Merge branch 'branch-24.08' into fix-block-size

rapidsai · Jun 26, 2024 · c74e053 · c74e053
2 parents 8e4a1d7 + b863f18
commit c74e053
Show file tree

Hide file tree

Showing 46 changed files with 1,005 additions and 107 deletions.
diff --git a/.devcontainer/cuda11.8-conda/devcontainer.json b/.devcontainer/cuda11.8-conda/devcontainer.json
@@ -11,7 +11,7 @@
   "runArgs": [
     "--rm",
     "--name",
-    "${localEnv:USER}-rapids-${localWorkspaceFolderBasename}-24.08-cuda11.8-conda"
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda11.8-conda"
   ],
   "hostRequirements": {"gpu": "optional"},
   "features": {

diff --git a/.devcontainer/cuda11.8-pip/devcontainer.json b/.devcontainer/cuda11.8-pip/devcontainer.json
@@ -11,7 +11,7 @@
   "runArgs": [
     "--rm",
     "--name",
-    "${localEnv:USER}-rapids-${localWorkspaceFolderBasename}-24.08-cuda11.8-pip"
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda11.8-pip"
   ],
   "hostRequirements": {"gpu": "optional"},
   "features": {

diff --git a/.devcontainer/cuda12.2-conda/devcontainer.json b/.devcontainer/cuda12.2-conda/devcontainer.json
@@ -11,7 +11,7 @@
   "runArgs": [
     "--rm",
     "--name",
-    "${localEnv:USER}-rapids-${localWorkspaceFolderBasename}-24.08-cuda12.2-conda"
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda12.2-conda"
   ],
   "hostRequirements": {"gpu": "optional"},
   "features": {

diff --git a/.devcontainer/cuda12.2-pip/devcontainer.json b/.devcontainer/cuda12.2-pip/devcontainer.json
@@ -11,7 +11,7 @@
   "runArgs": [
     "--rm",
     "--name",
-    "${localEnv:USER}-rapids-${localWorkspaceFolderBasename}-24.08-cuda12.2-pip"
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda12.2-pip"
   ],
   "hostRequirements": {"gpu": "optional"},
   "features": {

diff --git a/.github/CODEOWNERS b/.github/CODEOWNERS
@@ -17,7 +17,7 @@ build.sh           @rapidsai/raft-cmake-codeowners
 /.pre-commit-config.yaml @rapidsai/ci-codeowners
 
 #packaging code owners
-/.devcontainers/   @rapidsai/packaging-codeowners
+/.devcontainer/    @rapidsai/packaging-codeowners
 /conda/            @rapidsai/packaging-codeowners
 /dependencies.yaml @rapidsai/packaging-codeowners
 /build.sh          @rapidsai/packaging-codeowners

