This is an archive of the discontinued LLVM Phabricator instance.

Differential D125043

[mlir][sparse] add more zero-preserving unary ops to sparse compiler
ClosedPublic

Authored by aartbik on May 5 2022, 2:16 PM.

Details

Reviewers
bixia
Commits
rG952fa3018e73: [mlir][sparse] add more zero-preserving unary ops to sparse compiler
Summary

Although we now have semi-rings to deal with arbitrary ops,
it is still good to convey zero-preserving semantics of
ops to the sparse compiler.

Diff Detail

Event Timeline

aartbik created this revision.May 5 2022, 2:16 PM
Herald added a project: Restricted Project. · View Herald TranscriptMay 5 2022, 2:16 PM
Herald added subscribers: sdasgup3, wenzhicui, wrengr and 18 others. · View Herald Transcript
aartbik requested review of this revision.May 5 2022, 2:16 PM
Herald added a project: Restricted Project. · View Herald TranscriptMay 5 2022, 2:16 PM
Herald added subscribers: stephenneuendorffer, nicolasvasilache. · View Herald Transcript
Harbormaster completed remote builds in B163005: Diff 427446.May 5 2022, 3:10 PM
bixia accepted this revision.May 5 2022, 3:11 PM
This revision is now accepted and ready to land.May 5 2022, 3:11 PM
Closed by commit rG952fa3018e73: [mlir][sparse] add more zero-preserving unary ops to sparse compiler (authored by aartbik). · Explain WhyMay 5 2022, 3:35 PM
This revision was automatically updated to reflect the committed changes.
aartbik added a commit: rG952fa3018e73: [mlir][sparse] add more zero-preserving unary ops to sparse compiler.

Revision Contents

PathSize
mlir/
include/
mlir/
Dialect/
SparseTensor/
Utils/
5 lines
lib/
Dialect/
SparseTensor/
Utils/
50 lines
test/
Dialect/
SparseTensor/
459 lines

Diff 427479

mlir/include/mlir/Dialect/SparseTensor/Utils/Merger.h

Show All 27 Linesenum Kind {
// Leaf. // Leaf.
kTensor = 0, kTensor = 0,
kInvariant, kInvariant,
kIndex, kIndex,
// Unary operations. // Unary operations.
kAbsF, kAbsF,
kCeilF, kCeilF,
kFloorF, kFloorF,
kSqrtF,
kExpm1F,
kLog1pF,
kSinF,
kTanhF,
kNegF, kNegF,
kNegI, kNegI,
kTruncF, kTruncF,
kExtF, kExtF,
kCastFS, // signed kCastFS, // signed
kCastFU, // unsigned kCastFU, // unsigned
kCastSF, // signed kCastSF, // signed
kCastUF, // unsigned kCastUF, // unsigned
▲ Show 20 Lines • Show All 247 Lines • Show Last 20 Lines

