2.f);
if (f3.f != f4.f && (f3.ieee.exp != 0 || f4.ieee.exp != 0) && (f3.ieee.exp != 255 || (f4.ieee.exp != 0 && f4.ieee.exp != 255))) {
printf("%f*%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c2++;
}
f3.f = f1.f / f2.f;
f4.f = fdiv(f1.f,f2.f);
if (f3.f != f4.f && ((f3.ieee.exp != 0 && f3.ieee.exp != 255) || (f4.ieee.exp != 0 && f4.ieee.exp != 255))) {
printf("%f/%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c3++;
}
f1.ieee.sign = 0;
f3.f = sqrtf(f1.f);
f4.f = sqrt_m(f1.f);
if (f3.f != f4.f) {
printf("sqrt(%f)=%f\n",f1.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c4++;
}
i++;
}
printf("fadd:%d fmul:%d fdiv:%d sqrt:%d\n",c1,c2,c3,c4);
i = 0;
c1 = 0;
c2 = 0;
c3 = 0;
c4 = 0;
while(i < 1000000) { //0.0に対する計算
f1.ieee.sign = rand()%2;
f1.ieee.exp = (rand()%254+1)*(i%2);
f1.ieee.fraction = rand()%8388608;
f2.ieee.sign = rand()%2;
f2.ieee.exp = (rand()%254+1)*((i+1)%2);
f2.ieee.fraction = rand()%8388608;
if (i%2 == 1) f3.f = f1.f;
else f3.f = f2.f;
f4.f = fadd(f1.f,f2.f);
if (f3.f != f4.f) {
printf("%f+%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c1++;
}
f3.ieee.sign = (f1.ieee.sign + f2.ieee.sign)%2;
f3.ieee.exp = 0;
f3.ieee.fraction = f1.ieee.fraction;
f4.f = fmul(f1.f,f2.f);
if (f3.f != f4.f) {
printf("%f*%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c2++;
}
if (i%2 == 1) {
f3.ieee.sign = f1.ieee.sign + f2.ieee.sign;
f3.ieee.exp = 255;
f3.ieee.fraction = f1.ieee.fraction;
} else {
f3.f = 0.0 / f2.f;
}
f4.f = fdiv(f1.f,f2.f);
if ((f3.ieee.sign != f4.ieee.sign || f3.ieee.exp != f4.ieee.exp || f3.ieee.fraction != f4.ieee.fraction) && (i%2 == 1 || (f3.ieee.exp != 0 && f3.ieee.exp != 255) || (f4.ieee.exp != 0 && f4.ieee.exp != 255))) {
printf("%f/%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c3++;
}
f1.ieee.sign = 0;
if (i%2 == 1) f3.f = sqrtf(f1.f);
else f3.f = f1.f;
f4.f = sqrt_m(f1.f);
if (f3.f != f4.f) {
printf("sqrt(%f)=%f\n",f1.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c4++;
}
i++;
}
printf("コーナーケース\n");
printf("0.0(とみなされる数)の演算 fadd:%d fmul:%d fdiv:%d sqrt:%d\n",c1,c2,c3,c4);
i = 0;
c1 = 0;
c2 = 0;
c3 = 0;
c4 = 0;
while(i < 1000000) {
f1.ieee.sign = rand()%2;
f1.ieee.exp = (rand()%254+1);
f1.ieee.fraction = rand()%8388608;
f2.ieee.sign = rand()%2;
f2.ieee.exp = f1.ieee.exp + (rand()%3-1);
f2.ieee.fraction = rand()%8388608;
if (f2.ieee.exp == 255) f2.ieee.exp = 254;
if (f2.ieee.exp == 0) f2.ieee.exp = 1;
f3.f = f1.f + f2.f;
f4.f = fadd(f1.f,f2.f);
if (f3.f != f4.f && (f3.ieee.exp != 0 || f4.ieee.exp != 0) && (f3.ieee.exp != 255 || f4.ieee.exp != 255)) {
print_bit(f1);
print_bit(f2);
printf("%f+%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c1++;
}
f2.f = f1.f;
f2.ieee.sign = f1.ieee.sign + 1;
f3.f = f1.f + f2.f;
f4.f = fadd(f1.f,f2.f);
if (f3.f != f4.f && (f3.ieee.exp != 0 || f4.ieee.exp != 0) && (f3.ieee.exp != 255 || f4.ieee.exp != 255)) {
printf("%f+%f=%f\n",f1.f,f2.f,f3.f);
print_bit(f3);
print_bit(f4);
} else {
c2++;
}
i++;
}
printf("指数の差が高々1 fadd:%d\n",c1);
printf("絶対値が等しく符号が逆 fadd:%d\n",c2);
return 0;
}
static BOOL act(int p,int q)
{ /* act on selected key */
int k,n,c;
aprint(PRESSED,4+5*p,6+3*q,keys[q][p]);
switch(p+7*q)
{
case 0: if (degrees) fmul(x,radeg,x);
if (hyp) fsinh(x,x);
