[Bug target/58901] vax bootstrap fails on subreg reload
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901
--- Comment #1 from Martin Husemann ---
The real question is: why does memory_address_addr_space_p() return false for
this rtx. Stepping into it results in:
0x007618be in vax_legitimate_address_p (mode=HImode, x=0x7ea0fd2c,
strict=20, 5, 2124479788, 0) at ../../gcc-4.8.2/gcc/config/vax/vax.c:1801
1801vax_legitimate_address_p (enum machine_mode mode, rtx x, bool strict)
(gdb) n
1805 if (nonindexed_address_p (x, strict))
(gdb) print mode
$4 = HImode
(gdb) p debug_rtx(x)
(plus:SI (mult:SI (reg/v:SI 0 %r0 [orig:81 count ] [81])
(const_int 4 [0x4]))
(symbol_ref:SI ("DECPOWERS") [flags 0x40] ))
$5 = void
(gdb) p strict
$6 = false
(gdb) n
1808 if (GET_CODE (x) != PLUS)
(gdb)
1813 xfoo0 = XEXP (x, 0);
(gdb)
1814 xfoo1 = XEXP (x, 1);
(gdb)
1816 if (index_term_p (xfoo0, mode, strict)
(gdb)
1826 if (indexable_address_p (xfoo0, xfoo1, mode, strict)
(gdb)
1830 return false;
with a backtrace:
#0 vax_legitimate_address_p (mode=HImode, x=0x7ea0fd2c,
strict=false, 5, 2124479788, 0)
at ../../gcc-4.8.2/gcc/config/vax/vax.c:1830
#1 0x0052049a in default_addr_space_legitimate_address_p (mode=HImode,
mem=0x7ea0fd2c, strict=false, as=0 '\000', 5, 2124479788, 0, 0)
at ../../gcc-4.8.2/gcc/targhooks.c:1175
#2 0x004aa683 in memory_address_addr_space_p (mode=HImode, addr=0x7ea0fd2c,
as=0 '\000', 5, 2124479788, 0) at ../../gcc-4.8.2/gcc/recog.c:1299
#3 0x00295454 in change_address_1 (memref=0x7ea2d560, mode=HImode,
addr=0x7ea0fd2c, validate=1, 2124600672, 5, 2124479788, 1)
at ../../gcc-4.8.2/gcc/emit-rtl.c:2019
and, as expected:
(gdb) p debug_rtx(xfoo0)
(mult:SI (reg/v:SI 0 %r0 [orig:81 count ] [81])
(const_int 4 [0x4]))
$7 = void
(gdb) p debug_rtx(xfoo1)
(symbol_ref:SI ("DECPOWERS") [flags 0x40] )
$8 = void
[Bug target/58901] vax bootstrap fails on subreg reload
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901
--- Comment #2 from Martin Husemann ---
indexable_address_p() returns false for
(symbol_ref:SI ("DECPOWERS") [flags 0x40] )
because flag_pic is true and symbolic_operand (xfoo0, SImode)) returns true:
/* Return true if xfoo0 and xfoo1 constitute a valid indexed address. */
static bool
indexable_address_p (rtx xfoo0, rtx xfoo1, enum machine_mode mode, bool strict)
{
if (!CONSTANT_ADDRESS_P (xfoo0))
return false;
if (BASE_REGISTER_P (xfoo1, strict))
return !flag_pic || mode == QImode;
if (flag_pic && symbolic_operand (xfoo0, SImode))
return false;
return reg_plus_index_p (xfoo1, mode, strict);
}
I don't know vax addressing modes - help!
[Bug target/58901] vax bootstrap fails on subreg reload
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901
--- Comment #3 from Martin Husemann ---
Matt asked for the instruction involved - I think it is this:
(insn 245 244 246 51 (set (mem:HI (reg/v/f:SI 1 %r1 [orig:67 sup ] [67]) [2
*sup_104+0 S2 A16])
(plus:HI (subreg:HI (mem/u/c:SI (plus:SI (mult:SI (reg/v:SI 0 %r0
[orig:81 count ] [81])
(const_int 4 [0x4]))
(symbol_ref:SI ("DECPOWERS") [flags 0x40] )) [3 DECPOWERS S4 A32]) 0)
(const_int -1 [0x])))
../../gcc-4.8.2/libdecnumber/decNumber.c:7183 48 {addhi3}
(nil))
[Bug target/58901] vax bootstrap fails on subreg reload
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901
--- Comment #4 from Martin Husemann ---
I got a quick lesson in addressing modes for vax ;-)
It seems the mode = HImode passed to the upper functions in the call stack is
the problem - with HImode we can only use index operands with a factor of 2,
but this rtx has a factor of 4.
