Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 6, 2025

• * 14.0 toolchain update: Metal/Vulkan fixes + new CUDA support:

  Show description for 86a5f75
  a2flo

  committed
  86a5f75

  Copy full SHA for 86a5f75

Commits on Dec 5, 2025

• * added CUDA 13.1 and PTX 9.1 support

  a2flo

  committed
  f8d571d

  Copy full SHA for f8d571d

• * Metal: validate that tessellation evaluation shader control point count, pipeline vertex attribute count and run-time specified control point count all match

  a2flo

  committed
  38a760a

  Copy full SHA for 38a760a

Commits on Dec 3, 2025

• * 14.0 toolchain update: Metal/Vulkan/OpenCL fixes/improvements:

  Show description for c85288e
  a2flo

  committed
  c85288e

  Copy full SHA for c85288e

Commits on Nov 28, 2025

• * 14.0 toolchain update: Metal/Vulkan: added support for explicitly marking stage I/O as flat shaded/interpolated:

  Show description for 0f0a000
  a2flo

  committed
  0f0a000

  Copy full SHA for 0f0a000

• * 14.0 toolchain update: SimplifyCFG: disable switch lookup table creation, this is never a good idea for our use cases

  a2flo

  committed
  fc944e8

  Copy full SHA for fc944e8

Commits on Nov 25, 2025

• * 14.0 toolchain update: Vulkan/SPIR-V fixes:

  Show description for 2c192b3
  a2flo

  committed
  2c192b3

  Copy full SHA for 2c192b3

• * Vulkan device: removed the 16-/64-bit find_int_lsb intrinsics and emulate them instead (as these are not actually supported)

  Show description for a5c348d
  a2flo

  committed
  a5c348d

  Copy full SHA for a5c348d

Commits on Nov 13, 2025

• * device function/argument buffer arg handling improvements:

  Show description for aa66486
  a2flo

  committed
  aa66486

  Copy full SHA for aa66486

Commits on Nov 1, 2025

• * C++26 compat:

  Show description for 458dda9
  a2flo

  committed
  458dda9

  Copy full SHA for 458dda9

Commits on Oct 31, 2025

• * CMake: added initial support for building in C++26 mode (added WITH_CXX26 option, this is disabled by default)

  a2flo

  committed
  21e7c43

  Copy full SHA for 21e7c43

Commits on Oct 20, 2025

• * metal_program: argument buffers are allowed in indirect pipeines

  a2flo

  committed
  0739712

  Copy full SHA for 0739712

Commits on Oct 18, 2025

• * added HOST_READ_BACK_OPTIMIZE and HOST_READ_STAGING memory flags:

  Show description for 1686750
  a2flo

  committed
  1686750

  Copy full SHA for 1686750

Commits on Oct 15, 2025

• * device_buffer: read() is now const

  a2flo

  committed
  912b9b1

  Copy full SHA for 912b9b1

Commits on Oct 12, 2025

• * soft_f16: we can just always use __fp16 nowadays on all supported targets/compilers -> still retain the soft_f16 class/wrapper for now as we rely on its functionality and this is less invasive

  Show description for 52dbc9d
  a2flo

  committed
  52dbc9d

  Copy full SHA for 52dbc9d

Commits on Oct 11, 2025

• * 14.0 toolchain update: added missing include

  a2flo

  committed
  1b98890

  Copy full SHA for 1b98890

• * 14.0 toolchain update: added initial coherent global memory access support for Metal/Vulkan:

  Show description for 6522ffe
  a2flo

  committed
  6522ffe

  Copy full SHA for 6522ffe

• * added initial support for coherent global memory access:

  Show description for 731db10
  a2flo

  committed
  731db10

  Copy full SHA for 731db10

Commits on Oct 9, 2025

• * 14.0 toolchain update: Vulkan/SPIR-V fix:

  Show description for 4ee84e7
  a2flo

  committed
  4ee84e7

  Copy full SHA for 4ee84e7

Commits on Oct 8, 2025

• * 14.0 toolchain update: Vulkan/SPIR-V fixes/improvements:

  Show description for 991ab6f
  a2flo

  committed
  991ab6f

  Copy full SHA for 991ab6f

• * toolchain: added Vulkan "pointer_workarounds" option to enable generation of workarounds for certain pointer uses in SPIR-V (off by default)

  Show description for f9db90d
  a2flo

  committed
  f9db90d

  Copy full SHA for f9db90d

Commits on Oct 5, 2025

• * bbox: use ext::limits in invalid_intersection as well

  a2flo

  committed
  e185ff9

  Copy full SHA for e185ff9

• * bbox improvements:

  Show description for 760672d
  a2flo

  committed
  760672d

  Copy full SHA for 760672d

Commits on Oct 4, 2025

• * 14.0 toolchain update: Vulkan fix:

  Show description for af2d9a9
  a2flo

  committed
  af2d9a9

  Copy full SHA for af2d9a9

• * device atomics fixes:

  Show description for 7cb80db
  a2flo

  committed
  7cb80db

  Copy full SHA for 7cb80db

Commits on Oct 2, 2025

• * revert accidental -x on build.sh

  a2flo

  committed
  b0916ac

  Copy full SHA for b0916ac

• * aligned_ptr: on Windows, we need a minimum alignment of 64KiB when calling VirtualAlloc2

  Show description for 4517e10
  a2flo

  committed
  4517e10

  Copy full SHA for 4517e10

Commits on Oct 1, 2025

• * Host-Compute: implemented full subgroup support:

  Show description for 1e76733
  a2flo

  committed
  1e76733

  Copy full SHA for 1e76733

• * 14.0 toolchain update: Host-Compute/AArch64 and SPIR-V fixes:

  Show description for 38f1b8c
  a2flo

  committed
  38f1b8c

  Copy full SHA for 38f1b8c

Commits on Sep 29, 2025

• * 14.0 toolchain update: Metal/SPIR-V improvements:

  Show description for dc10d34
  a2flo

  committed
  dc10d34

  Copy full SHA for dc10d34

Commits on Sep 22, 2025

• * implemented initial subgroup support for Host-Compute:

  Show description for 51543d4
  a2flo

  committed
  51543d4

  Copy full SHA for 51543d4

• * 14.0 toolchain update: CUDA/OpenCL/SPIR-V improvements:

  Show description for b7b9ffb
  a2flo

  committed
  b7b9ffb

  Copy full SHA for b7b9ffb

Commits on Sep 21, 2025

• * 14.0 toolchain update: added initial SPIR-V SIMD/subgroup ballot support:

  Show description for a2a1380
  a2flo

  committed
  a2a1380

  Copy full SHA for a2a1380

• * compute algorithm: fixed type ambiguity in ternary

  a2flo

  committed
  ed5737a

  Copy full SHA for ed5737a

• * device backends / compute algorithm: added support for vector types in SIMD/shuffle ops and algorithms:

  Show description for a48773e
  a2flo

  committed
  a48773e

  Copy full SHA for a48773e

Pagination

Previous

Next