https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126899
Bug ID: 126899
Summary: backward threader rejects joiner-style threads on
statement count that DOM can thread
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: aldyh at gcc dot gnu.org
Blocks: 126010
Target Milestone: ---
For the following testcase with -O2 -S joiner-cost.c
--param=dom-jump-threading=0 \
-fdump-tree-threadfull1-details -fdump-tree-dom2-details
...
volatile int sink;
void f (int i, int *out)
{
if (i)
*out = 1;
sink = 1;
sink = 2;
sink = 3;
sink = 4;
sink = 5;
sink = 6;
sink = 7;
ink = 8;
if (i)
*out = 2;
}
threadfull1 finds and resolves the path, rejects on cost:
Checking profitability of path (backwards): bb:4 (10 insns) bb:2
Control statement insns: 2
Overall: 8 insns
FAIL: Did not thread around loop and would copy too many statements.
But dom2 threads the identical paths, duplicating bb4:
[1] Registering jump thread: (3, 4) incoming edge; (4, 5) normal;
[2] Registering jump thread: (2, 4) incoming edge; (4, 6) normal;
Removing basic block 4
With --param max-jump-thread-duplication-stmts=25, threadfull1 itself
registers both jump threads and dom2 has nothing to do, so it looks like
the problem here is the cost model.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126010
[Bug 126010] [meta-bug] Remove DOM
