Files
blender/intern/cycles/kernel/svm/math_util.h

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

259 lines
7.3 KiB
C
Raw Normal View History

/*
* Copyright 2011-2014 Blender Foundation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#pragma once
CCL_NAMESPACE_BEGIN
Cycles: Kernel address space changes for MSL This is the first of a sequence of changes to support compiling Cycles kernels as MSL (Metal Shading Language) in preparation for a Metal GPU device implementation. MSL requires that all pointer types be declared with explicit address space attributes (device, thread, etc...). There is already precedent for this with Cycles' address space macros (ccl_global, ccl_private, etc...), therefore the first step of MSL-enablement is to apply these consistently. Line-for-line this represents the largest change required to enable MSL. Applying this change first will simplify future patches as well as offering the emergent benefit of enhanced descriptiveness. The vast majority of deltas in this patch fall into one of two cases: - Ensuring ccl_private is specified for thread-local pointer types - Ensuring ccl_global is specified for device-wide pointer types Additionally, the ccl_addr_space qualifier can be removed. Prior to Cycles X, ccl_addr_space was used as a context-dependent address space qualifier, but now it is either redundant (e.g. in struct typedefs), or can be replaced by ccl_global in the case of pointer types. Associated function variants (e.g. lcg_step_float_addrspace) are also redundant. In cases where address space qualifiers are chained with "const", this patch places the address space qualifier first. The rationale for this is that the choice of address space is likely to have the greater impact on runtime performance and overall architecture. The final part of this patch is the addition of a metal/compat.h header. This is partially complete and will be extended in future patches, paving the way for the full Metal implementation. Ref T92212 Reviewed By: brecht Maniphest Tasks: T92212 Differential Revision: https://developer.blender.org/D12864
2021-10-14 13:53:40 +01:00
ccl_device void svm_vector_math(ccl_private float *value,
ccl_private float3 *vector,
2020-02-15 10:37:08 +11:00
NodeVectorMathType type,
float3 a,
float3 b,
float3 c,
float param1)
{
switch (type) {
case NODE_VECTOR_MATH_ADD:
*vector = a + b;
break;
case NODE_VECTOR_MATH_SUBTRACT:
*vector = a - b;
break;
case NODE_VECTOR_MATH_MULTIPLY:
*vector = a * b;
break;
case NODE_VECTOR_MATH_DIVIDE:
*vector = safe_divide_float3_float3(a, b);
break;
case NODE_VECTOR_MATH_CROSS_PRODUCT:
*vector = cross(a, b);
break;
case NODE_VECTOR_MATH_PROJECT:
*vector = project(a, b);
break;
case NODE_VECTOR_MATH_REFLECT:
*vector = reflect(a, b);
break;
case NODE_VECTOR_MATH_REFRACT:
*vector = refract(a, normalize(b), param1);
break;
case NODE_VECTOR_MATH_FACEFORWARD:
*vector = faceforward(a, b, c);
break;
case NODE_VECTOR_MATH_MULTIPLY_ADD:
*vector = a * b + c;
break;
case NODE_VECTOR_MATH_DOT_PRODUCT:
*value = dot(a, b);
break;
case NODE_VECTOR_MATH_DISTANCE:
*value = distance(a, b);
break;
case NODE_VECTOR_MATH_LENGTH:
*value = len(a);
break;
case NODE_VECTOR_MATH_SCALE:
*vector = a * param1;
break;
case NODE_VECTOR_MATH_NORMALIZE:
*vector = safe_normalize(a);
break;
case NODE_VECTOR_MATH_SNAP:
*vector = floor(safe_divide_float3_float3(a, b)) * b;
break;
case NODE_VECTOR_MATH_FLOOR:
*vector = floor(a);
break;
case NODE_VECTOR_MATH_CEIL:
*vector = ceil(a);
break;
case NODE_VECTOR_MATH_MODULO:
*vector = make_float3(safe_modulo(a.x, b.x), safe_modulo(a.y, b.y), safe_modulo(a.z, b.z));
break;
case NODE_VECTOR_MATH_WRAP:
2020-02-15 10:37:08 +11:00
*vector = make_float3(wrapf(a.x, b.x, c.x), wrapf(a.y, b.y, c.y), wrapf(a.z, b.z, c.z));
break;
case NODE_VECTOR_MATH_FRACTION:
*vector = a - floor(a);
break;
case NODE_VECTOR_MATH_ABSOLUTE:
*vector = fabs(a);
