I've tried that. No change. When using -Os or -O2 (or I think even -O3) loop unrolling isn't turned on (according to what I've found by looking in gcc-3.2.3/gcc/toplev.c)
-Chris Takahashi -----Original Message----- From: Dmitry [mailto:di...@mail.ru] Sent: Wednesday, August 27, 2003 3:13 AM To: mspgcc-users@lists.sourceforge.net Subject: Re: [Mspgcc-users] partial loop unrolling. use -fno-unroll-loops ~d On Wednesday 27 August 2003 03:01, Takahashi, Chris wrote: > The following code creates some interesting asm with the following flags: > -g -c -Os -fno-force-mem -mmcu=msp430x437 > > It looks like it unrolled the loop in 64 byte chunks and is instead > iterating over that. While this is faster this would hardly ever get run > so I would rather have it small. > > attached are list files with the -fno-force-mem on and off and the C file. > > If I change the span it iterates over to a nonmultiple of 64 I get the > expected asm. > > My question is what does no-force-mem have to do with this and why isn't > auto increment addressing being used? In this case it would save quite a > bit here. > > -Chris > > --- > > #include <msp430x44x.h> > > int test() > { > unsigned int sum = 0; > unsigned int *addr; > > for (addr = (int *)0x8000; addr != 0x0000; addr++) { > sum += *addr; > } > > return sum; > } > > ------- > > for (addr = (int *)0x8000; addr != 0x0000; addr++) {^M > 2: 3e 40 00 80 mov #-32768,r14 ;#0x8000 > sum += *addr;^M > 6: 2f 5e add @r14, r15 ; > 8: 1f 5e 02 00 add 2(r14), r15 ; > c: 1f 5e 04 00 add 4(r14), r15 ; > 10: 1f 5e 06 00 add 6(r14), r15 ; > 14: 1f 5e 08 00 add 8(r14), r15 ; > 18: 1f 5e 0a 00 add 10(r14),r15 ; > 1c: 1f 5e 0c 00 add 12(r14),r15 ; > 20: 1f 5e 0e 00 add 14(r14),r15 ; > 24: 1f 5e 10 00 add 16(r14),r15 ; > 28: 1f 5e 12 00 add 18(r14),r15 ; > 2c: 1f 5e 14 00 add 20(r14),r15 ; > 30: 1f 5e 16 00 add 22(r14),r15 ; > 34: 1f 5e 18 00 add 24(r14),r15 ; > 38: 1f 5e 1a 00 add 26(r14),r15 ; > 3c: 1f 5e 1c 00 add 28(r14),r15 ; > 40: 1f 5e 1e 00 add 30(r14),r15 ; > 44: 1f 5e 20 00 add 32(r14),r15 ; > 48: 1f 5e 22 00 add 34(r14),r15 ; > 4c: 1f 5e 24 00 add 36(r14),r15 ; > 50: 1f 5e 26 00 add 38(r14),r15 ; > 54: 1f 5e 28 00 add 40(r14),r15 ; > 58: 1f 5e 2a 00 add 42(r14),r15 ; > 5c: 1f 5e 2c 00 add 44(r14),r15 ; > 60: 1f 5e 2e 00 add 46(r14),r15 ; > 64: 1f 5e 30 00 add 48(r14),r15 ; > 68: 1f 5e 32 00 add 50(r14),r15 ; > 6c: 1f 5e 34 00 add 52(r14),r15 ; > 70: 1f 5e 36 00 add 54(r14),r15 ; > 74: 1f 5e 38 00 add 56(r14),r15 ; > 78: 1f 5e 3a 00 add 58(r14),r15 ; > 7c: 1f 5e 3c 00 add 60(r14),r15 ; > 80: 1f 5e 3e 00 add 62(r14),r15 ; > 84: 3e 50 40 00 add #64, r14 ;#0x0040 > 88: be 23 jnz $-130 ;abs 0x6 > } -- /***************************************************************** ("`-''-/").___..--''"`-._ (\ Dimmy the Wild UA1ACZ `6_ 6 ) `-. ( ).`-.__.`) State Polytechnical Univ. (_Y_.)' ._ ) `._ `. ``-..-' Radio-Physics Departament _..`--'_..-_/ /--'_.' ,' Saint Petersburg, Russia (il),-'' (li),' ((!.-' +7 (812) 5403923, 5585314 *****************************************************************/ ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Mspgcc-users mailing list Mspgcc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mspgcc-users
