diff options
Diffstat (limited to 'intern/cycles/util/util_transform.h')
| -rw-r--r-- | intern/cycles/util/util_transform.h | 38 |
1 files changed, 38 insertions, 0 deletions
diff --git a/intern/cycles/util/util_transform.h b/intern/cycles/util/util_transform.h index bfc8f55feed..ea5eb3b25b0 100644 --- a/intern/cycles/util/util_transform.h +++ b/intern/cycles/util/util_transform.h @@ -73,22 +73,60 @@ ccl_device_inline float3 transform_perspective(const Transform *t, const float3 ccl_device_inline float3 transform_point(const Transform *t, const float3 a) { + /* TODO(sergey): Disabled for now, causes crashes in certain cases. */ +#if defined(__KERNEL_SSE__) && defined(__KERNEL_SSE2__) && 0 + ssef x, y, z, w, aa; + aa = a.m128; + + x = _mm_loadu_ps(&t->x.x); + y = _mm_loadu_ps(&t->y.x); + z = _mm_loadu_ps(&t->z.x); + w = _mm_loadu_ps(&t->w.x); + + _MM_TRANSPOSE4_PS(x, y, z, w); + + ssef tmp = shuffle<0>(aa) * x; + tmp = madd(shuffle<1>(aa), y, tmp); + tmp = madd(shuffle<2>(aa), z, tmp); + tmp += w; + + return float3(tmp.m128); +#else float3 c = make_float3( a.x*t->x.x + a.y*t->x.y + a.z*t->x.z + t->x.w, a.x*t->y.x + a.y*t->y.y + a.z*t->y.z + t->y.w, a.x*t->z.x + a.y*t->z.y + a.z*t->z.z + t->z.w); return c; +#endif } ccl_device_inline float3 transform_direction(const Transform *t, const float3 a) { + /* TODO(sergey): Disabled for now, causes crashes in certain cases. */ +#if defined(__KERNEL_SSE__) && defined(__KERNEL_SSE2__) && 0 + ssef x, y, z, w, aa; + aa = a.m128; + x = _mm_loadu_ps(&t->x.x); + y = _mm_loadu_ps(&t->y.x); + z = _mm_loadu_ps(&t->z.x); + w = _mm_setzero_ps(); + + _MM_TRANSPOSE4_PS(x, y, z, w); + + ssef tmp = shuffle<0>(aa) * x; + tmp = madd(shuffle<1>(aa), y, tmp); + tmp = madd(shuffle<2>(aa), z, tmp); + + return float3(tmp.m128); +#else float3 c = make_float3( a.x*t->x.x + a.y*t->x.y + a.z*t->x.z, a.x*t->y.x + a.y*t->y.y + a.z*t->y.z, a.x*t->z.x + a.y*t->z.y + a.z*t->z.z); return c; +#endif } ccl_device_inline float3 transform_direction_transposed(const Transform *t, const float3 a) |