|
11 | 11 | (project-context |
12 | 12 | (name "futharkiser") |
13 | 13 | (description "Compile annotated array operations to GPU kernels via Futhark") |
14 | | - (status "scaffold — documentation bespoked") |
15 | | - (priority "—") |
| 14 | + (status "alpha") |
| 15 | + (priority "medium") |
16 | 16 | (ecosystem "-iser family (https://github.com/hyperpolymath/iseriser)")) |
17 | 17 |
|
18 | 18 | (current-position |
19 | | - (phase "initial-scaffold") |
20 | | - (completion-percentage 10) |
21 | | - (milestone "Architecture defined, CLI scaffolded, RSR template complete, Idris2 ABI types bespoked (SOAC, GPUBackend, ArrayShape, MemorySpace, ParallelPattern), Zig FFI bespoked with GPU buffer management, documentation updated")) |
| 19 | + (phase "phase-1-complete") |
| 20 | + (completion-percentage 45) |
| 21 | + (milestone "Phase 1 complete — pattern detection, Futhark codegen, build generator") |
| 22 | + (what-changed |
| 23 | + "2026-03-21: Phase 1 complete. Pattern detection, Futhark codegen, build generator, GPU buffer management stubs, bespoke manifest parsing. Integration tests passing.")) |
22 | 24 |
|
23 | 25 | (route-to-mvp |
24 | | - (step 1 "Implement pattern detection — identify map/reduce/scan/scatter in source code") |
25 | | - (step 2 "Implement Futhark codegen — generate .fut files from detected patterns") |
26 | | - (step 3 "Integrate Futhark compiler invocation (futhark opencl/cuda/multicore/c)") |
27 | | - (step 4 "GPU buffer allocation and host<->device memory transfer via Zig FFI") |
28 | | - (step 5 "Idris2 proofs of parallelism safety (associativity, purity, shape compatibility)") |
29 | | - (step 6 "First end-to-end example: vector addition on GPU via OpenCL backend") |
30 | | - (step 7 "Benchmarks against hand-written CUDA/OpenCL")) |
| 26 | + (step 1 "DONE — Pattern detection for map/reduce/scan/scatter") |
| 27 | + (step 2 "DONE — Futhark codegen generating .fut files") |
| 28 | + (step 3 "DONE — Build generator for opencl/cuda/multicore/c backends") |
| 29 | + (step 4 "DONE — GPU buffer management stubs via Zig FFI") |
| 30 | + (step 5 "DONE — Bespoke manifest parsing and CLI") |
| 31 | + (step 6 "TODO — End-to-end GPU kernel compilation and execution") |
| 32 | + (step 7 "TODO — Idris2 proofs for parallelism safety (associativity, purity)") |
| 33 | + (step 8 "TODO — Benchmarks against hand-written CUDA/OpenCL") |
| 34 | + (step 9 "TODO — PanLL panel and BoJ cartridge integration")) |
31 | 35 |
|
32 | 36 | (blockers-and-issues |
33 | | - (none "Project is in scaffold phase — no blockers yet")) |
| 37 | + (note "Futhark compiler not installed on dev machine — .fut output verified by structure") |
| 38 | + (note "GPU testing requires OpenCL/CUDA runtime")) |
34 | 39 |
|
35 | 40 | (critical-next-actions |
36 | | - (action "Implement source-level pattern detection for map/reduce/scan") |
37 | | - (action "Generate first .fut Futhark program from detected patterns") |
38 | | - (action "Invoke futhark compiler and link generated C library"))) |
| 41 | + (action "Install Futhark compiler and run end-to-end GPU test") |
| 42 | + (action "Write Idris2 proofs for parallelism safety properties") |
| 43 | + (action "Benchmark against hand-written CUDA for vector addition"))) |
0 commit comments