[Lldb-commits] [libclc] [libcxxabi] [compiler-rt] [mlir] [flang] [clang] [lld] [clang-tools-extra] [llvm] [libcxx] [libc] [libunwind] [lldb] [mlir][sparse][CRunnerUtils] Add shuffle in CRunnerUtils (P
@@ -176,6 +177,14 @@ extern "C" void rtdrand(void *g) {
delete generator;
}
+extern "C" void _mlir_ciface_shuffle(StridedMemRefType *mref,
+ void *g) {
+ std::mt19937 *generator = static_cast(g);
+ uint64_t s = mref->sizes[0];
aartbik wrote:
One last nit, we may want to ensure that the rank one tensor has the right
properties
assert(mref->strides[0] == 1); // consecutive
also,although offset is typically zero, I think we need to compute
uint64_t *data = mref->data + mref->offset
and then do
std::iota(data, data + s, 0);
https://github.com/llvm/llvm-project/pull/77124
___
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] refine sparse fusion with empty tensors materialization (PR #66563)
aartbik wrote: I have to fix a merge conflict on the test. Coming up. https://github.com/llvm/llvm-project/pull/66563 ___ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] Update Enum name for CompressedWithHigh (PR #67845)
https://github.com/aartbik approved this pull request. https://github.com/llvm/llvm-project/pull/67845 ___ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] Print new syntax (PR #68130)
@@ -533,7 +533,7 @@ func.func @sparse_compression(%tensor: tensor<8x8xf64,
#CSR>,
// CHECK: %[[A13:.*]]:4 = scf.for %[[A14:.*]] = %[[A11]] to %[[A7]]
step %[[A12]] iter_args(%[[A15:.*]] = %[[A0]], %[[A16:.*]] = %[[A1]],
%[[A17:.*]] = %[[A2]], %[[A18:.*]] = %[[A3]]) -> (memref,
memref, memref, !sparse_tensor.storage_specifier
// CHECK: %[[A19:.*]] = memref.load %[[A6]]{{\[}}%[[A14]]] :
memref
// CHECK: %[[A20:.*]] = memref.load %[[A4]]{{\[}}%[[A19]]] :
memref
-// CHECK: %[[A21:.*]]:4 = func.call
@_insert_dense_compressed_no_8_8_f64_0_0(%[[A15]], %[[A16]], %[[A17]],
%[[A18]], %[[A8]], %[[A19]], %[[A20]]) : (memref, memref,
memref, !sparse_tensor.storage_specifier
+// CHECK: %[[A21:.*]]:4 = func.call
@"_insert_dense_compressed(nonordered)_8_8_f64_0_0"(%[[A15]], %[[A16]],
%[[A17]], %[[A18]], %[[A8]], %[[A19]], %[[A20]]) : (memref,
memref, memref, !sparse_tensor.storage_specifier
aartbik wrote:
here and elsewhere, let's use compressed(nonordered) in the printing of type,
but not in the generation for methods names (replace ( ) with _ perhaps, so we
keep valid C identifiiers)
https://github.com/llvm/llvm-project/pull/68130
___
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] Print new syntax (PR #68130)
@@ -586,30 +586,56 @@ Attribute SparseTensorEncodingAttr::parse(AsmParser
&parser, Type type) {
}
void SparseTensorEncodingAttr::print(AsmPrinter &printer) const {
- // Print the struct-like storage in dictionary fashion.
- printer << "<{ lvlTypes = [ ";
- llvm::interleaveComma(getLvlTypes(), printer, [&](DimLevelType dlt) {
-printer << "\"" << toMLIRString(dlt) << "\"";
- });
- printer << " ]";
+ auto map = static_cast(getDimToLvl());
+ auto lvlTypes = getLvlTypes();
+ // Empty affine map indicates identity map
+ if (!map) {
+map = AffineMap::getMultiDimIdentityMap(getLvlTypes().size(),
getContext());
+ }
+ // Modified version of AsmPrinter::Impl::printAffineMap.