mlir/lib/Dialect/SparseTensor/Utils/Merger.cpp

Show All 32 Lines case kInvariant:
break; break;
case kIndex: case kIndex:
assert(x != -1u && y == -1u && !v && !o); assert(x != -1u && y == -1u && !v && !o);
index = x; index = x;
break; break;
case kAbsF: case kAbsF:
case kCeilF: case kCeilF:
case kFloorF: case kFloorF:
case kSqrtF:
case kExpm1F:
case kLog1pF:
case kSinF:
case kTanhF:
case kNegF: case kNegF:
case kNegI: case kNegI:
assert(x != -1u && y == -1u && !v && !o); assert(x != -1u && y == -1u && !v && !o);
children.e0 = x; children.e0 = x;
children.e1 = y; children.e1 = y;
break; break;
case kTruncF: case kTruncF:
case kExtF: case kExtF:
▲ Show 20 Lines • Show All 214 Lines • ▼ Show 20 Lines
bool Merger::isSingleCondition(unsigned t, unsigned e) const { bool Merger::isSingleCondition(unsigned t, unsigned e) const {
switch (tensorExps[e].kind) { switch (tensorExps[e].kind) {
case kTensor: case kTensor:
return tensorExps[e].tensor == t; return tensorExps[e].tensor == t;
case kAbsF: case kAbsF:
case kCeilF: case kCeilF:
case kFloorF: case kFloorF:
case kSqrtF:
case kExpm1F:
case kLog1pF:
case kSinF:
case kTanhF:
case kNegF: case kNegF:
case kNegI: case kNegI:
case kTruncF: case kTruncF:
case kExtF: case kExtF:
case kCastFS: case kCastFS:
case kCastFU: case kCastFU:
case kCastSF: case kCastSF:
case kCastUF: case kCastUF:
▲ Show 20 Lines • Show All 46 Lines • ▼ Show 20 Linesstatic const char *kindToOpSymbol(Kind kind) {
case kIndex: case kIndex:
return "index"; return "index";
case kAbsF: case kAbsF:
return "abs"; return "abs";
case kCeilF: case kCeilF:
return "ceil"; return "ceil";
case kFloorF: case kFloorF:
return "floor"; return "floor";
case kSqrtF:
return "sqrt";
case kExpm1F:
return "expm1";
case kLog1pF:
return "log1p";
case kSinF:
return "sin";
case kTanhF:
return "tanh";
case kNegF: case kNegF:
return "-"; return "-";
case kNegI: case kNegI:
return "-"; return "-";
case kTruncF: case kTruncF:
case kExtF: case kExtF:
case kCastFS: case kCastFS:
case kCastFU: case kCastFU:
▲ Show 20 Lines • Show All 58 Lines • ▼ Show 20 Lines case kInvariant:
llvm::dbgs() << "invariant"; llvm::dbgs() << "invariant";
break; break;
case kIndex: case kIndex:
llvm::dbgs() << "index_" << tensorExps[e].index; llvm::dbgs() << "index_" << tensorExps[e].index;
break; break;
case kAbsF: case kAbsF:
case kCeilF: case kCeilF:
case kFloorF: case kFloorF:
case kSqrtF:
case kExpm1F:
case kLog1pF:
case kSinF:
case kTanhF:
case kNegF: case kNegF:
case kNegI: case kNegI:
case kTruncF: case kTruncF:
case kExtF: case kExtF:
case kCastFS: case kCastFS:
case kCastFU: case kCastFU:
case kCastSF: case kCastSF:
case kCastUF: case kCastUF:
▲ Show 20 Lines • Show All 82 Lines • ▼ Show 20 Lines if (kind == kTensor) {
t = syntheticTensor; t = syntheticTensor;
} }
latSets[s].push_back(addLat(t, i, e)); latSets[s].push_back(addLat(t, i, e));
return s; return s;
} }
case kAbsF: case kAbsF:
case kCeilF: case kCeilF:
