https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102889

            Bug ID: 102889
           Summary: -funroll-loops generates incorrect codes from inline
                    assembly on aarch64
           Product: gcc
           Version: 11.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ariel at amazon dot com
  Target Milestone: ---

Created attachment 51651
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51651&action=edit
Source code that reproduces bug.

The attached code works correctly with both -O2 and -O3 on GCC7 and 11.2, all
clang versions on AARCH64.
But generates incorrect code with:

/usr/local/gcc11/bin/gcc -O2 -funroll-loops -Wall -Werror -Wextra aa1.c
-fsanitize=undefined -fno-strict-aliasing -fwrapv
-fno-aggressive-loop-optimizations -o aa11y

(and in -O3 as well, just adding -funroll-loops breaks).

on all versions of gcc (triggers printf("Mismatch %i != %i\n", res1, res2);)

Attaching code to reproduce.

---


Using built-in specs.
COLLECT_GCC=/usr/local/gcc11/bin/gcc
COLLECT_LTO_WRAPPER=/usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/lto-wrapper
Target: aarch64-unknown-linux-gnu
Configured with: ./configure --prefix=/usr/local/gcc11
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 11.2.0 (GCC)
COLLECT_GCC_OPTIONS='-v' '-save-temps' '-O3' '-funroll-loops' '-Wall' '-Werror'
'-Wextra' '-march=armv8.2-a+fp16+rcpc+dotprod+crypto' '-mtune=neoverse-n1'
'-fsanitize=undefined' '-fno-strict-aliasing' '-fwrapv'
'-fno-aggressive-loop-optimizations' '-o' 'aa11y' '-mlittle-endian'
'-mabi=lp64' '-dumpdir' 'aa11y-'
 /usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/cc1 -E -quiet -v
aa1.c -march=armv8.2-a+fp16+rcpc+dotprod+crypto -mtune=neoverse-n1
-mlittle-endian -mabi=lp64 -Wall -Werror -Wextra -funroll-loops
-fsanitize=undefined -fno-strict-aliasing -fwrapv
-fno-aggressive-loop-optimizations -O3 -fpch-preprocess -o aa11y-aa1.i
ignoring nonexistent directory
"/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/../../../../aarch64-unknown-linux-gnu/include"
#include "..." search starts here:
#include <...> search starts here:
 /usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/include
 /usr/local/include
 /usr/local/gcc11/include
 /usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/include-fixed
 /usr/include
End of search list.
COLLECT_GCC_OPTIONS='-v' '-save-temps' '-O3' '-funroll-loops' '-Wall' '-Werror'
'-Wextra' '-march=armv8.2-a+fp16+rcpc+dotprod+crypto' '-mtune=neoverse-n1'
'-fsanitize=undefined' '-fno-strict-aliasing' '-fwrapv'
'-fno-aggressive-loop-optimizations' '-o' 'aa11y' '-mlittle-endian'
'-mabi=lp64' '-dumpdir' 'aa11y-'
 /usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/cc1
-fpreprocessed aa11y-aa1.i -quiet -dumpdir aa11y- -dumpbase aa1.c -dumpbase-ext
.c -march=armv8.2-a+fp16+rcpc+dotprod+crypto -mtune=neoverse-n1 -mlittle-endian
-mabi=lp64 -O3 -Wall -Werror -Wextra -version -funroll-loops
-fsanitize=undefined -fno-strict-aliasing -fwrapv
-fno-aggressive-loop-optimizations -o aa11y-aa1.s
GNU C17 (GCC) version 11.2.0 (aarch64-unknown-linux-gnu)
        compiled by GNU C version 11.2.0, GMP version 6.2.1, MPFR version
4.1.0, MPC version 1.2.1, isl version none
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
GNU C17 (GCC) version 11.2.0 (aarch64-unknown-linux-gnu)
        compiled by GNU C version 11.2.0, GMP version 6.2.1, MPFR version
4.1.0, MPC version 1.2.1, isl version none
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
Compiler executable checksum: 3921e1b032be4ab2b700d43daf3de441
COLLECT_GCC_OPTIONS='-v' '-save-temps' '-O3' '-funroll-loops' '-Wall' '-Werror'
'-Wextra' '-march=armv8.2-a+fp16+rcpc+dotprod+crypto' '-mtune=neoverse-n1'
'-fsanitize=undefined' '-fno-strict-aliasing' '-fwrapv'
'-fno-aggressive-loop-optimizations' '-o' 'aa11y' '-mlittle-endian'
'-mabi=lp64' '-dumpdir' 'aa11y-'
 as -v -EL -march=armv8.2-a+fp16+rcpc+dotprod+crypto -mabi=lp64 -o aa11y-aa1.o
aa11y-aa1.s
GNU assembler version 2.29.1 (aarch64-redhat-linux) using BFD version version
2.29.1-30.amzn2
COMPILER_PATH=/usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/:/usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/:/usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/:/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/:/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/
LIBRARY_PATH=/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/:/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/../../../../lib64/:/lib/../lib64/:/usr/lib/../lib64/:/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/../../../:/lib/:/usr/lib/
COLLECT_GCC_OPTIONS='-v' '-save-temps' '-O3' '-funroll-loops' '-Wall' '-Werror'
'-Wextra' '-march=armv8.2-a+fp16+rcpc+dotprod+crypto' '-mtune=neoverse-n1'
'-fsanitize=undefined' '-fno-strict-aliasing' '-fwrapv'
'-fno-aggressive-loop-optimizations' '-o' 'aa11y' '-mlittle-endian'
'-mabi=lp64' '-dumpdir' 'aa11y.'
 /usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/collect2 -plugin
/usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/liblto_plugin.so
-plugin-opt=/usr/local/gcc11/libexec/gcc/aarch64-unknown-linux-gnu/11.2.0/lto-wrapper
-plugin-opt=-fresolution=aa11y.res -plugin-opt=-pass-through=-lgcc
-plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
-plugin-opt=-pass-through=-lgcc -plugin-opt=-pass-through=-lgcc_s
--eh-frame-hdr -dynamic-linker /lib/ld-linux-aarch64.so.1 -X -EL -maarch64linux
-o aa11y /lib/../lib64/crt1.o /lib/../lib64/crti.o
/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/crtbegin.o
-L/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0
-L/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/../../../../lib64
-L/lib/../lib64 -L/usr/lib/../lib64
-L/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/../../..
aa11y-aa1.o -lubsan -lgcc --push-state --as-needed -lgcc_s --pop-state -lc
-lgcc --push-state --as-needed -lgcc_s --pop-state
/usr/local/gcc11/lib/gcc/aarch64-unknown-linux-gnu/11.2.0/crtend.o
/lib/../lib64/crtn.o
COLLECT_GCC_OPTIONS='-v' '-save-temps' '-O3' '-funroll-loops' '-Wall' '-Werror'
'-Wextra' '-march=armv8.2-a+fp16+rcpc+dotprod+crypto' '-mtune=neoverse-n1'
'-fsanitize=undefined' '-fno-strict-aliasing' '-fwrapv'
'-fno-aggressive-loop-optimizations' '-o' 'aa11y' '-mlittle-endian'
'-mabi=lp64' '-dumpdir' 'aa11y.'

Reply via email to