From: "Adrian Francis" <adrian DOT francis AT cableinet DOT co DOT uk>
Subject: Re: Speed Test - TC vs. DJGPP
Newsgroups: comp.os.msdos.djgpp
References: <01bd492f$47fb8b40$929175c2 AT ctx> <34FFF92A DOT 3ECF AT cam DOT org>
Message-ID: <01bd4945$71bb18e0$8c9175c2@ctx>
NNTP-Posting-Host: usr130-wol.cableinet.co.uk
Date: 6 Mar 98 21:32:26 GMT
Organization: "Cable Internet (post doesn't reflect views of Cable Internet)"
Lines: 264
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

Vic <tudor AT cam DOT org> wrote in article <34FFF92A DOT 3ECF AT cam DOT org>...
> Adrian Francis wrote:
> > DJGPP -         tics:13345      sec:146
> > TC -            tics:5          sec:0
> > 
> > Can someone explain why this is happening?  I would expect DJGPP to be
> > twice as fast considering it is a 32bit compiler.  Something must be
>wrong!
> indeed, something is wrong. 146 seconds??
> Without optimising anything, I get 10 tics. I get 5 tics with
> optimisations on your program (O3)
> -- 
> --> http://www.cam.org/~tudor <--
> 

I get 5 tics with -O3 but I still have 13345 without optimizations!  If you
got 10 tics without, that means I must have a serious problem somewhere.

Maybe someone can work it out from the assembler dump below. I would
appreciate it.  I can make sense out of most of it, but I don't know what's
going on in the _sub function.  What's _mcount doing?

Thanks!

