forked from bartvdbraak/blender
Cycles: Cleanup, reduce indentation level
This commit is contained in:
Sergey Sharybin
parent
700330afe8
commit
65143542af
@ -87,7 +87,7 @@ ccl_device_inline int ray_triangle_intersect8(
|
||||
uint *num_hits,
|
||||
uint max_hits,
|
||||
int *num_hits_in_instance,
|
||||
float isec_t)
|
||||
float isect_t)
|
||||
{
|
||||
|
||||
const unsigned char prim_num_mask = (1 << prim_num) - 1;
|
||||
@ -217,156 +217,144 @@ ccl_device_inline int ray_triangle_intersect8(
|
||||
ccl_align(32) float den8[8], U8[8], V8[8], T8[8], sign_T8[8], xor_signmask8[8];
|
||||
ccl_align(32) unsigned int mask_minmaxUVW8[8];
|
||||
|
||||
if(visibility == PATH_RAY_SHADOW_OPAQUE){
|
||||
__m256i mask_final_256 = _mm256_cmpeq_epi32(mask_minmaxUVW_256, zero256);
|
||||
__m256i maskden256 = _mm256_cmpeq_epi32(_mm256_castps_si256(den_256), zero256);
|
||||
__m256i mask0 = _mm256_cmpgt_epi32(zero256, _mm256_castps_si256(sign_T_256));
|
||||
__m256 rayt_256 = _mm256_set1_ps((*isect)->t);
|
||||
__m256i mask1 = _mm256_cmpgt_epi32(_mm256_castps_si256(sign_T_256),
|
||||
_mm256_castps_si256(
|
||||
_mm256_mul_ps(_mm256_castsi256_ps(_mm256_xor_si256(_mm256_castps_si256(den_256), sign_den_256)), rayt_256)
|
||||
)
|
||||
);
|
||||
mask0 = _mm256_or_si256(mask1, mask0);
|
||||
mask_final_256 = _mm256_andnot_si256(mask0, mask_final_256); //(~mask_minmaxUVW_pos) &(~mask)
|
||||
mask_final_256 = _mm256_andnot_si256(maskden256, mask_final_256); //(~mask_minmaxUVW_pos) &(~mask) & (~maskden)
|
||||
unsigned char mask_final = _mm256_movemask_ps(_mm256_castsi256_ps(mask_final_256));
|
||||
if((mask_final & prim_num_mask) == 0) {
|
||||
return false;
|
||||
}
|
||||
unsigned long i = 0;
|
||||
if(visibility == PATH_RAY_SHADOW_OPAQUE) {
|
||||
__m256i mask_final_256 = _mm256_cmpeq_epi32(mask_minmaxUVW_256, zero256);
|
||||
__m256i maskden256 = _mm256_cmpeq_epi32(_mm256_castps_si256(den_256), zero256);
|
||||
__m256i mask0 = _mm256_cmpgt_epi32(zero256, _mm256_castps_si256(sign_T_256));
|
||||
__m256 rayt_256 = _mm256_set1_ps((*isect)->t);
|
||||
__m256i mask1 = _mm256_cmpgt_epi32(_mm256_castps_si256(sign_T_256),
|
||||
_mm256_castps_si256(
|
||||
_mm256_mul_ps(_mm256_castsi256_ps(_mm256_xor_si256(_mm256_castps_si256(den_256), sign_den_256)), rayt_256)
|
||||
)
|
||||
);
|
||||
mask0 = _mm256_or_si256(mask1, mask0);
|
||||
mask_final_256 = _mm256_andnot_si256(mask0, mask_final_256); //(~mask_minmaxUVW_pos) &(~mask)
|
||||
mask_final_256 = _mm256_andnot_si256(maskden256, mask_final_256); //(~mask_minmaxUVW_pos) &(~mask) & (~maskden)
|
||||
unsigned char mask_final = _mm256_movemask_ps(_mm256_castsi256_ps(mask_final_256));
|
||||
