[OpenXLA PIN]update openxla-pin to nov24

WORKSPACE

@@ -42,13 +42,12 @@ http_archive(
         "//openxla_patches:constexpr_return.diff",
         "//openxla_patches:gpu_race_condition.diff",
         "//openxla_patches:f16_abi_clang.diff",
-        "//openxla_patches:gpu_topk_rewriter.diff",
         "//openxla_patches:quant_dequant_converter.diff",
         "//openxla_patches:stablehlo_quant_seralization.diff",
     ],
-    strip_prefix = "xla-4f8381651977dff16b1d86bb4b198eb733c5f478",
+    strip_prefix = "xla-8744c9a94782cd7804f015e6d29df253437af3cb",
     urls = [
-        "https://github.com/openxla/xla/archive/4f8381651977dff16b1d86bb4b198eb733c5f478.tar.gz",
+        "https://github.com/openxla/xla/archive/8744c9a94782cd7804f015e6d29df253437af3cb.tar.gz",
     ],
 )
 

openxla_patches/cache_urls.diff

@@ -1,5 +1,5 @@
 diff --git a/xla/mlir_hlo/WORKSPACE b/xla/mlir_hlo/WORKSPACE
-index cc9eeb64f..b290eb455 100644
+index c3115e33d..d315ad745 100644
 --- a/xla/mlir_hlo/WORKSPACE
 +++ b/xla/mlir_hlo/WORKSPACE
 @@ -35,7 +35,10 @@ http_archive(
@@ -13,10 +13,9 @@ index cc9eeb64f..b290eb455 100644
 +    ],
  )
 
- load("@llvm-raw//utils/bazel:configure.bzl", "llvm_configure", "llvm_disable_optional_support_deps")
- load("@llvm-raw//utils/bazel:configure.bzl", "llvm_configure", "llvm_disable_optional_support_deps")
+ load("@llvm-raw//utils/bazel:configure.bzl", "llvm_configure")
 diff --git a/third_party/llvm/workspace.bzl b/third_party/llvm/workspace.bzl
-index a4574d75d..f9ce37094 100644
+index d7f3a8093..a7af9c68a 100644
 --- a/third_party/llvm/workspace.bzl
 +++ b/third_party/llvm/workspace.bzl
 @@ -13,7 +13,9 @@ def repo(name):
@@ -28,4 +27,4 @@ index a4574d75d..f9ce37094 100644
 +            "https://storage.googleapis.com/tpu-pytorch/llvm-raw/{commit}.tar.gz".format(commit = LLVM_COMMIT),
          ],
          build_file = "//third_party/llvm:llvm.BUILD",
-         patch_file = [
+         patch_file = [

openxla_patches/constexpr_return.diff

@@ -1,12 +1,12 @@
 diff --git a/xla/primitive_util.h b/xla/primitive_util.h
-index 696147844..dfea15a4d 100644
+index 63fa4e193..ab352626c 100644
 --- a/xla/primitive_util.h
 +++ b/xla/primitive_util.h
-@@ -748,6 +748,7 @@ inline bool FitsInIntegralType(int64_t x, PrimitiveType ty) {
+@@ -706,6 +706,7 @@ inline bool FitsInIntegralType(int64_t x, PrimitiveType ty) {
                   std::numeric_limits<NativeT>::max() >= x;
          }
          LOG(FATAL) << "Invalid primitive type " << PrimitiveType_Name(ty);
-+	return false;
++       return false;
        },
        ty);
  }

openxla_patches/gpu_race_condition.diff

@@ -1,14 +1,14 @@
 diff --git a/xla/service/gpu/gpu_executable.cc b/xla/service/gpu/gpu_executable.cc
-index 242961dd1..787275868 100644
+index 1f9903cb3..763b7fc23 100644
 --- a/xla/service/gpu/gpu_executable.cc
 +++ b/xla/service/gpu/gpu_executable.cc
-@@ -563,8 +563,7 @@ StatusOr<ExecutionOutput> GpuExecutable::ExecuteAsyncOnStreamImpl(
+@@ -589,8 +589,7 @@ StatusOr<ExecutionOutput> GpuExecutable::ExecuteAsyncOnStreamImpl(
    }
 