-ADRIAN-

	.file	"speed.c"
gcc2_compiled.:
___gnu_compiled_c:
	.space 4
.text
	.def	_clock_t;	.scl	13;	.type	04;	.endef
	.def	_size_t;	.scl	13;	.type	017;	.endef
	.def	_time_t;	.scl	13;	.type	016;	.endef
	.def	_tm;	.scl	10;	.type	010;	.size	44;	.endef
	.def	_tm_sec;	.val	0;	.scl	8;	.type	04;	.endef
	.def	_tm_min;	.val	4;	.scl	8;	.type	04;	.endef
	.def	_tm_hour;	.val	8;	.scl	8;	.type	04;	.endef
	.def	_tm_mday;	.val	12;	.scl	8;	.type	04;	.endef
	.def	_tm_mon;	.val	16;	.scl	8;	.type	04;	.endef
	.def	_tm_year;	.val	20;	.scl	8;	.type	04;	.endef
	.def	_tm_wday;	.val	24;	.scl	8;	.type	04;	.endef
	.def	_tm_yday;	.val	28;	.scl	8;	.type	04;	.endef
	.def	_tm_isdst;	.val	32;	.scl	8;	.type	04;	.endef
	.def	___tm_zone;	.val	36;	.scl	8;	.type	022;	.endef
	.def	___tm_gmtoff;	.val	40;	.scl	8;	.type	04;	.endef
	.def	.eos;	.val	44;	.scl	102;	.tag	_tm;	.size	44;	.endef
	.def	_timeval;	.scl	10;	.type	010;	.size	8;	.endef
	.def	_tv_sec;	.val	0;	.scl	8;	.type	016;	.endef
	.def	_tv_usec;	.val	4;	.scl	8;	.type	05;	.endef
	.def	.eos;	.val	8;	.scl	102;	.tag	_timeval;	.size	8;	.endef
	.def	_timezone;	.scl	10;	.type	010;	.size	8;	.endef
	.def	_tz_minuteswest;	.val	0;	.scl	8;	.type	04;	.endef
	.def	_tz_dsttime;	.val	4;	.scl	8;	.type	04;	.endef
	.def	.eos;	.val	8;	.scl	102;	.tag	_timezone;	.size	8;	.endef
	.def	_dev_t;	.scl	13;	.type	04;	.endef
	.def	_ino_t;	.scl	13;	.type	04;	.endef
	.def	_mode_t;	.scl	13;	.type	04;	.endef
	.def	_nlink_t;	.scl	13;	.type	04;	.endef
	.def	_gid_t;	.scl	13;	.type	04;	.endef
	.def	_off_t;	.scl	13;	.type	04;	.endef
	.def	_pid_t;	.scl	13;	.type	04;	.endef
	.def	_ssize_t;	.scl	13;	.type	04;	.endef
	.def	_uid_t;	.scl	13;	.type	04;	.endef
	.def	_fd_set;	.scl	10;	.type	010;	.size	32;	.endef
	.def	_fd_bits;	.val	0;	.scl	8;	.dim	32;	.size	32;	.type	074;	.endef
	.def	.eos;	.val	32;	.scl	102;	.tag	_fd_set;	.size	32;	.endef
	.def	_fd_set;	.scl	13;	.tag	_fd_set;	.size	32;	.type	010;	.endef
	.def	_uclock_t;	.scl	13;	.type	05;	.endef
	.align 4
	.def	_sub;	.val	_sub;	.scl	2;	.type	041;	.endef
.globl _sub
_sub:
	.def	.bf;	.val	.;	.scl	101;	.line	4;	.endef
	pushl %ebp
	movl %esp,%ebp
	cmpl $0,LPBX0
	jne LPBZ0
	pushl $LPBX0
	call ___bb_init_func
	addl $4,%esp
LPBZ0:
.data
	.align 2
LP0:
	.long 0
.text
	movl $LP0,%edx
	call _mcount
	.ln	2
L1:
LPB0:
	incl LPBX2
	.def	.ef;	.val	.;	.scl	101;	.line	2;	.endef
	movl %ebp,%esp
	popl %ebp
	ret
	.def	_sub;	.val	.;	.scl	-1;	.endef
LC0:
	.ascii "tics:%d\0"
LC1:
	.ascii "sec:%d\0"
	.align 4
	.def	_main;	.val	_main;	.scl	2;	.type	041;	.endef
.globl _main
_main:
	.def	.bf;	.val	.;	.scl	101;	.line	8;	.endef
	pushl %ebp
	movl %esp,%ebp
	subl $16,%esp
	pushl %ebx
	cmpl $0,LPBX0
	jne LPBZ1
	pushl $LPBX0
	call ___bb_init_func
	addl $4,%esp
LPBZ1:
.data
	.align 2
LP1:
	.long 0
.text
	movl $LP1,%edx
	call _mcount
LPB1:
	incl LPBX2+4
	call ___main
	.ln	2
	.def	_start;	.val	-4;	.scl	1;	.type	04;	.endef
	.def	_end;	.val	-8;	.scl	1;	.type	04;	.endef
	.def	_n;	.val	-12;	.scl	1;	.type	05;	.endef
	.def	_time;	.val	-16;	.scl	1;	.type	05;	.endef
	.ln	5
	call _clrscr
	.ln	6
	call _clock
	movl %eax,%eax
	movl %eax,-4(%ebp)
	.ln	7
	movl $0,-12(%ebp)
L3:
LPB2:
	incl LPBX2+8
	cmpl $999999,-12(%ebp)
	jle L6
LPB3:
	incl LPBX2+12
	jmp L4
	.align 4,0x90
L6:
	.ln	9
LPB4:
	incl LPBX2+16
	call _sub
	.ln	7
L5:
LPB5:
	incl LPBX2+20
	incl -12(%ebp)
	jmp L3
	.align 4,0x90
L4:
	.ln	11
LPB6:
	incl LPBX2+24
	call _clock
	movl %eax,%eax
	movl %eax,-8(%ebp)
	.ln	12
	movl -8(%ebp),%ebx
	subl -4(%ebp),%ebx
	movl %ebx,-16(%ebp)
	.ln	13
	movl -16(%ebp),%eax
	pushl %eax
	pushl $LC0
	call _printf
	addl $8,%esp
	.ln	14
	movl -16(%ebp),%ecx
	movl %ecx,%eax
	movl $91,%ebx
	cltd
	idivl %ebx
	movl %eax,-16(%ebp)
	.ln	15
	movl -16(%ebp),%eax
	pushl %eax
	pushl $LC1
	call _printf
	addl $8,%esp
	.ln	16
L2:
LPB7:
	incl LPBX2+28
	.def	.ef;	.val	.;	.scl	101;	.line	16;	.endef
	movl -20(%ebp),%ebx
	movl %ebp,%esp
	popl %ebp
	ret
	.def	_main;	.val	.;	.scl	-1;	.endef
.data
	.align 2
LPBX0:
	.long 0
	.long LPBX1
	.long LPBX2
	.long 8
	.long 0
	.long LPBX3
	.long 40
	.long LPBX4
	.long LPBX5
	.long LPBX6
LPBX1:
	.ascii "c:/djgpp/bin/../progs/speed.d\0"
.lcomm LPBX2,32
.text
	.align 2
LPBC0:
	.ascii "sub\0"
LPBC1:
	.ascii "../progs/speed.c\0"
LPBC2:
	.ascii "main\0"
	.align 2
LPBX3:
	.long LPB0
	.long LPB1
	.long LPB2
	.long LPB3
	.long LPB4
	.long LPB5
	.long LPB6
	.long LPB7
LPBX4:
	.long LPBC0
	.long LPBC2
	.long LPBC2
	.long LPBC2
	.long LPBC2
	.long LPBC2
	.long LPBC2
	.long LPBC2
LPBX5:
	.long 5
	.long 8
	.long 14
	.long 14
	.long 16
	.long 14
	.long 18
	.long 23
LPBX6:
	.long LPBC1
	.long LPBC1
	.long LPBC1
	.long LPBC1
	.long LPBC1
	.long LPBC1
	.long LPBC1
	.long LPBC1
	.long LPBX3