[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 Roger Sayle changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #9 from Roger Sayle --- This should now be fixed on mainline. Thanks HJ and Uros.
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 --- Comment #8 from CVS Commits --- The master branch has been updated by Roger Sayle : https://gcc.gnu.org/g:24a7980d0f48671ea13da18c9162a43420b5af58 commit r13-4872-g24a7980d0f48671ea13da18c9162a43420b5af58 Author: Roger Sayle Date: Fri Dec 23 09:50:18 2022 + PR target/106933: Limit TImode STV to SSA-like def-use chains on x86. With many thanks to H.J. for doing all the hard work, this patch resolves two P1 regressions; PR target/106933 and PR target/106959. Although superficially similar, the i386 backend's two scalar-to-vector (STV) passes perform their transformations in importantly different ways. The original pass converting SImode and DImode operations to V4SImode or V2DImode operations is "soft", allowing values to be maintained in both integer and vector hard registers. The newer pass converting TImode operations to V1TImode is "hard" (all or nothing) that converts all uses of a pseudo to vector form. To implement this it invokes powerful ju-ju calling SET_MODE on a reg_rtx, which due to RTL sharing, often updates this pseudo's mode everywhere in the RTL chain. Hence, TImode STV can only be performed when all uses of a pseudo are convertible to V1TImode form. To ensure this the STV passes currently use data-flow analysis to inspect all DEFs and USEs in a chain. This works fine for chains that are in the usual single assignment form, but the occurrence of uninitialized variables, or multiple assignments that split a pseudo's usage into several independent chains (lifetimes) can lead to situations where some but not all of a pseudo's occurrences need to be updated. This is safe for the SImode/DImode pass, but leads to the above bugs during the TImode pass. My one minor tweak to HJ's patch from comment #4 of bugzilla PR106959 is to only perform the new single_def_chain_p check for TImode STV; it turns out that STV of SImode/DImode min/max operates safely on multiple-def chains, and prohibiting this leads to testsuite regressions. We don't (yet) support V1TImode min/max, so this idiom isn't an issue during the TImode STV pass. For the record, the two alternate possible fixes are (i) make the TImode STV pass "soft", by eliminating use of SET_MODE, instead using replace_rtx with a new pseudo, or (ii) merging "chains" so that multiple DFA chains/lifetimes are considered a single STV chain. 2022-12-23 H.J. Lu Roger Sayle gcc/ChangeLog PR target/106933 PR target/106959 * config/i386/i386-features.cc (single_def_chain_p): New predicate function to check that a pseudo's use-def chain is in SSA form. (timode_scalar_to_vector_candidate_p): Check that TImode regs that are SET_DEST or SET_SRC of an insn match/are single_def_chain_p. gcc/testsuite/ChangeLog PR target/106933 PR target/106959 * gcc.target/i386/pr106933-1.c: New test case. * gcc.target/i386/pr106933-2.c: Likewise. * gcc.target/i386/pr106959-1.c: Likewise. * gcc.target/i386/pr106959-2.c: Likewise. * gcc.target/i386/pr106959-3.c: Likewise.
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 Jakub Jelinek changed: What|Removed |Added CC||jakub at gcc dot gnu.org URL||https://gcc.gnu.org/piperma ||il/gcc-patches/2022-October ||/604743.html --- Comment #7 from Jakub Jelinek --- Have you pinged your patch (and CCed Uros)?
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933
--- Comment #6 from H.J. Lu ---
This
diff --git a/gcc/config/i386/i386-features.cc
b/gcc/config/i386/i386-features.cc
index e357294bc4e..aca34d730a8 100644
--- a/gcc/config/i386/i386-features.cc
+++ b/gcc/config/i386/i386-features.cc
@@ -1449,6 +1449,9 @@ timode_scalar_chain::compute_convert_gain ()
break;
case COMPARE:
+if (REG_P (XEXP (src, 0))
+&& GET_MODE (XEXP (src, 0)) == V1TImode)
+ return 1;
if (XEXP (src, 1) == const0_rtx)
{
if (GET_CODE (XEXP (src, 0)) == AND)
also works.