else fsin(x,x);
newx=TRUE;
break;
case 1: if (degrees) fmul(x,radeg,x);
if (hyp) fcosh(x,x);
else fcos(x,x);
newx=TRUE;
break;
case 2: if (degrees) fmul(x,radeg,x);
if (hyp) ftanh(x,x);
else ftan(x,x);
newx=TRUE;
break;
case 3: if (lgbase>0)
{
n=size(x);
if (abs(n)<MR_TOOBIG)
{
convert(lgbase,x);
if (n<0) frecip(x,x);
fpower(x,abs(n),x);
newx=TRUE;
break;
}
if (lgbase==2) fmul(x,loge2,x);
if (lgbase==10) fmul(x,loge10,x);
}
fexp(x,x);
newx=TRUE;
break;
case 4: mip->RPOINT=!mip->RPOINT;
newx=TRUE;
break;
case 5: clrall();
newx=TRUE;
break;
case 6: return TRUE;
case 7: if (hyp) fasinh(x,x);
else fasin(x,x);
if (degrees) fdiv(x,radeg,x);
newx=TRUE;
break;
case 8: if (hyp) facosh(x,x);
else facos(x,x);
if (degrees) fdiv(x,radeg,x);
newx=TRUE;
break;
case 9: if (hyp) fatanh(x,x);
else fatan(x,x);
if (degrees) fdiv(x,radeg,x);
newx=TRUE;
break;
case 10: flog(x,x);
if (lgbase==2) fdiv(x,loge2,x);
if (lgbase==10) fdiv(x,loge10,x);
newx=TRUE;
break;
case 11: newx=TRUE;
k=3;
forever
{
aprint(INVER,2+stptr[k],2,settings[k][option[k]]);
curser(2+stptr[k],2);
c=arrow(gethit());
if (c==1)
{
if (option[k]==nops[k]) option[k]=0;
else option[k]+=1;
continue;
}
aprint(STATCOL,2+stptr[k],2,settings[k][option[k]]);
if (c==0 || c==2) break;
if (c==4 && k>0) k--;
if (c==3 && k<3) k++;
}
setopts();
break;
case 12: chekit(7);
break;
case 13: result=FALSE;
if (ipt==0) break;
ipt--;
mybuff[ipt]='\0';
if (ipt==0) clr();
just(mybuff);
cinstr(x,mybuff);
newx=TRUE;
break;
case 14: if (!next('7')) putchar(BELL);
break;
case 15: if (!next('8')) putchar(BELL);
break;
case 16: if (!next('9')) putchar(BELL);
break;
case 17: chekit(6);
break;
case 18: chekit(5);
break;
case 19: chekit(4);
break;
case 20: copy(m,x);
newx=TRUE;
break;
case 21: if (!next('4')) putchar(BELL);
break;
case 22: if (!next('5')) putchar(BELL);
break;
case 23: if (!next('6')) putchar(BELL);
break;
case 24: fmul(x,x,x);
newx=TRUE;
break;
case 25: froot(x,2,x);
newx=TRUE;
break;
case 26: chekit(3);
break;
case 27: brkt=0;
chekit(0);
flag=OFF;
fadd(m,x,m);
newx=TRUE;
break;
case 28: if (!next('1')) putchar(BELL);
break;
case 29: if (!next('2')) putchar(BELL);
break;
case 30: if (!next('3')) putchar(BELL);
break;
case 31: frecip(x,x);
newx=TRUE;
break;
case 32: fpi(x);
newx=TRUE;
break;
case 33: chekit(2);
break;
case 34: negify(x,x);
newx=TRUE;
break;
case 35: if (!next('0')) putchar(BELL);
break;
case 36: if (!next('/')) putchar(BELL);
break;
case 37: if (!next('.')) putchar(BELL);
break;
case 38: if (ipt>0)
{
putchar(BELL);
result=FALSE;
}
else
{
zero(x);
brkt+=1;
newx=TRUE;
}
break;
case 39: if (brkt>0)
{
chekit(0);
brkt-=1;
}
else
{
putchar(BELL);
result=FALSE;
}
break;
case 40: chekit(1);
break;
case 41: brkt=0;
equals(0);
flag=OFF;
break;
}
return FALSE;
}
static int
math_emulate(struct trapframe * info)
{
unsigned short code;
temp_real tmp;
char * address;
u_long oldeip;
/* ever used fp? */
if ((((struct pcb *)curproc->p_addr)->pcb_flags & FP_SOFTFP) == 0) {
((struct pcb *)curproc->p_addr)->pcb_flags |= FP_SOFTFP;
I387.cwd = 0x037f;
I387.swd = 0x0000;
I387.twd = 0x0000;
}
if (I387.cwd & I387.swd & 0x3f)
I387.swd |= 0x8000;
else
I387.swd &= 0x7fff;
oldeip = info->tf_eip;
/* 0x001f means user code space */
