Skip to content

Commit 7fdd03b

Browse files
boomanaiden154Google-ML-Automation
boomanaiden154
authored and
Google-ML-Automation
committed
Updates LLVM usage to match [4f39a4ff0ada](llvm/llvm-project@4f39a4ff0ada) PiperOrigin-RevId: 837108380
1 parent 6bb9119 commit 7fdd03b

File tree

4 files changed

+80
-151
lines changed

4 files changed

+80
-151
lines changed

‎third_party/llvm/generated.patch‎

Lines changed: 27 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -1,17 +1,29 @@
11
Auto generated patch. Do not edit or delete it, even if empty.
2-
diff -ruN --strip-trailing-cr a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
3-
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
4-
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
5-
@@ -4095,6 +4095,12 @@
6-
llvm::SmallVector<size_t> occludedChildren;
7-
llvm::sort(
8-
indices.begin(), indices.end(), [&](const size_t a, const size_t b) {
9-
+ // Bail early if we are asked to look at the same index. If we do not
10-
+ // bail early, we can end up mistakenly adding indices to
11-
+ // occludedChildren. This can occur with some types of libc++ hardening.
12-
+ if (a == b)
13-
+ return false;
14-
+
15-
auto memberIndicesA = cast<ArrayAttr>(indexAttr[a]);
16-
auto memberIndicesB = cast<ArrayAttr>(indexAttr[b]);
2+
diff -ruN --strip-trailing-cr a/clang/lib/Analysis/ThreadSafety.cpp b/clang/lib/Analysis/ThreadSafety.cpp
3+
--- a/clang/lib/Analysis/ThreadSafety.cpp
4+
+++ b/clang/lib/Analysis/ThreadSafety.cpp
5+
@@ -2820,7 +2820,7 @@
6+
case CFGElement::AutomaticObjectDtor: {
7+
CFGAutomaticObjDtor AD = BI.castAs<CFGAutomaticObjDtor>();
8+
const auto *DD = AD.getDestructorDecl(AC.getASTContext());
9+
- if (!DD->hasAttrs())
10+
+ if (!DD || !DD->hasAttrs())
11+
break;
1712

13+
LocksetBuilder.handleCall(
14+
diff -ruN --strip-trailing-cr a/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp b/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp
15+
--- a/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp
16+
+++ b/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp
17+
@@ -0,0 +1,12 @@
18+
+// RUN: %clang_cc1 -fsyntax-only -verify -std=c++11 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=0 %s
19+
+// RUN: %clang_cc1 -fsyntax-only -verify -std=c++11 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=1 %s
20+
+// RUN: %clang_cc1 -fsyntax-only -verify -std=c++17 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=0 %s
21+
+// RUN: %clang_cc1 -fsyntax-only -verify -std=c++17 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=1 %s
22+
+// expected-no-diagnostics
23+
+
24+
+struct foo {
25+
+ ~foo();
26+
+};
27+
+struct bar : foo {};
28+
+struct baz : bar {};
29+
+baz foobar(baz a) { return a; }

‎third_party/llvm/workspace.bzl‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -4,8 +4,8 @@ load("//third_party:repo.bzl", "tf_http_archive")
44

55
def repo(name):
66
"""Imports LLVM."""
7-
LLVM_COMMIT = "26362c68579dd4375198aae4651b4d5f8a36c715"
8-
LLVM_SHA256 = "1b81809d98940d0a6d4f19ef9e0bf72cd5847b9bbed47bc3517fcf8a40d38fd9"
7+
LLVM_COMMIT = "4f39a4ff0ada92870ca1c2dccad382ea04947da8"
8+
LLVM_SHA256 = "264c7cc3e166b840911494a2e94cff2ae8730b56239bd91dc8f65c8dc9468262"
99

