http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59429
Bug ID: 59429
Summary: Missed optimization opportunity in qsort-style
comparison functions
Product: gcc
Version: 4.8.1
Status: UNCONFIRMED
Severity: minor
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: jengelh at inai dot de
The following code:
extern int comLG(int, int);
extern int comLE(int, int);
extern int comEL(int, int);
extern int comEG(int, int);
extern int comGL(int, int);
extern int comGE(int, int);
int comLG(int a, int b) { return (a < b) ? -1 : (a > b) ? 1 : 0; }
int comLE(int a, int b) { return (a < b) ? -1 : (a == b) ? 0 : 1; }
int comEL(int a, int b) { return (a == b) ? 0 : (a < b) ? -1 : 1; }
int comEG(int a, int b) { return (a == b) ? 0 : (a > b) ? 1 : -1; }
int comGL(int a, int b) { return (a > b) ? 1 : (a < b) ? -1 : 0; }
int comGE(int a, int b) { return (a > b) ? 1 : (a == b) ? 0 : -1; }
when compiled with:
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib64/gcc/x86_64-suse-linux/4.8/lto-wrapper
Target: x86_64-suse-linux
Configured with: ../configure --prefix=/usr --infodir=/usr/share/info
--mandir=/usr/share/man --libdir=/usr/lib64 --libexecdir=/usr/lib64
--enable-languages=c,c++,objc,fortran,obj-c++,java,ada
--enable-checking=release --with-gxx-include-dir=/usr/include/c++/4.8
--enable-ssp --disable-libssp --disable-plugin
--with-bugurl=http://bugs.opensuse.org/ --with-pkgversion='SUSE Linux'
--disable-libgcj --disable-libmudflap --with-slibdir=/lib64 --with-system-zlib
--enable-__cxa_atexit --enable-libstdcxx-allocator=new --disable-libstdcxx-pch
--enable-version-specific-runtime-libs --enable-linker-build-id
--program-suffix=-4.8 --enable-linux-futex --without-system-libunwind
--with-arch-32=i586 --with-tune=generic --build=x86_64-suse-linux
Thread model: posix
gcc version 4.8.1 20130909 [gcc-4_8-branch revision 202388] (SUSE Linux)
yields object code with different realizations and sizes.
What I observe:
$ gcc -Os -c cmp.c
(also shows with -O3)
$ readelf -s cmp.o
Num: Value Size Type Bind Vis Ndx Name
8: 0000000000000000 16 FUNC GLOBAL DEFAULT 1 comLG
9: 0000000000000010 16 FUNC GLOBAL DEFAULT 1 comLE
10: 0000000000000020 18 FUNC GLOBAL DEFAULT 1 comEL
11: 0000000000000032 18 FUNC GLOBAL DEFAULT 1 comEG
12: 0000000000000044 18 FUNC GLOBAL DEFAULT 1 comGL
13: 0000000000000056 18 FUNC GLOBAL DEFAULT 1 comGE
$ objdump -d cmp.o -Mintel
0000000000000000 <comLG>:
0: 31 d2 xor edx,edx
2: 39 f7 cmp edi,esi
4: b8 ff ff ff ff mov eax,0xffffffff
9: 0f 9f c2 setg dl
c: 0f 4d c2 cmovge eax,edx
f: c3 ret
0000000000000010 <comLE>:
10: 31 d2 xor edx,edx
12: 39 f7 cmp edi,esi
14: b8 ff ff ff ff mov eax,0xffffffff
19: 0f 95 c2 setne dl
1c: 0f 4d c2 cmovge eax,edx
1f: c3 ret
0000000000000020 <comEL>:
20: 39 f7 cmp edi,esi
22: 74 0b je 2f <comEL+0xf>
24: 0f 9d c0 setge al
27: 0f b6 c0 movzx eax,al
2a: 8d 44 00 ff lea eax,[rax+rax*1-0x1]
2e: c3 ret
2f: 31 c0 xor eax,eax
31: c3 ret
0000000000000032 <comEG>:
32: 39 f7 cmp edi,esi
34: 74 0b je 41 <comEG+0xf>
36: 0f 9f c0 setg al
39: 0f b6 c0 movzx eax,al
3c: 8d 44 00 ff lea eax,[rax+rax*1-0x1]
40: c3 ret
41: 31 c0 xor eax,eax
43: c3 ret
0000000000000044 <comGL>:
44: 39 f7 cmp edi,esi
46: b8 01 00 00 00 mov eax,0x1
4b: 7f 08 jg 55 <comGL+0x11>
4d: 0f 9c c0 setl al
50: 0f b6 c0 movzx eax,al
53: f7 d8 neg eax
55: c3 ret
0000000000000056 <comGE>:
56: 39 f7 cmp edi,esi
58: b8 01 00 00 00 mov eax,0x1
5d: 7f 08 jg 67 <comGE+0x11>
5f: 0f 95 c0 setne al
62: 0f b6 c0 movzx eax,al
65: f7 d8 neg eax
67: c3 ret
What I expected instead:
All functions to have the same asm.
