gh-106290: Fix edge cases around uops (#106319)

- Tweak uops debugging output
- Fix the bug from gh-106290
- Rename `SET_IP` to `SAVE_IP` (per https://github.com/faster-cpython/ideas/issues/558)
- Add a `SAVE_IP` uop at the start of the trace (ditto)
- Allow `unbound_local_error`; this gives us uops for `LOAD_FAST_CHECK`, `LOAD_CLOSURE`, and `DELETE_FAST`
- Longer traces
- Support `STORE_FAST_LOAD_FAST`, `STORE_FAST_STORE_FAST`
- Add deps on pycore_uops.h to Makefile(.pre.in)

Tools/cases_generator/generate_cases.py

@@ -308,8 +308,7 @@ class ActiveCacheEffect:
 
 
 FORBIDDEN_NAMES_IN_UOPS = (
-    "resume_with_error",  # Proxy for "goto", which isn't an IDENTIFIER
-    "unbound_local_error",
+    "resume_with_error",
     "kwnames",
     "next_instr",
     "oparg1",  # Proxy for super-instructions like LOAD_FAST_LOAD_FAST
@@ -401,20 +400,25 @@ class Instruction:
     def is_viable_uop(self) -> bool:
         """Whether this instruction is viable as a uop."""
         if self.always_exits:
+            # print(f"Skipping {self.name} because it always exits")
             return False
         if self.instr_flags.HAS_ARG_FLAG:
             # If the instruction uses oparg, it cannot use any caches
             if self.active_caches:
+                # print(f"Skipping {self.name} because it uses oparg and caches")
                 return False
         else:
             # If it doesn't use oparg, it can have one cache entry
             if len(self.active_caches) > 1:
+                # print(f"Skipping {self.name} because it has >1 cache entries")
                 return False
+        res = True
         for forbidden in FORBIDDEN_NAMES_IN_UOPS:
             # TODO: Don't check in '#ifdef ENABLE_SPECIALIZATION' regions
             if variable_used(self.inst, forbidden):
-                return False
-        return True
+                # print(f"Skipping {self.name} because it uses {forbidden}")
+                res = False
+        return res
 
     def write(self, out: Formatter, tier: Tiers = TIER_ONE) -> None:
         """Write one instruction, sans prologue and epilogue."""
@@ -1323,7 +1327,7 @@ class Analyzer:
             self.out.emit(make_text(name, counter))
             counter += 1
         add("EXIT_TRACE")
-        add("SET_IP")
+        add("SAVE_IP")
         for instr in self.instrs.values():
             if instr.kind == "op" and instr.is_viable_uop():
                 add(instr.name)