spcl / open-earth-compiler

development repository for the open earth compiler

https://arxiv.org/abs/2005.13014

spcl/open-earth-compiler Issues

status of project
Updated a year ago3
stencil-shape-inference is not working with tridiagonal example
Updated a year ago3
problem about compiling fastwaves.mlir
Updated 2 years ago2
[mlir]the ptx compilation flow
Updated 2 years ago2
problem about laplace example in readme
Updated 3 years ago2
cmake error: too many arguments to function createGpuToLLVMConversionPass
Updated 3 years ago2
String-based OpBuilder removed
Closed 3 years ago4
if/else optimization
Updated 4 years ago
explicit intermediate storage to prevent fusion of long stencil chains
Closed 4 years ago1
add stencil access that support dynamic offsets given some constant bound
Closed 4 years ago1
Interleaved unrolling
Updated 4 years ago
subdomains
Updated 4 years ago1
add the possiblity to load and store from and to the same field
Updated 4 years ago1
Change stencil.assert to stencil.cast with a return value
Closed 4 years ago
operation to access a lower dimensional slice
Updated 4 years ago
Multi-dimensional Unrolling
Updated 4 years ago1
register scheduling optimization
Updated 4 years ago1
Instruction count difference compared to reference
Closed 4 years ago1
Multi-threading bug in compiler
Updated 4 years ago3
Vector loads for stencils
Updated 4 years ago
Refactoring ideas for the stencil dialect
Closed 4 years ago2
Write outputs with different range
Closed 4 years ago
Loop.if vs cmpf/select during canonicalization
Closed 4 years ago1
LLVM CSE vs (PTX/ROCM) CSE
Closed 4 years ago
Performance comparison to dawn
Closed 4 years ago
Support any unroll factor
Closed 4 years ago
tridiagonal solvers
Updated 4 years ago
implement additional stencils and analyze the performance of our compiler
Closed 4 years ago
CMake needs to include CUDA directories
Closed 4 years ago5
Temporary + Output Fields are Shifted
Closed 4 years ago
update the argument list format of the stencil dialect to match the standards of mlir
Closed 4 years ago
Add continuous integration
Updated 4 years ago2
test different cu jit compiler flags
Closed 4 years ago1
lower dimensional arrays
Closed 4 years ago
Inlining when stencil.applies use the same inputs
Closed 4 years ago
Test if 'if-else' statements are lowered correctly
Closed 4 years ago
Abstracted Stencil DSL
Updated 4 years ago
lower to parallel loop instead to affine loops
Closed 4 years ago
loop unrolling
Closed 4 years ago1
build system issues
Closed 4 years ago
support bare ptr calling convention in the GPU backend
Closed 4 years ago1