aartbik wrote:
I would remove this. This is diverged sufficiently to no longer refer to
printAffineMap
https://github.com/llvm/llvm-project/pull/68130
___
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] Print new syntax (PR #68130)
@@ -472,8 +472,11 @@ class SparseInsertGenerator
llvm::raw_svector_ostream nameOstream(nameBuffer);
nameOstream << kInsertFuncNamePrefix;
const Level lvlRank = stt.getLvlRank();
-for (Level l = 0; l < lvlRank; l++)
- nameOstream << toMLIRString(stt.getLvlType(l)) << "_";
+for (Level l = 0; l < lvlRank; l++) {
+ std::string lvlType = toMLIRString(stt.getLvlType(l));
+ replaceWithUnderscore(lvlType);
aartbik wrote:
std::string::replace ?
https://github.com/llvm/llvm-project/pull/68130
___
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] introduce MapRef, unify conversion/codegen for reader (PR #68360)
https://github.com/aartbik updated
https://github.com/llvm/llvm-project/pull/68360
>From 6094912685a0cfa5c13e023e8ec97238a84fca2f Mon Sep 17 00:00:00 2001
From: Aart Bik
Date: Thu, 5 Oct 2023 13:22:28 -0700
Subject: [PATCH 1/4] [mlir][sparse] introduce MapRef, unify conversion/codegen
for reader
This revision introduces a MapRef, which will support a future
generalization beyond permutations (e.g. block sparsity). This
revision also unifies the conversion/codegen paths for the
sparse_tensor.new operation from file (eg. the readers). Note
that more unification is planned as well as general affine
dim2lvl and lvl2dim (all marked with TODOs).
---
.../mlir/ExecutionEngine/SparseTensor/File.h | 156 ++--
.../ExecutionEngine/SparseTensor/MapRef.h | 96 ++
.../ExecutionEngine/SparseTensor/Storage.h| 108 +--
.../ExecutionEngine/SparseTensorRuntime.h | 8 -
.../SparseTensor/Transforms/CodegenUtils.cpp | 89 +
.../SparseTensor/Transforms/CodegenUtils.h| 18 ++
.../Transforms/SparseTensorCodegen.cpp| 73 ++--
.../Transforms/SparseTensorConversion.cpp | 111 ++-
.../SparseTensor/CMakeLists.txt | 1 +
.../ExecutionEngine/SparseTensor/MapRef.cpp | 52 ++
.../ExecutionEngine/SparseTensorRuntime.cpp | 60 +++---
mlir/test/Dialect/SparseTensor/codegen.mlir | 172 +-
.../test/Dialect/SparseTensor/conversion.mlir | 18 +-
13 files changed, 475 insertions(+), 487 deletions(-)
create mode 100644 mlir/include/mlir/ExecutionEngine/SparseTensor/MapRef.h
create mode 100644 mlir/lib/ExecutionEngine/SparseTensor/MapRef.cpp
diff --git a/mlir/include/mlir/ExecutionEngine/SparseTensor/File.h
b/mlir/include/mlir/ExecutionEngine/SparseTensor/File.h
index 78c1a0544e3a521..9157bfa7e773239 100644
--- a/mlir/include/mlir/ExecutionEngine/SparseTensor/File.h
+++ b/mlir/include/mlir/ExecutionEngine/SparseTensor/File.h
@@ -20,6 +20,7 @@
#ifndef MLIR_EXECUTIONENGINE_SPARSETENSOR_FILE_H
#define MLIR_EXECUTIONENGINE_SPARSETENSOR_FILE_H
+#include "mlir/ExecutionEngine/SparseTensor/MapRef.h"
#include "mlir/ExecutionEngine/SparseTensor/Storage.h"
#include
@@ -75,6 +76,10 @@ inline V readValue(char **linePtr, bool isPattern) {
} // namespace detail
+//===--===//
+//
+// Reader class.
+//
//===--===//
/// This class abstracts over the information stored in file headers,
@@ -132,6 +137,7 @@ class SparseTensorReader final {
/// Reads and parses the file's header.
void readHeader();
+ /// Returns the stored value kind.
ValueKind getValueKind() const { return valueKind_; }
/// Checks if a header has been successfully read.
