From 2449f88f2e19ef9561d82ffab697484c3a5bfa46 Mon Sep 17 00:00:00 2001 From: Alessandro Bellina Date: Thu, 13 Apr 2023 18:05:01 -0500 Subject: [PATCH 1/2] Change exception text ensuring we mention GPU OOM (#1074) Signed-off-by: Alessandro Bellina --- src/main/cpp/src/SparkResourceAdaptorJni.cpp | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/src/main/cpp/src/SparkResourceAdaptorJni.cpp b/src/main/cpp/src/SparkResourceAdaptorJni.cpp index 63aeb506630..173d3d0e1ae 100644 --- a/src/main/cpp/src/SparkResourceAdaptorJni.cpp +++ b/src/main/cpp/src/SparkResourceAdaptorJni.cpp @@ -642,7 +642,7 @@ class spark_resource_adaptor final : public rmm::mr::device_memory_resource { // In testing it looks like it is a few ms if in a tight loop, not including spill // overhead if (state.num_times_retried + 1 > 500) { - throw_java_exception(JAVA_OOM_CLASS, "retry limit exceeded"); + throw_java_exception(JAVA_OOM_CLASS, "GPU OutOfMemory: retry limit exceeded"); } state.num_times_retried++; } @@ -651,14 +651,14 @@ class spark_resource_adaptor final : public rmm::mr::device_memory_resource { const std::unique_lock &lock) { state.num_times_retry_throw++; check_before_oom(state, lock); - throw_java_exception(RETRY_OOM_CLASS, "task should retry operation"); + throw_java_exception(RETRY_OOM_CLASS, "GPU OutOfMemory"); } void throw_split_n_retry_oom(const char *msg, full_thread_state &state, const std::unique_lock &lock) { state.num_times_split_retry_throw++; check_before_oom(state, lock); - throw_java_exception(SPLIT_AND_RETRY_OOM_CLASS, "task should split input and retry operation"); + throw_java_exception(SPLIT_AND_RETRY_OOM_CLASS, "GPU OutOfMemory"); } bool is_blocked(thread_state state) { @@ -702,7 +702,7 @@ class spark_resource_adaptor final : public rmm::mr::device_memory_resource { break; case SHUFFLE_THROW: transition(thread->second, thread_state::SHUFFLE_RUNNING); - throw_java_exception(JAVA_OOM_CLASS, "could not allocate enough for shuffle"); + throw_java_exception(JAVA_OOM_CLASS, "GPU OutOfMemory: could not allocate enough for shuffle"); break; case TASK_BUFN_THROW: transition(thread->second, thread_state::TASK_BUFN_WAIT); From e2382ae118584ece85ca8e9f4813a2f9f4a8fe39 Mon Sep 17 00:00:00 2001 From: spark-rapids automation <70000568+nvauto@users.noreply.github.com> Date: Thu, 13 Apr 2023 23:05:15 +0000 Subject: [PATCH 2/2] Auto-merge use submodule in BASE ref Signed-off-by: spark-rapids automation <70000568+nvauto@users.noreply.github.com> --- thirdparty/cudf | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/thirdparty/cudf b/thirdparty/cudf index 4d31a6f62f3..5764ba5146f 160000 --- a/thirdparty/cudf +++ b/thirdparty/cudf @@ -1 +1 @@ -Subproject commit 4d31a6f62f3b05dfcbf9af87044c46ce46134913 +Subproject commit 5764ba5146fd33c31ce5e1a7588f555e738bd731