void vector_accumulate(Vector * accumulator, Vector v, float factor)
{
#ifdef USE_FMA
accumulator->x = fmaf(v.x, factor, accumulator->x);
accumulator->y = fmaf(v.y, factor, accumulator->y);
accumulator->z = fmaf(v.z, factor, accumulator->z);
#else
accumulator->x += v.x * factor;
accumulator->y += v.y * factor;
accumulator->z += v.z * factor;
#endif
}
float myexp(float x)
{
int sign = +1;
if (x < 0)
sign = -1;
x = abs(x);
const float z = x * M_LOG2E;
const int m = floorf(z);
const float w = z - m;
float r = 0.0;
const float u = w * M_LN2;
float r1 = ldexpf(1, m);
float r2 = 1.0/5040;;
//r2 = fmaf(r2, x, 1.0/5040);
r2 = fmaf(r2, x, 1.0/720);
r2 = fmaf(r2, x, 1.0/120);
r2 = fmaf(r2, x, 1.0/24);
r2 = fmaf(r2, x, 1.0/6);
r2 = fmaf(r2, x, 0.5);
r2 = fmaf(r2, x, 1);
r2 = fmaf(r2, x, 1);
r = fmaf(r1, r2, 0);
//r = r1 * r2;
return (sign < 0)? 1.0/r : r;
}
/**
* @brief Main program
* @param None
* @retval None
*/
int main(void)
{
/*!< At this stage the microcontroller clock setting is already configured,
this is done through SystemInit() function which is called from startup
file (startup_stm32f30x.s) before to branch to application main.
To reconfigure the default setting of SystemInit() function, refer to
system_stm32f30x.c file
*/
/* Initialize Leds mounted on STM32F3-discovery */
STM_EVAL_LEDInit(LED4);
/* all operations are repeated in infinite loop, at the beginning of operation
the LED4 will be toggled, the time spend between two toggling operations
is the time of mathematical operation processing : this time is different
if FPU is enabled or not */
while(1)
{
/* Toggle LD4 */
STM_EVAL_LEDToggle(LED4);
/* Initialze float variables */
Data0 = 1.0 ;
Data1 = 2.0 ;
/* Start a set of mathematical operations */
Result = fabsf(Data0) ;
Result = -Data0 ;
Result = Data1 + Data2 ;
Result = Data1 - Data2 ;
Result = Data0 + Data1 * Data2 ;
Result = Data0 - Data1 * Data2 ;
Result = -Data0 + Data1 * Data2 ;
Result = -Data0 - Data1 * Data2 ;
Result = fmaf(Data1,Data2,Data0) ;
Result = fmaf(-Data1,Data2,Data0) ;
Result = fmaf(Data1,Data2,-Data0) ;
Result = fmaf(-Data1,Data2,-Data0) ;
Result = Data1 / Data2 ;
Result = sqrtf(Data1) ;
Result = sqrt(Data1) ;
Result = fabs(Data1) ;
}
}
void test_fma()
{
static_assert((std::is_same<decltype(fma((double)0, (double)0, (double)0)), double>::value), "");
static_assert((std::is_same<decltype(fmaf(0,0,0)), float>::value), "");
static_assert((std::is_same<decltype(fmal(0,0,0)), long double>::value), "");
assert(fma(1,1,1) == 2);
}
CrankVertexP3N3*
crank_make_v_p3n3_sphere_uv (const guint uc,
const guint vc,
guint *len)
{
guint mlen;
CrankVertexP3N3V *vertices;
CrankVertexP3N3V *sides;
gfloat vang_inc = G_PI / (vc + 1);
gfloat vang_base = - G_PI_2 + vang_inc;
gfloat uang_inc = G_PI * 2 / uc;
guint top_vert;
guint j;
guint i;
mlen = 2 + uc * vc;
vertices = g_new (CrankVertexP3N3V, mlen);
sides = vertices + 1;
crank_vec_float3_init (& vertices[0].position , 0, 0, -1);
crank_vec_float3_copy (& vertices[0].position, & vertices[0].normal);
for (j = 0; j < vc; j++)
{
CrankVertexP3N3V *row = sides + (uc * j);
gfloat vang = fmaf (j, vang_inc, vang_base);
gfloat vcos = cosf (vang);