case kFloorF: case kFloorF:
case kSqrtF:
case kExpm1F:
case kLog1pF:
case kSinF:
case kTanhF:
case kNegF: case kNegF:
case kNegI: case kNegI:
case kTruncF: case kTruncF:
case kExtF: case kExtF:
case kCastFS: case kCastFS:
case kCastFU: case kCastFU:
case kCastSF: case kCastSF:
case kCastUF: case kCastUF:
▲ Show 20 Lines • Show All 194 Lines • ▼ Show 20 Lines if (def->getNumOperands() == 1) {
if (x.hasValue()) { if (x.hasValue()) {
unsigned e = x.getValue(); unsigned e = x.getValue();
if (isa<math::AbsOp>(def)) if (isa<math::AbsOp>(def))
return addExp(kAbsF, e); return addExp(kAbsF, e);
if (isa<math::CeilOp>(def)) if (isa<math::CeilOp>(def))
return addExp(kCeilF, e); return addExp(kCeilF, e);
if (isa<math::FloorOp>(def)) if (isa<math::FloorOp>(def))
return addExp(kFloorF, e); return addExp(kFloorF, e);
if (isa<math::SqrtOp>(def))
return addExp(kSqrtF, e);
if (isa<math::ExpM1Op>(def))
return addExp(kExpm1F, e);
if (isa<math::Log1pOp>(def))
return addExp(kLog1pF, e);
if (isa<math::SinOp>(def))
return addExp(kSinF, e);
if (isa<math::TanhOp>(def))
return addExp(kTanhF, e);
if (isa<arith::NegFOp>(def)) if (isa<arith::NegFOp>(def))
return addExp(kNegF, e); // no negi in std return addExp(kNegF, e); // no negi in std
if (isa<arith::TruncFOp>(def)) if (isa<arith::TruncFOp>(def))
return addExp(kTruncF, e, v); return addExp(kTruncF, e, v);
if (isa<arith::ExtFOp>(def)) if (isa<arith::ExtFOp>(def))
return addExp(kExtF, e, v); return addExp(kExtF, e, v);
if (isa<arith::FPToSIOp>(def)) if (isa<arith::FPToSIOp>(def))
return addExp(kCastFS, e, v); return addExp(kCastFS, e, v);
▲ Show 20 Lines • Show All 118 Lines • ▼ Show 20 Lines case kIndex:
llvm_unreachable("unexpected non-op"); llvm_unreachable("unexpected non-op");
// Unary ops. // Unary ops.
case kAbsF: case kAbsF:
return rewriter.create<math::AbsOp>(loc, v0); return rewriter.create<math::AbsOp>(loc, v0);
case kCeilF: case kCeilF:
return rewriter.create<math::CeilOp>(loc, v0); return rewriter.create<math::CeilOp>(loc, v0);
case kFloorF: case kFloorF:
return rewriter.create<math::FloorOp>(loc, v0); return rewriter.create<math::FloorOp>(loc, v0);
case kSqrtF:
return rewriter.create<math::SqrtOp>(loc, v0);
case kExpm1F:
return rewriter.create<math::ExpM1Op>(loc, v0);
case kLog1pF:
return rewriter.create<math::Log1pOp>(loc, v0);
case kSinF:
return rewriter.create<math::SinOp>(loc, v0);
case kTanhF:
return rewriter.create<math::TanhOp>(loc, v0);
case kNegF: case kNegF:
return rewriter.create<arith::NegFOp>(loc, v0); return rewriter.create<arith::NegFOp>(loc, v0);
case kNegI: // no negi in std case kNegI: // no negi in std
return rewriter.create<arith::SubIOp>( return rewriter.create<arith::SubIOp>(
loc, loc,
rewriter.create<arith::ConstantOp>(loc, v0.getType(), rewriter.create<arith::ConstantOp>(loc, v0.getType(),
rewriter.getZeroAttr(v0.getType())), rewriter.getZeroAttr(v0.getType())),
v0); v0);
▲ Show 20 Lines • Show All 67 Lines • Show Last 20 Lines

