Remove DIRTY NIF flag

https://www.erlang.org/doc/man/erl_nif.html documentation indicates the threshold for being a dirty NIF should be about 1ms. Generally compress/decompress is much quicker than this. Testing indicates a significant overhead from using this flag when compress/decompress is in fact very fast.
OpenRiak · Jan 12, 2024 · 3a02e16 · 3a02e16
1 parent 0ca9cbe
commit 3a02e16
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/c_src/zstd_nif.c b/c_src/zstd_nif.c
@@ -405,8 +405,8 @@ static int zstd_on_upgrade(ErlNifEnv *env, void **priv, void **old, ERL_NIF_TERM
 }
 
 static ErlNifFunc nif_funcs[] = {
-  { "compress"                    , 2, zstd_nif_compress                   , ERL_DIRTY_JOB_CPU_BOUND },
-  { "decompress"                  , 1, zstd_nif_decompress                 , ERL_DIRTY_JOB_CPU_BOUND },
+  { "compress"                    , 2, zstd_nif_compress                                             },
+  { "decompress"                  , 1, zstd_nif_decompress                                           },
 
   { "new_compression_stream"      , 0, zstd_nif_new_compression_stream                               },
   { "new_decompression_stream"    , 0, zstd_nif_new_decompression_stream                             },