Fixed a regression in 20.0.611 where subsurface scatter might output slightly different amount of energy on arm64 platforms.

Fixes a bug in Poly Extrude 2.0 SOP where 64-bit attribute data is incorrectly read when supplied as normals.

XPU: Non-32bit POD primvars now supported (ie f16, f64, i64 etc...). This change also fixes a rare internal data-alignment crash bug.

Subdividing curves with 64-bit P will work the first time now.

The Vellum Solver fluid pipeline now supports 64-bit P and v attributes for position and velocity, enabling fluid simulations far from the origin. 64-precision can also be important at extremely high substeps with low Time Scale, where the timestep becomes so small that 32-bit point positions don't properly update.

The minimum processor requirement on Linux is now SSE4.2 / x64-64-v2.

The Houdini Engine Unity plugin is now available in the Houdini macOS Apple Silicon (arm64) build.

The Houdini Engine Unity plugin is now available in the Houdini macOS Apple Silicon (arm64) build.

The Enumerate SOP can now create 64-bit attributes if the geometry's preferred precision is set to 64-bit.

The Enumerate SOP can now create 64-bit attributes if the geometry's preferred precision is set to 64-bit.

Fixed a bug where the Python Pillow package could not be imported in the macOS arm64 Houdini build.

Fixed a bug where the Python Pillow package could not be imported in the macOS arm64 Houdini build.

XPU: fixed a crash with motion-blurred diced meshes (regression from 19.5.364)

Fixed a bug in the HOM methods for manipulating 64-bit floating point geometry attributes that unexpectedly lowered precision to 32-bit floating point.

Fixed an issue with service jobs ocassionally failing due to unexpected characters in the service JSON object. Previously work item data was saved as plain text, however it is now base64 encoded to avoid that issue.

The Extract Transform SOP now preserves the input detail's preferred precision and supports 64-bit inputs.

The Extract Centroid SOP now preserves the input detail's preferred precision (e.g. 64-bit) when the output mode is Points, and uses this precision for P in the new detail. When the output mode is Attribute, the new attribute follows the detail's preferred precision.

Added 32bit and 64bit position precision options to the LidarImport SOP.

Added 32bit and 64bit position precision options to the LidarImport SOP.

Fixed Intel Denoiser not working in IPR under non-pathtraced convergence modes. This bug was introduced in 19.5.164.

Fixed Intel Denoiser not working in IPR under non-pathtraced convergence modes. This bug was introduced in 19.5.164.

Fixed a bug in RGB -> HSV conversion in MaterialX that was introduced in 19.0.464.

The Sort SOP will now properly sort 64-bit integer and float attributes rather than casting them to 32-bit versions first.

Fixed issue with 'uint' and 'uint64' types in the USD shader translator.