mlir/test/Dialect/SparseTensor/sparse_fp_ops.mlir

Show All 25 Lines#traitc = {
indexing_maps = [ indexing_maps = [
affine_map<(i) -> (i)>, // a affine_map<(i) -> (i)>, // a
affine_map<(i) -> (i)> // x (out) affine_map<(i) -> (i)> // x (out)
], ],
iterator_types = ["parallel"], iterator_types = ["parallel"],
doc = "x(i) = a(i) OP c" doc = "x(i) = a(i) OP c"
} }
// CHECK-LABEL: func @abs( // CHECK-LABEL: func @abs(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK: %[[VAL_2:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_2:.*]] = arith.constant 0 : index
// CHECK: %[[VAL_3:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64> // CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64>
// CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex> // CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
// CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex> // CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex>
// CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] { // CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] {
// CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex> // CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex>
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64> // CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64>
// CHECK: %[[VAL_13:.*]] = math.abs %[[VAL_12]] : f64 // CHECK: %[[VAL_13:.*]] = math.abs %[[VAL_12]] : f64
// CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64> // CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64> // CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64>
// CHECK: return %[[VAL_14]] : tensor<32xf64> // CHECK: return %[[VAL_14]] : tensor<32xf64>
// CHECK: }
func.func @abs(%arga: tensor<32xf64, #SV>, func.func @abs(%arga: tensor<32xf64, #SV>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%0 = linalg.generic #trait1 %0 = linalg.generic #trait1
ins(%arga: tensor<32xf64, #SV>) ins(%arga: tensor<32xf64, #SV>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %x: f64): ^bb(%a: f64, %x: f64):
%0 = math.abs %a : f64 %0 = math.abs %a : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @ceil( // CHECK-LABEL: func @ceil(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK: %[[VAL_2:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_2:.*]] = arith.constant 0 : index
// CHECK: %[[VAL_3:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64> // CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64>
// CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex> // CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
// CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex> // CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex>
// CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] { // CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] {
// CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex> // CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex>
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64> // CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64>
// CHECK: %[[VAL_13:.*]] = math.ceil %[[VAL_12]] : f64 // CHECK: %[[VAL_13:.*]] = math.ceil %[[VAL_12]] : f64
// CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64> // CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64> // CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64>
// CHECK: return %[[VAL_14]] : tensor<32xf64> // CHECK: return %[[VAL_14]] : tensor<32xf64>
// CHECK: } // CHECK: }
func.func @ceil(%arga: tensor<32xf64, #SV>, func.func @ceil(%arga: tensor<32xf64, #SV>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%0 = linalg.generic #trait1 %0 = linalg.generic #trait1
ins(%arga: tensor<32xf64, #SV>) ins(%arga: tensor<32xf64, #SV>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %x: f64): ^bb(%a: f64, %x: f64):
%0 = math.ceil %a : f64 %0 = math.ceil %a : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @floor( // CHECK-LABEL: func @floor(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK: %[[VAL_2:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_2:.*]] = arith.constant 0 : index
// CHECK: %[[VAL_3:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64> // CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64>
// CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex> // CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
// CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex> // CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex>
// CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] { // CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] {
// CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex> // CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex>
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64> // CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64>
// CHECK: %[[VAL_13:.*]] = math.floor %[[VAL_12]] : f64 // CHECK: %[[VAL_13:.*]] = math.floor %[[VAL_12]] : f64
// CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64> // CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64> // CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64>
// CHECK: return %[[VAL_14]] : tensor<32xf64> // CHECK: return %[[VAL_14]] : tensor<32xf64>
// CHECK: } // CHECK: }
