Define and use xss_get_num_threads()

Author: Raghuveer Devulapalli

src/avx512-16bit-qsort.hpp

@@ -552,9 +552,7 @@ avx512_qsort_fp16_helper(uint16_t *arr, arrsize_t arrsize)
     bool use_parallel = arrsize > 100000;
 
     if (use_parallel) {
-        // This thread limit was determined experimentally; it may be better for it to be the number of physical cores on the system
-        constexpr int thread_limit = 8;
-        int thread_count = std::min(thread_limit, omp_get_max_threads());
+        int thread_count = xss_get_num_threads();
         arrsize_t task_threshold = std::max((arrsize_t)100000, arrsize / 100);
 
         // We use omp parallel and then omp single to setup the threads that will run the omp task calls in qsort_

src/xss-common-argsort.h

@@ -629,9 +629,7 @@ X86_SIMD_SORT_INLINE void xss_argsort(T *arr,
         bool use_parallel = arrsize > 10000;
 
         if (use_parallel) {
-            // This thread limit was determined experimentally; it may be better for it to be the number of physical cores on the system
-            constexpr int thread_limit = 8;
-            int thread_count = std::min(thread_limit, omp_get_max_threads());
+            int thread_count = xss_get_num_threads();
             arrsize_t task_threshold
                     = std::max((arrsize_t)10000, arrsize / 100);
 

src/xss-common-includes.h

@@ -85,6 +85,14 @@
 #if defined(XSS_USE_OPENMP) && defined(_OPENMP)
 #define XSS_COMPILE_OPENMP
 #include <omp.h>
+
+X86_SIMD_SORT_INLINE int xss_get_num_threads()
+{
+    // Get the number of physical cores: works only when hyperthreading is
+    // enabled on all cores
+    int num_physical_cores = omp_get_num_procs() / 2;
+    return std::min(num_physical_cores, (int)omp_get_max_threads());
+}
 #endif
 
 template <class... T>

src/xss-common-keyvaluesort.hpp

@@ -610,9 +610,7 @@ X86_SIMD_SORT_INLINE void xss_qsort_kv(
         bool use_parallel = arrsize > 10000;
 
         if (use_parallel) {
-            // This thread limit was determined experimentally; it may be better for it to be the number of physical cores on the system
-            constexpr int thread_limit = 8;
-            int thread_count = std::min(thread_limit, omp_get_max_threads());
+            int thread_count = xss_get_num_threads();
             arrsize_t task_threshold
                     = std::max((arrsize_t)10000, arrsize / 100);
 

src/xss-common-qsort.h

@@ -672,9 +672,7 @@ X86_SIMD_SORT_INLINE void xss_qsort(T *arr, arrsize_t arrsize, bool hasnan)
         bool use_parallel = arrsize > 100000;
 
         if (use_parallel) {
-            // This thread limit was determined experimentally; it may be better for it to be the number of physical cores on the system
-            constexpr int thread_limit = 8;
-            int thread_count = std::min(thread_limit, omp_get_max_threads());
+            int thread_count = xss_get_num_threads();
             arrsize_t task_threshold
                     = std::max((arrsize_t)100000, arrsize / 100);