Repository navigation
Releases: mscf/OrdinaryShade
Release list
OrdinaryShade v0.1.0a17
Part of Ordinary v0.1.11.
Validated commit: 71c9f27eb639dbc2c0b59db821f6c2e1ad32bc4e.
OrdinaryShade v0.1.0a15
Part of Ordinary v0.1.10.
Validated commit: 72e743e7ce84148bb7595de367c8e2676de03ad5.
OrdinaryShade v0.1.0a14
Part of Ordinary v0.1.9.
Validated commit: 88a9ee93f3b614c4b7e204bf9f8e6304bab6bd9c.
OrdinaryShade v0.1.0a13
Part of Ordinary v0.1.8.
Validated commit: 927e589adb90ba7a10763ce0fb0e3b0c03149528.
OrdinaryShade v0.1.0a12
Part of Ordinary v0.1.7.
Validated commit: 7b9b39dfa439094e032118a4ca4fe95ce09437c6.
OrdinaryShade v0.1.0a11
Part of Ordinary v0.1.6.
Validated commit: 2656f8c4ec50d0cb0f9883c20750ea0776d1b597.
OrdinaryShade v0.1.0a10
Part of Ordinary v0.1.5.
Validated commit: 5ab06e381905a3fb9c0ebacb7e9032bb9be3cb46.
OrdinaryShade v0.1.0a9
Part of Ordinary v0.1.4.
Validated commit: cf16445f71db8fa35c168c4dc577815033b5cbdf.
OrdinaryShade v0.1.0a8
OrdinaryShade 0.1.0a8
Add typed GPU primitives used by resident geometry and voxel-output pipelines.
- Fixed-size workgroup arrays through
osh.shared(osh.array(type, count)), with GLSL and WGSL lowering. Applications provide initialization and barriers. - Explicit
buffer_float32_atomic_addcapability and scalar storage-bufferosh.atomic_addsupport in GLSL. Reject missing capabilities, mismatched scalar types, non-buffer targets, and unsupported WGSL compilation. - Ray-query triangle candidate confirmation through
query.confirm_intersection().
Floating-point atomic addition requires the corresponding Vulkan device features;
it does not enable shared-memory or image atomics. Sum rounding depends on GPU
execution order. This remains a pre-alpha compiler release.
OrdinaryShade v0.1.0a7
OrdinaryShade 0.1.0a7
Add typed integer atomic_compare_exchange(target, comparator, replacement)
for GPU hash tables and other lock-free integer algorithms. GLSL/SPIR-V lowers
to atomicCompSwap and returns the observed old value. Comparator and replacement
must match the scalar signed/unsigned integer target. WGSL is explicitly rejected
until equivalent backend support is implemented.
Verified by compiler type/arity tests and vxl8r GPU visible-face averaging parity,
including repeated frames, invalid identities, HDR precision and no frame-stage
uploads/allocations/waits.