aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Fix a few places in examples where C reference implementaion had a double-precision

Browse Source 
fp constant undesirably causing computation to be done in double precision.

Makes C scalar versions of the options pricing models, rt, and aobench 3-5% faster.
Makes scalar version of noise about 15% faster.
Others are unchanged.
This commit is contained in:
Matt Pharr
2011-09-01 16:31:22 -07:00
parent eb7913f1dd
commit 99221f7d17
6 changed files with 24 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
examples/aobench/ao_serial.cpp
View File

@@ -140,7 +140,7 @@ ray_plane_intersect(Isect &isect, Ray &ray,
float d = -dot(plane.p, plane.n);
float v = dot(ray.dir, plane.n);
if (fabsf(v) < 1.0e-17)
if (fabsf(v) < 1.0e-17f)
return;
else {
float t = -(dot(ray.org, plane.n) + d) / v;
@@ -183,11 +183,11 @@ orthoBasis(vec basis[3], const vec &n) {
basis[2] = n;
basis[1].x = 0.0; basis[1].y = 0.0; basis[1].z = 0.0;
if ((n.x < 0.6) && (n.x > -0.6)) {
if ((n.x < 0.6f) && (n.x > -0.6f)) {
basis[1].x = 1.0;
} else if ((n.y < 0.6) && (n.y > -0.6)) {
} else if ((n.y < 0.6f) && (n.y > -0.6f)) {
basis[1].y = 1.0;
} else if ((n.z < 0.6) && (n.z > -0.6)) {
} else if ((n.z < 0.6f) && (n.z > -0.6f)) {
basis[1].z = 1.0;
} else {
basis[1].x = 1.0;
@@ -224,7 +224,7 @@ ambient_occlusion(Isect &isect, Plane &plane,
float phi = 2.0f * M_PI * drand48();
float x = cosf(phi) * theta;
float y = sinf(phi) * theta;
float z = sqrtf(1.0 - theta * theta);
float z = sqrtf(1.0f - theta * theta);
// local . global
float rx = x * basis[0].x + y * basis[1].x + z * basis[2].x;
@@ -236,14 +236,14 @@ ambient_occlusion(Isect &isect, Plane &plane,
ray.dir.y = ry;
ray.dir.z = rz;
occIsect.t = 1.0e+17;
occIsect.t = 1.0e+17f;
occIsect.hit = 0;
for (int snum = 0; snum < 3; ++snum)
ray_sphere_intersect(occIsect, ray, spheres[snum]);
ray_plane_intersect (occIsect, ray, plane);
if (occIsect.hit) occlusion += 1.0;
if (occIsect.hit) occlusion += 1.f;
}
}
@@ -280,10 +280,10 @@ static void ao_scanlines(int y0, int y1, int w, int h, int nsubsamples,
ray.dir.x = px;
ray.dir.y = py;
ray.dir.z = -1.0;
ray.dir.z = -1.0f;
vnormalize(ray.dir);
isect.t = 1.0e+17;
isect.t = 1.0e+17f;
isect.hit = 0;
for (int snum = 0; snum < 3; ++snum)

2
examples/mandelbrot/mandelbrot_serial.cpp
View File

@@ -36,7 +36,7 @@ static int mandel(float c_re, float c_im, int count) {
float z_re = c_re, z_im = c_im;
int i;
for (i = 0; i < count; ++i) {
if (z_re * z_re + z_im * z_im > 4.)
if (z_re * z_re + z_im * z_im > 4.f)
break;
float new_re = z_re*z_re - z_im*z_im;

2
examples/mandelbrot_tasks/mandelbrot_serial.cpp
View File

@@ -36,7 +36,7 @@ static int mandel(float c_re, float c_im, int count) {
float z_re = c_re, z_im = c_im;
int i;
for (i = 0; i < count; ++i) {
if (z_re * z_re + z_im * z_im > 4.)
if (z_re * z_re + z_im * z_im > 4.f)
break;
float new_re = z_re*z_re - z_im*z_im;