if ((u_short)info->tf_cs != 0x001F) {
printf("math_emulate: %04x:%08lx\n", (u_short)info->tf_cs,
oldeip);
panic("?Math emulation needed in kernel?");
}
code = get_fs_word((unsigned short *) oldeip);
bswapw(code);
code &= 0x7ff;
I387.fip = oldeip;
*(unsigned short *) &I387.fcs = (u_short) info->tf_cs;
*(1+(unsigned short *) &I387.fcs) = code;
info->tf_eip += 2;
switch (code) {
case 0x1d0: /* fnop */
return(0);
case 0x1d1: case 0x1d2: case 0x1d3: /* fst to 32-bit mem */
case 0x1d4: case 0x1d5: case 0x1d6: case 0x1d7:
math_abort(info,SIGILL);
case 0x1e0: /* fchs */
ST(0).exponent ^= 0x8000;
return(0);
case 0x1e1: /* fabs */
ST(0).exponent &= 0x7fff;
return(0);
case 0x1e2: case 0x1e3:
math_abort(info,SIGILL);
case 0x1e4: /* ftst */
ftst(PST(0));
return(0);
case 0x1e5: /* fxam */
printf("fxam not implemented\n");
math_abort(info,SIGILL);
case 0x1e6: case 0x1e7: /* fldenv */
math_abort(info,SIGILL);
case 0x1e8: /* fld1 */
fpush();
ST(0) = CONST1;
return(0);
case 0x1e9: /* fld2t */
fpush();
ST(0) = CONSTL2T;
return(0);
case 0x1ea: /* fld2e */
fpush();
ST(0) = CONSTL2E;
return(0);
case 0x1eb: /* fldpi */
fpush();
ST(0) = CONSTPI;
return(0);
case 0x1ec: /* fldlg2 */
fpush();
ST(0) = CONSTLG2;
return(0);
case 0x1ed: /* fldln2 */
fpush();
ST(0) = CONSTLN2;
return(0);
case 0x1ee: /* fldz */
fpush();
ST(0) = CONSTZ;
return(0);
case 0x1ef:
math_abort(info,SIGILL);
case 0x1f0: /* f2xm1 */
case 0x1f1: /* fyl2x */
case 0x1f2: /* fptan */
case 0x1f3: /* fpatan */
case 0x1f4: /* fxtract */
case 0x1f5: /* fprem1 */
case 0x1f6: /* fdecstp */
case 0x1f7: /* fincstp */
case 0x1f8: /* fprem */
case 0x1f9: /* fyl2xp1 */
case 0x1fa: /* fsqrt */
case 0x1fb: /* fsincos */
case 0x1fe: /* fsin */
case 0x1ff: /* fcos */
uprintf(
"math_emulate: instruction %04x not implemented\n",
code + 0xd800);
math_abort(info,SIGILL);
case 0x1fc: /* frndint */
frndint(PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 0x1fd: /* fscale */
/* incomplete and totally inadequate -wfj */
Fscale(PST(0), PST(1), &tmp);
real_to_real(&tmp,&ST(0));
return(0); /* 19 Sep 92*/
case 0x2e9: /* ????? */
/* if this should be a fucomp ST(0),ST(1) , it must be a 0x3e9 ATS */
fucom(PST(1),PST(0));
fpop(); fpop();
return(0);
case 0x3d0: case 0x3d1: /* fist ?? */
return(0);
case 0x3e2: /* fclex */
I387.swd &= 0x7f00;
return(0);
case 0x3e3: /* fninit */
I387.cwd = 0x037f;
I387.swd = 0x0000;
I387.twd = 0x0000;
return(0);
case 0x3e4:
return(0);
case 0x6d9: /* fcompp */
fcom(PST(1),PST(0));
fpop(); fpop();
return(0);
case 0x7e0: /* fstsw ax */
*(short *) &info->tf_eax = I387.swd;
return(0);
}
switch (code >> 3) {
case 0x18: /* fadd */
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 0x19: /* fmul */
fmul(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 0x1a: /* fcom */
fcom(PST(code & 7),PST(0));
return(0);
case 0x1b: /* fcomp */
fcom(PST(code & 7),PST(0));
fpop();
return(0);
case 0x1c: /* fsubr */
real_to_real(&ST(code & 7),&tmp);
tmp.exponent ^= 0x8000;
fadd(PST(0),&tmp,&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 0x1d: /* fsub */
ST(0).exponent ^= 0x8000;
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 0x1e: /* fdivr */
fdiv(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 0x1f: /* fdiv */