if((mask_final & prim_num_mask) == 0) {
|
||||
return false;
|
||||
}
|
||||
unsigned long i = 0;
|
||||
#if defined(_MSC_VER)
|
||||
unsigned char res = _BitScanForward(&i, (unsigned long)mask_final);
|
||||
unsigned char res = _BitScanForward(&i, (unsigned long)mask_final);
|
||||
#else
|
||||
i = __builtin_ffs(mask_final)-1;
|
||||
i = __builtin_ffs(mask_final)-1;
|
||||
#endif
|
||||
|
||||
__m256 inv_den_256 = _mm256_rcp_ps(den_256);
|
||||
U_256 = _mm256_mul_ps(U_256, inv_den_256);
|
||||
V_256 = _mm256_mul_ps(V_256, inv_den_256);
|
||||
T_256 = _mm256_mul_ps(T_256, inv_den_256);
|
||||
|
||||
_mm256_store_ps(U8, U_256);
|
||||
_mm256_store_ps(V8, V_256);
|
||||
_mm256_store_ps(T8, T_256);
|
||||
|
||||
/* NOTE: Here we assume visibility for all triangles in the node is
|
||||
* the same. */
|
||||
|
||||
(*isect)->u = U8[i];
|
||||
(*isect)->v = V8[i];
|
||||
(*isect)->t = T8[i];
|
||||
|
||||
(*isect)->prim = (prim_addr + i);
|
||||
(*isect)->object = object;
|
||||
(*isect)->type = PRIMITIVE_TRIANGLE;
|
||||
|
||||
return true;
|
||||
}
|
||||
__m256 inv_den_256 = _mm256_rcp_ps(den_256);
|
||||
U_256 = _mm256_mul_ps(U_256, inv_den_256);
|
||||
V_256 = _mm256_mul_ps(V_256, inv_den_256);
|
||||
T_256 = _mm256_mul_ps(T_256, inv_den_256);
|
||||
_mm256_store_ps(U8, U_256);
|
||||
_mm256_store_ps(V8, V_256);
|
||||
_mm256_store_ps(T8, T_256);
|
||||
/* NOTE: Here we assume visibility for all triangles in the node is
|
||||
* the same. */
|
||||
(*isect)->u = U8[i];
|
||||
(*isect)->v = V8[i];
|
||||
(*isect)->t = T8[i];
|
||||
(*isect)->prim = (prim_addr + i);
|
||||
(*isect)->object = object;
|
||||
(*isect)->type = PRIMITIVE_TRIANGLE;
|
||||
return true;
|
||||
}
|
||||
else {
|
||||
_mm256_store_ps(den8, den_256);
|
||||
_mm256_store_ps(U8, U_256);
|
||||
_mm256_store_ps(V8, V_256);
|
||||
_mm256_store_ps(T8, T_256);
|
||||
_mm256_store_ps(den8, den_256);
|
||||
_mm256_store_ps(U8, U_256);
|
||||
_mm256_store_ps(V8, V_256);
|
||||
_mm256_store_ps(T8, T_256);
|
||||
|
||||
_mm256_store_ps(sign_T8, sign_T_256);
|
||||
_mm256_store_ps(xor_signmask8, xor_signmask_256);
|
||||
_mm256_store_si256((__m256i*)mask_minmaxUVW8, mask_minmaxUVW_256);
|
||||
_mm256_store_ps(sign_T8, sign_T_256);
|
||||
_mm256_store_ps(xor_signmask8, xor_signmask_256);
|
||||
_mm256_store_si256((__m256i*)mask_minmaxUVW8, mask_minmaxUVW_256);
|
||||
|
||||
int ret = false;
|
||||
int ret = false;
|
||||
|
||||
if(visibility == PATH_RAY_SHADOW) {
|
||||
for(int i = 0; i < prim_num; i++) {
|
||||
if(!mask_minmaxUVW8[i]) {
|
||||
#ifdef __VISIBILITY_FLAG__
|
||||
if(kernel_tex_fetch(__prim_visibility, (prim_addr + i)) & visibility)
|
||||
#endif
|
||||
{
|
||||