@@ -185,58 +191,37 @@ class SparseTensorReader final {
/// valid after parsing the header.
void assertMatchesShape(uint64_t rank, const uint64_t *shape) const;
- /// Reads a sparse tensor element from the next line in the input file and
- /// returns the value of the element. Stores the coordinates of the element
- /// to the `dimCoords` array.
- template
- V readElement(uint64_t dimRank, uint64_t *dimCoords) {
-assert(dimRank == getRank() && "rank mismatch");
-char *linePtr = readCoords(dimCoords);
-return detail::readValue(&linePtr, isPattern());
- }
-
- /// Allocates a new COO object for `lvlSizes`, initializes it by reading
- /// all the elements from the file and applying `dim2lvl` to their
- /// dim-coordinates, and then closes the file. Templated on V only.
- template
- SparseTensorCOO *readCOO(uint64_t lvlRank, const uint64_t *lvlSizes,
- const uint64_t *dim2lvl);
-
/// Allocates a new sparse-tensor storage object with the given encoding,
/// initializes it by reading all the elements from the file, and then
/// closes the file. Templated on P, I, and V.
template
SparseTensorStorage *
readSparseTensor(uint64_t lvlRank, const uint64_t *lvlSizes,
- const DimLevelType *lvlTypes, const uint64_t *lvl2dim,
- const uint64_t *dim2lvl) {
-auto *lvlCOO = readCOO(lvlRank, lvlSizes, dim2lvl);
+ const DimLevelType *lvlTypes, const uint64_t *dim2lvl,
+ const uint64_t *lvl2dim) {
+const uint64_t dimRank = getRank();
+MapRef map(dimRank, lvlRank, dim2lvl, lvl2dim);
+auto *coo = readCOO(map, lvlSizes);
auto *tensor = SparseTensorStorage::newFromCOO(
-getRank(), getDimSizes(), lvlRank, lvlTypes, lvl2dim, *lvlCOO);
-delete lvlCOO;
+dimRank, getDimSizes(), lvlRank, lvlTypes, lvl2dim, *coo);
+delete coo;
return tensor;
}
/// Reads the COO tensor from the file, stores the coordinates and values to
/// the given buffers, returns a boolean value to indicate whether the COO
/// elements are sorted.
- /// Precondition: the buffers sho
[Lldb-commits] [lldb] [mlir][sparse] introduce MapRef, unify conversion/codegen for reader (PR #68360)
https://github.com/aartbik closed https://github.com/llvm/llvm-project/pull/68360 ___ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] add a forwarding insertion to SparseTensorStorage (PR #68939)
https://github.com/aartbik closed https://github.com/llvm/llvm-project/pull/68939 ___ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
[Lldb-commits] [lldb] [mlir][sparse] remove unused sparse tensor iterator (PR #68951)
https://github.com/aartbik created
https://github.com/llvm/llvm-project/pull/68951
None
>From 64011a22c3e9b97515e34139d0fbb03e0b1fe330 Mon Sep 17 00:00:00 2001
From: Aart Bik
Date: Thu, 12 Oct 2023 21:24:05 -0700
Subject: [PATCH] [mlir][sparse] remove unused sparse tensor iterator
---
.../mlir/Dialect/SparseTensor/IR/Enums.h | 3 -
.../ExecutionEngine/SparseTensorRuntime.h | 22 +---
.../ExecutionEngine/SparseTensorRuntime.cpp | 109 +-
3 files changed, 7 insertions(+), 127 deletions(-)
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/Enums.h
b/mlir/include/mlir/Dialect/SparseTensor/IR/Enums.h
index f1643d66c26a195..1434c649acd29b4 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/Enums.h
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/Enums.h
@@ -146,11 +146,8 @@ enum class Action : uint32_t {
kEmptyForward = 1,
kFromCOO = 2,
kSparseToSparse = 3,
- kFuture = 4, // not used
kToCOO = 5,
- kToIterator = 6,
kPack = 7,
- // Sort an unordered COO in place.
kSortCOOInPlace = 8,
};
diff --git a/mlir/include/mlir/ExecutionEngine/SparseTensorRuntime.h
b/mlir/include/mlir/ExecutionEngine/SparseTensorRuntime.h
index f9312c866f36317..e8dd50d6730c784 100644
--- a/mlir/include/mlir/ExecutionEngine/SparseTensorRuntime.h
+++ b/mlir/include/mlir/ExecutionEngine/SparseTensorRuntime.h
@@ -39,9 +39,9 @@ extern "C" {
/// This is the "swiss army knife" method for materializing sparse
/// tensors into the computation. The types of the `ptr` argument and
-/// the result depend on the action, as explained in the following table
-/// (where "STS" means a sparse-tensor-storage object, "COO" means
-/// a coordinate-scheme object, and "Iterator" means an iterator object).
