Few fixes microbench_low_level

Author: paugier

bench/microbench_low_level/README.md

@@ -18,4 +18,8 @@ We measure the performance for functions containing low level Python code.
 - `instantiate` (command `make bench_instantiate`): dominated by the
   instantiation/deletion of small arrays of 4 floats.
 
+- `element_wise` (command `make bench_element_wise`): dominated by the
+  instantiation/deletion of small arrays of 4 floats and calling element-wise
+  operations.
+
 The files result_*.txt contain few results.

bench/microbench_low_level/bench.py

@@ -152,7 +152,7 @@ def element_wise(arr):
 compute_from_arr = locals()[name_bench]
 
 if size is None:
-    if method.startswith("sum_loop") or method == "cort":
+    if name_bench.startswith("sum_loop") or name_bench == "cort":
         size = 10000
     else:
         size = 4

bench/microbench_low_level/julia/bench_element_wise.jl

@@ -27,4 +27,4 @@ for irun in 1:nb_runs
     times[irun] = @elapsed compute_from_arr(arr)
 end
 
-println(median(times))
+println(median(times))

bench/microbench_low_level/result_sum_loop.md

@@ -56,7 +56,7 @@ piconumpy.cpython_capi        : 2.90e-03 s (219.1 * Julia)
 
 ## Summary
 
-- PyPy is fast with list (1.3 * Julia, same order of magnitude that with Julia)
+- PyPy is fast with list (1.3 * Julia, same order of magnitude than with Julia)
 and as fast for a piconumpy array based on a list ("piconumpy.purepy", zero
 cost abstraction!)