diff --git a/README.md b/README.md
@@ -369,7 +369,7 @@ If citing the k-selection routines, please consider the following bibtex:
     isbn = {9798400701092},
     publisher = {Association for Computing Machinery},
     address = {New York, NY, USA},
-    location = {Denver, CO, USA}
+    location = {Denver, CO, USA},
     series = {SC '23}
 }
 ```

diff --git a/ci/build_cpp.sh b/ci/build_cpp.sh
@@ -13,10 +13,8 @@ export CMAKE_GENERATOR=Ninja
 
 rapids-print-env
 
-version=$(rapids-generate-version)
-
 rapids-logger "Begin cpp build"
 
-RAPIDS_PACKAGE_VERSION=${version} rapids-conda-retry mambabuild conda/recipes/libraft
+RAPIDS_PACKAGE_VERSION=$(rapids-generate-version) rapids-conda-retry mambabuild conda/recipes/libraft
 
 rapids-upload-conda-to-s3 cpp
diff --git a/ci/build_docs.sh b/ci/build_docs.sh
@@ -8,7 +8,7 @@ rapids-logger "Create test conda environment"
 
 rapids-dependency-file-generator \
   --output conda \
-  --file_key docs \
+  --file-key docs \
   --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" | tee env.yaml
 
 rapids-mamba-retry env create --yes -f env.yaml -n docs

diff --git a/ci/check_style.sh b/ci/check_style.sh
@@ -8,7 +8,7 @@ rapids-logger "Create checks conda environment"
 
 rapids-dependency-file-generator \
   --output conda \
-  --file_key checks \
+  --file-key checks \
   --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" | tee env.yaml
 
 rapids-mamba-retry env create --yes -f env.yaml -n checks

diff --git a/ci/test_cpp.sh b/ci/test_cpp.sh
@@ -11,7 +11,7 @@ cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../
 rapids-logger "Generate C++ testing dependencies"
 rapids-dependency-file-generator \
   --output conda \
-  --file_key test_cpp \
+  --file-key test_cpp \
   --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch)" | tee env.yaml
 
 rapids-mamba-retry env create --yes -f env.yaml -n test

diff --git a/ci/test_python.sh b/ci/test_python.sh
@@ -11,7 +11,7 @@ cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../
 rapids-logger "Generate Python testing dependencies"
 rapids-dependency-file-generator \
   --output conda \
-  --file_key test_python \
+  --file-key test_python \
   --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" | tee env.yaml
 
 rapids-mamba-retry env create --yes -f env.yaml -n test

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
@@ -302,6 +302,8 @@ if(RAFT_COMPILE_LIBRARY)
     src/distance/detail/pairwise_matrix/dispatch_correlation_float_float_float_int.cu
     src/distance/detail/pairwise_matrix/dispatch_cosine_double_double_double_int.cu
     src/distance/detail/pairwise_matrix/dispatch_cosine_float_float_float_int.cu
+    src/distance/detail/pairwise_matrix/dispatch_dice_double_double_double_int.cu
+    src/distance/detail/pairwise_matrix/dispatch_dice_float_float_float_int.cu
     src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_double_double_double_int.cu
     src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_float_float_float_int.cu
     src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_double_double_double_int.cu

diff --git a/cpp/bench/ann/CMakeLists.txt b/cpp/bench/ann/CMakeLists.txt
@@ -58,10 +58,6 @@ if(BUILD_CPU_ONLY)
   set(RAFT_ANN_BENCH_USE_RAFT_BRUTE_FORCE OFF)
   set(RAFT_ANN_BENCH_USE_RAFT_CAGRA_HNSWLIB OFF)
   set(RAFT_ANN_BENCH_USE_GGNN OFF)
-elseif(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0.0)
-  # Disable faiss benchmarks on CUDA 12 since faiss is not yet CUDA 12-enabled.
-  # https://github.com/rapidsai/raft/issues/1627
-  set(RAFT_FAISS_ENABLE_GPU OFF)
 endif()
 
 set(RAFT_ANN_BENCH_USE_RAFT OFF)

diff --git a/cpp/bench/ann/src/common/benchmark.hpp b/cpp/bench/ann/src/common/benchmark.hpp
@@ -459,8 +459,14 @@ void register_search(std::shared_ptr<const Dataset<T>> dataset,
                    */
                   ->MeasureProcessCPUTime()
                   ->UseRealTime();
-
-      if (metric_objective == Objective::THROUGHPUT) { b->ThreadRange(threads[0], threads[1]); }
+      if (metric_objective == Objective::THROUGHPUT) {
+        if (index.algo.find("faiss_gpu") != std::string::npos) {
+          log_warn(
+            "FAISS GPU does not work in throughput mode because the underlying "
+            "StandardGpuResources object is not thread-safe. This will cause unexpected results");
+        }
+        b->ThreadRange(threads[0], threads[1]);
+      }
     }
   }
 }

diff --git a/cpp/bench/ann/src/faiss/faiss_cpu_benchmark.cpp b/cpp/bench/ann/src/faiss/faiss_cpu_benchmark.cpp
@@ -51,10 +51,10 @@ void parse_build_param(const nlohmann::json& conf,
 {
   parse_base_build_param<T>(conf, param);
   param.M = conf.at("M");
-  if (conf.contains("usePrecomputed")) {
-    param.usePrecomputed = conf.at("usePrecomputed");
+  if (conf.contains("use_precomputed_table")) {
+    param.use_precomputed_table = conf.at("use_precomputed_table");
   } else {
-    param.usePrecomputed = false;
+    param.use_precomputed_table = false;
   }
   if (conf.contains("bitsPerCode")) {
     param.bitsPerCode = conf.at("bitsPerCode");

diff --git a/cpp/bench/ann/src/faiss/faiss_cpu_wrapper.h b/cpp/bench/ann/src/faiss/faiss_cpu_wrapper.h
@@ -229,7 +229,7 @@ class FaissCpuIVFPQ : public FaissCpu<T> {
   struct BuildParam : public FaissCpu<T>::BuildParam {
     int M;
     int bitsPerCode;
-    bool usePrecomputed;
+    bool use_precomputed_table;
   };
 
   FaissCpuIVFPQ(Metric metric, int dim, const BuildParam& param) : FaissCpu<T>(metric, dim, param)

diff --git a/cpp/bench/ann/src/faiss/faiss_gpu_benchmark.cu b/cpp/bench/ann/src/faiss/faiss_gpu_benchmark.cu
@@ -45,6 +45,11 @@ void parse_build_param(const nlohmann::json& conf,
                        typename raft::bench::ann::FaissGpuIVFFlat<T>::BuildParam& param)
 {
   parse_base_build_param<T>(conf, param);
+  if (conf.contains("use_raft")) {
+    param.use_raft = conf.at("use_raft");
+  } else {
+    param.use_raft = false;
+  }
 }
 
 template <typename T>
@@ -63,6 +68,16 @@ void parse_build_param(const nlohmann::json& conf,
   } else {
     param.useFloat16 = false;
   }
+  if (conf.contains("use_raft")) {
+    param.use_raft = conf.at("use_raft");
+  } else {
+    param.use_raft = false;
+  }
+  if (conf.contains("bitsPerCode")) {
+    param.bitsPerCode = conf.at("bitsPerCode");
+  } else {
+    param.bitsPerCode = 8;
+  }
 }
 
 template <typename T>
@@ -160,5 +175,18 @@ REGISTER_ALGO_INSTANCE(std::uint8_t);
 
 #ifdef ANN_BENCH_BUILD_MAIN
 #include "../common/benchmark.hpp"
-int main(int argc, char** argv) { return raft::bench::ann::run_main(argc, argv); }
+int main(int argc, char** argv)
+{
+  rmm::mr::cuda_memory_resource cuda_mr;
+  // Construct a resource that uses a coalescing best-fit pool allocator
+  // and is initially sized to half of free device memory.
+  rmm::mr::pool_memory_resource<rmm::mr::cuda_memory_resource> pool_mr{
+    &cuda_mr, rmm::percent_of_free_device_memory(50)};
+  // Updates the current device resource pointer to `pool_mr`
+  auto old_mr = rmm::mr::set_current_device_resource(&pool_mr);
+  auto ret    = raft::bench::ann::run_main(argc, argv);
+  // Restores the current device resource pointer to its previous value
+  rmm::mr::set_current_device_resource(old_mr);
+  return ret;
+}
 #endif