RE-BASELINE (v6 bench). The split view ('s') was not covered at all, and it turns out to be the most expensive thing in the app: 16.2s of a 33.5s session (lg_split 10189 + sm_split 6034), or 500ms per function on echo and 850ms on bash, just to open it.

Result: {"status":"keep","total_ms":33502.3,"lg_boot_ms":772.2,"lg_decomp_ms":2546.9,"lg_graph_ms":903.4,"lg_hex_ms":434.1,"lg_index_ms":97.7,"lg_listing_cold_ms":420.2,"lg_listing_warm_ms":397.4,"lg_nav_ms":6501.1,"lg_palette_ms":4.9,"lg_render_ms":212.4,"lg_search_ms":760.4,"lg_split_ms":10189.2,"pure_graph_ms":212.3,"sm_boot_ms":443.8,"sm_decomp_ms":1275.3,"sm_graph_ms":720.5,"sm_hex_ms":451,"sm_index_ms":2.5,"sm_listing_cold_ms":263.3,"sm_listing_warm_ms":264.7,"sm_nav_ms":302.8,"sm_palette_ms":0.3,"sm_render_ms":248.4,"sm_search_ms":43.9,"sm_split_ms":6033.6,"fails":0}

1 files changed, 49 insertions, 0 deletions

diff --git a/.auto/wip-decompmap.patch b/.auto/wip-decompmap.patch
new file mode 100644
index 0000000..a11db98
--- /dev/null
+++ b/.auto/wip-decompmap.patch
@@ -0,0 +1,49 @@
+diff --git a/server/patch_server.py b/server/patch_server.py
+index fe16ede..b8f5453 100644
+--- a/server/patch_server.py
++++ b/server/patch_server.py
+@@ -884,16 +884,39 @@ def decomp_map(
+         return {"error": f"decompile failed: {e}"}
+     if cfunc is None:
+         return {"error": "decompile failed"}
++    import ida_lines
++    # Three things this loop must not do, each measured on real functions (the 25
++    # largest of bash went 68.3s -> 6.5s; echo's 60 largest 5.4s -> 0.6s, with
++    # byte-identical output):
++    #
++    #  * allocate ctree_item_t's per COLUMN. They are SWIG objects and this is
++    #    the innermost loop; one per call is enough, and head/tail are never
++    #    read, so don't ask for them at all.
++    #  * sweep the TAGGED length. ``x`` is a screen column but ``sl.line`` still
++    #    carries IDA's colour tags, so a 23-column line was swept 124 times.
++    #  * call dstr() per column. It formats a whole 'EA: description' string, and
++    #    consecutive columns are nearly always the same ctree item -- so ask the
++    #    item for its id first and only format when it changes. (The result is
++    #    deduped by ``seen`` anyway, so skipping a repeat cannot change it.)
++    item = ida_hexrays.ctree_item_t()
++    tag_remove = ida_lines.tag_remove
++    get_line_item = cfunc.get_line_item
+     lines = []
+     for sl in cfunc.get_pseudocode():
+         line = sl.line
+         eas, seen = [], set()
+-        for x in range(len(line) + 1):
+-            head = ida_hexrays.ctree_item_t()
+-            item = ida_hexrays.ctree_item_t()
+-            tail = ida_hexrays.ctree_item_t()
+-            if not cfunc.get_line_item(line, x, False, head, item, tail):
++        prev_id = None
++        for x in range(len(tag_remove(line)) + 1):
++            if not get_line_item(line, x, False, None, item, None):
+                 continue
++            it = item.it
++            if it is not None:
++                oid = it.obj_id
++                if oid == prev_id:
++                    continue
++                prev_id = oid
++            else:
++                prev_id = None
+             # Match the /*ea*/ marker's source (decompile_function_safe): the
+             # item's dstr() is 'EA: description'; get_ea() reports a different ea.
+             dstr = item.dstr()