Internal compiler error

[eunoia]

I have been trying to compile the cilkplus branch of the gcc 4.7 compiler on a 32 bit machine running ubuntu 10.10 compiling with
gcc version 4.4.5 (Ubuntu/Linaro 4.4.4-14ubuntu5)

It goes a long way into the comple until it reaches the libcilkrts compile in stage 3 at which point it gives the following error:

../.././libcilkrts/runtime/metacall_impl.c: In function __cilkrts_metacall:
../.././libcilkrts/runtime/metacall_impl.c:45:23: internal compiler error: in expand_builtin_cilk_metadata, at cilk.c:1071

Any idea what I am doing wrong?

Thanks

[Balaji_I_Intel]

Hello,
Can you please tell me what command you are using to configure the compiler?

Thanks,

Balaji V. Iyer.

[eunoia]

Hello,

I used the following

./configure --prefix=/usr/local/cilk --enable-shared --enable-languages=c,c++ LD_LIBRARY_PATH=/usr/lib:/usr/local/lib --with-mpfr-lib=$(pwd)/mpfr/.libs --with-mpfr-include=$(pwd)/mpfr/include --with-gmp-include=$(pwd)/gmp --with-gmp-lib=$(pwd)/gmp/.libs LD_RUN_PATH=/usr/lib

I also have the following versions of bison, flex, gmp, mpfr and mpc installed:

bison 2.5
flex 2.5.35
gmp 4.3.2
mpfr 2.42
mpc 0.8.1

Thanks

[patrick_marlier]

Hello,

Just to report that I see the same thing on i686 (32bits) / fedora so probably you can reproduce this with 32 bits.

Some info (revision 184130):

(gdb) bt

#2 0x087902c5 in expand_builtin_cilk_metadata (annotation=0xb7c7527c "cilkscreen_metacall", exp=0xb7c754a4) at ../../cilkplus/gcc/cilk.c:1071

#3 0x081ee504 in expand_builtin (exp=0xb7c754a4, target=0xb7bed33c, subtarget=0x0, mode=VOIDmode, ignore=1)

at ../../cilkplus/gcc/builtins.c:6964

#4 0x082e7602 in expand_expr_real_1 (exp=0xb7c754a4, target=0x0, tmode=VOIDmode, modifier=EXPAND_NORMAL, alt_rtl=0x0)

at ../../cilkplus/gcc/expr.c:9998

#5 0x08220b3c in expand_call_stmt (stmt=0xb7caa000) at ../../cilkplus/gcc/cfgexpand.c:2077

#6 expand_gimple_stmt_1 (stmt=0xb7caa000) at ../../cilkplus/gcc/cfgexpand.c:2116

#7 expand_gimple_stmt (stmt=0xb7caa000) at ../../cilkplus/gcc/cfgexpand.c:2271

#8 0x082224e2 in expand_gimple_basic_block (bb=0xb7bf161c) at ../../cilkplus/gcc/cfgexpand.c:4020

(gdb)

#2 0x087902c5 in expand_builtin_cilk_metadata (annotation=0xb7c7527c "cilkscreen_metacall", exp=0xb7c754a4) at ../../cilkplus/gcc/cilk.c:1071

1071 gcc_assert (loc_ref);

(gdb) l

1066 }

1067

1068 loc_ref = loc_descriptor (metadata_info.reg_rtx, VOIDmode,

1069 VAR_INIT_STATUS_UNKNOWN);

1070

1071 gcc_assert (loc_ref);

1072 size = (unsigned short) size_of_locs (loc_ref);

1073

1074 metadata_info.dwarf_expr = (unsigned short)(size & 0xFF) |

1075 (unsigned short) (loc_ref->dw_loc_opc << 8);

(gdb) p debug_tree(exp)

type

align 8 symtab 0 alias set -1 canonical type 0xb7bf676c

pointer_to_this >

side-effects nothrow

fn

type

unsigned SI

size

unit size

align 32 symtab 0 alias set -1 canonical type 0xb7ca5258>

constant

arg 0

addressable used nothrow public external built-in decl_5 QI file line 0 col 0

align 8 built-in BUILT_IN_NORMAL:BUILT_IN_NOTIFY_INTRINSIC

(mem:QI (symbol_ref:SI ("__notify_intrinsic") [flags 0x41] ) [0 S1 A8]) chain >>

arg 0

type

unsigned SI size unit size

align 32 symtab 0 alias set 4 canonical type 0xb7ca512c>

readonly constant

arg 0

readonly constant static "cilkscreen_metacall\000">

../../../cilkplus/libcilkrts/runtime/metacall_impl.c:45:24>

arg 1

type

public unsigned SI size unit size

align 32 symtab 0 alias set -1 canonical type 0xb7ca51f4>

arg 0

addressable used BLK file ../../../cilkplus/libcilkrts/runtime/metacall_impl.c line 38 col 21

size

unit size

align 32 context

(mem/c:BLK (plus:SI (reg/f:SI 54 virtual-stack-vars)

(const_int -12 [0xfffffffffffffff4])) [2 d+0 S12 A32])>

../../../cilkplus/libcilkrts/runtime/metacall_impl.c:45:47>

../../../cilkplus/libcilkrts/runtime/metacall_impl.c:45:23>

I hope this can help you. Thanks for your work on Cilk/GCC!

[Balaji_I_Intel]

Hello Patrick and Eunoia,
It seem to work fine on a 64 bit machine and I think it has something to do with the 32 bit machine description of x86. I will investigate this and will let you know if I get more information.

Patrick, Thank you for giving me the additional information.

thanks,

Balaji V. Iyer.

[Balaji_I_Intel]

Hello Everyone,
I have fixed this bug and submitted a patch to GCC. Here is the snapshot for the updated version (http://gcc.gnu.org/git/?p=gcc.git;a=snapshot;h=9772b731380e25c3250e1d445c143854526c7d60;sf=tgz).

Please let me know if you have any additional problems.

Thanks,

Balaji V. Iyer.

[patrick_marlier]

Works for me.

Thanks a lot ;)