fdiv(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 0x38: /* fld */
fpush();
ST(0) = ST((code & 7)+1); /* why plus 1 ????? ATS */
return(0);
case 0x39: /* fxch */
fxchg(&ST(0),&ST(code & 7));
return(0);
case 0x3b: /* ??? ??? wrong ???? ATS */
ST(code & 7) = ST(0);
fpop();
return(0);
case 0x98: /* fadd */
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return(0);
case 0x99: /* fmul */
fmul(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return(0);
case 0x9a: /* ???? , my manual don't list a direction bit
for fcom , ??? ATS */
fcom(PST(code & 7),PST(0));
return(0);
case 0x9b: /* same as above , ATS */
fcom(PST(code & 7),PST(0));
fpop();
return(0);
case 0x9c: /* fsubr */
ST(code & 7).exponent ^= 0x8000;
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return(0);
case 0x9d: /* fsub */
real_to_real(&ST(0),&tmp);
tmp.exponent ^= 0x8000;
fadd(PST(code & 7),&tmp,&tmp);
real_to_real(&tmp,&ST(code & 7));
return(0);
case 0x9e: /* fdivr */
fdiv(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
return(0);
case 0x9f: /* fdiv */
fdiv(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
return(0);
case 0xb8: /* ffree */
printf("ffree not implemented\n");
math_abort(info,SIGILL);
case 0xb9: /* fstp ???? where is the pop ? ATS */
fxchg(&ST(0),&ST(code & 7));
return(0);
case 0xba: /* fst */
ST(code & 7) = ST(0);
return(0);
case 0xbb: /* ????? encoding of fstp to mem ? ATS */
ST(code & 7) = ST(0);
fpop();
return(0);
case 0xbc: /* fucom */
fucom(PST(code & 7),PST(0));
return(0);
case 0xbd: /* fucomp */
fucom(PST(code & 7),PST(0));
fpop();
return(0);
case 0xd8: /* faddp */
fadd(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return(0);
case 0xd9: /* fmulp */
fmul(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return(0);
case 0xda: /* ??? encoding of ficom with 16 bit mem ? ATS */
fcom(PST(code & 7),PST(0));
fpop();
return(0);
case 0xdc: /* fsubrp */
ST(code & 7).exponent ^= 0x8000;
fadd(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return(0);
case 0xdd: /* fsubp */
real_to_real(&ST(0),&tmp);
tmp.exponent ^= 0x8000;
fadd(PST(code & 7),&tmp,&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return(0);
case 0xde: /* fdivrp */
fdiv(PST(0),PST(code & 7),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return(0);
case 0xdf: /* fdivp */
fdiv(PST(code & 7),PST(0),&tmp);
real_to_real(&tmp,&ST(code & 7));
fpop();
return(0);
case 0xf8: /* fild 16-bit mem ???? ATS */
printf("ffree not implemented\n");
math_abort(info,SIGILL);
fpop();
return(0);
case 0xf9: /* ????? ATS */
fxchg(&ST(0),&ST(code & 7));
return(0);
case 0xfa: /* fist 16-bit mem ? ATS */
case 0xfb: /* fistp 16-bit mem ? ATS */
ST(code & 7) = ST(0);
fpop();
return(0);
}
switch ((code>>3) & 0xe7) {
case 0x22:
put_short_real(PST(0),info,code);
return(0);
case 0x23:
put_short_real(PST(0),info,code);
fpop();
return(0);
case 0x24:
address = ea(info,code);
for (code = 0 ; code < 7 ; code++) {
((long *) & I387)[code] =
get_fs_long((unsigned long *) address);
address += 4;
}
return(0);
case 0x25:
address = ea(info,code);
*(unsigned short *) &I387.cwd =
get_fs_word((unsigned short *) address);
return(0);
case 0x26:
address = ea(info,code);
/*verify_area(address,28);*/
for (code = 0 ; code < 7 ; code++) {