The problem starts in alter_subreg(), where the call to adjust_address_1 is
done:
#6 0x002d1a06 in alter_subreg (xp=0x7f0e8bc8, final_p=true, 2131659720, 1)
at ../../gcc-4.8.2/gcc/final.c:3072
3072*xp = adjust_address (y, GET_MODE (x), offset);
and x has (due to subreg) HImode:
(gdb) p debug_rtx(x)
(subreg:HI (mem/u/c:SI (plus:SI (mult:SI (reg/v:SI 0 %r0 [orig:81 count ] [81])
(const_int 4 [0x4]))
(symbol_ref:SI ("DECPOWERS") [flags 0x40] )) [3 DECPOWERS S4 A32]) 0)
$12 = void
(gdb) p debug_rtx(y)
(mem/u/c:SI (plus:SI (mult:SI (reg/v:SI 0 %r0 [orig:81 count ] [81])
(const_int 4 [0x4]))
(symbol_ref:SI ("DECPOWERS") [flags 0x40] )) [3 DECPOWERS S4 A32])
[Bug target/58901] vax bootstrap fails on subreg reload
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901 --- Comment #5 from Martin Husemann --- Created attachment 31275 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=31275&action=edit .i file from the failing compile
[Bug target/58901] vax bootstrap fails on subreg reload
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901 --- Comment #6 from Martin Husemann --- This is beyound my gcc capabilities: the rtx in question is wrong for vax, but the bug seems to be at a higher level: an indexed memory access can not be wrapped into a subreg with offset in the same statement - the rtx needs to be split into a register load and then subreg extraction (or so I think, not an expert in vax addressing modes)
[Bug target/58901] vax bootstrap fails on subreg reload
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901 coypu changed: What|Removed |Added CC||coypu at sdf dot org --- Comment #7 from coypu --- Created attachment 46884 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=46884&action=edit matt's changes, synced by other matt, against trunk. Confirming as still occurring in gcc-trunk as of: xgcc (GCC) 10.0.0 20190907 (experimental) NetBSD has a local diff to avoid this crash. I attached it (updated for trunk) for reference. The commit message for it (by Matt Thomas): https://github.com/NetBSD/src/commit/e437e96750193b86d0464965661f616e011056fa "Fix a problem with reloading the inner reg of a subreg when it's a mode dependent address."
[Bug target/58901] vax bootstrap fails on subreg reload
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901 Andrew Pinski changed: What|Removed |Added Target Milestone|--- |11.0
[Bug target/58901] vax bootstrap fails on subreg reload
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901
--- Comment #8 from CVS Commits ---
The master branch has been updated by Maciej W. Rozycki :
https://gcc.gnu.org/g:a27d5f9a73978f20cfef1796a94f6a1a82438146
commit r11-5761-ga27d5f9a73978f20cfef1796a94f6a1a82438146
Author: Matt Thomas
Date: Sat Dec 5 18:26:23 2020 +
PR target/58901: reload: Handle SUBREG of MEM with a mode-dependent address
Fix an ICE with the handling of RTL expressions like:
(subreg:QI (mem/c:SI (plus:SI (plus:SI (mult:SI (reg/v:SI 0 %r0 [orig:67 i
] [67])
(const_int 4 [0x4]))
(reg/v/f:SI 7 %r7 [orig:59 doacross ] [59]))
(const_int 40 [0x28])) [1 MEM[(unsigned int *)doacross_63 + 40B
+ i_106 * 4]+0 S4 A32]) 0)
that causes the compilation of libgomp to fail:
during RTL pass: reload
.../libgomp/ordered.c: In function 'GOMP_doacross_wait':
.../libgomp/ordered.c:507:1: internal compiler error: in change_address_1,
at emit-rtl.c:2275
507 | }
| ^
0x10a3462b change_address_1
.../gcc/emit-rtl.c:2275
0x10a353a7 adjust_address_1(rtx_def*, machine_mode, poly_int<1u, long>,
int, int, int, poly_int<1u, long>)
.../gcc/emit-rtl.c:2409
0x10ae2993 alter_subreg(rtx_def**, bool)
.../gcc/final.c:3368
0x10ae25cf cleanup_subreg_operands(rtx_insn*)
.../gcc/final.c:3322
0x110922a3 reload(rtx_insn*, int)
.../gcc/reload1.c:1232
0x10de2bf7 do_reload
.../gcc/ira.c:5812
0x10de3377 execute
.../gcc/ira.c:5986
in a `vax-netbsdelf' build, where an attempt is made to change the mode
of the contained memory reference to the mode of the containing SUBREG.
