https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110460
Bug ID: 110460
Summary: [14 Regression] ft32 ICE on 931110-1.c with new
TYPE_PRECISION checking
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: law at gcc dot gnu.org
Target Milestone: ---
commit fe48f2651334bc4d96b6df6b2bb6b29fcb732a83
Author: Richard Biener <rguent...@suse.de>
Date: Fri Jun 9 09:31:14 2023 +0200
Prevent TYPE_PRECISION on VECTOR_TYPEs
The following makes sure that using TYPE_PRECISION on VECTOR_TYPE
ICEs when tree checking is enabled. This should avoid wrong-code
in cases like PR110182 and instead ICE.
It also introduces a TYPE_PRECISION_RAW accessor and adjusts
places I found that are eligible to use that.
* tree.h (TYPE_PRECISION): Check for non-VECTOR_TYPE.
(TYPE_PRECISION_RAW): Provide raw access to the precision
field.
* tree.cc (verify_type_variant): Compare TYPE_PRECISION_RAW.
(gimple_canonical_types_compatible_p): Likewise.
* tree-streamer-out.cc (pack_ts_type_common_value_fields):
Stream TYPE_PRECISION_RAW.
* tree-streamer-in.cc (unpack_ts_type_common_value_fields):
Likewise.
* lto-streamer-out.cc (hash_tree): Hash TYPE_PRECISION_RAW.
gcc/lto/
* lto-common.cc (compare_tree_sccs_1): Use TYPE_PRECISION_RAW.
One example on ft32-elf:
Tests that now fail, but worked before (13 tests):
ft32-sim: gcc.c-torture/execute/931110-1.c -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions (test for excess
errors)
ft32-sim: gcc.c-torture/execute/931110-1.c -O3 -g (test for excess errors)
ft32-sim: gcc.dg/pr108095.c (test for excess errors)
And if you dig into the 931110-1.c failure you find:
ft32-sim: gcc.c-torture/execute/931110-1.c -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions (internal compiler
error: tree check: expected none of vector_type, have vector_type in
type_has_mode_precision_p, at tree.h:6644)
ft32-sim: gcc.c-torture/execute/931110-1.c -O3 -g (internal compiler error:
tree check: expected none of vector_type, have vector_type in
type_has_mode_precision_p, at tree.h:6644)
It looks like SCALAR_DEST in vectorizable_operation is actually a vector type
-- meaning that STMT was already vectorized.
This is the patch I'm testing. There are other failures that don't seem to be
fixed by this patch. Anyway, the whole point of the change is to find these
lurking bugs.
diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc
index d642d3c257f..3dd8a284577 100644
--- a/gcc/tree-vect-stmts.cc
+++ b/gcc/tree-vect-stmts.cc
@@ -6481,6 +6481,10 @@ vectorizable_operation (vec_info *vinfo,
scalar_dest = gimple_assign_lhs (stmt);
vectype_out = STMT_VINFO_VECTYPE (stmt_info);
+ /* STMT may have already been vectorized. */
+ if (VECTOR_TYPE_P (TREE_TYPE (scalar_dest)))
+ return false;
+
/* Most operations cannot handle bit-precision types without extra
truncations. */
bool mask_op_p = VECTOR_BOOLEAN_TYPE_P (vectype_out);
