Use fast compresssion for arange bench

Author: FrancescAlted

bench/ndarray/linear-constructor.py

@@ -12,32 +12,38 @@
 
 import blosc2
 
-dtype = np.float64
+dtype = np.int64
 shape = (10_000, 10_000)
+start, stop = 1, 2
+cparams = blosc2.CParams(codec=blosc2.Codec.BLOSCLZ, clevel=1)
 
 @blosc2.dsl_kernel
-def kernel_ramp(start):
-    return start + _flat_idx  # noqa: F821  # DSL index/shape symbols resolved by miniexpr
+def kernel_ramp(start, stop, nitems):  # noqa
+    step = (float(stop) - float(start)) / float(nitems - 1)
+    return float(start) + _flat_idx * step  # noqa: F821  # DSL index/shape symbols resolved by miniexpr
 
 t0 = time()
-npa = np.arange(np.prod(shape), dtype=dtype).reshape(shape)
+npa = np.linspace(start, stop, np.prod(shape), dtype=dtype).reshape(shape)
 print("NumPy arange:", round(time() - t0, 3), "s")
+#print(npa)
 
 t0 = time()
-a1 = blosc2.arange(0, np.prod(shape), dtype=dtype, shape=shape)
+a1 = blosc2.linspace(start, stop, np.prod(shape), dtype=dtype, shape=shape, cparams=cparams)
 print("Blosc2 arange:", round(time() - t0, 3), "s")
 
+np.testing.assert_array_equal(a1, npa)
+
 t0 = time()
-a2 = blosc2.lazyudf(kernel_ramp, (0, ), dtype=dtype, shape=shape)
+a2 = blosc2.lazyudf(kernel_ramp, (start, stop, np.prod(shape)), dtype=dtype, shape=shape)
 # a2 = blosc2.lazyudf(kernel_ramp, (0, ), dtype=dtype, shape=shape, jit_backend="cc")
-a3 = a2.compute(cparams=dict(clevel=1, codec=blosc2.Codec.LZ4))
+a3 = a2.compute(cparams=cparams)
 print("Blosc2 with DSL kernel (tcc jit backend):", round(time() - t0, 3), "s")
 
-np.testing.assert_array_equal(npa, a3)
+np.testing.assert_array_equal(a3, npa)
 
 t0 = time()
-a2 = blosc2.lazyudf(kernel_ramp, (0, ), dtype=dtype, shape=shape, jit_backend="cc")
-a3 = a2.compute(cparams=dict(clevel=1, codec=blosc2.Codec.LZ4))
+a2 = blosc2.lazyudf(kernel_ramp, (start, stop, np.prod(shape)), dtype=dtype, shape=shape, jit_backend="cc")
+a3 = a2.compute(cparams=cparams)
 print("Blosc2 with DSL kernel (cc jit backend):", round(time() - t0, 3), "s")
 
-np.testing.assert_array_equal(npa, a3)
+np.testing.assert_array_equal(a3, npa)