if((sign_T8[i] >= 0.0f) &&
|
||||
(sign_T8[i] <= (*isect)->t * xor_signmask8[i]))
|
||||
{
|
||||
if(den8[i]) {
|
||||
const float inv_den = 1.0f / den8[i];
|
||||
|
||||
(*isect)->u = U8[i] * inv_den;
|
||||
(*isect)->v = V8[i] * inv_den;
|
||||
(*isect)->t = T8[i] * inv_den;
|
||||
|
||||
(*isect)->prim = (prim_addr + i);
|
||||
(*isect)->object = object;
|
||||
(*isect)->type = PRIMITIVE_TRIANGLE;
|
||||
|
||||
int prim = kernel_tex_fetch(__prim_index, (*isect)->prim);
|
||||
int shader = 0;
|
||||
|
||||
#ifdef __HAIR__
|
||||
if(kernel_tex_fetch(__prim_type, (*isect)->prim) & PRIMITIVE_ALL_TRIANGLE)
|
||||
#endif
|
||||
{
|
||||
shader = kernel_tex_fetch(__tri_shader, prim);
|
||||
}
|
||||
#ifdef __HAIR__
|
||||
else {
|
||||
float4 str = kernel_tex_fetch(__curves, prim);
|
||||
shader = __float_as_int(str.z);
|
||||
}
|
||||
#endif
|
||||
int flag = kernel_tex_fetch(__shaders, (shader & SHADER_MASK)).flags;
|
||||
|
||||
/* if no transparent shadows, all light is blocked. */
|
||||
if(!(flag & SD_HAS_TRANSPARENT_SHADOW)) {
|
||||
return 2;
|
||||
}
|
||||
/* if maximum number of hits reached, block all light. */
|
||||
else if(*num_hits == max_hits) {
|
||||
return 2;
|
||||
}
|
||||
/* move on to next entry in intersections array */
|
||||
ret = true;
|
||||
|
||||
(*isect)++;
|
||||
(*num_hits)++;
|
||||
|
||||
(*num_hits_in_instance)++;
|
||||
|
||||
(*isect)->t = isec_t;
|
||||
|
||||
} //den
|
||||
} //if sign
|
||||
} //vis
|
||||
}//if mask
|
||||
} //for
|
||||
}
|
||||
else { //default case
|
||||
if(visibility == PATH_RAY_SHADOW) {
|
||||
for(int i = 0; i < prim_num; i++) {
|
||||
if(!mask_minmaxUVW8[i]) {
|
||||
if(mask_minmaxUVW8[i]) {
|
||||
continue;
|
||||
}
|
||||
#ifdef __VISIBILITY_FLAG__
|
||||
if(kernel_tex_fetch(__prim_visibility, (prim_addr + i)) & visibility)
|
||||
if((kernel_tex_fetch(__prim_visibility, (prim_addr + i)) & visibility) == 0) {
|
||||
continue;
|
||||
}
|
||||
#endif
|
||||
{
|
||||
if((sign_T8[i] >= 0.0f) &&
|
||||
(sign_T8[i] <= (*isect)->t * xor_signmask8[i]))
|
||||
{
|
||||
if(den8[i]) {
|
||||
const float inv_den = 1.0f / den8[i];
|
||||
|
||||
(*isect)->u = U8[i] * inv_den;
|
||||
(*isect)->v = V8[i] * inv_den;
|
||||
(*isect)->t = T8[i] * inv_den;
|
||||
|
||||
(*isect)->prim = (prim_addr + i);
|
||||
(*isect)->object = object;
|
||||
(*isect)->type = PRIMITIVE_TRIANGLE;
|
||||
|
||||
ret = true;
|
||||
} //den
|
||||
} //if sign
|
||||
} //vis
|
||||
}//if mask
|
||||
} //for
|
||||
} //default
|
||||
if((sign_T8[i] < 0.0f) ||
|
||||
(sign_T8[i] > (*isect)->t * xor_signmask8[i]))
|
||||
{
|
||||
continue;
|
||||
}
|
||||
if(!den8[i]) {
|
||||
continue;
|
||||
}
|
||||
