* [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 @ 2003-05-20 16:17 Andrew Pinski 2003-05-20 20:59 ` Andrew Pinski 2003-06-12 15:22 ` Diego Novillo 0 siblings, 2 replies; 9+ messages in thread From: Andrew Pinski @ 2003-05-20 16:17 UTC (permalink / raw) To: gcc; +Cc: Andrew Pinski Fails on stage3 compiling libgcc testcase: #define TEMP (!0?4:8) #if TEMP > 4 int temp1 = TEMP; #else int temp2 = TEMP; #endif Should output temp2, not temp1. stage1/2 and mainline of gcc: [zhivago2:objdir/gcc/stage1] pinskia% !./ ./cc1 test1.c -E -o - # 1 "test1.c" # 1 "<built-in>" # 1 "<command line>" # 1 "test1.c" # 10 "test1.c" int temp2 = (!0?4:8); Execution times (seconds) TOTAL : 0.05 0.00 0.28 stage3: ./cc1 test1.c -E -o - # 1 "test1.c" # 1 "<built-in>" # 1 "<command line>" # 1 "test1.c" int temp1 = (!0?4:8); Execution times (seconds) TOTAL : 0.02 0.00 0.03 Thanks, Andrew Pinski ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-05-20 16:17 [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 Andrew Pinski @ 2003-05-20 20:59 ` Andrew Pinski 2003-05-20 21:16 ` Andrew Pinski 2003-06-12 15:22 ` Diego Novillo 1 sibling, 1 reply; 9+ messages in thread From: Andrew Pinski @ 2003-05-20 20:59 UTC (permalink / raw) To: Andrew Pinski; +Cc: gcc Looks like this is from a mis compiling of cppexp.c. On Tuesday, May 20, 2003, at 11:17 US/Eastern, Andrew Pinski wrote: > Fails on stage3 compiling libgcc > > testcase: > > #define TEMP (!0?4:8) > #if TEMP > 4 > int temp1 = TEMP; > #else > int temp2 = TEMP; > #endif > > > Should output temp2, not temp1. > > stage1/2 and mainline of gcc: > [zhivago2:objdir/gcc/stage1] pinskia% !./ > ./cc1 test1.c -E -o - > # 1 "test1.c" > # 1 "<built-in>" > # 1 "<command line>" > # 1 "test1.c" > # 10 "test1.c" > int temp2 = (!0?4:8); > > Execution times (seconds) > TOTAL : 0.05 0.00 0.28 > > > > stage3: > ./cc1 test1.c -E -o - > # 1 "test1.c" > # 1 "<built-in>" > # 1 "<command line>" > # 1 "test1.c" > > > > > > int temp1 = (!0?4:8); > > Execution times (seconds) > TOTAL : 0.02 0.00 0.03 > > Thanks, > Andrew Pinski > > > ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-05-20 20:59 ` Andrew Pinski @ 2003-05-20 21:16 ` Andrew Pinski 2003-05-21 16:07 ` Andrew Pinski 0 siblings, 1 reply; 9+ messages in thread From: Andrew Pinski @ 2003-05-20 21:16 UTC (permalink / raw) To: Andrew Pinski; +Cc: gcc Ok, there are two functions that differ in cppexp.c: num_bitwise_op num_inequality <-- this is the one that is causing this failure, most likely There are two functions from resource.c that differ also: find_dead_or_set_register mark_set_resources Thanks, Andrew Pinski On Tuesday, May 20, 2003, at 16:51 US/Eastern, Andrew Pinski