1010
tf_http_archive(
1111
name = name,

‎third_party/shardy/temporary.patch‎

Lines changed: 49 additions & 132 deletions
Original file line numberDiff line numberDiff line change
@@ -1,147 +1,64 @@
1-
diff --git a/shardy/dialect/sdy/transforms/export/explicit_reshards_util.cc b/shardy/dialect/sdy/transforms/export/explicit_reshards_util.cc
2-
index f8e9369..f1bece8 100644
3-
--- a/shardy/dialect/sdy/transforms/export/explicit_reshards_util.cc
4-
+++ b/shardy/dialect/sdy/transforms/export/explicit_reshards_util.cc
5-
@@ -330,8 +330,7 @@ int64_t getCommunicationCost(const ShardingProjection& shardingProjection,
6-
OpShardingRuleAttr shardingRule,
7-
ArrayRef<int64_t> tensorSizes,
8-
ArrayRef<int64_t> localTensorSizes, MeshAttr mesh,
9-
- const FactorAxesPair& factorAxesPair,
10-
- const int64_t expandedShardingSize) {
11-
+ const FactorAxesPair& factorAxesPair) {
12-
// The relative cost of collective operations.
13-
constexpr int64_t allToAllCost = 1;
14-
constexpr int64_t collectivePermuteCost = 2;
15-
@@ -408,14 +407,10 @@ int64_t getCommunicationCost(const ShardingProjection& shardingProjection,
16-
// If the result contains this factor, we need
17-
// 1. all-to-all to move AX from this factor to other factors.
18-
// 2. all-gather to shrink the sharding size after the all-to-all above.
19-
- for (const auto& [localTensorSize, tensorFactorSharding] : llvm::zip_equal(
20-
+ for (const auto& [tensorSize, tensorFactorSharding] : llvm::zip_equal(
21-
localTensorSizes.drop_front(shardingProjection.getNumOperands()),
22-
shardingProjection.getResults())) {
23-
- // A candidate factor axes (factorAxesPair) is guaranteed to be an expansion
24-
- // of its existing sharding and `localTensorSize` has already taken into its
25-
- // existing sharding. In order to avoid double counting, it needs to shard
26-
- // further on the expanded sharding size only.
27-
- int64_t shardedTensorSize = localTensorSize / expandedShardingSize;
28-
+ int64_t shardedTensorSize = tensorSize / axesXSize;
29-
auto [axesA, axesB] = getShardingAxesInOtherAndThisFactor(
30-
tensorFactorSharding, factorAxesPair.factorIndex);
31-
32-
@@ -533,16 +528,9 @@ class FactorAxesCandidateBag {
33-
34-
FactorAxesCandidate bestCandidate;
35-
for (FactorAxesCandidate& candidate : candidates) {
36-
- // NOTE: The axes on replication factors are distributed to batching
37-
- // dimensions after the common axes are found for all non-replication
38-
- // factors. The communication cost calculation does not take this into
39-
- // account yet and hence is not ready for cases that sharding rule has
40-
- // replication factors.
41-
- if (shardingRule.getNeedReplicationFactors().empty()) {
42-
- candidate.communicationCost = getCommunicationCost(
43-
- shardingProjection, shardingRule, tensorSizes, localTensorSizes,
44-
- mesh, candidate.factorAxes, candidate.shardingSize);
45-
- }
46-
+ candidate.communicationCost =
47-
+ getCommunicationCost(shardingProjection, shardingRule, tensorSizes,
48-
+ localTensorSizes, mesh, candidate.factorAxes);
49-
if (isValid(candidate)) {
50-
bestCandidate = std::max(bestCandidate, candidate);
51-
}
52-
diff --git a/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/cholesky_triangular_solve.mlir b/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/cholesky_triangular_solve.mlir
53-
index 8b9401e..bb14074 100644
54-
--- a/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/cholesky_triangular_solve.mlir
55-
+++ b/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/cholesky_triangular_solve.mlir
56-
@@ -147,12 +147,12 @@ func.func @cholesky_cholesky_dims_shardings_can_merge(%arg0: tensor<16x8x8x8xf32
57-
return %0 : tensor<16x8x8x8xf32>
58-
}
59-
60-
+// TODO(zixuanjiang). We may want to keep 'x' due to its larger size.
61-
// CHECK-LABEL: func @cholesky_sharded_cholesky_dim_input_only_batch_dim_both_but_input_sharding_larger
62-