6
examples/noise/noise_serial.cpp
View File

@@ -104,7 +104,7 @@ inline float NoiseWeight(float t) {
inline float Lerp(float t, float low, float high) {
return (1. - t) * low + t * high;
return (1.f - t) * low + t * high;
}
@@ -147,7 +147,7 @@ static float Turbulence(float x, float y, float z, int octaves) {
lambda *= 1.99f;
o *= omega;
}
return sum * 0.5;
return sum * 0.5f;
}
@@ -163,7 +163,7 @@ void noise_serial(float x0, float y0, float x1, float y1,
float y = y0 + j * dy;
int index = (j * width + i);
output[index] = Turbulence(x, y, 0.6, 8);
output[index] = Turbulence(x, y, 0.6f, 8);
}
}
}

6
examples/options/options_serial.cpp
View File

@@ -47,7 +47,7 @@ static inline float
CND(float X) {
float L = fabsf(X);
float k = 1.0 / (1.0 + 0.2316419 * L);
float k = 1.f / (1.f + 0.2316419f * L);
float k2 = k*k;
float k3 = k2*k;
float k4 = k2*k2;
@@ -59,7 +59,7 @@ CND(float X) {
w *= invSqrt2Pi * expf(-L * L * .5f);
if (X > 0.f)
w = 1.0 - w;
w = 1.f - w;
return w;
}
@@ -94,7 +94,7 @@ binomial_put_serial(float Sa[], float Xa[], float Ta[],
float dt = T / BINOMIAL_NUM;
float u = expf(v * sqrtf(dt));
float d = 1. / u;
float d = 1.f / u;
float disc = expf(r * dt);
float Pu = (disc - d) / (u - d);

14
examples/volume_rendering/volume_serial.cpp
View File

@@ -104,7 +104,7 @@ Inside(float3 p, float3 pMin, float3 pMax) {
static bool
IntersectP(const Ray &ray, float3 pMin, float3 pMax, float *hit0, float *hit1) {
float t0 = -1e30, t1 = 1e30;
float t0 = -1e30f, t1 = 1e30f;
float3 tNear = (pMin - ray.origin) / ray.dir;
float3 tFar = (pMax - ray.origin) / ray.dir;
@@ -213,7 +213,7 @@ transmittance(float3 p0, float3 p1, float3 pMin,
float tau = 0;
float rayLength = sqrtf(ray.dir.x * ray.dir.x + ray.dir.y * ray.dir.y +
ray.dir.z * ray.dir.z);
float stepDist = 0.2;
float stepDist = 0.2f;
float stepT = stepDist / rayLength;
float t = rayT0;
@@ -239,8 +239,8 @@ distanceSquared(float3 a, float3 b) {
static float
raymarch(float density[], int nVoxels[3], const Ray &ray) {
float rayT0, rayT1;
float3 pMin(.3, -.2, .3), pMax(1.8, 2.3, 1.8);
float3 lightPos(-1, 4, 1.5);
float3 pMin(.3f, -.2f, .3f), pMax(1.8f, 2.3f, 1.8f);
float3 lightPos(-1f, 4f, 1.5f);
if (!IntersectP(ray, pMin, pMax, &rayT0, &rayT1))
return 0.;
@@ -249,10 +249,10 @@ raymarch(float density[], int nVoxels[3], const Ray &ray) {
// Parameters that define the volume scattering characteristics and
// sampling rate for raymarching
float Le = .25; // Emission coefficient
float Le = .25f; // Emission coefficient
float sigma_a = 10; // Absorption coefficient
float sigma_s = 10; // Scattering coefficient
float stepDist = 0.025; // Ray step amount
float stepDist = 0.025f; // Ray step amount
float lightIntensity = 40; // Light source intensity
float tau = 0.f; // accumulated beam transmittance
@@ -269,7 +269,7 @@ raymarch(float density[], int nVoxels[3], const Ray &ray) {
// terminate once attenuation is high
float atten = expf(-tau);
if (atten < .005)
if (atten < .005f)
break;
// direct lighting