YafaRay 0.1.5 for Blender 2.67 release notes.

  • Bug fixed in the blend material that caused normals to not be blended (bug report).
  • Camera clipping corrected using planes (bug report).
  • Fix for issue "Blendermat issue: fatal error LNK1120: 1 link extern not resolved" Reported by Povmaniaco.
  • Fix to add fsqrt masm version for msvc builds.
  • Fix for target path installing python and ruby bindings.
  • Fix for DarkSky a.k.a Sunsky 2 (bug report):

    - Sunlight calculation algorithm improved
    - Gamma correction fix
    - Color clamping forced to avoid stray samples
  • Removed local definition of fast_math and fast_trig macros.
  • Output color alpha premultiplication corrected for Blender compatibility (bug report).
  • Improved glass internal reflection (bug report). Increasing the internal limit for glass internal reflection gives more realistic detail in glass. As discussed with DarkTide, it will cause a small performance penalty, but the added realism is worth it and the performance penalty is not so big.

Comparison of rendering of internal reflection between YafaRay 012 (left) and YafaRay 015 (right).

  • New parameters for Background Transparency.
  • color_t: Added isNaN isInf methods useful for debuggin.
  • MathOptimizations: Removed fTan, results too far from the right precision.
  • Improved glossy materials (bug report, forum discussion).

Glossy Rendering comparison between YafaRay 012 (left) and YafaRay 015 (right)

  • Fixes for Black and White Dots (bug report, scenes, commit log)
  • Fixed hangs and crashes in Linux (bug report).
  • ImageFilm: Replaced doubles with floats.
  • PFLOAT removed from matrix4x4.
  • SPPM: removed incompatible long long type.
  • Microfacet: Fixed out of domain divisors and added firefly rejection value.
  • MathOptimizations: Removed double overloads.
  • ImageFilm: Fixed filter calculation for floats instead of doubles.
  • CMake: added headers to source lists. FAST_MATH and FAST_TRIG enabled by default.
  • SampleUtils: Removed doubles from cone sampling code.
  • Microfacet: fixed the tan cutoff.
  • SPPM: added int64_t to the big variables.
  • Fixes for white dots.
  • Clamping for Halton screening.
  • pdf1D_t: Throws a warning when index is under or over bounds.
  • RoughGlass: full glossy reflection/refraction achieved.
  • MC_QMCUtils: clamping on all the sample generation algorithms.
  • RoughGlass: full reflections added (Warning, this affects all the glossy materials, testing is required).
  • Sunsky2: Sun color calculation fixed.
  • Fix to avoid NaN in roughglass when IOR=1.000
  • Extension of normal_t class.
  • Removed redundant operations and checks.
  • New "Mean Weighted by Angle" (MWA) smoothing algorithm that fits better on highly subdivided meshes. No more reflect/transmit distorsion. Reference paper: Comparison of Algorithms for Vertex Normal Computation. Shuangshuang Jin, Robert R. Lewis, David West. Washington State University. School of Electrical Engineering and Computer Science.
  • Calculation of barycentyric coordinate w (b0 in our case) moved.
  • Fixed transparency iossue on debug integrator.
  • Refactoring of material library names.
  • Algorithm for autosmoothing normals with angle threshold below 180 degrees fixed.
  • Returned checks for invalid vertex normal indices.
  • Depth map export support on python interface.
  • Nea, Far clipping distances stored and made available on cameras.
  • Initial support for camera clipping distance based depth map.
  • Depth buffer feature, rendering normalized and unnormalized values to be used inside a z-buffer that supports absolute values like blender's.
  • Depth fix for hits on background on unnormalized depth map rendering.