put_fs_long( ((long *) & I387)[code],
(unsigned long *) address);
address += 4;
}
return(0);
case 0x27:
address = ea(info,code);
/*verify_area(address,2);*/
put_fs_word(I387.cwd,(short *) address);
return(0);
case 0x62:
put_long_int(PST(0),info,code);
return(0);
case 0x63:
put_long_int(PST(0),info,code);
fpop();
return(0);
case 0x65:
fpush();
get_temp_real(&tmp,info,code);
real_to_real(&tmp,&ST(0));
return(0);
case 0x67:
put_temp_real(PST(0),info,code);
fpop();
return(0);
case 0xa2:
put_long_real(PST(0),info,code);
return(0);
case 0xa3:
put_long_real(PST(0),info,code);
fpop();
return(0);
case 0xa4:
address = ea(info,code);
for (code = 0 ; code < 27 ; code++) {
((long *) & I387)[code] =
get_fs_long((unsigned long *) address);
address += 4;
}
return(0);
case 0xa6:
address = ea(info,code);
/*verify_area(address,108);*/
for (code = 0 ; code < 27 ; code++) {
put_fs_long( ((long *) & I387)[code],
(unsigned long *) address);
address += 4;
}
I387.cwd = 0x037f;
I387.swd = 0x0000;
I387.twd = 0x0000;
return(0);
case 0xa7:
address = ea(info,code);
/*verify_area(address,2);*/
put_fs_word(I387.swd,(short *) address);
return(0);
case 0xe2:
put_short_int(PST(0),info,code);
return(0);
case 0xe3:
put_short_int(PST(0),info,code);
fpop();
return(0);
case 0xe4:
fpush();
get_BCD(&tmp,info,code);
real_to_real(&tmp,&ST(0));
return(0);
case 0xe5:
fpush();
get_longlong_int(&tmp,info,code);
real_to_real(&tmp,&ST(0));
return(0);
case 0xe6:
put_BCD(PST(0),info,code);
fpop();
return(0);
case 0xe7:
put_longlong_int(PST(0),info,code);
fpop();
return(0);
}
switch (code >> 9) {
case 0:
get_short_real(&tmp,info,code);
break;
case 1:
get_long_int(&tmp,info,code);
break;
case 2:
get_long_real(&tmp,info,code);
break;
case 4:
get_short_int(&tmp,info,code);
}
switch ((code>>3) & 0x27) {
case 0:
fadd(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 1:
fmul(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 2:
fcom(&tmp,PST(0));
return(0);
case 3:
fcom(&tmp,PST(0));
fpop();
return(0);
case 4:
tmp.exponent ^= 0x8000;
fadd(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 5:
ST(0).exponent ^= 0x8000;
fadd(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 6:
fdiv(PST(0),&tmp,&tmp);
real_to_real(&tmp,&ST(0));
return(0);
case 7:
fdiv(&tmp,PST(0),&tmp);
real_to_real(&tmp,&ST(0));
return(0);
}
if ((code & 0x138) == 0x100) {
fpush();
real_to_real(&tmp,&ST(0));
return(0);
}
printf("Unknown math-insns: %04x:%08x %04x\n",(u_short)info->tf_cs,
info->tf_eip,code);
math_abort(info,SIGFPE);
}
static void equals(int no)
{ /* perform binary operation */
BOOL pop;
newx=FALSE;
pop=TRUE;
while (pop)
{
if (prec(no)>prop[sp])
{ /* if higher precedence, keep this one pending */
if (sp==top)
{
mip->ERNUM=(-1);
result=FALSE;
newx=TRUE;
}
else sp++;
return;
}
newx=TRUE;
if (flag && op[sp]!=3 && op[sp]!=0) swap();
switch (op[sp])
{
case 7: fdiv(x,y[sp],t);
ftrunc(t,t,t);
fmul(t,y[sp],t);
fsub(x,t,x);
break;
case 6: fpowf(x,y[sp],x);
break;
case 5: frecip(y[sp],t);
fpowf(x,t,x);
break;
case 4: fdiv(x,y[sp],x);
break;
case 3: fmul(x,y[sp],x);
break;
case 2: fsub(x,y[sp],x);
break;
case 1: fadd(x,y[sp],x);
break;
case 0: break;
}
if (sp>0 && (prec(no)<=prop[sp-1])) sp--;
else pop=FALSE;
}
}