gfloat vsin = sinf (vang);
for (i = 0; i < uc; i++)
{
gfloat uang = i * uang_inc;
gfloat ucos = cosf (uang);
gfloat usin = sinf (uang);
crank_vec_float3_init (& row[i].position, ucos * vcos, usin * vcos, vsin);
crank_vec_float3_copy (& row[i].position, & row[i].normal);
}
}
top_vert = uc * vc + 1;
crank_vec_float3_init (& vertices[top_vert].position , 0, 0, 1);
crank_vec_float3_copy (& vertices[top_vert].position , & vertices[top_vert].normal);
if (len != NULL)
*len = mlen;
return (CrankVertexP3N3*)vertices;
}
Vector vector_add_scaled(Vector a, Vector b, float f)
{
Vector v =
{
#ifdef USE_FMA
fmaf(b.x, f, a.x),
fmaf(b.y, f, a.y),
fmaf(b.z, f, a.z)
#else
a.x + b.x * f,
a.y + b.y * f,
a.z + b.z * f
#endif
};
return v;
}
/*++
Function:
fmaf
See MSDN.
--*/
PALIMPORT float __cdecl PAL_fmaf(float x, float y, float z)
{
float ret;
PERF_ENTRY(fmaf);
ENTRY("fmaf (x=%f, y=%f, z=%f)\n", x, y, z);
ret = fmaf(x, y, z);
LOGEXIT("fma returns float %f\n", ret);
PERF_EXIT(fmaf);
return ret;
}
// CHECK-YES-LABEL: define void @test_fma
// CHECK-NO-LABEL: define void @test_fma
void test_fma(float a0, double a1, long double a2) {
// CHECK-YES: call float @llvm.fma.f32
// CHECK-NO: call float @llvm.fma.f32
float l0 = fmaf(a0, a0, a0);
// CHECK-YES: call double @llvm.fma.f64
// CHECK-NO: call double @llvm.fma.f64
double l1 = fma(a1, a1, a1);
// CHECK-YES: call x86_fp80 @llvm.fma.f80
// CHECK-NO: call x86_fp80 @llvm.fma.f80
long double l2 = fmal(a2, a2, a2);
}
__global__ void FloatMathPrecise() {
int iX;
float fX, fY;
acosf(1.0f);
acoshf(1.0f);
asinf(0.0f);
asinhf(0.0f);
atan2f(0.0f, 1.0f);
atanf(0.0f);
atanhf(0.0f);
cbrtf(0.0f);
fX = ceilf(0.0f);
fX = copysignf(1.0f, -2.0f);
cosf(0.0f);
coshf(0.0f);
cospif(0.0f);
cyl_bessel_i0f(0.0f);
cyl_bessel_i1f(0.0f);
erfcf(0.0f);
erfcinvf(2.0f);
erfcxf(0.0f);
erff(0.0f);
erfinvf(1.0f);
exp10f(0.0f);
exp2f(0.0f);
expf(0.0f);
expm1f(0.0f);
fX = fabsf(1.0f);
fdimf(1.0f, 0.0f);
fdividef(0.0f, 1.0f);
fX = floorf(0.0f);
fmaf(1.0f, 2.0f, 3.0f);
fX = fmaxf(0.0f, 0.0f);
fX = fminf(0.0f, 0.0f);
fmodf(0.0f, 1.0f);
frexpf(0.0f, &iX);
hypotf(1.0f, 0.0f);
ilogbf(1.0f);
isfinite(0.0f);
fX = isinf(0.0f);
fX = isnan(0.0f);
j0f(0.0f);
j1f(0.0f);
jnf(-1.0f, 1.0f);
ldexpf(0.0f, 0);
lgammaf(1.0f);
llrintf(0.0f);
llroundf(0.0f);
log10f(1.0f);
log1pf(-1.0f);
log2f(1.0f);
logbf(1.0f);
logf(1.0f);
lrintf(0.0f);
lroundf(0.0f);
modff(0.0f, &fX);
fX = nanf("1");
fX = nearbyintf(0.0f);
nextafterf(0.0f, 0.0f);
norm3df(1.0f, 0.0f, 0.0f);
norm4df(1.0f, 0.0f, 0.0f, 0.0f);
normcdff(0.0f);
normcdfinvf(1.0f);
fX = 1.0f;
normf(1, &fX);
powf(1.0f, 0.0f);
rcbrtf(1.0f);
remainderf(2.0f, 1.0f);
remquof(1.0f, 2.0f, &iX);
rhypotf(0.0f, 1.0f);
fY = rintf(1.0f);
rnorm3df(0.0f, 0.0f, 1.0f);
rnorm4df(0.0f, 0.0f, 0.0f, 1.0f);
fX = 1.0f;
rnormf(1, &fX);
fY = roundf(0.0f);
rsqrtf(1.0f);
scalblnf(0.0f, 1);
scalbnf(0.0f, 1);
signbit(1.0f);
sincosf(0.0f, &fX, &fY);
sincospif(0.0f, &fX, &fY);
sinf(0.0f);
sinhf(0.0f);
sinpif(0.0f);
sqrtf(0.0f);
tanf(0.0f);
tanhf(0.0f);
tgammaf(2.0f);
fY = truncf(0.0f);
y0f(1.0f);
y1f(1.0f);
ynf(1, 1.0f);
}
address TemplateInterpreterGenerator::generate_math_entry(AbstractInterpreter::MethodKind kind) {
