Cr8-xyz/blender
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Cycles: Tweaks for 32bit CUDA binaries

Browse Source 
Tweak some inline policies. Not totally crazy yet, and in fact we now
have one less ifdef statement now.
This commit is contained in:
Sergey Sharybin
2016-02-15 19:11:02 +01:00
parent 1336e97b12
commit c93069083e
4 changed files with 12 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
intern/cycles/kernel/geom/geom_motion_triangle.h
View File

@@ -178,7 +178,12 @@ ccl_device_inline float3 motion_triangle_refine(KernelGlobals *kg, ShaderData *s
/* Same as above, except that isect->t is assumed to be in object space for instancing */ /* Same as above, except that isect->t is assumed to be in object space for instancing */
#ifdef __SUBSURFACE__ #ifdef __SUBSURFACE__
ccl_device_inline float3 motion_triangle_refine_subsurface(KernelGlobals *kg, ShaderData *sd, const Intersection *isect, const Ray *ray, float3 verts[3]) # if defined(__KERNEL_CUDA__) && (defined(i386) || defined(_M_IX86))
ccl_device_noinline
# else
ccl_device_inline
# endif
float3 motion_triangle_refine_subsurface(KernelGlobals *kg, ShaderData *sd, const Intersection *isect, const Ray *ray, float3 verts[3])
{ {
float3 P = ray->P; float3 P = ray->P;
float3 D = ray->D; float3 D = ray->D;

9
intern/cycles/kernel/geom/geom_triangle_intersect.h
View File

@@ -49,13 +49,12 @@ typedef struct IsectPrecalc {
float Sx, Sy, Sz; float Sx, Sy, Sz;
} IsectPrecalc; } IsectPrecalc;
#if defined(__KERNEL_CUDA__) #if (defined(__KERNEL_OPENCL_APPLE__)) || \
ccl_device_inline (defined(__KERNEL_CUDA__) && (defined(i386) || defined(_M_IX86)))
#elif defined(__KERNEL_OPENCL_APPLE__)
ccl_device_noinline ccl_device_noinline
#else /* defined(__KERNEL_OPENCL_APPLE__) */ #else
ccl_device_inline ccl_device_inline
#endif /* defined(__KERNEL_OPENCL_APPLE__) */ #endif
void triangle_intersect_precalc(float3 dir, void triangle_intersect_precalc(float3 dir,
IsectPrecalc *isect_precalc) IsectPrecalc *isect_precalc)
{ {

2
intern/cycles/kernel/kernel_accumulate.h
View File

@@ -55,7 +55,7 @@ ccl_device_inline void bsdf_eval_init(BsdfEval *eval, ClosureType type, float3 v
} }
/* TODO(sergey): This is just a workaround for annoying 6.5 compiler bug. */ /* TODO(sergey): This is just a workaround for annoying 6.5 compiler bug. */
#if !defined(__KERNEL_CUDA__) || __CUDA_ARCH__ < 500 #if !defined(__KERNEL_CUDA__) || __CUDA_ARCH__ < 520
ccl_device_inline ccl_device_inline
#else #else
ccl_device_noinline ccl_device_noinline

7
intern/cycles/kernel/kernel_light.h
View File

@@ -620,12 +620,7 @@ ccl_device void lamp_light_sample(KernelGlobals *kg, int lamp,
} }
} }
#if defined(__KERNEL_CUDA__) && (__CUDA_ARCH__ >= 500) && (defined(i386) || defined(_M_IX86)) ccl_device bool lamp_light_eval(KernelGlobals *kg, int lamp, float3 P, float3 D, float t, LightSample *ls)
ccl_device_noinline
#else
ccl_device
#endif
bool lamp_light_eval(KernelGlobals *kg, int lamp, float3 P, float3 D, float t, LightSample *ls)
{ {
float4 data0 = kernel_tex_fetch(__light_data, lamp*LIGHT_SIZE + 0); float4 data0 = kernel_tex_fetch(__light_data, lamp*LIGHT_SIZE + 0);
float4 data1 = kernel_tex_fetch(__light_data, lamp*LIGHT_SIZE + 1); float4 data1 = kernel_tex_fetch(__light_data, lamp*LIGHT_SIZE + 1);