break;
case NODE_VECTOR_MATH_MINIMUM:
*vector = min(a, b);
break;
case NODE_VECTOR_MATH_MAXIMUM:
*vector = max(a, b);
break;
case NODE_VECTOR_MATH_SINE:
*vector = make_float3(sinf(a.x), sinf(a.y), sinf(a.z));
break;
case NODE_VECTOR_MATH_COSINE:
*vector = make_float3(cosf(a.x), cosf(a.y), cosf(a.z));
break;
case NODE_VECTOR_MATH_TANGENT:
*vector = make_float3(tanf(a.x), tanf(a.y), tanf(a.z));
break;
default:
*vector = zero_float3();
*value = 0.0f;
}
}
ccl_device float svm_math(NodeMathType type, float a, float b, float c)
{
switch (type) {
case NODE_MATH_ADD:
return a + b;
case NODE_MATH_SUBTRACT:
return a - b;
case NODE_MATH_MULTIPLY:
return a * b;
case NODE_MATH_DIVIDE:
return safe_divide(a, b);
case NODE_MATH_POWER:
return safe_powf(a, b);
case NODE_MATH_LOGARITHM:
return safe_logf(a, b);
case NODE_MATH_SQRT:
return safe_sqrtf(a);
case NODE_MATH_INV_SQRT:
return inversesqrtf(a);
case NODE_MATH_ABSOLUTE:
return fabsf(a);
case NODE_MATH_RADIANS:
return a * (M_PI_F / 180.0f);
case NODE_MATH_DEGREES:
return a * (180.0f / M_PI_F);
case NODE_MATH_MINIMUM:
return fminf(a, b);
case NODE_MATH_MAXIMUM:
return fmaxf(a, b);
case NODE_MATH_LESS_THAN:
return a < b;
case NODE_MATH_GREATER_THAN:
return a > b;
case NODE_MATH_ROUND:
return floorf(a + 0.5f);
case NODE_MATH_FLOOR:
return floorf(a);
case NODE_MATH_CEIL:
return ceilf(a);
case NODE_MATH_FRACTION:
return a - floorf(a);
case NODE_MATH_MODULO:
return safe_modulo(a, b);
case NODE_MATH_TRUNC:
return a >= 0.0f ? floorf(a) : ceilf(a);
case NODE_MATH_SNAP:
return floorf(safe_divide(a, b)) * b;
case NODE_MATH_WRAP:
return wrapf(a, b, c);
case NODE_MATH_PINGPONG:
return pingpongf(a, b);
case NODE_MATH_SINE:
return sinf(a);
case NODE_MATH_COSINE:
return cosf(a);
case NODE_MATH_TANGENT:
return tanf(a);
case NODE_MATH_SINH:
return sinhf(a);
case NODE_MATH_COSH:
return coshf(a);
case NODE_MATH_TANH:
return tanhf(a);
case NODE_MATH_ARCSINE:
return safe_asinf(a);
case NODE_MATH_ARCCOSINE:
return safe_acosf(a);
case NODE_MATH_ARCTANGENT:
return atanf(a);
case NODE_MATH_ARCTAN2:
return atan2f(a, b);
case NODE_MATH_SIGN:
return compatible_signf(a);
case NODE_MATH_EXPONENT:
return expf(a);
case NODE_MATH_COMPARE:
return ((a == b) || (fabsf(a - b) <= fmaxf(c, FLT_EPSILON))) ? 1.0f : 0.0f;
case NODE_MATH_MULTIPLY_ADD:
return a * b + c;
case NODE_MATH_SMOOTH_MIN:
return smoothminf(a, b, c);
case NODE_MATH_SMOOTH_MAX:
return -smoothminf(-a, -b, c);
default:
return 0.0f;
}
}
ccl_device float3 svm_math_blackbody_color(float t)
{
/* TODO(lukas): Reimplement in XYZ. */
/* Calculate color in range 800..12000 using an approximation
* a/x+bx+c for R and G and ((at + b)t + c)t + d) for B
* Max absolute error for RGB is (0.00095, 0.00077, 0.00057),
* which is enough to get the same 8 bit/channel color.
*/
if (t >= 12000.0f) {
return make_float3(0.826270103f, 0.994478524f, 1.56626022f);
}
else if (t < 965.0f) {
/* For 800 <= t < 965 color does not change in OSL implementation, so keep color the same */
return make_float3(4.70366907f, 0.0f, 0.0f);
}
2021-03-26 12:28:49 +11:00
/* Manually align for readability. */
/* clang-format off */
int i = (t >= 6365.0f) ? 5 :
(t >= 3315.0f) ? 4 :
(t >= 1902.0f) ? 3 :
(t >= 1449.0f) ? 2 :
(t >= 1167.0f) ? 1 :
0;
2021-03-26 12:28:49 +11:00
/* clang-format on */
ccl_constant float *r = blackbody_table_r[i];
ccl_constant float *g = blackbody_table_g[i];
ccl_constant float *b = blackbody_table_b[i];
const float t_inv = 1.0f / t;
return make_float3(r[0] * t_inv + r[1] * t + r[2],
g[0] * t_inv + g[1] * t + g[2],
((b[0] * t + b[1]) * t + b[2]) * t + b[3]);
}
ccl_device_inline float3 svm_math_gamma_color(float3 color, float gamma)
{
if (gamma == 0.0f)
return make_float3(1.0f, 1.0f, 1.0f);
if (color.x > 0.0f)
color.x = powf(color.x, gamma);
if (color.y > 0.0f)
color.y = powf(color.y, gamma);
if (color.z > 0.0f)
color.z = powf(color.z, gamma);
return color;
}
CCL_NAMESPACE_END