func.func @cholesky_sharded_cholesky_dim_input_only_batch_dim_both_but_input_sharding_larger(%arg0: tensor<8x4x8x8xf32> {sdy.sharding = #sdy.sharding<@mesh_xyz, [{"x"}, {}, {}, {"z"}]>}) -> (tensor<8x4x8x8xf32> {sdy.sharding = #sdy.sharding<@mesh_xyz, [{"y"}, {}, {}, {}]>}){
63-
- // CHECK-NEXT: %[[RESHARD1:.*]] = sdy.reshard %arg0 <@mesh_xyz, [{"x"}, {}, {}, {}]> : tensor<8x4x8x8xf32>
64-
- // CHECK-NEXT: %[[CHOLESKY:.*]] = stablehlo.cholesky %[[RESHARD1]], lower = true {sdy.sharding = #sdy.sharding_per_value<[<@mesh_xyz, [{"x"}, {}, {}, {}]>]>} : tensor<8x4x8x8xf32>
65-
- // CHECK-NEXT: %[[RESHARD2:.*]] = sdy.reshard %[[CHOLESKY]] <@mesh_xyz, [{"y"}, {}, {}, {}]> : tensor<8x4x8x8xf32>
66-
- // CHECK-NEXT: return %[[RESHARD2]] : tensor<8x4x8x8xf32>
67-
+ // CHECK: %[[RESHARD1:.*]] = sdy.reshard %arg0 <@mesh_xyz, [{"y"}, {}, {}, {}]> : tensor<8x4x8x8xf32>
68-
+ // CHECK-NEXT: %[[CHOLESKY:.*]] = stablehlo.cholesky %[[RESHARD1]], lower = true {sdy.sharding = #sdy.sharding_per_value<[<@mesh_xyz, [{"y"}, {}, {}, {}]>]>} : tensor<8x4x8x8xf32>
69-
+ // CHECK-NEXT: return %[[CHOLESKY]] : tensor<8x4x8x8xf32>
70-
%0 = stablehlo.cholesky %arg0, lower = true {sdy.sharding = #sdy.sharding_per_value<[<@mesh_xyz, [{"y"}, {}, {}, {}]>]>} : (tensor<8x4x8x8xf32>) -> tensor<8x4x8x8xf32>
71-
return %0 : tensor<8x4x8x8xf32>
72-
}
73-
diff --git a/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/dynamic_slice_dynamic_update_slice.mlir b/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/dynamic_slice_dynamic_update_slice.mlir
74-
index 904d776..c12086a 100644
75-
--- a/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/dynamic_slice_dynamic_update_slice.mlir
76-
+++ b/shardy/dialect/sdy/transforms/export/test/insert_explicit_reshards/dynamic_slice_dynamic_update_slice.mlir
77-
@@ -42,10 +42,10 @@ func.func @dynamic_update_slice(%arg0: tensor<32x4x8xf32> {sdy.sharding = #sdy.s
78-
79-
// CHECK-LABEL: func @dynamic_update_slice_different_input_and_output_sharding
80-
func.func @dynamic_update_slice_different_input_and_output_sharding(%arg0: tensor<32x4x8xf32> {sdy.sharding = #sdy.sharding<@mesh, [{}, {"x"}, {"y"}]>}, %arg1: tensor<32x1x2xf32> {sdy.sharding = #sdy.sharding<@mesh, [{}, {}, {"y"}]>}, %arg2: tensor<i32>, %arg3: tensor<i32>, %arg4: tensor<i32>) -> (tensor<32x4x8xf32> {sdy.sharding = #sdy.sharding<@mesh, [{}, {"y"}, {"x"}]>}){
81-
- // CHECK-NEXT: %[[RESHARD0:.*]] = sdy.reshard %arg0 <@mesh, [{}, {"y"}, {"x"}]> : tensor<32x4x8xf32>
82-
- // CHECK-NEXT: %[[RESHARD1:.*]] = sdy.reshard %arg1 <@mesh, [{}, {}, {}]> : tensor<32x1x2xf32>
83-
- // CHECK-NEXT: %[[DYNAMIC_UPDATE_SLICE:.*]] = stablehlo.dynamic_update_slice %[[RESHARD0]], %[[RESHARD1]], %arg2, %arg3, %arg4 {sdy.sharding = #sdy.sharding_per_value<[<@mesh, [{}, {"y"}, {"x"}]>]>} : (tensor<32x4x8xf32>, tensor<32x1x2xf32>, tensor<i32>, tensor<i32>, tensor<i32>) -> tensor<32x4x8xf32>
84-
- // CHECK-NEXT: return %[[DYNAMIC_UPDATE_SLICE]] : tensor<32x4x8xf32>
85-
+ // CHECK-NEXT: %[[REPLICATED_UPDATE:.*]] = sdy.reshard %arg1 <@mesh, [{}, {}, {}]> : tensor<32x1x2xf32>
86-
+ // CHECK-NEXT: %[[DYNAMIC_UPDATE_SLICE:.*]] = stablehlo.dynamic_update_slice %arg0, %[[REPLICATED_UPDATE]], %arg2, %arg3, %arg4 {sdy.sharding = #sdy.sharding_per_value<[<@mesh, [{}, {"x"}, {"y"}]>]>} : (tensor<32x4x8xf32>, tensor<32x1x2xf32>, tensor<i32>, tensor<i32>, tensor<i32>) -> tensor<32x4x8xf32>
87-
+ // CHECK-NEXT: %[[RESHARD:.*]] = sdy.reshard %[[DYNAMIC_UPDATE_SLICE]] <@mesh, [{}, {"y"}, {"x"}]> : tensor<32x4x8xf32>
88-
+ // CHECK-NEXT: return %[[RESHARD]] : tensor<32x4x8xf32>
89-
%0 = stablehlo.dynamic_update_slice %arg0, %arg1, %arg2, %arg3, %arg4 {sdy.sharding = #sdy.sharding_per_value<[<@mesh, [{}, {"y"}, {"x"}]>]>} : (tensor<32x4x8xf32>, tensor<32x1x2xf32>, tensor<i32>, tensor<i32>, tensor<i32>) -> tensor<32x4x8xf32>
90-
return %0 : tensor<32x4x8xf32>
91-
}
921
diff --git a/third_party/llvm/generated.patch b/third_party/llvm/generated.patch
93-
index 8327093..e25178e 100644
2+
index e25178e..37a7256 100644
943
--- a/third_party/llvm/generated.patch
954
+++ b/third_party/llvm/generated.patch
96-
@@ -1,20 +1,17 @@
5+
@@ -1,17 +1,29 @@
976
