Hi,
When I compile source code with GCC 4.4.1 (-m68060 -O3 -fomit-frame-pointer):
#include <stdint.h>
#define umul_ppmm(xh, xl, a, b) \
__asm__ ("| Inlined umul_ppmm\n" \
" move%.l %2,%/d0\n" \
" move%.l %3,%/d1\n" \
" move%.l %/d0,%/d2\n" \
" swap %/d0\n" \
" move%.l %/d1,%/d3\n" \
" swap %/d1\n" \
" move%.w %/d2,%/d4\n" \
" mulu %/d3,%/d4\n" \
" mulu %/d1,%/d2\n" \
" mulu %/d0,%/d3\n" \
" mulu %/d0,%/d1\n" \
" move%.l %/d4,%/d0\n" \
" eor%.w %/d0,%/d0\n" \
" swap %/d0\n" \
" add%.l %/d0,%/d2\n" \
" add%.l %/d3,%/d2\n" \
" jcc 1f\n" \
" add%.l %#65536,%/d1\n" \
"1: swap %/d2\n" \
" moveq %#0,%/d0\n" \
" move%.w %/d2,%/d0\n" \
" move%.w %/d4,%/d2\n" \
" move%.l %/d2,%1\n" \
" add%.l %/d1,%/d0\n" \
" move%.l %/d0,%0" \
: "=g" ((uint32_t) (xh)), \
"=g" ((uint32_t) (xl)) \
: "g" ((uint32_t) (a)), \
"g" ((uint32_t) (b)) \
: "d0", "d1", "d2", "d3", "d4")
int64_t MUL64(int a, int b)
{
uint32_t au = a;
uint32_t bu = b;
uint32_t resh, resl;
uint64_t res;
umul_ppmm(resh, resl, au, bu);
if (a < 0)
resh -= bu;
if (b < 0)
resh -= au;
res = ((uint64_t)resh << 32) | resl;
return res;
}
I get this asm output:
#NO_APP
.text
.even
.globl _MUL64
_MUL64:
movem.l #16128,-(sp)
move.l 28(sp),a0
move.l 32(sp),a1
#APP
| Inlined umul_ppmm
move.l a0,d0
move.l a1,d1
move.l d0,d2
swap d0
move.l d1,d3
swap d1
move.w d2,d4
mulu d3,d4
mulu d1,d2
mulu d0,d3
mulu d0,d1
move.l d4,d0
eor.w d0,d0
swap d0
add.l d0,d2
add.l d3,d2
jcc 1f
add.l #65536,d1
1: swap d2
moveq #0,d0
move.w d2,d0
move.w d4,d2
move.l d2,d6
add.l d1,d0
move.l d0,d5
#NO_APP
tst.l a0
jlt L6
tst.l a1
jlt L7
L3:
move.l d5,d0
clr.l d1
move.l d0,d2
move.l d1,d3
or.l d6,d3
move.l d2,d6
move.l d3,d7
move.l d2,d0
move.l d7,d1
movem.l (sp)+,#252
rts
L7:
sub.l a0,d5
move.l d5,d0
clr.l d1
move.l d0,d2
move.l d1,d3
or.l d6,d3
move.l d2,d6
move.l d3,d7
move.l d2,d0
move.l d7,d1
movem.l (sp)+,#252
rts
L6:
sub.l a1,d5
tst.l a1
jge L3
jra L7
The asm output is not good (not optimized), because GCC 4.4.1 have problems
with this part of the code:
res = ((uint64_t)resh << 32) | resl;
GCC 3.4.0 generates better asm output (-m68060 -O3 -fomit-frame-ppointer):
#NO_APP
.text
.even
.globl _MUL64
_MUL64:
moveml #0x3e00,s...@-
movel sp@(24),a1
movel sp@(28),a0
#APP
| Inlined umul_ppmm
movel a1,d0
movel a0,d1
movel d0,d2
swap d0
movel d1,d3
swap d1
movew d2,d4
mulu d3,d4
mulu d1,d2
mulu d0,d3
mulu d0,d1
movel d4,d0
eorw d0,d0
swap d0
addl d0,d2
addl d3,d2
