Update on the mspgcc version: $ msp430-gcc --version msp430-gcc (GCC) 4.5.3 20110428 (msp430) LTS_20110716_0813
> -----Original Message----- > From: Michiel Konstapel [mailto:m.konsta...@sownet.nl] > Sent: woensdag 14 september 2011 14:38 > To: mspgcc-users@lists.sourceforge.net > Subject: [Mspgcc-users] uint64_t issue? > > I think I've run into a compiler bug, but I'm not sure how to pin it > down. > > I'm using the TinyOS RandomMlcg random generator: > > /* Return the next 32 bit random number */ > async command uint32_t Random.rand32() { > uint32_t mlcg,p,q; > uint64_t tmpseed; > atomic > { > tmpseed = (uint64_t)33614U * (uint64_t)seed; > q = tmpseed; /* low */ > q = q >> 1; > p = tmpseed >> 32 ; /* hi */ > mlcg = p + q; > if (mlcg & 0x80000000) { > mlcg = mlcg & 0x7FFFFFFF; > mlcg++; > } > seed = mlcg; > } > return mlcg; > } > > Now, in my program, I have a function that takes one argument. After > the > call to Random.rand16(), this argument is overwritten by the value > 0x834e - which is 33614U. This happens with mspgcc4 (20110813) but not > with 3.2.3. My trivial test program failed to reproduce this; I'll see > if I can find a way to trigger it. > > The context in which it's called (noinlined so I can find it in the > disassembly): > > void backoff(uint16_t fixed) __attribute__((noinline)) { > uint16_t delay = 0; > printf("fixed=%u\n", fixed); // here, fixed is > correct > delay = call Random.rand16(); > printf("fixed=%u\n", fixed); // here, fixed == 0x834e > ... > > For reference, the disassembly of the working version (3.2.3) followed > by the broken one, up to the second call to printf: > > 3.2.3: > 00005204 <SendP__backoff>: > 5204: 1b 15 .word 0x151b; ???? > 5206: 0a 4f mov r15, r10 > 5208: 0f 12 push r15 > 520a: 30 12 fa 51 push #20986 ;#0x51fa > 520e: b0 13 .word 0x13b0; ???? > 5210: 6a 7d subc.b @r13, r10 > 5212: 21 52 add #4, r1 ;r2 As==10 > 5214: b0 13 .word 0x13b0; ???? > 5216: 5e 52 0b 4e add.b &0x4e0b,r14 > 521a: 0a 12 push r10 > 521c: 30 12 fa 51 push #20986 ;#0x51fa > 5220: b0 13 .word 0x13b0; ???? > 5222: 6a 7d subc.b @r13, r10 > 5224: 21 52 add #4, r1 ;r2 As==10 > 5226: 0a 4b mov r11, r10 > 5228: 0b 43 clr r11 > 522a: b0 13 .word 0x13b0; ???? > 522c: f6 34 jge $+494 ;abs 0x541a > 522e: 53 12 push.b #1 ;r3 As==01 > 5230: 0b 12 push r11 > 5232: 0a 12 push r10 > 5234: 0d 4e mov r14, r13 > 5236: 0e 4f mov r15, r14 > 5238: 7f 40 03 00 mov.b #3, r15 ;#0x0003 > 523c: b0 13 .word 0x13b0; ???? > 523e: 4a 42 mov.b r2, r10 > 5240: 31 50 06 00 add #6, r1 ;#0x0006 > 5244: 1f 42 70 14 mov &0x1470,r15 > 5248: 0f 5f rla r15 > 524a: 82 4f 70 14 mov r15, &0x1470 > 524e: 3f 90 11 00 cmp #17, r15 ;#0x0011 > 5252: 03 28 jnc $+8 ;abs 0x525a > 5254: b2 40 10 00 mov #16, &0x1470 ;#0x0010 > 5258: 70 14 > 525a: 1a 17 .word 0x171a; ???? > 525c: 10 01 .word 0x0110; ???? > > 0000525e <RandomMlcgC__Random__rand32>: > 525e: 4b 15 .word 0x154b; ???? > 5260: b0 13 .word 0x13b0; ???? > 5262: 8a 31 jn $+790 ;abs 0x5578 > 5264: 47 4f mov.b r15, r7 > 5266: 1c 42 c8 13 mov &0x13c8,r12 > 526a: 1d 42 ca 13 mov &0x13ca,r13 > 526e: 0e 43 clr r14 > 5270: 0f 43 clr r15 > 5272: 03 12 push #0 ;r3 As==00 > 5274: 03 12 push #0 ;r3 As==00 > 5276: 03 12 push #0 ;r3 As==00 > 5278: 30 12 4e 83 push #-31922 ;#0x834e > 527c: b0 13 .word 