Aart Bik ee42e23614 [mlir][sparse][gpu] first implementation of the GPU libgen approach ...

The sparse compiler now has two prototype strategies for GPU acceleration:

* CUDA codegen: this converts sparsified code to CUDA threads
* CUDA libgen: this converts pre-sparsified code to cuSPARSE library calls

This revision introduces the first steps required for the second approach.

Reviewed By: ThomasRaoux

Differential Revision: https://reviews.llvm.org/D150170

2023-05-15 08:49:38 -07:00

..

GPU

[mlir][sparse][gpu] first implementation of the GPU libgen approach

2023-05-15 08:49:38 -07:00

buffer_rewriting.mlir

[mlir][sparse] Modify the pivot selection method for quick sort.

2023-03-15 13:53:00 -07:00

codegen_buffer_initialization.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

codegen_sparse_alloc.mlir

[mlir][sparse] support sparse bufferization.alloc_tensor with copy argument.

2023-03-31 22:27:23 +00:00

codegen_sparse_dealloc.mlir

[mlir][sparse] add create-sparse-deallocs options to match the create-deallocs in BufferizationOption.

2023-03-27 23:18:32 +00:00

codegen_to_llvm.mlir

[mlir][sparse] add test that combines sparse codegen and lowering to llvm struct

2023-01-04 12:12:31 -08:00

codegen.mlir

[mlir][sparse] Fix a problem in handling data type conversion.

2023-03-30 14:54:53 -07:00

constant_index_map.mlir

[mlir][sparse] fix crash when using pure constant index in indexing mapping (fixes #61530)

2023-03-21 23:45:20 +00:00

conversion.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

convert_dense2sparse.mlir

[mlir][sparse] fix bugs when convert coo to coo but with different dim ordering

2023-03-09 20:55:03 +00:00

convert_sparse2dense.mlir

[mlir][sparse] Cleaning up the dim/lvl distinction in SparseTensorConversion

2022-12-05 16:59:42 -08:00

convert_sparse2sparse_element.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

convert_sparse2sparse.mlir

[mlir][sparse] convert a sparse tensor slice to sparse tensor correctly.

2023-03-28 21:39:31 +00:00

dense.mlir

…

fold.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

invalid_encoding.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

invalid.mlir

[mlir][sparse] extend unpack operation to support unpacking a batched COO type

2023-05-01 18:17:29 +00:00

one_shot_bufferize_tensor_copy_insertion.mlir

[mlir][bufferization] Make TensorCopyInsertionPass a test pass

2022-12-02 15:38:02 +01:00

one_trip.mlir

…

post_rewriting.mlir

[mlir][sparse] relax constraints on tensor.cast with pre-rewriting

2023-05-01 16:03:44 -07:00

pre_rewriting.mlir

[mlir][sparse] relax constraints on tensor.cast with pre-rewriting

2023-05-01 16:03:44 -07:00

rejected.mlir

[mlir][sparse] reject kernels with non-sparsfiable reduction expression.

2022-12-08 23:36:30 +00:00

rewriting_for_codegen.mlir

[mlir][sparse] Improve the implementation of sparse_tensor.new for the codegen path.

2023-03-01 07:29:49 -08:00

roundtrip_encoding.mlir

[mlir][sparse] introduce a new compressed(hi) dimension level type

2023-04-18 23:26:11 +00:00

roundtrip.mlir

[mlir][sparse] extend unpack operation to support unpacking a batched COO type

2023-05-01 18:17:29 +00:00

scf_1_N_conversion.mlir

[mlir][sparse] use sparse_tensor::StorageSpecifier to store dim/memSizes

2022-12-23 00:47:36 +00:00

sorted_coo.mlir

[mlir][sparse] deduplicate non-unique coordinates unconditionally

2023-03-09 21:59:57 +00:00

sparse_1d.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_2d.mlir

[mlir][sparse] extend unpack operation to support unpacking a batched COO type

2023-05-01 18:17:29 +00:00

sparse_3d.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_affine.mlir

[mlir][sparse] fix crash when using pure constant index in indexing mapping (fixes #61530)

2023-03-21 23:45:20 +00:00

sparse_broadcast.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_concat_codegen.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_concat.mlir

[mlir][sparse] Improve concatenate operation conversion for the case with annotated all dense result.

2022-12-07 12:06:50 -08:00

sparse_conv_2d_slice_based.mlir

[mlir][sparse] extend loop emitter to emit slice driven loops

2023-04-13 03:29:40 +00:00

sparse_expand.mlir

[mlir][sparse] Cleaning up the dim/lvl distinction in SparseTensorConversion

2022-12-05 16:59:42 -08:00

sparse_extract_slice.mlir

[mlir][sparse] support dynamic sparse tensor slices.

2023-03-10 23:12:41 +00:00

sparse_fill_zero.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_foreach.mlir

[mlir][sparse] extend unpack operation to support unpacking a batched COO type

2023-05-01 18:17:29 +00:00

sparse_fp_ops.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_index.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_int_ops.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_kernels.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_lower_col.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_lower_inplace.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_lower.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_matmul_codegen.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_nd.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_out.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_outbuf.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_pack.mlir

[mlir][sparse] extend unpack operation to support unpacking a batched COO type

2023-05-01 18:17:29 +00:00

sparse_parallel_reduce.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_parallel.mlir

[mlir][sparse] support parallel for/reduction in sparsification.

2022-11-07 18:04:46 +00:00

sparse_perm_lower.mlir

[mlir][sparse] Cleaning up the dim/lvl distinction in SparseTensorConversion

2022-12-05 16:59:42 -08:00

sparse_perm.mlir

[mlir][sparse] use loop emitter to generate loop in sparsification

2022-10-26 00:27:56 +00:00

sparse_reshape_dot.mlir

[mlir][sparse] deduplicate non-unique coordinates unconditionally

2023-03-09 21:59:57 +00:00

sparse_reshape.mlir

[mlir][sparse] avoid creating temporary unordered COO buffer when reshape sparse tensor.

2023-03-30 01:29:55 +00:00

sparse_scalars.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_sddmm_org.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_sddmm.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_storage.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_transpose.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_vector_chain.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_vector_concat.mlir

[mlir][sparse][simd] only accept proper unit stride subscripts

2022-12-14 13:07:14 -08:00

sparse_vector_index.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_vector_mv.mlir

[mlir][sparse] introduce sparse vectorization to the sparse compiler pipeline

2022-12-07 16:06:53 -08:00

sparse_vector_ops.mlir

[sparse][mlir][vectorization] add support for shift-by-invariant

2022-12-27 11:07:13 -08:00

sparse_vector_peeled.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

sparse_vector.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

specifier_to_llvm.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00

vectorize_reduction.mlir

[mlir][sparse] Renaming "pointer/index" to "position/coordinate"

2023-03-06 12:23:33 -08:00