split_heuristic_estimate_performance config option; defaults to True

slothy/core/config.py

@@ -581,6 +581,15 @@ def split_heuristic_preprocess_naive_interleaving_by_latency(self):
                 "read config.split_heuristic_preprocess_naive_interleaving_by_latency otherwise.")
         return self._split_heuristic_preprocess_naive_interleaving_by_latency
 
+    @property
+    def split_heuristic_estimate_performance(self):
+        """After applying the split heuristic, run SLOTHY again on the entire code to estimate the
+           performance and display un-used issue slots in the output."""
+        if not self.split_heuristic:
+            raise InvalidConfig("Did you forget to set config.split_heuristic=True? Shouldn't"    \
+                "read config.split_heuristic_estimate_performance otherwise.")
+        return self._split_heuristic_estimate_performance
+
     @property
     def flexible_lifetime_start(self):
         """Internal property indicating whether the lifetime interval of a register
@@ -1063,6 +1072,7 @@ def __init__(self, Arch, Target):
         self._split_heuristic_repeat = 1
         self._split_heuristic_preprocess_naive_interleaving = False
         self._split_heuristic_preprocess_naive_interleaving_by_latency = False
+        self._split_heuristic_estimate_performance = True
 
         self._compiler_binary = "gcc"
         self._compiler_include_paths = None
@@ -1247,15 +1257,12 @@ def split_heuristic_region(self, val):
     @split_heuristic_preprocess_naive_interleaving.setter
     def split_heuristic_preprocess_naive_interleaving(self, val):
         self._split_heuristic_preprocess_naive_interleaving = val
-    @split_heuristic_preprocess_naive_interleaving.setter
-    def split_heuristic_preprocess_naive_interleaving(self, val):
-        self._split_heuristic_preprocess_naive_interleaving = val
-    @split_heuristic_preprocess_naive_interleaving_by_latency.setter
-    def split_heuristic_preprocess_naive_interleaving_by_latency(self, val):
-        self._split_heuristic_preprocess_naive_interleaving_by_latency = val
     @split_heuristic_preprocess_naive_interleaving_by_latency.setter
     def split_heuristic_preprocess_naive_interleaving_by_latency(self, val):
         self._split_heuristic_preprocess_naive_interleaving_by_latency = val
+    @split_heuristic_estimate_performance.setter
+    def split_heuristic_estimate_performance(self, val):
+        self._split_heuristic_estimate_performance = val
     @split_heuristic_repeat.setter
     def split_heuristic_repeat(self, val):
         self._split_heuristic_repeat = val

slothy/core/heuristics.py

@@ -773,29 +773,30 @@ def not_empty(x):
         res.selfcheck(log.getChild("split_heuristic_full"))
 
         # Estimate performance of final code
-        conf2 = conf.copy()
-        conf2.constraints.allow_renaming = False
-        conf2.constraints.allow_reordering = False
-        conf2.variable_size = True
-        stall_res = Heuristics.optimize_binsearch(res.code,
-            logger.getChild("split_estimtate_perf"), conf2)
-        if stall_res.success is False:
-            log.error("Stall-estimate for final code after split heuristic failed -- should not happen? Maybe increase timeout? Just returning the result without stall-estimate.")
-        else:
-            res2 = Result(conf2)
-            res2.orig_code = orig_body
-            res2.code = res.code_raw
-            res2.codesize_with_bubbles = stall_res.codesize_with_bubbles
-            res2.success = True
-            # Compose actual code reordering from split heuristic with bubble-introducing (order-preserving) map
-            res2.reordering_with_bubbles = { i: stall_res.reordering_with_bubbles[
-                res.reordering_with_bubbles[i]] for i in range(res.codesize) }
-            res2.input_renamings = { s:s for s in inputs }
-            res2.output_renamings = { s:s for s in outputs }
-            res2.valid = True
-            res2.selfcheck(log.getChild("split_heuristic_full_with_stalls"))
-
-            res = res2
+        if conf.split_heuristic_estimate_performance:
+            conf2 = conf.copy()
+            conf2.constraints.allow_renaming = False
+            conf2.constraints.allow_reordering = False
+            conf2.variable_size = True
+            stall_res = Heuristics.optimize_binsearch(res.code,
+                logger.getChild("split_estimtate_perf"), conf2)
+            if stall_res.success is False:
+                log.error("Stall-estimate for final code after split heuristic failed -- should not happen? Maybe increase timeout? Just returning the result without stall-estimate.")
+            else:
+                res2 = Result(conf2)
+                res2.orig_code = orig_body
+                res2.code = res.code_raw
+                res2.codesize_with_bubbles = stall_res.codesize_with_bubbles
+                res2.success = True
+                # Compose actual code reordering from split heuristic with bubble-introducing (order-preserving) map
+                res2.reordering_with_bubbles = { i: stall_res.reordering_with_bubbles[
+                    res.reordering_with_bubbles[i]] for i in range(res.codesize) }
+                res2.input_renamings = { s:s for s in inputs }
+                res2.output_renamings = { s:s for s in outputs }
+                res2.valid = True
+                res2.selfcheck(log.getChild("split_heuristic_full_with_stalls"))
+
+                res = res2
 
         return res