Commits · 0d7526c07b744f3cd1c7a692b794bd6e7f8bfc37 · pycodegen / pystencils · GitLab

Snippets Groups Projects

Mar 20, 2025
- HIP Target and Platform · 214c345d
  Frederik Hennig authored 2 months ago
  
  214c345d
Mar 18, 2025
- Optimization for GPU block size determination · 570223e8
  Richard Angersbach authored 2 months ago
  
  570223e8
Mar 12, 2025
- Fix header incl · 974cf848
  Richard Angersbach authored 2 months ago
  
  974cf848
- Move manual atomic op implementations to new header · c7f1518e
  Richard Angersbach authored 2 months ago
  
  c7f1518e
Mar 11, 2025
- Fix Linear3D Indexing if Iteration Space Rank < 3 · 082e4c04
  Frederik Hennig authored 2 months ago
  
  082e4c04
Mar 07, 2025
- Some cleanup to the pystencils runtime headers · fcd5bfb6
  Frederik Hennig authored 2 months ago
  
  fcd5bfb6
Mar 05, 2025
- Introduction of structural ast nodes · 5a616930
  Richard Angersbach authored 3 months ago
  
  5a616930
Feb 27, 2025
- Fix lint · 4a031fc1
  Richard Angersbach authored 3 months ago
  
  4a031fc1
Feb 26, 2025
- Fix getter for thread exec condition for dense/sparse iteration spaces in cuda.py · a972d759
  Richard Angersbach authored 3 months ago
  
  a972d759
Feb 21, 2025
- Fix typecheck · 02be4d5e
  Richard Angersbach authored 3 months ago
  
  02be4d5e
- Fix lint · 89a6f36a
  Richard Angersbach authored 3 months ago
  
  89a6f36a
- Refactor reductionfunction mechanism · a8479afa
  Richard Angersbach authored 3 months ago
  
  a8479afa
- Use full mask for CUDA reductions · bcd83842
  Richard Angersbach authored 3 months ago
  
  bcd83842
Feb 19, 2025
- Introduce masks for warp reductions and fix errors when shuffling warp results · 2b6589b8
  Richard Angersbach authored 3 months ago
  
  2b6589b8
- Add initial version of warp-level reduction for CUDA · ad292c2b
  Richard Angersbach authored 3 months ago
  
  ad292c2b
- Encapsulate fetching of kernel conditions for iteration spaces in separate function · ce816539
  Richard Angersbach authored 3 months ago
  
  ce816539
- Clean up kernel creation code in `codegen.driver`. Fix and dry-test SYCL codegen. · b132e39a
  Frederik Hennig authored 3 months ago
  
  b132e39a
- Relax type annotation in PsBlock __init__ · f149f823
  Daniel Bauer authored 3 months ago and Frederik Hennig committed 3 months ago
  
  f149f823
Feb 18, 2025
- GPU Indexing Schemes and Launch Configurations · 6f541733
  Frederik Hennig authored 3 months ago and Richard Angersbach committed 3 months ago
  
  6f541733
- Extend symbol canonicalization · 2a23ec52
  Daniel Bauer authored 3 months ago and Frederik Hennig committed 3 months ago
  
  2a23ec52
- Avoid duplicate definition of atomicMin/Max for HIP · 77a22268
  Richard Angersbach authored 3 months ago
  
  77a22268
- Fix order of arguments in x86 set intrinsics · d3a713b9
  Daniel Bauer authored 3 months ago and Frederik Hennig committed 3 months ago
  
  d3a713b9
Feb 17, 2025
- Fix typecheck · d1489837
  Richard Angersbach authored 3 months ago
  
  d1489837
- Fix lint · 13569a61
  Richard Angersbach authored 3 months ago
  
  13569a61
- Add generator for SIMD horizontal operations and the emitted code. · fe3cd6cd
  Richard Angersbach authored 3 months ago
  
  fe3cd6cd
- Fix bug with doubly inverted sign for subtraction reductions · 58cdb792
  Richard Angersbach authored 3 months ago
  
  58cdb792
Feb 12, 2025
- Minor fix · 7306f4dd
  Richard Angersbach authored 3 months ago
  
  7306f4dd
- Refactor PsVecHorizontal as PsBinOp · 8e0a7478
  Richard Angersbach authored 3 months ago
  
  8e0a7478
- Minor refactor of reduction ops · eb7823a5
  Richard Angersbach authored 3 months ago
  
  eb7823a5
Feb 11, 2025
- Add initial implementation for horizontal reductions for vectorization · d4b7e78f
  Richard Angersbach authored 3 months ago
  
  d4b7e78f
Feb 10, 2025
- Support atomic sub, min, max for fp reductions using custom implementations with CAS mechanism · 60a348f1
  Richard Angersbach authored 3 months ago
  
  60a348f1
- Temporarily change default CUDA block size for CUDA jit · a71e0d31
  Richard Angersbach authored 3 months ago
  
  a71e0d31
- Add guard for INFINITY numeric limit macro used by cuda backend · 5caafdd0
  Richard Angersbach authored 3 months ago
  
  5caafdd0
Feb 04, 2025
- Wrap statement around generated atomic call [skip ci] · a2a59d40
  Richard Angersbach authored 3 months ago
  
  a2a59d40
- Use predefined macro values for numeric limits in cuda backend · d8ae9002
  Richard Angersbach authored 3 months ago
  
  d8ae9002
- Try supporting pointer dtypes for reductions in cupy gpu jit · 826ee8e2
  Richard Angersbach authored 3 months ago
  
  826ee8e2
Jan 31, 2025
- Object-Oriented CPU JIT API and Prototype Implementation · fa2683ce
  Frederik Hennig authored 4 months ago
  
  fa2683ce
- Fix lint [skip ci] · 4c7fd409
  Richard Angersbach authored 4 months ago
  
  4c7fd409
- Fix lint [skip ci] · 616f609f
  Richard Angersbach authored 4 months ago
  
  616f609f
Jan 30, 2025
- Add CUDA backend for numeric limits · 0fb11858
  Richard Angersbach authored 4 months ago
  
  0fb11858