// rbx,: Method*
// rcx: scratrch
// r13: sender sp
if (!InlineIntrinsics) return NULL; // Generate a vanilla entry
address entry_point = __ pc();
// These don't need a safepoint check because they aren't virtually
// callable. We won't enter these intrinsics from compiled code.
// If in the future we added an intrinsic which was virtually callable
// we'd have to worry about how to safepoint so that this code is used.
// mathematical functions inlined by compiler
// (interpreter must provide identical implementation
// in order to avoid monotonicity bugs when switching
// from interpreter to compiler in the middle of some
// computation)
//
// stack: [ ret adr ] <-- rsp
// [ lo(arg) ]
// [ hi(arg) ]
//
if (kind == Interpreter::java_lang_math_fmaD) {
__ movdbl(xmm0, Address(rsp, wordSize));
__ movdbl(xmm1, Address(rsp, 3 * wordSize));
__ movdbl(xmm2, Address(rsp, 5 * wordSize));
__ fmad(xmm0, xmm1, xmm2, xmm0);
} else if (kind == Interpreter::java_lang_math_fmaF) {
__ movflt(xmm0, Address(rsp, wordSize));
__ movflt(xmm1, Address(rsp, 2 * wordSize));
__ movflt(xmm2, Address(rsp, 3 * wordSize));
__ fmaf(xmm0, xmm1, xmm2, xmm0);
} else if (kind == Interpreter::java_lang_math_sqrt) {
__ sqrtsd(xmm0, Address(rsp, wordSize));
} else if (kind == Interpreter::java_lang_math_exp) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dexp() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dexp())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dexp));
}
} else if (kind == Interpreter::java_lang_math_log) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dlog() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dlog())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dlog));
}
} else if (kind == Interpreter::java_lang_math_log10) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dlog10() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dlog10())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dlog10));
}
} else if (kind == Interpreter::java_lang_math_sin) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dsin() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dsin())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dsin));
}
} else if (kind == Interpreter::java_lang_math_cos) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dcos() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dcos())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dcos));
}
} else if (kind == Interpreter::java_lang_math_pow) {
__ movdbl(xmm1, Address(rsp, wordSize));
__ movdbl(xmm0, Address(rsp, 3 * wordSize));
if (StubRoutines::dpow() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dpow())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dpow));
}
} else if (kind == Interpreter::java_lang_math_tan) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dtan() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dtan())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dtan));
}
} else {
__ fld_d(Address(rsp, wordSize));
switch (kind) {
case Interpreter::java_lang_math_abs:
__ fabs();
break;
default:
ShouldNotReachHere();
}
// return double result in xmm0 for interpreter and compilers.
