Improved profiler result printing. (#7709)

* Fixed the regularization for BGU.

* Improved profiler result printing.

* Clang-format ain't liking pretty code.

* Clang-tidy ain't liking pretty code.

---------

Co-authored-by: Steven Johnson <[email protected]>

src/runtime/profiler_common.cpp

@@ -366,6 +366,15 @@ WEAK void halide_profiler_report_unlocked(void *user_context, halide_profiler_st
         }
 
         if (print_f_states) {
+            int max_func_name_length = 0;
+            for (int i = 0; i < p->num_funcs; i++) {
+                halide_profiler_func_stats *fs = p->funcs + i;
+                int name_len = strlen(fs->name);
+                if (name_len > max_func_name_length) {
+                    max_func_name_length = name_len;
+                }
+            }
+
             for (int i = 0; i < p->num_funcs; i++) {
                 size_t cursor = 0;
                 sstr.clear();
@@ -378,17 +387,29 @@ WEAK void halide_profiler_report_unlocked(void *user_context, halide_profiler_st
                 }
 
                 sstr << "  " << fs->name << ": ";
-                cursor += 25;
+                cursor += max_func_name_length + 5;
                 while (sstr.size() < cursor) {
                     sstr << " ";
                 }
 
                 float ft = fs->time / (p->runs * 1000000.0f);
+                if (ft < 10000) {
+                    sstr << " ";
+                }
+                if (ft < 1000) {
+                    sstr << " ";
+                }
+                if (ft < 100) {
+                    sstr << " ";
+                }
+                if (ft < 10) {
+                    sstr << " ";
+                }
                 sstr << ft;
                 // We don't need 6 sig. figs.
                 sstr.erase(3);
                 sstr << "ms";
-                cursor += 10;
+                cursor += 12;
                 while (sstr.size() < cursor) {
                     sstr << " ";
                 }