const float inv_den = 1.0f / den8[i];
|
||||
(*isect)->u = U8[i] * inv_den;
|
||||
(*isect)->v = V8[i] * inv_den;
|
||||
(*isect)->t = T8[i] * inv_den;
|
||||
(*isect)->prim = (prim_addr + i);
|
||||
(*isect)->object = object;
|
||||
(*isect)->type = PRIMITIVE_TRIANGLE;
|
||||
const int prim = kernel_tex_fetch(__prim_index, (*isect)->prim);
|
||||
int shader = 0;
|
||||
#ifdef __HAIR__
|
||||
if(kernel_tex_fetch(__prim_type, (*isect)->prim) & PRIMITIVE_ALL_TRIANGLE)
|
||||
#endif
|
||||
{
|
||||
shader = kernel_tex_fetch(__tri_shader, prim);
|
||||
}
|
||||
#ifdef __HAIR__
|
||||
else {
|
||||
float4 str = kernel_tex_fetch(__curves, prim);
|
||||
shader = __float_as_int(str.z);
|
||||
}
|
||||
#endif
|
||||
const int flag = kernel_tex_fetch(__shaders, (shader & SHADER_MASK)).flags;
|
||||
/* If no transparent shadows, all light is blocked. */
|
||||
if(!(flag & SD_HAS_TRANSPARENT_SHADOW)) {
|
||||
return 2;
|
||||
}
|
||||
/* If maximum number of hits reached, block all light. */
|
||||
else if(*num_hits == max_hits) {
|
||||
return 2;
|
||||
}
|
||||
/* Move on to next entry in intersections array. */
|
||||
ret = true;
|
||||
(*isect)++;
|
||||
(*num_hits)++;
|
||||
(*num_hits_in_instance)++;
|
||||
(*isect)->t = isect_t;
|
||||
}
|
||||
}
|
||||
else {
|
||||
for(int i = 0; i < prim_num; i++) {
|
||||
if(mask_minmaxUVW8[i]) {
|
||||
continue;
|
||||
}
|
||||
#ifdef __VISIBILITY_FLAG__
|
||||
if((kernel_tex_fetch(__prim_visibility, (prim_addr + i)) & visibility) == 0) {
|
||||
continue;
|
||||
}
|
||||
#endif
|
||||
if((sign_T8[i] < 0.0f) ||
|
||||
(sign_T8[i] > (*isect)->t * xor_signmask8[i]))
|
||||
{
|
||||
continue;
|
||||
}
|
||||
if(!den8[i]) {
|
||||
continue;
|
||||
}
|
||||
const float inv_den = 1.0f / den8[i];
|
||||
(*isect)->u = U8[i] * inv_den;
|
||||
(*isect)->v = V8[i] * inv_den;
|
||||
(*isect)->t = T8[i] * inv_den;
|
||||
(*isect)->prim = (prim_addr + i);
|
||||
(*isect)->object = object;
|
||||
(*isect)->type = PRIMITIVE_TRIANGLE;
|
||||
ret = true;
|
||||
}
|
||||
}
|
||||
return ret;
|
||||
}// else PATH_RAY_SHADOW_OPAQUE
|
||||
}
|
||||
}
|
||||
|
||||
ccl_device_inline int triangle_intersect8(
|
||||
@ -381,7 +369,7 @@ ccl_device_inline int triangle_intersect8(
|
||||
uint *num_hits,
|
||||
uint max_hits,
|
||||
int *num_hits_in_instance,
|
||||
float isec_t)
|
||||
float isect_t)
|
||||
{
|
||||
__m128 tri_a[8], tri_b[8], tri_c[8];
|
||||
__m256 tritmp[12], tri[12];
|
||||
@ -483,7 +471,7 @@ ccl_device_inline int triangle_intersect8(
|
||||
num_hits,
|
||||
max_hits,
|
||||
num_hits_in_instance,
|
||||
isec_t);
|
||||
isect_t);
|
||||
return result;
|
||||
}
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user