Auto generated patch. Do not edit or delete it, even if empty.
98-
-diff -ruN --strip-trailing-cr a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
99-
---- a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
100-
-+++ b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
101-
-@@ -1563,7 +1563,6 @@
102-
- ":basic",
103-
- ":config",
104-
- ":driver_options_inc_gen",
105-
-- ":frontend",
106-
- ":lex",
107-
- ":options",
108-
- ":parse",
109-
-@@ -1719,6 +1718,7 @@
110-
- ":ast",
111-
- ":basic",
112-
- ":config",
113-
-+ ":driver",
114-
- ":driver_options_inc_gen",
115-
- ":edit",
116-
- ":lex",
117-
+diff -ruN --strip-trailing-cr a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
118-
+--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
119-
++++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
120-
+@@ -4095,6 +4095,12 @@
121-
+ llvm::SmallVector<size_t> occludedChildren;
122-
+ llvm::sort(
123-
+ indices.begin(), indices.end(), [&](const size_t a, const size_t b) {
124-
++ // Bail early if we are asked to look at the same index. If we do not
125-
++ // bail early, we can end up mistakenly adding indices to
126-
++ // occludedChildren. This can occur with some types of libc++ hardening.
127-
++ if (a == b)
128-
++ return false;
7+
-diff -ruN --strip-trailing-cr a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
8+
---- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
9+
-+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
10+
-@@ -4095,6 +4095,12 @@
11+
- llvm::SmallVector<size_t> occludedChildren;
12+
- llvm::sort(
13+
- indices.begin(), indices.end(), [&](const size_t a, const size_t b) {
14+
-+ // Bail early if we are asked to look at the same index. If we do not
15+
-+ // bail early, we can end up mistakenly adding indices to
16+
-+ // occludedChildren. This can occur with some types of libc++ hardening.
17+
-+ if (a == b)
18+
-+ return false;
19+
-+
20+
- auto memberIndicesA = cast<ArrayAttr>(indexAttr[a]);
21+
- auto memberIndicesB = cast<ArrayAttr>(indexAttr[b]);
22+
+diff -ruN --strip-trailing-cr a/clang/lib/Analysis/ThreadSafety.cpp b/clang/lib/Analysis/ThreadSafety.cpp
23+
+--- a/clang/lib/Analysis/ThreadSafety.cpp
24+
++++ b/clang/lib/Analysis/ThreadSafety.cpp
25+
+@@ -2820,7 +2820,7 @@
26+
+ case CFGElement::AutomaticObjectDtor: {
27+
+ CFGAutomaticObjDtor AD = BI.castAs<CFGAutomaticObjDtor>();
28+
+ const auto *DD = AD.getDestructorDecl(AC.getASTContext());
29+
+- if (!DD->hasAttrs())
30+
++ if (!DD || !DD->hasAttrs())
31+
+ break;
32+
33+
+ LocksetBuilder.handleCall(
34+
+diff -ruN --strip-trailing-cr a/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp b/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp
35+
+--- a/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp
36+
++++ b/clang/test/SemaCXX/no-warn-thread-safety-analysis.cpp
37+
+@@ -0,0 +1,12 @@
38+
++// RUN: %clang_cc1 -fsyntax-only -verify -std=c++11 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=0 %s
39+
++// RUN: %clang_cc1 -fsyntax-only -verify -std=c++11 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=1 %s
40+
++// RUN: %clang_cc1 -fsyntax-only -verify -std=c++17 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=0 %s
41+
++// RUN: %clang_cc1 -fsyntax-only -verify -std=c++17 -Wthread-safety -Wthread-safety-pointer -Wthread-safety-beta -Wno-thread-safety-negative -fcxx-exceptions -DUSE_CAPABILITY=1 %s
42+
++// expected-no-diagnostics
12943
++
130-
+ auto memberIndicesA = cast<ArrayAttr>(indexAttr[a]);
131-
+ auto memberIndicesB = cast<ArrayAttr>(indexAttr[b]);
132-
+
44+
++struct foo {
45+
++ ~foo();
46+
++};
47+
++struct bar : foo {};
48+
++struct baz : bar {};
49+
++baz foobar(baz a) { return a; }
13350
diff --git a/third_party/llvm/workspace.bzl b/third_party/llvm/workspace.bzl
134-
index 8e6cbc9..215fe72 100644
51+
index 215fe72..3b35979 100644
13552
--- a/third_party/llvm/workspace.bzl
13653
+++ b/third_party/llvm/workspace.bzl
13754
@@ -4,8 +4,8 @@ load("//third_party:repo.bzl", "tf_http_archive")
13855

