tommi/kfd
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Initial extraction from isis: direct /dev/kfd GPU compute

Browse Source 
AMD RDNA3 GPU driver via raw KFD ioctls. No ROCm, no OpenCL.

- Device discovery and topology queries
- VRAM/GTT/Userptr memory with shared address space
- PM4 compute queue and kernel dispatch
- Hand-written RDNA3 ASM kernels: matmul (3100 GFLOP/s), matvec, superlinear
- Tile-safe buffer padding for OOB protection
- Event-based interrupt wait (no CPU polling)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Tommi Niemi
2026-04-06 20:38:10 +07:00
commit 66a9edb898
75 changed files with 5858 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
target/debug/deps/kfd-2447e9870f0e8752.d Normal file
View File

@@ -0,0 +1,16 @@
/steam/rotko/kfd/target/debug/deps/kfd-2447e9870f0e8752.d: src/lib.rs src/ioctl.rs src/memory.rs src/queue.rs src/dispatch.rs src/compute.rs src/device.rs src/kernels/test_store.co src/kernels/matvec_asm.co src/kernels/matmul_blocked.co src/kernels/matmul_small.co src/kernels/superlinear.co
/steam/rotko/kfd/target/debug/deps/libkfd-2447e9870f0e8752.rmeta: src/lib.rs src/ioctl.rs src/memory.rs src/queue.rs src/dispatch.rs src/compute.rs src/device.rs src/kernels/test_store.co src/kernels/matvec_asm.co src/kernels/matmul_blocked.co src/kernels/matmul_small.co src/kernels/superlinear.co
src/lib.rs:
src/ioctl.rs:
src/memory.rs:
src/queue.rs:
src/dispatch.rs:
src/compute.rs:
src/device.rs:
src/kernels/test_store.co:
src/kernels/matvec_asm.co:
src/kernels/matmul_blocked.co:
src/kernels/matmul_small.co:
src/kernels/superlinear.co: