ENH, SIMD: improve argmax/argmin performance

  for all integers, f32 and f64 data types on all
  supported architectures via universal intrinsics.

1 files changed, 13 insertions, 1 deletions

diff --git a/benchmarks/benchmarks/bench_reduce.py b/benchmarks/benchmarks/bench_reduce.py
index 81316c492..ca07bd180 100644
--- a/benchmarks/benchmarks/bench_reduce.py
+++ b/benchmarks/benchmarks/bench_reduce.py
@@ -73,7 +73,8 @@ class FMinMax(Benchmark):
         np.fmax.reduce(self.d)
 
 class ArgMax(Benchmark):
-    params = [np.float32, np.float64, bool]
+    params = [np.int8, np.uint8, np.int16, np.uint16, np.int32, np.uint32,
+              np.int64, np.uint64, np.float32, np.float64, bool]
     param_names = ['dtype']
 
     def setup(self, dtype):
@@ -82,6 +83,17 @@ class ArgMax(Benchmark):
     def time_argmax(self, dtype):
         np.argmax(self.d)
 
+class ArgMin(Benchmark):
+    params = [np.int8, np.uint8, np.int16, np.uint16, np.int32, np.uint32,
+              np.int64, np.uint64, np.float32, np.float64, bool]
+    param_names = ['dtype']
+
+    def setup(self, dtype):
+        self.d = np.ones(200000, dtype=dtype)
+
+    def time_argmin(self, dtype):
+        np.argmin(self.d)
+
 class SmallReduction(Benchmark):
     def setup(self):
         self.d = np.ones(100, dtype=np.float32)