+/// the result depend on the action, as explained in the following table,
+/// where "STS" means a sparse-tensor-storage object and "COO" means
+/// a coordinate-scheme object.
///
/// Action: `ptr`: Returns:
/// kEmpty - STS, empty
@@ -49,8 +49,8 @@ extern "C" {
/// kFromCOOCOO STS, copied from the COO source
/// kSparseToSparse STS STS, copied from the STS source
/// kToCOO STS COO, copied from the STS source
-/// kToIterator STS Iterator
(@getNext/@delSparseTensorIterator)
/// kPack buffers STS, from level buffers
+/// kSortCOOInPlace STS STS, sorted in place
MLIR_CRUNNERUTILS_EXPORT void *_mlir_ciface_newSparseTensor( // NOLINT
StridedMemRefType *dimSizesRef,
StridedMemRefType *lvlSizesRef,
@@ -90,14 +90,6 @@ MLIR_SPARSETENSOR_FOREVERY_O(DECL_SPARSECOORDINATES)
MLIR_SPARSETENSOR_FOREVERY_V(DECL_FORWARDINGINSERT)
#undef DECL_FORWARDINGINSERT
-/// Coordinate-scheme method for getting the next element while iterating.
-#define DECL_GETNEXT(VNAME, V)
\
- MLIR_CRUNNERUTILS_EXPORT bool _mlir_ciface_getNext##VNAME(
\
- void *iter, StridedMemRefType *cref,
\
- StridedMemRefType *vref);
-MLIR_SPARSETENSOR_FOREVERY_V(DECL_GETNEXT)
-#undef DECL_GETNEXT
-
/// Tensor-storage method to insert elements in lexicographical
/// level-coordinate order.
#define DECL_LEXINSERT(VNAME, V)
\
@@ -201,12 +193,6 @@ MLIR_CRUNNERUTILS_EXPORT void delSparseTensor(void
*tensor);
MLIR_SPARSETENSOR_FOREVERY_V(DECL_DELCOO)
#undef DECL_DELCOO
-/// Releases the memory for an iterator object.
-#define DECL_DELITER(VNAME, V)
\
- MLIR_CRUNNERUTILS_EXPORT void delSparseTensorIterator##VNAME(void *iter);
-MLIR_SPARSETENSOR_FOREVERY_V(DECL_DELITER)
-#undef DECL_DELITER
-
/// Helper function to read a sparse tensor filename from the environment,
/// defined with the naming convention ${TENSOR0}, ${TENSOR1}, etc.
MLIR_CRUNNERUTILS_EXPORT char *getTensorFilename(index_type id);
diff --git a/mlir/lib/ExecutionEngine/SparseTensorRuntime.cpp
b/mlir/lib/ExecutionEngine/SparseTensorRuntime.cpp
index cd1b663578a48ce..ae33a869497a01c 100644
--- a/mlir/lib/ExecutionEngine/SparseTensorRuntime.cpp
+++ b/mlir/lib/ExecutionEngine/SparseTensorRuntime.cpp
@@ -63,71 +63,18 @@ using namespace mlir::sparse_tensor;
//===--===//
//
-// Implementation details for public functions, which don't have a good
-// place to live in the C++ library this file is wrapping.
+// Utilities for manipulating `StridedMemRefType`.
//
//===--===//
namespace {
-/// Wrapper class to avoid memory leakage issues. The `SparseTensorCOO`
-/// class provides a standard C++ iterator interface, where the iterator
-/// is implemented as per `std::vector`'s iterator. However, for MLIR's
-/// usage we need to have an iterator which also holds onto the underlying