    // Force synchronous execution if the allocator requires it.
 -  const bool block_host_until_done =
 -      !memory_allocator->AllowsAsynchronousDeallocation();
 +  const bool block_host_until_done = true;
 
-
-   // Lock the GPU with a shared lock so that we don't interfere with autotuning
+   // Lock the GPU with a shared lock so that we don't interfere with autotuning
+   // that may be running during JIT compilation while allowing multiple XLA

openxla_patches/quant_dequant_converter.diff

@@ -2,7 +2,7 @@
 // stablehlo.uniform_quantize/dequantize to be converted to stablehlo.uniform_quantize/dequantize.
 // The patch can be removed after quantize/dequantize, quantized dtype support is added to HLO.
 diff --git a/xla/translate/hlo_to_mhlo/BUILD b/xla/translate/hlo_to_mhlo/BUILD
-index f74973ae1..8e3f0e06b 100644
+index 0f0c5e842..59a30c585 100644
 --- a/xla/translate/hlo_to_mhlo/BUILD
 +++ b/xla/translate/hlo_to_mhlo/BUILD
 @@ -67,6 +67,7 @@ cc_library(
@@ -14,7 +14,7 @@ index f74973ae1..8e3f0e06b 100644
          "@llvm-project//mlir:SparseTensorDialect",
          "@tsl//tsl/platform:statusor",
 diff --git a/xla/translate/hlo_to_mhlo/hlo_function_importer.cc b/xla/translate/hlo_to_mhlo/hlo_function_importer.cc
-index 08d5f49c8..2f9ad1e0b 100644
+index cc7aa9e9e..c24f24b50 100644
 --- a/xla/translate/hlo_to_mhlo/hlo_function_importer.cc
 +++ b/xla/translate/hlo_to_mhlo/hlo_function_importer.cc
 @@ -664,6 +664,70 @@ StatusOr<mlir::Operation*> HloFunctionImporter::ImportInstruction(
@@ -115,7 +115,7 @@ index 08d5f49c8..2f9ad1e0b 100644
        } else {
          attributes.push_back(builder_->getNamedAttr(
 diff --git a/xla/translate/hlo_to_mhlo/hlo_module_importer.cc b/xla/translate/hlo_to_mhlo/hlo_module_importer.cc
-index 9f05992c8..03cf4840d 100644
+index 1494efd9e..dcb3d9e89 100644
 --- a/xla/translate/hlo_to_mhlo/hlo_module_importer.cc
 +++ b/xla/translate/hlo_to_mhlo/hlo_module_importer.cc
 @@ -19,6 +19,8 @@ limitations under the License.

openxla_patches/stablehlo_quant_seralization.diff

@@ -33,7 +33,7 @@ index 000000000..24e23b67d
 + }  // namespace
 + 
 diff --git a/third_party/stablehlo/workspace.bzl b/third_party/stablehlo/workspace.bzl
-index 9f4494aac..64fa072bb 100644
+index 80ab0e479..caaa11080 100644
 --- a/third_party/stablehlo/workspace.bzl
 +++ b/third_party/stablehlo/workspace.bzl
 @@ -15,5 +15,6 @@ def repo():

setup.py

@@ -72,7 +72,7 @@
 
 base_dir = os.path.dirname(os.path.abspath(__file__))
 
-_libtpu_version = '0.1.dev20231022'
+_libtpu_version = '0.1.dev20231125'
 _libtpu_storage_path = f'https://storage.googleapis.com/cloud-tpu-tpuvm-artifacts/wheels/libtpu-nightly/libtpu_nightly-{_libtpu_version}-py3-none-any.whl'