float64 helper_fmul_DT(CPUSH4State *env, float64 t0, float64 t1)
{
set_float_exception_flags(0, &env->fp_status);
t0 = float64_mul(t0, t1, &env->fp_status);
update_fpscr(env, GETPC());
return t0;
}
int
main ()
{
int main_result;
int i;
float64 x1, x2;
main_result = 0;
for (i = 0; i < N; i++)
{
float64 result;
x1 = a_input[i];
x2 = b_input[i];
result = float64_mul (x1, x2);
main_result += (result == z_output[i]);
printf
("a_input=%016llx b_input=%016llx expected=%016llx output=%016llx\n",
a_input[i], b_input[i], z_output[i], result);
}
//printf ("Result: %d\n", main_result);
//if (main_result == 20) {
// printf("RESULT: PASS\n");
//} else {
// printf("RESULT: FAIL\n");
//}
return main_result;
}
/* 64-bit FP multiply and subtract RR */
void HELPER(msdbr)(CPUS390XState *env, uint32_t f1, uint32_t f3, uint32_t f2)
{
HELPER_LOG("%s: f1 %d f2 %d f3 %d\n", __func__, f1, f2, f3);
env->fregs[f1].d = float64_sub(float64_mul(env->fregs[f2].d,
env->fregs[f3].d,
&env->fpu_status),
env->fregs[f1].d, &env->fpu_status);
}
uint64_t helper_fmul_DT(uint64_t t0, uint64_t t1)
{
CPU_DoubleU d0, d1;
d0.ll = t0;
d1.ll = t1;
d0.d = float64_mul(d0.d, d1.d, &env->fp_status);
return d0.ll;
}
static float64 float64_ftsmul(float64 op1, uint64_t op2, float_status *stat)
{
float64 result = float64_mul(op1, op1, stat);
if (!float64_is_any_nan(result)) {
result = float64_set_sign(result, op2 & 1);
}
return result;
}
float64 float64_sin(float64 rad) {
float64 app;
float64 diff;
float64 m_rad2;
int inc;
app = diff = rad;
inc = 1;
m_rad2 = float64_neg(float64_mul(rad, rad, 3));
do {
diff = float64_div(float64_mul(diff, m_rad2, 3),
int32_to_float64((2 * inc) * (2 * inc + 1)), 3);
app = float64_add(app, diff, 3);
inc++;
} while (
float64_ge(float64_abs(diff), 0x3ee4f8b588e368f1ULL, 3)); /* 0.00001 */
return app;
}
/* 64-bit FP multiplication RM */
void HELPER(mdb)(CPUS390XState *env, uint32_t f1, uint64_t a2)
{
float64 v1 = env->fregs[f1].d;
CPU_DoubleU v2;
v2.ll = cpu_ldq_data(env, a2);
HELPER_LOG("%s: multiplying 0x%lx from f%d and 0x%ld\n", __func__,
v1, f1, v2.d);
env->fregs[f1].d = float64_mul(v1, v2.d, &env->fpu_status);
}
/* 64-bit FP multiply and add RM */
void HELPER(madb)(CPUS390XState *env, uint32_t f1, uint64_t a2, uint32_t f3)
{
CPU_DoubleU v2;
HELPER_LOG("%s: f1 %d a2 0x%lx f3 %d\n", __func__, f1, a2, f3);
v2.ll = cpu_ldq_data(env, a2);
env->fregs[f1].d = float64_add(env->fregs[f1].d,
float64_mul(v2.d, env->fregs[f3].d,
&env->fpu_status),
&env->fpu_status);
}
float64 HELPER(vfp_mulxd)(float64 a, float64 b, void *fpstp)
{
float_status *fpst = fpstp;
a = float64_squash_input_denormal(a, fpst);
b = float64_squash_input_denormal(b, fpst);
if ((float64_is_zero(a) && float64_is_infinity(b)) ||
(float64_is_infinity(a) && float64_is_zero(b))) {
/* 2.0 with the sign bit set to sign(A) XOR sign(B) */
return make_float64((1ULL << 62) |
((float64_val(a) ^ float64_val(b)) & (1ULL << 63)));
}
return float64_mul(a, b, fpst);
}
void *mult() {
int i, main_result = 0;
//#pragma omp parallel for num_threads(OMP_ACCEL) private(i)
for (i = 0; i < mulN; i++) {
float64 result;
// tid = omp_get_thread_num();
float64 x1 = mul_a_input[i];
float64 x2 = mul_b_input[i];
result = float64_mul(x1, x2, 1);
if (result != mul_z_output[i])
printf("result = %lld, expected = %lld\n", result, mul_z_output[i]);
main_result += (result == mul_z_output[i]);
// temp[tid] += (result == mul_z_output[i]);
}
printf("mult result = %d\n", main_result);
pthread_exit((void *)main_result);
}
int
main ()
{
int main_result;
int i,m,n,a;
m= 16; n= 8;
main_result = 0;
for (i = 0; i < N; i++)
{
buff[i*2+0] = a_input[i];
buff[i*2+1] = b_input[i];
}
*((volatile unsigned int *)0xf000f008) = (unsigned int)(&buff[0]);
*((volatile unsigned int *)0xf000f00C) = 2*N*8;
for(i=0;i<N;i++)
{ a = float64_mul (&buff[2*i], m, &result[i], n);
}
for(i=0;i<N;i++)
{
//printf("%llx ",result[i]);
main_result += (result[i] != z_output[i]);
}
// printf
// ("a_input=%016llx b_input=%016llx expected=%016llx output=%016llx (%lf)\n",
// a_input[i], b_input[i], z_output[i], result,
// ullong_to_double (result));
printf ("%d\n", main_result);
return main_result;
}
int
main ()
{
int main_result;
int i;
float64 x1, x2;
main_result = 0;
for (i = 0; i < N; i++)
{
float64 result;
x1 = a_input[i];
x2 = b_input[i];
result = float64_mul (x1, x2);
main_result += (result != z_output[i]);
printf
("a_input=%016llx b_input=%016llx expected=%016llx output=%016llx (%lf)\n",
a_input[i], b_input[i], z_output[i], result,
ullong_to_double (result));
}
printf ("%d\n", main_result);
return main_result;
}
/* 64-bit FP multiplication RR */
void HELPER(mdbr)(CPUS390XState *env, uint32_t f1, uint32_t f2)
{
env->fregs[f1].d = float64_mul(env->fregs[f1].d, env->fregs[f2].d,
&env->fpu_status);
}
float64 __muldf3(float64 A, float64 B)
{
return float64_mul(A, B);
}
unsigned int DoubleCPDO(const unsigned int opcode)
{
float64 rFm, rFn;
unsigned int Fd, Fm, Fn, nRc = 1;
//printk("DoubleCPDO(0x%08x)\n",opcode);
Fm = getFm(opcode);
if (CONSTANT_FM(opcode))
{
rFm = getDoubleConstant(Fm);
}
else
{
switch (fpa11->fType[Fm])
{
case typeSingle:
rFm = float32_to_float64(fpa11->fpreg[Fm].fSingle);
break;
case typeDouble:
rFm = fpa11->fpreg[Fm].fDouble;
break;
case typeExtended:
// !! patb
//printk("not implemented! why not?\n");
//!! ScottB
// should never get here, if extended involved
// then other operand should be promoted then
// ExtendedCPDO called.
