Tune radix sort keys for U8/U16 and 64-bit offsets

gevtushenko · May 13, 2023 · 2afa84d · 2afa84d
1 parent aea1bff
commit 2afa84d
Showing 1 changed file with 2 additions and 0 deletions.
diff --git a/cub/device/dispatch/dispatch_radix_sort.cuh b/cub/device/dispatch/dispatch_radix_sort.cuh
@@ -629,7 +629,9 @@ struct sm90_small_key_tuning
 
 // keys
 template <> struct sm90_small_key_tuning<1,  0, 4> { static constexpr int threads = 512; static constexpr int items = 19; };
+template <> struct sm90_small_key_tuning<1,  0, 8> { static constexpr int threads = 512; static constexpr int items = 19; };
 template <> struct sm90_small_key_tuning<2,  0, 4> { static constexpr int threads = 512; static constexpr int items = 19; };
+template <> struct sm90_small_key_tuning<2,  0, 8> { static constexpr int threads = 512; static constexpr int items = 19; };
 
 // pairs  8:xx
 template <> struct sm90_small_key_tuning<1,  1, 4> { static constexpr int threads = 512; static constexpr int items = 15; };