[ExecuTorch] Arm Ethos: Always verbose all for Vela

Pull Request resolved: #8406

Better to have the output then to have to rerun it.
ghstack-source-id: 266835195
@exported-using-ghexport

Differential Revision: [D69503726](https://our.internmc.facebook.com/intern/diff/D69503726/)

Author: digantdesai

backends/arm/arm_vela.py

@@ -39,7 +39,10 @@ def vela_bin_pack_io(prefix, data, shape_order=None):
 # Output via Vela to binary stream for ArmBackendEthosU
 # WARNING: Do not change this without changing VelaBinStream.cpp as that
 #          function consumes this format and the two need to align.
-def vela_compile(tosa_flatbuffer: bytes, args: List[str], shape_order=None):
+def vela_compile(tosa_flatbuffer: bytes, args: List[str], shape_order=None, verbose: bool = False):
+    """
+    Compile a TOSA graph to a binary stream for ArmBackendEthosU using Vela.
+    """
     with tempfile.TemporaryDirectory() as tmpdir:
         tosaname = "out.tosa"
         tosa_path = os.path.join(tmpdir, tosaname)
@@ -50,6 +53,8 @@ def vela_compile(tosa_flatbuffer: bytes, args: List[str], shape_order=None):
         output_dir = os.path.join(tmpdir, "output")
         args.append(f"--output-dir={output_dir}")
         args.append(tosa_path)
+        if verbose:
+            args.append("--verbose-all")
         vela.main(" ".join(args).split(" "))
 
         if any("ethos-u85" in arg for arg in args) or any(

backends/arm/ethosu_backend.py

@@ -58,7 +58,12 @@ def _compile_tosa_flatbuffer(
             )
 
         # Pass on the TOSA flatbuffer to the vela compiler.
-        binary = vela_compile(tosa_flatbuffer, compile_flags, input_order)
+        binary = vela_compile(
+            tosa_flatbuffer,
+            compile_flags,
+            input_order,
+            verbose=logger.getEffectiveLevel() == logging.INFO,
+        )
         return binary
 
     @staticmethod