Such RTL expressions are produced by the VAX shift and rotate patterns
(`ashift', `ashiftrt', `rotate', `rotatert') where the count operand
always has the QI mode regardless of the mode, either SI or DI, of the
datum shifted or rotated.
Such a mode change cannot work where the memory reference uses the
indexed addressing mode, where a multiplier is implied that in the VAX
ISA depends on the width of the memory access requested and therefore
changing the machine mode would change the address calculation as well.
Avoid the attempt then by forcing the reload of any SUBREGs containing
a mode-dependent memory reference, also fixing these regressions:
FAIL: gcc.c-torture/compile/pr46883.c -Os (internal compiler error)
FAIL: gcc.c-torture/compile/pr46883.c -Os (test for excess errors)
FAIL: gcc.c-torture/execute/20120808-1.c -O2 (internal compiler error)
FAIL: gcc.c-torture/execute/20120808-1.c -O2 (test for excess errors)
FAIL: gcc.c-torture/execute/20120808-1.c -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions (internal compiler
error)
FAIL: gcc.c-torture/execute/20120808-1.c -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions (test for excess
errors)
FAIL: gcc.c-torture/execute/20120808-1.c -O3 -g (internal compiler
error)
FAIL: gcc.c-torture/execute/20120808-1.c -O3 -g (test for excess errors)
FAIL: gcc.c-torture/execute/20120808-1.c -O2 -flto -fno-use-linker-plugin
-flto-partition=none (internal compiler error)
FAIL: gcc.c-torture/execute/20120808-1.c -O2 -flto -fno-use-linker-plugin
-flto-partition=none (test for excess errors)
FAIL: gcc.c-torture/execute/20120808-1.c -O2 -flto -fuse-linker-plugin
-fno-fat-lto-objects (internal compiler error)
FAIL: gcc.c-torture/execute/20120808-1.c -O2 -flto -fuse-linker-plugin
-fno-fat-lto-objects (test for excess errors)
FAIL: gcc.dg/20050629-1.c (internal compiler error)
FAIL: gcc.dg/20050629-1.c (test for excess errors)
FAIL: c-c++-common/torture/pr53505.c -Os (internal compiler error)
FAIL: c-c++-common/torture/pr53505.c -Os (test for excess errors)
FAIL: gfortran.dg/coarray_failed_images_1.f08 -Os (internal compiler
error)
FAIL: gfortran.dg/coarray_stopped_images_1.f08 -Os (internal compiler
error)
With test case #0 included it causes a reload with:
(insn 15 14 16 4 (set (reg:SI 31)
(ashift:SI (const_int 1 [0x1])
(subreg:QI (reg:SI 30 [ MEM[(int *)s_8(D) + 4B + _5 * 4] ])
0))) "pr58901-0.c":15:12 94 {ashlsi3}
(expr_list:REG_DEAD (reg:SI 30 [ MEM[(int *)s_8(D) + 4B + _5 * 4] ])
(nil)))
as follows:
Reloads for insn # 15
Reload 0: reload_in (SI) = (reg:SI 30 [ MEM[(int *)s_8(D) + 4B + _5 * 4] ])
ALL_REGS, RELOAD_FOR_INPUT (opnum = 2)
reload_in_reg: (reg:SI 30 [ MEM[(int *)s_8(D) + 4B + _5 * 4] ])
reload_reg_rtx: (reg:SI 5 %r5)
resulting in:
(insn 37 14 15 4 (set (reg:SI 5 %r5)
(mem/c:SI (plus:SI (plus:SI (mult:SI (reg/v:SI 1 %r1 [orig:25 i ]
[25])
(const_int 4 [0x4]))
[Bug target/58901] vax bootstrap fails on subreg reload
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58901 Maciej W. Rozycki changed: What|Removed |Added Assignee|unassigned at gcc dot gnu.org |ma...@linux-mips.org Resolution|--- |FIXED CC||ma...@linux-mips.org Status|UNCONFIRMED |RESOLVED --- Comment #9 from Maciej W. Rozycki --- Fixed as per the commit message automatically entered.