0x13b0; ???? > 527e: f6 86 31 52 sub.b @r6+, 21041(r6);0x5231(r6) > 5282: 0a 4c mov r12, r10 > 5284: 0b 4d mov r13, r11 > 5286: 12 c3 clrc > 5288: 0b 10 rrc r11 > 528a: 0a 10 rrc r10 > 528c: 0c 4e mov r14, r12 > 528e: 0d 4f mov r15, r13 > 5290: 0e 43 clr r14 > 5292: 0f 43 clr r15 > 5294: 08 4c mov r12, r8 > 5296: 09 4d mov r13, r9 > 5298: 08 5a add r10, r8 > 529a: 09 6b addc r11, r9 > 529c: 39 b0 00 80 bit #-32768,r9 ;#0x8000 > 52a0: 04 24 jz $+10 ;abs 0x52aa > 52a2: 39 f0 ff 7f and #32767, r9 ;#0x7fff > 52a6: 18 53 inc r8 > 52a8: 09 63 adc r9 > 52aa: 82 48 c8 13 mov r8, &0x13c8 > 52ae: 82 49 ca 13 mov r9, &0x13ca > 52b2: 4f 47 mov.b r7, r15 > 52b4: b0 13 .word 0x13b0; ???? > 52b6: 9e 31 jn $+830 ;abs 0x55f4 > 52b8: 0e 48 mov r8, r14 > 52ba: 0f 49 mov r9, r15 > 52bc: 47 17 .word 0x1747; ???? > 52be: 10 01 .word 0x0110; ???? > > > 20110813 inlines the call: > > 00004c90 <SendP__backoff>: > 4c90: 0b 12 push r11 > 4c92: 0a 12 push r10 > 4c94: 09 12 push r9 > 4c96: 08 12 push r8 > 4c98: 07 12 push r7 > 4c9a: 06 12 push r6 > 4c9c: 05 12 push r5 > 4c9e: 04 12 push r4 > 4ca0: 21 83 decd r1 > 4ca2: 0f 12 push r15 > 4ca4: 0f 12 push r15 > 4ca6: 30 12 80 4c push #19584 ;#0x4c80 > 4caa: b0 12 b8 76 call #0x76b8 > 4cae: 31 50 06 00 add #6, r1 ;#0x0006 > 4cb2: b0 12 06 3a call #0x3a06 > 4cb6: c1 4f 00 00 mov.b r15, 0(r1) ;0x0000(r1) > 4cba: 18 42 cc 13 mov &0x13cc,r8 > 4cbe: 19 42 ce 13 mov &0x13ce,r9 > 4cc2: 34 40 4e 83 mov #-31922,r4 ;#0x834e > 4cc6: 05 43 clr r5 > 4cc8: b0 12 36 7c call #0x7c36 > 4ccc: 0a 4c mov r12, r10 > 4cce: 0b 4d mov r13, r11 > 4cd0: 12 c3 clrc > 4cd2: 0b 10 rrc r11 > 4cd4: 0a 10 rrc r10 > 4cd6: 0a 5e add r14, r10 > 4cd8: 0b 6f addc r15, r11 > 4cda: 0b 93 tst r11 > 4cdc: 05 34 jge $+12 ;abs 0x4ce8 > 4cde: 3a f3 and #-1, r10 ;r3 As==11 > 4ce0: 3b f0 ff 7f and #32767, r11 ;#0x7fff > 4ce4: 1a 53 inc r10 > 4ce6: 0b 63 adc r11 > 4ce8: 82 4a cc 13 mov r10, &0x13cc > 4cec: 82 4b ce 13 mov r11, &0x13ce > 4cf0: 6f 41 mov.b @r1, r15 > 4cf2: b0 12 1a 3a call #0x3a1a > 4cf6: 04 12 push r4 > 4cf8: 04 12 push r4 > 4cfa: 30 12 80 4c push #19584 ;#0x4c80 > 4cfe: b0 12 b8 76 call #0x76b8 > > I'm afraid my assembly-fu isn't very strong, but I hope someone can > glean... something from it. > > Best, > Michiel > > > ----------------------------------------------------------------------- > ------- > BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA > Learn about the latest advances in developing for the > BlackBerry® mobile platform with sessions, labs & more. > See new tools and technologies. Register for BlackBerry® DevCon > today! > http://p.sf.net/sfu/rim-devcon-copy1 > _______________________________________________ > Mspgcc-users mailing list > Mspgcc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/mspgcc-users ------------------------------------------------------------------------------ BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA Learn about the latest advances in developing for the BlackBerry® mobile platform with sessions, labs & more. See new tools and technologies. Register for BlackBerry® DevCon today! http://p.sf.net/sfu/rim-devcon-copy1 _______________________________________________ Mspgcc-users mailing list Mspgcc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mspgcc-users
