huge refactor for high RewritePatterns usage and less ad-hoc cpp code

remove Spatial many ops in favor of tensor ops like in pim

validation/validate.py

@@ -27,7 +27,9 @@ def print_validation_error(reporter, rel, exc):
           file=sys.stderr, flush=True)
     if isinstance(exc, subprocess.CalledProcessError):
         print(format_return_status(exc.returncode), file=sys.stderr, flush=True)
-        if exc.output:
+        if getattr(exc, "output_already_streamed", False):
+            print("Failure log already printed above.", file=sys.stderr, flush=True)
+        elif exc.output:
             output_text = exc.output.decode("utf-8", errors="replace") if isinstance(exc.output, bytes) else str(exc.output)
             if output_text:
                 print(output_text, file=sys.stderr, end="" if output_text.endswith("\n") else "\n", flush=True)
@@ -160,12 +162,13 @@ def main():
                 Fore.RED + plain_status.ljust(status_width) + Style.RESET_ALL
             print(f"| {rel.ljust(path_width)} | {status} |")
         print(separator)
-    print_average_pim_pass_timings(
-        pass_timing_sums,
-        pass_timing_counts,
-        total_timing_sum,
-        timed_benchmark_count,
-    )
+    if a.verbose:
+        print_average_pim_pass_timings(
+            pass_timing_sums,
+            pass_timing_counts,
+            total_timing_sum,
+            timed_benchmark_count,
+        )
 
     sys.exit(0 if n_passed == n_total else 1)