__ subptr(rsp, 2*wordSize);
// Round to 64bit precision
__ fstp_d(Address(rsp, 0));
__ movdbl(xmm0, Address(rsp, 0));
__ addptr(rsp, 2*wordSize);
}
__ pop(rax);
__ mov(rsp, r13);
__ jmp(rax);
return entry_point;
}
std::pair<float,float> TwoMul(float a,float b)
{
float res = a * b;
float err = fmaf(a,b,-res);
return std::make_pair(res,err);
}
sl_def(do_test, void)
{
#define x values[p].d
#define xf values[p].f
#define y values[p+1].d
#define yf values[p+1].f
#define n values[p+2].i
#define nl values[p+2].l
#define call1(F) \
values[p].desc = #F; \
values[p].d = F(x); \
++p; \
values[p].desc = #F "f"; \
values[p].f = F ## f (xf); \
++p
#define call1i(F) \
values[p].desc = #F; \
values[p].i = F(x); \
++p
#define call2(F) \
values[p].desc = #F; \
values[p].d = F(x, y); \
p += 2; \
values[p].desc = #F "f"; \
values[p].f = F ## f (xf, yf); \
p += 2
#define call2i(F) \
values[p].desc = #F; \
values[p].i = F(x, y); \
p += 2; \
values[p].desc = #F "f"; \
values[p].i = F(xf, yf); \
p += 2
/* classify */
call1i(fpclassify);
call1i(signbit);
call1i(isfinite);
call1i(isnormal);
call1i(isnan);
call1i(isinf);
/* trig */
call1(acos);
call1(asin);
call1(atan);
call2(atan2);
call1(cos);
call1(sin);
call1(tan);
/* hyperbolic */
call1(acosh);
call1(asinh);
call1(atanh);
call1(cosh);
call1(sinh);
call1(tanh);
/* exp/log */
call1(exp);
call1(exp2);
call1(expm1);
values[p].desc = "frexp";
values[p].d = frexp(x, &values[p+1].i);
p += 2;
values[p].desc = "frexpf";
values[p].f = frexpf(xf, &values[p+1].i);
p += 2;
values[p].desc = "ilogb";
values[p].i = ilogb(x); p++;
values[p].desc = "ilogbf";
values[p].i = ilogbf(xf); p++;
values[p].desc = "ldexp";
values[p].d = ldexp(x, n); p+=3;
values[p].desc = "ldexpf";
values[p].f = ldexpf(xf, n); p+=3;
call1(log);
call1(log10);
call1(log1p);
call1(log2);
call1(logb);
values[p].desc = "modf";
values[p].d = modf(x, &y);
p += 2;
values[p].desc = "modff";
values[p].f = modff(xf, &yf);
p += 2;
values[p].desc = "scalbn";
values[p].d = scalbn(x, n); p+=3;
values[p].desc = "scalbnf";
values[p].f = scalbnf(xf, n); p+=3;
values[p].desc = "scalbln";
values[p].d = scalbln(x, nl); p+=3;
values[p].desc = "scalblnf";
values[p].f = scalblnf(xf, nl); p+=3;
/* power/abs */
call1(cbrt);
call1(fabs);
call2(hypot);
call2(pow);
call1(sqrt);
/* error/gamma */
call1(erf);
call1(erfc);
call1(lgamma);
call1(tgamma);
call1(ceil);
call1(floor);
call1(nearbyint);
call1(rint);
call1(lrint);
values[p].desc = "lrint";
values[p].l = lrint(x); p++;
values[p].desc = "lrintf";
values[p].l = lrintf(xf); p++;
values[p].desc = "llrint";
values[p].ll = llrint(x); p++;
values[p].desc = "llrintf";
values[p].ll = llrintf(xf); p++;
call1(round);