func.func @floor(%arga: tensor<32xf64, #SV>, func.func @floor(%arga: tensor<32xf64, #SV>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%0 = linalg.generic #trait1 %0 = linalg.generic #trait1
ins(%arga: tensor<32xf64, #SV>) ins(%arga: tensor<32xf64, #SV>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %x: f64): ^bb(%a: f64, %x: f64):
%0 = math.floor %a : f64 %0 = math.floor %a : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @neg( // CHECK-LABEL: func @neg(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK: %[[VAL_2:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_2:.*]] = arith.constant 0 : index
// CHECK: %[[VAL_3:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64> // CHECK: %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64>
// CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex> // CHECK: %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
// CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex> // CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex>
// CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] { // CHECK: scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] {
// CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex> // CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex>
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64> // CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf64>
// CHECK: %[[VAL_13:.*]] = arith.negf %[[VAL_12]] : f64 // CHECK: %[[VAL_13:.*]] = arith.negf %[[VAL_12]] : f64
// CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64> // CHECK: memref.store %[[VAL_13]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64> // CHECK: %[[VAL_14:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<32xf64>
// CHECK: return %[[VAL_14]] : tensor<32xf64> // CHECK: return %[[VAL_14]] : tensor<32xf64>
// CHECK: } // CHECK: }
func.func @neg(%arga: tensor<32xf64, #SV>, func.func @neg(%arga: tensor<32xf64, #SV>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%0 = linalg.generic #trait1 %0 = linalg.generic #trait1
ins(%arga: tensor<32xf64, #SV>) ins(%arga: tensor<32xf64, #SV>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %x: f64): ^bb(%a: f64, %x: f64):
%0 = arith.negf %a : f64 %0 = arith.negf %a : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @add( // CHECK-LABEL: func @add(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64>, // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64>,
// CHECK-SAME: %[[VAL_2:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_2:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 32 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 32 : index
// CHECK-DAG: %[[VAL_4:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_4:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[VAL_5:.*]] = arith.constant true // CHECK-DAG: %[[VAL_5:.*]] = arith.constant true
// CHECK-DAG: %[[VAL_6:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_6:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf64> // CHECK: %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf64>
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex> // CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex>
// CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex> // CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex>
// CHECK: %[[VAL_14:.*]]:2 = scf.while (%[[VAL_15:.*]] = %[[VAL_12]], %[[VAL_16:.*]] = %[[VAL_4]]) : (index, index) -> (index, index) { // CHECK: %[[VAL_14:.*]]:2 = scf.while (%[[VAL_15:.*]] = %[[VAL_12]], %[[VAL_16:.*]] = %[[VAL_4]]) : (index, index) -> (index, index) {
// CHECK: %[[VAL_17:.*]] = arith.cmpi ult, %[[VAL_15]], %[[VAL_13]] : index // CHECK: %[[VAL_17:.*]] = arith.cmpi ult, %[[VAL_15]], %[[VAL_13]] : index
// CHECK: scf.condition(%[[VAL_17]]) %[[VAL_15]], %[[VAL_16]] : index, index // CHECK: scf.condition(%[[VAL_17]]) %[[VAL_15]], %[[VAL_16]] : index, index
// CHECK: } do { // CHECK: } do {
// CHECK: ^bb0(%[[VAL_18:.*]]: index, %[[VAL_19:.*]]: index): // CHECK: ^bb0(%[[VAL_18:.*]]: index, %[[VAL_19:.*]]: index):
// CHECK: %[[VAL_20:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_18]]] : memref<?xindex> // CHECK: %[[VAL_20:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_18]]] : memref<?xindex>
// CHECK: %[[VAL_21:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index // CHECK: %[[VAL_21:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index
// CHECK: scf.if %[[VAL_21]] { // CHECK: scf.if %[[VAL_21]] {
// CHECK: %[[VAL_22:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_18]]] : memref<?xf64> // CHECK: %[[VAL_22:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_18]]] : memref<?xf64>
// CHECK: %[[VAL_23:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: %[[VAL_23:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: %[[VAL_24:.*]] = arith.addf %[[VAL_22]], %[[VAL_23]] : f64 // CHECK: %[[VAL_24:.*]] = arith.addf %[[VAL_22]], %[[VAL_23]] : f64
// CHECK: memref.store %[[VAL_24]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: memref.store %[[VAL_24]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: } else { // CHECK: } else {
// CHECK: scf.if %[[VAL_5]] { // CHECK: scf.if %[[VAL_5]] {
// CHECK: %[[VAL_25:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: %[[VAL_25:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: memref.store %[[VAL_25]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: memref.store %[[VAL_25]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: } else { // CHECK: } else {