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 Richard Biener changed: What|Removed |Added Priority|P3 |P1
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933
H.J. Lu changed:
What|Removed |Added
CC||hjl.tools at gmail dot com
--- Comment #5 from H.J. Lu ---
This fixes ICE:
diff --git a/gcc/config/i386/i386-features.cc
b/gcc/config/i386/i386-features.cc
index fd212262f50..cc690a6064c 100644
--- a/gcc/config/i386/i386-features.cc
+++ b/gcc/config/i386/i386-features.cc
@@ -1975,6 +1975,9 @@ timode_scalar_to_vector_candidate_p (rtx_insn *insn)
switch (GET_CODE (src))
{
case REG:
+ /* Return false if the source is uninitialized. */
+ return DF_REG_DEF_CHAIN (REGNO (src)) != nullptr;
+
case CONST_WIDE_INT:
return true;
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933
--- Comment #4 from Arseny Solokha ---
JFTR, I've just got an ICE that I presume to be another manifestation of the
same issue. The current gcc trunk snapshots ICE the way reported in comment 0
on the following testcase w/ -msse4 -Os:
__int128 n;
__int128
empty (void)
{
}
int
foo (void)
{
n = empty ();
return n == 0;
}
while replacing the last return statement w/ "return n == 1;" gives the
following:
% x86_64-unknown-linux-gnu-gcc-13.0.0 -msse4 -Os -c lr4kukze.c
during RTL pass: cse2
lr4kukze.c: In function 'foo':
lr4kukze.c:14:1: internal compiler error: in as_a, at machmode.h:365
14 | }
| ^
0x73665b scalar_int_mode as_a(machine_mode)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/machmode.h:365
0x736783 scalar_mode as_a(machine_mode)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/rtl.h:2288
0x736783 wi::int_traits
>::get_precision(std::pair const&)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/rtl.h:2273
0x736783 unsigned int wi::get_precision
>(std::pair const&)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/wide-int.h:1794
0x736783 wide_int_ref_storage::wide_int_ref_storage
>(std::pair const&)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/wide-int.h:1024
0x736783 generic_wide_int
>::generic_wide_int >(std::pair const&)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/wide-int.h:782
0x736783 poly_int<1u, generic_wide_int >
>::poly_int >(std::pair const&)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/poly-int.h:670
0x736783 wi::to_poly_wide(rtx_def const*, machine_mode)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/rtl.h:2373
0x736783 neg_poly_int_rtx
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/simplify-rtx.cc:57
0xec311e simplify_context::simplify_binary_operation_1(rtx_code, machine_mode,
rtx_def*, rtx_def*, rtx_def*, rtx_def*)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/simplify-rtx.cc:3118
0xec501d simplify_context::simplify_binary_operation(rtx_code, machine_mode,
rtx_def*, rtx_def*)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/simplify-rtx.cc:2626
0xed00c2 simplify_binary_operation(rtx_code, machine_mode, rtx_def*, rtx_def*)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/rtl.h:3475
0xed00c2 simplify_const_relational_operation(rtx_code, machine_mode, rtx_def*,
rtx_def*)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/simplify-rtx.cc:6112
0xec948b simplify_context::simplify_relational_operation(rtx_code,
machine_mode, machine_mode, rtx_def*, rtx_def*)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/simplify-rtx.cc:5704
0x1cfed28 simplify_relational_operation(rtx_code, machine_mode, machine_mode,
rtx_def*, rtx_def*)
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/rtl.h:3490
0x1cfed28 fold_rtx
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/cse.cc:3385
0x1cfffb3 cse_insn
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/cse.cc:4669
0x1d04e24 cse_extended_basic_block
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/cse.cc:6566
0x1d04e24 cse_main
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/cse.cc:6711
0x1d05750 rest_of_handle_cse2
/var/tmp/portage/sys-devel/gcc-13.0.0_p20220925/work/gcc-13-20220925/gcc/cse.cc:7600
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 --- Comment #3 from Roger Sayle --- Ignore all of Comment 2. The reason the instruction isn't getting split, is that (unlike the other clones), the pseudo's mode is V1TI (not TI), hence doesn't match the constraint. The problem is STV (again), where uses of a pseudo are split into different chains, and some chains are considered profitable, and other chains aren't supported in V1TImode. Searching for mode conversion candidates... insn 13 is marked as a TImode candidate insn 16 is marked as a TImode candidate Created a new instruction chain #7 Building chain #7... Adding insn 13 to chain #7 Collected chain #7... insns: 13 Computing gain for chain #7... Instruction gain 4 for13: [r91:DI]=r82:TI REG_DEAD r91:DI Total gain: 5 Converting chain #7... deferring rescan insn with uid = 13. Created a new instruction chain #8 Building chain #8... Adding insn 16 to chain #8 Collected chain #8... insns: 16 Computing gain for chain #8... Instruction gain -4 for16: flags:CCZ=cmp(r82:V1TI,0x1) REG_DEAD r82:V1TI Total gain: -3 Chain #8 conversion is not profitable Total insns converted: 1 The issue is that you can't convert insn 13 to V1TImode if you don't convert insn 16, i.e. all uses of a pseudo should have the same mode. [Calling PUT_MODE on the first use, implicitly updates the second]. I believe (this time) that the uninitialized value from "empty()" confuses that DF analysis, and insns 13 and 16 end up on different (dependency) chains, because the lack on initialization makes them look independent.
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 Roger Sayle changed: What|Removed |Added CC||roger at nextmovesoftware dot com --- Comment #2 from Roger Sayle --- Something very odd is going on. foo.simdclone.2 contains a *cmpti_doubleword instruction, that is defined using define_insn_and_split to be split pre-reload. For some reason, this isn't getting split during the split1 pass, but continues through reload until things finally fail in cprop_hardreg. Interestingly the corresponding instruction is correctly split in simdclone.0 and simdclone.1. The .293e.split1 file contains (much redacted)... ;; Function foo.simdclone.0 Splitting with gen_split_2 (i386.md:1511) deleting insn with uid = 16. ;; Function foo.simdclone.1 Splitting with gen_split_2 (i386.md:1511) deleting insn with uid = 22. ;; Function foo.simdclone.2 i.e. no splits are performed on foo.simdclone.2, but they are on simdclone.0 and simdclone.1. Perhaps an off-by-one error? I'll investigate further, but I doubt this is an issue in the i386.md machine description; the ((simd)) attribute is interacting strangely with the split pass.
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 Richard Biener changed: What|Removed |Added Target Milestone|--- |13.0
[Bug target/106933] [13 Regression] ICE in extract_insn, at recog.cc:2791 (error: unrecognizable insn) since r13-2049-g6f94923dea21bd92
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106933 Martin Liška changed: What|Removed |Added CC||marxin at gcc dot gnu.org, ||sayle at gcc dot gnu.org Summary|[13 Regression] ICE in |[13 Regression] ICE in |extract_insn, at|extract_insn, at |recog.cc:2791 (error: |recog.cc:2791 (error: |unrecognizable insn)|unrecognizable insn) since ||r13-2049-g6f94923dea21bd92 Last reconfirmed||2022-09-13 Ever confirmed|0 |1 Status|UNCONFIRMED |NEW --- Comment #1 from Martin Liška --- Started since r13-2049-g6f94923dea21bd92.