jcc 1f
addl #65536,d1
1: swap d2
moveq #0,d0
movew d2,d0
movew d4,d2
movel d2,d6
addl d1,d0
movel d0,d5
#NO_APP
tstl a1
jlt L5
tstl a0
jge L3
jra L6
.even
L5:
subl a0,d5
tstl a0
jge L3
.even
L6:
subl a1,d5
.even
L3:
movel d5,d0
clrl d1
orl d6,d1
moveml s...@+,#0x7c
rts
For GCC 4.4.1 the best asm code is generaded from this C code:
typedef unsigned int uint32_t;
typedef unsigned long long uint64_t;
#define umul_ppmm(xh, xl, a, b) \
__asm__ ("| Inlined umul_ppmm\n" \
" move%.l %2,%/d5\n" \
" move%.l %3,%/d6\n" \
" move%.l %/d5,%/d2\n" \
" swap %/d5\n" \
" move%.l %/d6,%/d3\n" \
" swap %/d6\n" \
" move%.w %/d2,%/d4\n" \
" mulu %/d3,%/d4\n" \
" mulu %/d6,%/d2\n" \
" mulu %/d5,%/d3\n" \
" mulu %/d5,%/d6\n" \
" move%.l %/d4,%/d5\n" \
" eor%.w %/d5,%/d5\n" \
" swap %/d5\n" \
" add%.l %/d5,%/d2\n" \
" add%.l %/d3,%/d2\n" \
" jcc 1f\n" \
" add%.l %#65536,%/d6\n" \
"1: swap %/d2\n" \
" moveq %#0,%/d5\n" \
" move%.w %/d2,%/d5\n" \
" move%.w %/d4,%/d2\n" \
" move%.l %/d2,%1\n" \
" add%.l %/d6,%/d5\n" \
" move%.l %/d5,%0" \
: "=g" ((uint32_t) (xh)), \
"=g" ((uint32_t) (xl)) \
: "g" ((uint32_t) (a)), \
"g" ((uint32_t) (b)) \
: "d2", "d3", "d4", "d5", "d6")
inline uint64_t MUL64(int a, int b)
{
uint32_t au = a;
uint32_t bu = b;
union
{
struct
{
uint32_t h, l;
} parts;
uint64_t whole;
} res;
umul_ppmm(res.parts.h, res.parts.l, au, bu);
if (a < 0)
res.parts.h -= bu;
if (b < 0)
res.parts.h -= au;
return res.whole;
}
Asm output:
#NO_APP
.text
.even
.globl _MUL64
_xxMULH:
movem.l #15872,-(sp)
move.l 24(sp),a0
move.l 28(sp),a1
#APP
| Inlined umul_ppmm
move.l a0,d5
move.l a1,d6
move.l d5,d2
swap d5
move.l d6,d3
swap d6
move.w d2,d4
mulu d3,d4
mulu d6,d2
mulu d5,d3
mulu d5,d6
move.l d4,d5
eor.w d5,d5
swap d5
add.l d5,d2
add.l d3,d2
jcc 1f
add.l #65536,d6
1: swap d2
moveq #0,d5
move.w d2,d5
move.w d4,d2
move.l d2,d1
add.l d6,d5
move.l d5,d0
#NO_APP
tst.l a0
jlt L6
tst.l a1
jlt L7
L3:
movem.l (sp)+,#124
rts
L7:
sub.l a0,d0
movem.l (sp)+,#124
rts
L6:
sub.l a1,d0
tst.l a1
jge L3
Maybe someone will teach GCC 4.4.x to use optimized code with:
res = ((uint64_t)resh << 32) | resl;
?
Regards
--
Summary: Problem with code like this: res = ((uint64_t)resh <<
32) | resl;
Product: gcc
Version: 4.4.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: ami_stuff at o2 dot pl
GCC host triplet: i686-cygwin
GCC target triplet: m68k-amigaos
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40977