// CHECK: } // CHECK: }
// CHECK: } // CHECK: }
// CHECK: %[[VAL_26:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index // CHECK: %[[VAL_26:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index
// CHECK: %[[VAL_27:.*]] = arith.addi %[[VAL_18]], %[[VAL_6]] : index // CHECK: %[[VAL_27:.*]] = arith.addi %[[VAL_18]], %[[VAL_6]] : index
// CHECK: %[[VAL_28:.*]] = arith.select %[[VAL_26]], %[[VAL_27]], %[[VAL_18]] : index // CHECK: %[[VAL_28:.*]] = arith.select %[[VAL_26]], %[[VAL_27]], %[[VAL_18]] : index
// CHECK: %[[VAL_29:.*]] = arith.addi %[[VAL_19]], %[[VAL_6]] : index // CHECK: %[[VAL_29:.*]] = arith.addi %[[VAL_19]], %[[VAL_6]] : index
// CHECK: scf.yield %[[VAL_28]], %[[VAL_29]] : index, index // CHECK: scf.yield %[[VAL_28]], %[[VAL_29]] : index, index
// CHECK: } // CHECK: }
// CHECK: scf.for %[[VAL_30:.*]] = %[[VAL_31:.*]]#1 to %[[VAL_3]] step %[[VAL_6]] { // CHECK: scf.for %[[VAL_30:.*]] = %[[VAL_31:.*]]#1 to %[[VAL_3]] step %[[VAL_6]] {
// CHECK: %[[VAL_32:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_30]]] : memref<32xf64> // CHECK: %[[VAL_32:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_30]]] : memref<32xf64>
// CHECK: memref.store %[[VAL_32]], %[[VAL_11]]{{\[}}%[[VAL_30]]] : memref<32xf64> // CHECK: memref.store %[[VAL_32]], %[[VAL_11]]{{\[}}%[[VAL_30]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_33:.*]] = bufferization.to_tensor %[[VAL_11]] : memref<32xf64> // CHECK: %[[VAL_33:.*]] = bufferization.to_tensor %[[VAL_11]] : memref<32xf64>
// CHECK: return %[[VAL_33]] : tensor<32xf64> // CHECK: return %[[VAL_33]] : tensor<32xf64>
// CHECK: } // CHECK: }
func.func @add(%arga: tensor<32xf64, #SV>, func.func @add(%arga: tensor<32xf64, #SV>,
%argb: tensor<32xf64>, %argb: tensor<32xf64>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%0 = linalg.generic #trait2 %0 = linalg.generic #trait2
ins(%arga, %argb: tensor<32xf64, #SV>, tensor<32xf64>) ins(%arga, %argb: tensor<32xf64, #SV>, tensor<32xf64>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %b: f64, %x: f64): ^bb(%a: f64, %b: f64, %x: f64):
%0 = arith.addf %a, %b : f64 %0 = arith.addf %a, %b : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @sub( // CHECK-LABEL: func @sub(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64>, // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64>,
// CHECK-SAME: %[[VAL_2:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_2:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 32 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 32 : index
// CHECK-DAG: %[[VAL_4:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_4:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[VAL_5:.*]] = arith.constant true // CHECK-DAG: %[[VAL_5:.*]] = arith.constant true
// CHECK-DAG: %[[VAL_6:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_6:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex> // CHECK: %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64> // CHECK: %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64>
// CHECK: %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf64> // CHECK: %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf64>
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex> // CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex>
// CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex> // CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex>
// CHECK: %[[VAL_14:.*]]:2 = scf.while (%[[VAL_15:.*]] = %[[VAL_12]], %[[VAL_16:.*]] = %[[VAL_4]]) : (index, index) -> (index, index) { // CHECK: %[[VAL_14:.*]]:2 = scf.while (%[[VAL_15:.*]] = %[[VAL_12]], %[[VAL_16:.*]] = %[[VAL_4]]) : (index, index) -> (index, index) {
// CHECK: %[[VAL_17:.*]] = arith.cmpi ult, %[[VAL_15]], %[[VAL_13]] : index // CHECK: %[[VAL_17:.*]] = arith.cmpi ult, %[[VAL_15]], %[[VAL_13]] : index
// CHECK: scf.condition(%[[VAL_17]]) %[[VAL_15]], %[[VAL_16]] : index, index // CHECK: scf.condition(%[[VAL_17]]) %[[VAL_15]], %[[VAL_16]] : index, index
// CHECK: } do { // CHECK: } do {
// CHECK: ^bb0(%[[VAL_18:.*]]: index, %[[VAL_19:.*]]: index): // CHECK: ^bb0(%[[VAL_18:.*]]: index, %[[VAL_19:.*]]: index):
// CHECK: %[[VAL_20:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_18]]] : memref<?xindex> // CHECK: %[[VAL_20:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_18]]] : memref<?xindex>
// CHECK: %[[VAL_21:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index // CHECK: %[[VAL_21:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index
// CHECK: scf.if %[[VAL_21]] { // CHECK: scf.if %[[VAL_21]] {
// CHECK: %[[VAL_22:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_18]]] : memref<?xf64> // CHECK: %[[VAL_22:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_18]]] : memref<?xf64>
// CHECK: %[[VAL_23:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: %[[VAL_23:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: %[[VAL_24:.*]] = arith.subf %[[VAL_22]], %[[VAL_23]] : f64 // CHECK: %[[VAL_24:.*]] = arith.subf %[[VAL_22]], %[[VAL_23]] : f64
// CHECK: memref.store %[[VAL_24]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: memref.store %[[VAL_24]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: } else { // CHECK: } else {
// CHECK: scf.if %[[VAL_5]] { // CHECK: scf.if %[[VAL_5]] {