break;
default: return 0;
}
}
if (!MONADIC_INSTRUCTION(opcode))
{
Fn = getFn(opcode);
switch (fpa11->fType[Fn])
{
case typeSingle:
rFn = float32_to_float64(fpa11->fpreg[Fn].fSingle);
break;
case typeDouble:
rFn = fpa11->fpreg[Fn].fDouble;
break;
default: return 0;
}
}
Fd = getFd(opcode);
/* !! this switch isn't optimized; better (opcode & MASK_ARITHMETIC_OPCODE)>>24, sort of */
switch (opcode & MASK_ARITHMETIC_OPCODE)
{
/* dyadic opcodes */
case ADF_CODE:
fpa11->fpreg[Fd].fDouble = float64_add(rFn,rFm);
break;
case MUF_CODE:
case FML_CODE:
fpa11->fpreg[Fd].fDouble = float64_mul(rFn,rFm);
break;
case SUF_CODE:
fpa11->fpreg[Fd].fDouble = float64_sub(rFn,rFm);
break;
case RSF_CODE:
fpa11->fpreg[Fd].fDouble = float64_sub(rFm,rFn);
break;
case DVF_CODE:
case FDV_CODE:
fpa11->fpreg[Fd].fDouble = float64_div(rFn,rFm);
break;
case RDF_CODE:
case FRD_CODE:
fpa11->fpreg[Fd].fDouble = float64_div(rFm,rFn);
break;
#if 0
case POW_CODE:
fpa11->fpreg[Fd].fDouble = float64_pow(rFn,rFm);
break;
case RPW_CODE:
fpa11->fpreg[Fd].fDouble = float64_pow(rFm,rFn);
break;
#endif
case RMF_CODE:
fpa11->fpreg[Fd].fDouble = float64_rem(rFn,rFm);
break;
#if 0
case POL_CODE:
fpa11->fpreg[Fd].fDouble = float64_pol(rFn,rFm);
break;
#endif
/* monadic opcodes */
case MVF_CODE:
fpa11->fpreg[Fd].fDouble = rFm;
break;
case MNF_CODE:
{
unsigned int *p = (unsigned int*)&rFm;
p[1] ^= 0x80000000;
fpa11->fpreg[Fd].fDouble = rFm;
}
break;
case ABS_CODE:
{
unsigned int *p = (unsigned int*)&rFm;
p[1] &= 0x7fffffff;
fpa11->fpreg[Fd].fDouble = rFm;
}
break;
case RND_CODE:
case URD_CODE:
fpa11->fpreg[Fd].fDouble =
int32_to_float64(float64_to_int32(rFm));
break;
case SQT_CODE:
fpa11->fpreg[Fd].fDouble = float64_sqrt(rFm);
break;
#if 0
case LOG_CODE:
fpa11->fpreg[Fd].fDouble = float64_log(rFm);
break;
case LGN_CODE:
fpa11->fpreg[Fd].fDouble = float64_ln(rFm);
break;
case EXP_CODE:
fpa11->fpreg[Fd].fDouble = float64_exp(rFm);
break;
case SIN_CODE:
fpa11->fpreg[Fd].fDouble = float64_sin(rFm);
break;
case COS_CODE:
fpa11->fpreg[Fd].fDouble = float64_cos(rFm);
break;
case TAN_CODE:
fpa11->fpreg[Fd].fDouble = float64_tan(rFm);
break;
case ASN_CODE:
fpa11->fpreg[Fd].fDouble = float64_arcsin(rFm);
break;
case ACS_CODE:
fpa11->fpreg[Fd].fDouble = float64_arccos(rFm);
break;
case ATN_CODE:
fpa11->fpreg[Fd].fDouble = float64_arctan(rFm);
break;
#endif
case NRM_CODE:
break;
default:
{
nRc = 0;
}
}
if (0 != nRc) fpa11->fType[Fd] = typeDouble;
return nRc;
}
float64 float64_pow(float64 rFn,float64 rFm)
{
return float64_exp(float64_mul(rFm,float64_ln(rFn)));
}