values[p].desc = "lround";
values[p].l = lround(x); p++;
values[p].desc = "lroundf";
values[p].l = lroundf(xf); p++;
values[p].desc = "llround";
values[p].ll = llround(x); p++;
values[p].desc = "llroundf";
values[p].ll = llroundf(xf); p++;
call1(trunc);
/* rem/mod */
call2(fmod);
call2(remainder);
values[p].desc = "remquo";
values[p].d = remquo(x, y, &values[p+1].i); p+=2;
values[p].desc = "remquof";
values[p].f = remquof(xf, yf, &values[p+1].i); p+=2;
call2(copysign);
/* nan */
values[p].desc = "nan";
values[p].d = nan(""); ++p;
values[p].desc = "nanf";
values[p].f = nanf(""); ++p;
call2(nextafter);
/* min/max/dim */
call2(fdim);
call2(fmax);
call2(fmin);
values[p].desc = "fma";
values[p].d = fma(x, y, values[p+2].d); p+=3;
values[p].desc = "fmaf";
values[p].d = fmaf(xf, yf, values[p+2].f); p+=3;
/* comp */
call2i(isgreater);
call2i(isgreaterequal);
call2i(isless);
call2i(islessequal);
call2i(islessgreater);
call2i(isunordered);
#undef x
#undef xf
#undef y
#undef n
}
void
domathf (void)
{
#ifndef NO_FLOAT
float f1;
float f2;
int i1;
f1 = acosf(0.0);
fprintf( stdout, "acosf : %f\n", f1);
f1 = acoshf(0.0);
fprintf( stdout, "acoshf : %f\n", f1);
f1 = asinf(1.0);
fprintf( stdout, "asinf : %f\n", f1);
f1 = asinhf(1.0);
fprintf( stdout, "asinhf : %f\n", f1);
f1 = atanf(M_PI_4);
fprintf( stdout, "atanf : %f\n", f1);
f1 = atan2f(2.3, 2.3);
fprintf( stdout, "atan2f : %f\n", f1);
f1 = atanhf(1.0);
fprintf( stdout, "atanhf : %f\n", f1);
f1 = cbrtf(27.0);
fprintf( stdout, "cbrtf : %f\n", f1);
f1 = ceilf(3.5);
fprintf( stdout, "ceilf : %f\n", f1);
f1 = copysignf(3.5, -2.5);
fprintf( stdout, "copysignf : %f\n", f1);
f1 = cosf(M_PI_2);
fprintf( stdout, "cosf : %f\n", f1);
f1 = coshf(M_PI_2);
fprintf( stdout, "coshf : %f\n", f1);
f1 = erff(42.0);
fprintf( stdout, "erff : %f\n", f1);
f1 = erfcf(42.0);
fprintf( stdout, "erfcf : %f\n", f1);
f1 = expf(0.42);
fprintf( stdout, "expf : %f\n", f1);
f1 = exp2f(0.42);
fprintf( stdout, "exp2f : %f\n", f1);
f1 = expm1f(0.00042);
fprintf( stdout, "expm1f : %f\n", f1);
f1 = fabsf(-1.123);
fprintf( stdout, "fabsf : %f\n", f1);
f1 = fdimf(1.123, 2.123);
fprintf( stdout, "fdimf : %f\n", f1);
f1 = floorf(0.5);
fprintf( stdout, "floorf : %f\n", f1);
f1 = floorf(-0.5);
fprintf( stdout, "floorf : %f\n", f1);
f1 = fmaf(2.1, 2.2, 3.01);
fprintf( stdout, "fmaf : %f\n", f1);
f1 = fmaxf(-0.42, 0.42);
fprintf( stdout, "fmaxf : %f\n", f1);
f1 = fminf(-0.42, 0.42);
fprintf( stdout, "fminf : %f\n", f1);
f1 = fmodf(42.0, 3.0);
fprintf( stdout, "fmodf : %f\n", f1);
/* no type-specific variant */
i1 = fpclassify(1.0);
fprintf( stdout, "fpclassify : %d\n", i1);
f1 = frexpf(42.0, &i1);
fprintf( stdout, "frexpf : %f\n", f1);