// CHECK: %[[VAL_25:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: %[[VAL_25:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: %[[VAL_26:.*]] = arith.negf %[[VAL_25]] : f64 // CHECK: %[[VAL_26:.*]] = arith.negf %[[VAL_25]] : f64
// CHECK: memref.store %[[VAL_26]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64> // CHECK: memref.store %[[VAL_26]], %[[VAL_11]]{{\[}}%[[VAL_19]]] : memref<32xf64>
// CHECK: } else { // CHECK: } else {
// CHECK: } // CHECK: }
// CHECK: } // CHECK: }
// CHECK: %[[VAL_27:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index // CHECK: %[[VAL_27:.*]] = arith.cmpi eq, %[[VAL_20]], %[[VAL_19]] : index
// CHECK: %[[VAL_28:.*]] = arith.addi %[[VAL_18]], %[[VAL_6]] : index // CHECK: %[[VAL_28:.*]] = arith.addi %[[VAL_18]], %[[VAL_6]] : index
// CHECK: %[[VAL_29:.*]] = arith.select %[[VAL_27]], %[[VAL_28]], %[[VAL_18]] : index // CHECK: %[[VAL_29:.*]] = arith.select %[[VAL_27]], %[[VAL_28]], %[[VAL_18]] : index
// CHECK: %[[VAL_30:.*]] = arith.addi %[[VAL_19]], %[[VAL_6]] : index // CHECK: %[[VAL_30:.*]] = arith.addi %[[VAL_19]], %[[VAL_6]] : index
// CHECK: scf.yield %[[VAL_29]], %[[VAL_30]] : index, index // CHECK: scf.yield %[[VAL_29]], %[[VAL_30]] : index, index
// CHECK: } // CHECK: }
// CHECK: scf.for %[[VAL_31:.*]] = %[[VAL_32:.*]]#1 to %[[VAL_3]] step %[[VAL_6]] { // CHECK: scf.for %[[VAL_31:.*]] = %[[VAL_32:.*]]#1 to %[[VAL_3]] step %[[VAL_6]] {
// CHECK: %[[VAL_33:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_31]]] : memref<32xf64> // CHECK: %[[VAL_33:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_31]]] : memref<32xf64>
// CHECK: %[[VAL_34:.*]] = arith.negf %[[VAL_33]] : f64 // CHECK: %[[VAL_34:.*]] = arith.negf %[[VAL_33]] : f64
// CHECK: memref.store %[[VAL_34]], %[[VAL_11]]{{\[}}%[[VAL_31]]] : memref<32xf64> // CHECK: memref.store %[[VAL_34]], %[[VAL_11]]{{\[}}%[[VAL_31]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_35:.*]] = bufferization.to_tensor %[[VAL_11]] : memref<32xf64> // CHECK: %[[VAL_35:.*]] = bufferization.to_tensor %[[VAL_11]] : memref<32xf64>
// CHECK: return %[[VAL_35]] : tensor<32xf64> // CHECK: return %[[VAL_35]] : tensor<32xf64>
// CHECK: } // CHECK: }
func.func @sub(%arga: tensor<32xf64, #SV>, func.func @sub(%arga: tensor<32xf64, #SV>,
%argb: tensor<32xf64>, %argb: tensor<32xf64>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%0 = linalg.generic #trait2 %0 = linalg.generic #trait2
ins(%arga, %argb: tensor<32xf64, #SV>, tensor<32xf64>) ins(%arga, %argb: tensor<32xf64, #SV>, tensor<32xf64>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %b: f64, %x: f64): ^bb(%a: f64, %b: f64, %x: f64):
%0 = arith.subf %a, %b : f64 %0 = arith.subf %a, %b : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @mul( // CHECK-LABEL: func @mul(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64>, // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64>,
// CHECK-SAME: %[[VAL_2:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_2:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[VAL_4:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_4:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf64> // CHECK: %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf64>
// CHECK: %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex> // CHECK: %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
// CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex> // CHECK: %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
// CHECK: scf.for %[[VAL_12:.*]] = %[[VAL_10]] to %[[VAL_11]] step %[[VAL_4]] { // CHECK: scf.for %[[VAL_12:.*]] = %[[VAL_10]] to %[[VAL_11]] step %[[VAL_4]] {
// CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_12]]] : memref<?xindex> // CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_12]]] : memref<?xindex>
// CHECK: %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_12]]] : memref<?xf64> // CHECK: %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_12]]] : memref<?xf64>
// CHECK: %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_13]]] : memref<32xf64> // CHECK: %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_13]]] : memref<32xf64>
// CHECK: %[[VAL_16:.*]] = arith.mulf %[[VAL_14]], %[[VAL_15]] : f64 // CHECK: %[[VAL_16:.*]] = arith.mulf %[[VAL_14]], %[[VAL_15]] : f64
// CHECK: memref.store %[[VAL_16]], %[[VAL_9]]{{\[}}%[[VAL_13]]] : memref<32xf64> // CHECK: memref.store %[[VAL_16]], %[[VAL_9]]{{\[}}%[[VAL_13]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_17:.*]] = bufferization.to_tensor %[[VAL_9]] : memref<32xf64> // CHECK: %[[VAL_17:.*]] = bufferization.to_tensor %[[VAL_9]] : memref<32xf64>
// CHECK: return %[[VAL_17]] : tensor<32xf64> // CHECK: return %[[VAL_17]] : tensor<32xf64>
// CHECK: } // CHECK: }
func.func @mul(%arga: tensor<32xf64, #SV>, func.func @mul(%arga: tensor<32xf64, #SV>,
%argb: tensor<32xf64>, %argb: tensor<32xf64>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%0 = linalg.generic #trait2 %0 = linalg.generic #trait2
ins(%arga, %argb: tensor<32xf64, #SV>, tensor<32xf64>) ins(%arga, %argb: tensor<32xf64, #SV>, tensor<32xf64>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %b: f64, %x: f64): ^bb(%a: f64, %b: f64, %x: f64):
%0 = arith.mulf %a, %b : f64 %0 = arith.mulf %a, %b : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @divbyc( // CHECK-LABEL: func @divbyc(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, // CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>,
// CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { // CHECK-SAME: %[[VAL_1:.*]]: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
// CHECK-DAG: %[[VAL_2:.*]] = arith.constant 2.000000e+00 : f64 // CHECK-DAG: %[[VAL_2:.*]] = arith.constant 2.000000e+00 : f64
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 0 : index // CHECK-DAG: %[[VAL_3:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[VAL_4:.*]] = arith.constant 1 : index // CHECK-DAG: %[[VAL_4:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> // CHECK: %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64> // CHECK: %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf64>
// CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex> // CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
// CHECK: %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex> // CHECK: %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
// CHECK: scf.for %[[VAL_11:.*]] = %[[VAL_9]] to %[[VAL_10]] step %[[VAL_4]] { // CHECK: scf.for %[[VAL_11:.*]] = %[[VAL_9]] to %[[VAL_10]] step %[[VAL_4]] {
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_11]]] : memref<?xindex> // CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_11]]] : memref<?xindex>
// CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<?xf64> // CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<?xf64>
// CHECK: %[[VAL_14:.*]] = arith.divf %[[VAL_13]], %[[VAL_2]] : f64 // CHECK: %[[VAL_14:.*]] = arith.divf %[[VAL_13]], %[[VAL_2]] : f64
// CHECK: memref.store %[[VAL_14]], %[[VAL_8]]{{\[}}%[[VAL_12]]] : memref<32xf64> // CHECK: memref.store %[[VAL_14]], %[[VAL_8]]{{\[}}%[[VAL_12]]] : memref<32xf64>
// CHECK: } // CHECK: }
// CHECK: %[[VAL_15:.*]] = bufferization.to_tensor %[[VAL_8]] : memref<32xf64> // CHECK: %[[VAL_15:.*]] = bufferization.to_tensor %[[VAL_8]] : memref<32xf64>
// CHECK: return %[[VAL_15]] : tensor<32xf64> // CHECK: return %[[VAL_15]] : tensor<32xf64>
// CHECK: } // CHECK: }
func.func @divbyc(%arga: tensor<32xf64, #SV>, func.func @divbyc(%arga: tensor<32xf64, #SV>,
%argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> { %argx: tensor<32xf64> {linalg.inplaceable = true}) -> tensor<32xf64> {
%c = arith.constant 2.0 : f64 %c = arith.constant 2.0 : f64
%0 = linalg.generic #traitc %0 = linalg.generic #traitc
ins(%arga: tensor<32xf64, #SV>) ins(%arga: tensor<32xf64, #SV>)
outs(%argx: tensor<32xf64>) { outs(%argx: tensor<32xf64>) {
^bb(%a: f64, %x: f64): ^bb(%a: f64, %x: f64):
%0 = arith.divf %a, %c : f64 %0 = arith.divf %a, %c : f64
linalg.yield %0 : f64 linalg.yield %0 : f64
} -> tensor<32xf64> } -> tensor<32xf64>
return %0 : tensor<32xf64> return %0 : tensor<32xf64>
} }
// CHECK-LABEL: func @zero_preserving_math(
// CHECK-SAME: %[[VAL_0:.*]]: tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> {
// CHECK-DAG: %[[VAL_1:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[VAL_2:.*]] = arith.constant 1 : index
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 32 : index
// CHECK: %[[VAL_4:.*]] = sparse_tensor.init{{\[}}%[[VAL_3]]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_1]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_1]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
// CHECK: %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xf64>
// CHECK: %[[VAL_8:.*]] = memref.alloca(%[[VAL_2]]) : memref<?xindex>
// CHECK: %[[VAL_9:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_1]]] : memref<?xindex>
// CHECK: %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_2]]] : memref<?xindex>
// CHECK: scf.for %[[VAL_11:.*]] = %[[VAL_9]] to %[[VAL_10]] step %[[VAL_2]] {
// CHECK: %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_11]]] : memref<?xindex>
// CHECK: memref.store %[[VAL_12]], %[[VAL_8]]{{\[}}%[[VAL_1]]] : memref<?xindex>
// CHECK: %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<?xf64>
// CHECK: %[[VAL_14:.*]] = math.abs %[[VAL_13]] : f64
// CHECK: %[[VAL_15:.*]] = math.ceil %[[VAL_14]] : f64
// CHECK: %[[VAL_16:.*]] = math.floor %[[VAL_15]] : f64
// CHECK: %[[VAL_17:.*]] = math.sqrt %[[VAL_16]] : f64
// CHECK: %[[VAL_18:.*]] = math.expm1 %[[VAL_17]] : f64
// CHECK: %[[VAL_19:.*]] = math.log1p %[[VAL_18]] : f64
// CHECK: %[[VAL_20:.*]] = math.sin %[[VAL_19]] : f64
// CHECK: %[[VAL_21:.*]] = math.tanh %[[VAL_20]] : f64
// CHECK: sparse_tensor.lex_insert %[[VAL_4]], %[[VAL_8]], %[[VAL_21]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>, memref<?xindex>, f64
// CHECK: }
// CHECK: %[[VAL_22:.*]] = sparse_tensor.load %[[VAL_4]] hasInserts : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: return %[[VAL_22]] : tensor<32xf64, #sparse_tensor.encoding<{{{.*}}}>>
// CHECK: }
func.func @zero_preserving_math(%arga: tensor<32xf64, #SV>) -> tensor<32xf64, #SV> {
%c32 = arith.constant 32 : index
%xinp = sparse_tensor.init [%c32] : tensor<32xf64, #SV>
%0 = linalg.generic #trait1
ins(%arga: tensor<32xf64, #SV>)
outs(%xinp: tensor<32xf64, #SV>) {
^bb(%a: f64, %x: f64):
%0 = math.abs %a : f64
%1 = math.ceil %0 : f64
%2 = math.floor %1 : f64
%3 = math.sqrt %2 : f64
%4 = math.expm1 %3 : f64
%5 = math.log1p %4 : f64
%6 = math.sin %5 : f64
%7 = math.tanh %6 : f64
linalg.yield %7 : f64
} -> tensor<32xf64, #SV>
return %0 : tensor<32xf64, #SV>
}