13956
def repo(name):
14057
"""Imports LLVM."""
141-
- LLVM_COMMIT = "dea330b38d9c18b68219abdb52baaa72c9f1103d"
142-
- LLVM_SHA256 = "0f00dd4e0d61e49051b09169450af0c5ca364bf7e3f015794089455ae8c8555c"
143-
+ LLVM_COMMIT = "26362c68579dd4375198aae4651b4d5f8a36c715"
144-
+ LLVM_SHA256 = "1b81809d98940d0a6d4f19ef9e0bf72cd5847b9bbed47bc3517fcf8a40d38fd9"
58+
- LLVM_COMMIT = "26362c68579dd4375198aae4651b4d5f8a36c715"
59+
- LLVM_SHA256 = "1b81809d98940d0a6d4f19ef9e0bf72cd5847b9bbed47bc3517fcf8a40d38fd9"
60+
+ LLVM_COMMIT = "4f39a4ff0ada92870ca1c2dccad382ea04947da8"
61+
+ LLVM_SHA256 = "264c7cc3e166b840911494a2e94cff2ae8730b56239bd91dc8f65c8dc9468262"
14562

14663
tf_http_archive(
14764
name = name,

‎third_party/shardy/workspace.bzl‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -3,8 +3,8 @@
33
load("//third_party:repo.bzl", "tf_http_archive", "tf_mirror_urls")
44

55
def repo():
6-
SHARDY_COMMIT = "3f7332e771ff5eba89bb0e1113aceeef8116a181"
7-
SHARDY_SHA256 = "7535e2823590036ee898324a0d360e2bbd8dd741cb383f84b089c56d66b0b562"
6+
SHARDY_COMMIT = "1f65a631b69e6053f18bdb2d01b5cde0cb2e5f89"
7+
SHARDY_SHA256 = "19c5a4902d6c0317dd0362e45ef95ffd877ed0c4d98a7ed7a86fac517d594725"
88

99
tf_http_archive(
1010
name = "shardy",

0 commit comments

Comments
 (0)