f1 = hypotf(42.0, 42.0);
fprintf( stdout, "hypotf : %f\n", f1);
i1 = ilogbf(42.0);
fprintf( stdout, "ilogbf : %d\n", i1);
/* no type-specific variant */
i1 = isfinite(3.0);
fprintf( stdout, "isfinite : %d\n", i1);
/* no type-specific variant */
i1 = isgreater(3.0, 3.1);
fprintf( stdout, "isgreater : %d\n", i1);
/* no type-specific variant */
i1 = isgreaterequal(3.0, 3.1);
fprintf( stdout, "isgreaterequal : %d\n", i1);
/* no type-specific variant */
i1 = isinf(3.0);
fprintf( stdout, "isinf : %d\n", i1);
/* no type-specific variant */
i1 = isless(3.0, 3.1);
fprintf( stdout, "isless : %d\n", i1);
/* no type-specific variant */
i1 = islessequal(3.0, 3.1);
fprintf( stdout, "islessequal : %d\n", i1);
/* no type-specific variant */
i1 = islessgreater(3.0, 3.1);
fprintf( stdout, "islessgreater : %d\n", i1);
/* no type-specific variant */
i1 = isnan(0.0);
fprintf( stdout, "isnan : %d\n", i1);
/* no type-specific variant */
i1 = isnormal(3.0);
fprintf( stdout, "isnormal : %d\n", i1);
/* no type-specific variant */
f1 = isunordered(1.0, 2.0);
fprintf( stdout, "isunordered : %d\n", i1);
f1 = j0f(1.2);
fprintf( stdout, "j0f : %f\n", f1);
f1 = j1f(1.2);
fprintf( stdout, "j1f : %f\n", f1);
f1 = jnf(2,1.2);
fprintf( stdout, "jnf : %f\n", f1);
f1 = ldexpf(1.2,3);
fprintf( stdout, "ldexpf : %f\n", f1);
f1 = lgammaf(42.0);
fprintf( stdout, "lgammaf : %f\n", f1);
f1 = llrintf(-0.5);
fprintf( stdout, "llrintf : %f\n", f1);
f1 = llrintf(0.5);
fprintf( stdout, "llrintf : %f\n", f1);
f1 = llroundf(-0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = llroundf(0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = logf(42.0);
fprintf( stdout, "logf : %f\n", f1);
f1 = log10f(42.0);
fprintf( stdout, "log10f : %f\n", f1);
f1 = log1pf(42.0);
fprintf( stdout, "log1pf : %f\n", f1);
f1 = log2f(42.0);
fprintf( stdout, "log2f : %f\n", f1);
f1 = logbf(42.0);
fprintf( stdout, "logbf : %f\n", f1);
f1 = lrintf(-0.5);
fprintf( stdout, "lrintf : %f\n", f1);
f1 = lrintf(0.5);
fprintf( stdout, "lrintf : %f\n", f1);
f1 = lroundf(-0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = lroundf(0.5);
fprintf( stdout, "lroundf : %f\n", f1);
f1 = modff(42.0,&f2);
fprintf( stdout, "lmodff : %f\n", f1);
f1 = nanf("");
fprintf( stdout, "nanf : %f\n", f1);
f1 = nearbyintf(1.5);
fprintf( stdout, "nearbyintf : %f\n", f1);
f1 = nextafterf(1.5,2.0);
fprintf( stdout, "nextafterf : %f\n", f1);
f1 = powf(3.01, 2.0);
fprintf( stdout, "powf : %f\n", f1);
f1 = remainderf(3.01,2.0);
fprintf( stdout, "remainderf : %f\n", f1);
f1 = remquof(29.0,3.0,&i1);
fprintf( stdout, "remquof : %f\n", f1);
f1 = rintf(0.5);
fprintf( stdout, "rintf : %f\n", f1);
f1 = rintf(-0.5);
fprintf( stdout, "rintf : %f\n", f1);
