WIP: Add suggestion to discuss maros and aliases

examples/naive/aarch64/aarch64_simple0_macros.s

@@ -0,0 +1,50 @@
+qdata0   .req q8
+qdata1   .req q9
+qdata2   .req q10
+qdata3   .req q11
+
+qtwiddle .req q0
+
+data0    .req v8
+data1    .req v9
+data2    .req v10
+data3    .req v11
+
+twiddle  .req v0
+modulus  .req v1
+
+tmp      .req v12
+
+data_ptr      .req x0
+twiddle_ptr   .req x1
+
+.macro barmul out, in, twiddle, modulus
+    mul      \out.8h,   \in.8h, \twiddle.h[0]
+    sqrdmulh \in.8h,    \in.8h, \twiddle.h[1]
+    mls      \out.8h,   \in.8h, \modulus.h[0]
+.endm
+
+.macro butterfly data0, data1, tmp, twiddle, modulus
+    barmul \tmp, \data1, \twiddle, \modulus
+    sub    \data1.8h, \data0.8h, \tmp.8h
+    add    \data0.8h, \data0.8h, \tmp.8h
+.endm
+
+start:
+
+    ldr qtwiddle, [twiddle_ptr, #0]
+
+    ldr qdata0, [data_ptr, #0*16]
+    ldr qdata1, [data_ptr, #1*16]
+    ldr qdata2, [data_ptr, #2*16]
+    ldr qdata3, [data_ptr, #3*16]
+
+    butterfly data0, data1, tmp, twiddle, modulus
+    butterfly data2, data3, tmp, twiddle, modulus
+
+    str qdata0, [data_ptr], #4*16
+    str qdata1, [data_ptr, #-3*16]
+    str qdata2, [data_ptr, #-2*16]
+    str qdata3, [data_ptr, #-1*16]
+
+end:

tutorial/README.md

@@ -244,7 +244,10 @@ To make use of these issue slots one would have to mix in scalar instructions (o
 
 Also note the registers used: In the original code `v24` was as a temporary register in both computation streams preventing to effectively interleave them.
 SLOTHY renamed those registers to be able to interleave both computations. Other registers have also been arbitrarily renamed, but without any specific reason.
-Even with this small Neon example, you can see that understanding the input code is much easier than the output code which is the reason why we believe SLOTHY can help with writing auditable code.
+Even with this small Neon example, you can see that understanding the input code is much easier than the output code
+which is the reason why we believe SLOTHY can help with writing auditable code.
+
+**TODO: Add section on taking readability one step further by making extensive use of aliases and macros? See `aarch64_simple0_macros.s`**
 
 ## 3. Writing your own calling code
 

tutorial/tutorial-3-macros.py

@@ -0,0 +1,23 @@
+import logging
+import sys
+
+sys.path.append("../")
+from slothy import Slothy
+
+import slothy.targets.aarch64.aarch64_neon as AArch64_Neon
+import slothy.targets.aarch64.cortex_a55 as Target_CortexA55
+
+logging.basicConfig(stream=sys.stdout, level=logging.INFO)
+
+arch = AArch64_Neon
+target = Target_CortexA55
+
+slothy = Slothy(arch, target)
+
+# example
+slothy.load_source_from_file("../examples/naive/aarch64/aarch64_simple0_macros.s")
+slothy.config.variable_size=True
+slothy.config.constraints.stalls_first_attempt=32
+
+slothy.optimize(start="start", end="end")
+slothy.write_source_to_file("../examples/opt/aarch64/aarch64_simple0_macros_a55.s")