f1 = roundf(0.5);
fprintf( stdout, "roundf : %f\n", f1);
f1 = roundf(-0.5);
fprintf( stdout, "roundf : %f\n", f1);
f1 = scalblnf(1.2,3);
fprintf( stdout, "scalblnf : %f\n", f1);
f1 = scalbnf(1.2,3);
fprintf( stdout, "scalbnf : %f\n", f1);
/* no type-specific variant */
i1 = signbit(1.0);
fprintf( stdout, "signbit : %i\n", i1);
f1 = sinf(M_PI_4);
fprintf( stdout, "sinf : %f\n", f1);
f1 = sinhf(M_PI_4);
fprintf( stdout, "sinhf : %f\n", f1);
f1 = sqrtf(9.0);
fprintf( stdout, "sqrtf : %f\n", f1);
f1 = tanf(M_PI_4);
fprintf( stdout, "tanf : %f\n", f1);
f1 = tanhf(M_PI_4);
fprintf( stdout, "tanhf : %f\n", f1);
f1 = tgammaf(2.1);
fprintf( stdout, "tgammaf : %f\n", f1);
f1 = truncf(3.5);
fprintf( stdout, "truncf : %f\n", f1);
f1 = y0f(1.2);
fprintf( stdout, "y0f : %f\n", f1);
f1 = y1f(1.2);
fprintf( stdout, "y1f : %f\n", f1);
f1 = ynf(3,1.2);
fprintf( stdout, "ynf : %f\n", f1);
#endif
}
static int testf(float float_x, long double long_double_x, /*float complex float_complex_x,*/ int int_x, long long_x)
{
int r = 0;
r += acosf(float_x);
r += acoshf(float_x);
r += asinf(float_x);
r += asinhf(float_x);
r += atan2f(float_x, float_x);
r += atanf(float_x);
r += atanhf(float_x);
/*r += cargf(float_complex_x); - will fight with complex numbers later */
r += cbrtf(float_x);
r += ceilf(float_x);
r += copysignf(float_x, float_x);
r += cosf(float_x);
r += coshf(float_x);
r += erfcf(float_x);
r += erff(float_x);
r += exp2f(float_x);
r += expf(float_x);
r += expm1f(float_x);
r += fabsf(float_x);
r += fdimf(float_x, float_x);
r += floorf(float_x);
r += fmaf(float_x, float_x, float_x);
r += fmaxf(float_x, float_x);
r += fminf(float_x, float_x);
r += fmodf(float_x, float_x);
r += frexpf(float_x, &int_x);
r += gammaf(float_x);
r += hypotf(float_x, float_x);
r += ilogbf(float_x);
r += ldexpf(float_x, int_x);
r += lgammaf(float_x);
r += llrintf(float_x);
r += llroundf(float_x);
r += log10f(float_x);
r += log1pf(float_x);
r += log2f(float_x);
r += logbf(float_x);
r += logf(float_x);
r += lrintf(float_x);
r += lroundf(float_x);
r += modff(float_x, &float_x);
r += nearbyintf(float_x);
r += nexttowardf(float_x, long_double_x);
r += powf(float_x, float_x);
r += remainderf(float_x, float_x);
r += remquof(float_x, float_x, &int_x);
r += rintf(float_x);
r += roundf(float_x);
#ifdef __UCLIBC_SUSV3_LEGACY__
r += scalbf(float_x, float_x);
#endif
r += scalblnf(float_x, long_x);
r += scalbnf(float_x, int_x);
r += significandf(float_x);
r += sinf(float_x);
r += sinhf(float_x);
r += sqrtf(float_x);
r += tanf(float_x);
r += tanhf(float_x);
r += tgammaf(float_x);
r += truncf(float_x);
return r;
}
__host__ void single_precision_math_functions()
{
int iX;
float fX, fY;
acosf(1.0f);
acoshf(1.0f);
asinf(0.0f);
asinhf(0.0f);
atan2f(0.0f, 1.0f);
atanf(0.0f);
atanhf(0.0f);
cbrtf(0.0f);
ceilf(0.0f);
copysignf(1.0f, -2.0f);
cosf(0.0f);
coshf(0.0f);
//cospif(0.0f);
//cyl_bessel_i0f(0.0f);
//cyl_bessel_i1f(0.0f);
erfcf(0.0f);
//erfcinvf(2.0f);
//erfcxf(0.0f);
erff(0.0f);
//erfinvf(1.0f);
exp10f(0.0f);
exp2f(0.0f);
expf(0.0f);
expm1f(0.0f);
fabsf(1.0f);
fdimf(1.0f, 0.0f);
//fdividef(0.0f, 1.0f);
floorf(0.0f);
fmaf(1.0f, 2.0f, 3.0f);
fmaxf(0.0f, 0.0f);
fminf(0.0f, 0.0f);
fmodf(0.0f, 1.0f);
frexpf(0.0f, &iX);
hypotf(1.0f, 0.0f);
ilogbf(1.0f);
isfinite(0.0f);
isinf(0.0f);
isnan(0.0f);
///j0f(0.0f);
///j1f(0.0f);
///jnf(-1.0f, 1.0f);
ldexpf(0.0f, 0);
///lgammaf(1.0f);
///llrintf(0.0f);
///llroundf(0.0f);
log10f(1.0f);
log1pf(-1.0f);
log2f(1.0f);
logbf(1.0f);
logf(1.0f);
///lrintf(0.0f);
///lroundf(0.0f);
modff(0.0f, &fX);
///nanf("1");
nearbyintf(0.0f);
//nextafterf(0.0f);
//norm3df(1.0f, 0.0f, 0.0f);
//norm4df(1.0f, 0.0f, 0.0f, 0.0f);
//normcdff(0.0f);
//normcdfinvf(1.0f);
//fX = 1.0f; normf(1, &fX);
powf(1.0f, 0.0f);
//rcbrtf(1.0f);
remainderf(2.0f, 1.0f);
remquof(1.0f, 2.0f, &iX);
//rhypotf(0.0f, 1.0f);
///rintf(1.0f);
//rnorm3df(0.0f, 0.0f, 1.0f);
//rnorm4df(0.0f, 0.0f, 0.0f, 1.0f);
//fX = 1.0f; rnormf(1, &fX);
roundf(0.0f);
//rsqrtf(1.0f);
///scalblnf(0.0f, 1);
scalbnf(0.0f, 1);
signbit(1.0f);
sincosf(0.0f, &fX, &fY);
//sincospif(0.0f, &fX, &fY);
sinf(0.0f);
sinhf(0.0f);
//sinpif(0.0f);
sqrtf(0.0f);
tanf(0.0f);
tanhf(0.0f);
tgammaf(2.0f);
truncf(0.0f);
///y0f(1.0f);
///y1f(1.0f);
///ynf(1, 1.0f);
}
int main(int argc, char *argv[])
{
float x = 0.0;
if (argv) x = fmaf((float) argc, (float) argc, (float) argc);
return 0;
}
CoglVertexP3*
crank_make_v_p3_sphere_uv (const guint uc,
const guint vc,
guint *len)
{
guint mlen;
CoglVertexP3 *vertices;
CoglVertexP3 *sides;
gfloat vang_inc = G_PI / (vc + 1);
gfloat vang_base = - G_PI_2 + vang_inc;
gfloat uang_inc = G_PI * 2 / uc;
guint top_vert;
guint j;
guint i;
mlen = 2 + uc * vc;
vertices = g_new (CoglVertexP3, mlen);
sides = vertices + 1;
vertices[0].x = 0;
vertices[0].y = 0;
vertices[0].z = -1;
for (j = 0; j < vc; j++)
{
CoglVertexP3 *row = sides + (uc * j);
gfloat vang = fmaf (j, vang_inc, vang_base);
gfloat vcos = cosf (vang);
gfloat vsin = sinf (vang);
for (i = 0; i < uc; i++)
{
gfloat uang = i * uang_inc;
gfloat ucos = cosf (uang);
gfloat usin = sinf (uang);
row[i].x = ucos * vcos;
row[i].y = usin * vcos;
row[i].z = vsin;
}
}
top_vert = uc * vc + 1;
vertices[top_vert].x = 0;
vertices[top_vert].y = 0;
vertices[top_vert].z = -1;
if (len != NULL)
*len = mlen;
return vertices;
}
TEST(math, fmaf) {
ASSERT_FLOAT_EQ(10.0f, fmaf(2.0f, 3.0f, 4.0f));
}