wrote: > Looks like this is from a mis compiling of cppexp.c. > > On Tuesday, May 20, 2003, at 11:17 US/Eastern, Andrew Pinski wrote: > >> Fails on stage3 compiling libgcc >> >> testcase: >> >> #define TEMP (!0?4:8) >> #if TEMP > 4 >> int temp1 = TEMP; >> #else >> int temp2 = TEMP; >> #endif >> >> >> Should output temp2, not temp1. >> >> stage1/2 and mainline of gcc: >> [zhivago2:objdir/gcc/stage1] pinskia% !./ >> ./cc1 test1.c -E -o - >> # 1 "test1.c" >> # 1 "<built-in>" >> # 1 "<command line>" >> # 1 "test1.c" >> # 10 "test1.c" >> int temp2 = (!0?4:8); >> >> Execution times (seconds) >> TOTAL : 0.05 0.00 0.28 >> >> >> >> stage3: >> ./cc1 test1.c -E -o - >> # 1 "test1.c" >> # 1 "<built-in>" >> # 1 "<command line>" >> # 1 "test1.c" >> >> >> >> >> >> int temp1 = (!0?4:8); >> >> Execution times (seconds) >> TOTAL : 0.02 0.00 0.03 >> >> Thanks, >> Andrew Pinski >> >> >> > > > ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-05-20 21:16 ` Andrew Pinski @ 2003-05-21 16:07 ` Andrew Pinski 2003-05-21 16:17 ` Andrew Pinski 0 siblings, 1 reply; 9+ messages in thread From: Andrew Pinski @ 2003-05-21 16:07 UTC (permalink / raw) To: Andrew Pinski; +Cc: gcc Here is the diff of the two asm (of cppexp.c where the problem is found), temp is the one compiled with stage2, and temp1 is the one compiled with stage1: --- temp/cppexp.s Wed May 21 02:07:54 2003 +++ temp1/cppexp.s Wed May 21 02:08:40 2003 @@ -3648,12 +3648,11 @@ stw r7,40(r1) stw r9,48(r1) stw r10,52(r1) - lwz r11,64(r1) + lwz r11,56(r1) .stabd 68,0,1140 bne- cr7,L597 - lwz r0,56(r1) - addic r7,r0,-1 - subfe r2,r7,r0 + addic r7,r11,-1 + subfe r2,r7,r11 L597: .stabd 68,0,1144 cmpwi cr7,r11,9 @@ -3722,7 +3721,7 @@ .stabs "pfile:P(1,2)",64,0,1135,4 .stabs "lhs:p(60,56)",160,0,1136,32 .stabs "rhs:p(60,56)",160,0,1136,48 - .stabs "op:p(60,18)",160,0,1137,64 + .stabs "op:p(60,18)",160,0,1137,56 .stabs "op:r(60,18)",64,0,1137,11 .align 2 _num_inequality_op: @@ -3748,16 +3747,14 @@ stw r8,140(r1) stw r9,144(r1) stw r10,148(r1) + lwz r30,152(r1) .stabd 68,0,1170 lwz r0,440(r4) mr r4,r6 - .stabd 68,0,1169 - lwz r30,160(r1) - .stabd 68,0,1170 mr r6,r8 - lwz r9,152(r1) mr r8,r10 lwz r10,156(r1) + mr r9,r30 stw r0,56(r1) bl _num_greater_eq .stabd 68,0,1172 @@ -3873,7 +3870,7 @@ .stabs "pfile:P(1,2)",64,0,1166,4 .stabs "lhs:p(60,56)",160,0,1167,128 .stabs "rhs:p(60,56)",160,0,1167,144 - .stabs "op:p(60,18)",160,0,1168,160 + .stabs "op:p(60,18)",160,0,1168,152 .stabs "gte:r(0,20)",64,0,1170,3 .stabs "op:r(60,18)",64,0,1168,30 .align 2 @@ -3896,7 +3893,7 @@ .stabd 68,0,1195 bne- cr0,L630 .stabd 68,0,1196 - lwz r0,64(r1) + lwz r0,56(r1) .stabd 68,0,1195 mr r5,r4 .stabd 68,0,1196 @@ -3922,7 +3919,7 @@ subfic r0,r4,0 adde r4,r0,r4 .stabd 68,0,1196 - lwz r0,64(r1) + lwz r0,56(r1) .stabd 68,0,1195 mr r5,r4 .stabd 68,0,1196 @@ -3949,7 +3946,7 @@ .stabs "pfile:P(1,2)",64,0,1190,4 .stabs "lhs:p(60,56)",160,0,1191,32 .stabs "rhs:p(60,56)",160,0,1191,48 - .stabs "op:p(60,18)",160,0,1192,64 + .stabs "op:p(60,18)",160,0,1192,56 .stabs "eq:r(0,20)",64,0,1195,5 .stabs "op:r(60,18)",64,0,1192,0 .align 2 @@ -4448,7 +4445,7 @@ mr r29,r6 stwu r1,-128(r1) mr r11,r7 - lwz r30,192(r1) + lwz r30,184(r1) mr r28,r8 stw r9,176(r1) .stabd 68,0,1406 @@ -4817,7 +4814,7 @@ .stabs "pfile:P(1,2)",64,0,1330,4 .stabs "lhs:p(60,56)",160,0,1331,160 .stabs "rhs:p(60,56)",160,0,1331,176 - .stabs "op:p(60,18)",160,0,1332,192 + .stabs "op:p(60,18)",160,0,1332,184 .stabs "result:(60,56)",128,0,1334,80 .stabs "precision:r(7,2)",64,0,1335,26 .stabs "gte:r(0,20)",64,0,1336,3 @@ -5373,7 +5370,7 @@ bne+ cr7,L826 L808: .stabd 68,0,1572 - lwz r0,224(r1) + lwz r0,216(r1) cmpwi cr7,r0,7 beq- cr7,L827 .stabd 68,0,1590 @@ -5551,7 +5548,7 @@ .stabs "pfile:P(1,2)",64,0,1498,28 .stabs "lhs:p(60,56)",160,0,1499,192 .stabs "rhs:p(60,56)",160,0,1499,208 - .stabs "op:p(60,18)",160,0,1500,224 + .stabs "op:p(60,18)",160,0,1500,216 .stabs "result:(60,56)",128,0,1502,80 .stabs "sub:(60,56)",128,0,1502,96 .stabs "mask:r(60,55)",64,0,1503,0 As you can see that it is really only loads that are different, why I do not know, can someone help me here, I have the rtl dumps, as the tree dumps are the same (minus different names for the labels). Thanks, Andrew Pinski ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-05-21 16:07 ` Andrew Pinski @ 2003-05-21 16:17 ` Andrew Pinski 0 siblings, 0 replies; 9+ messages in thread From: Andrew Pinski @ 2003-05-21 16:17 UTC (permalink / raw) To: Andrew Pinski; +Cc: gcc The rtl right after it gcc converts simple to rtl differers for some reason (- is the correct one): (insn 11 10 12 0x0 (set (reg/v:SI 120 [ op ]) (mem/f:SI (plus:SI (reg/f:SI 113 virtual-incoming-args) - (const_int 40 [0x28])) [61 op+0 S4 A32])) -1 (nil) + (const_int 32 [0x20])) [61 op+0 S4 A32])) -1 (nil) (expr_list:REG_EQUIV (mem/f:SI (plus:SI (reg/f:SI 113 virtual-incoming-args) - (const_int 40 [0x28])) [61 op+0 S4 A32]) + (const_int 32 [0x20])) [61 op+0 S4 A32]) (nil))) Thanks, Andrew Pinski ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-05-20 16:17 [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 Andrew Pinski 2003-05-20 20:59 ` Andrew Pinski @ 2003-06-12 15:22 ` Diego Novillo 2003-06-12 15:42 ` Andrew Pinski 1 sibling, 1 reply; 9+ messages in thread From: Diego Novillo @ 2003-06-12 15:22 UTC (permalink / raw) To: Andrew Pinski; +Cc: gcc On Tue, 2003-05-20 at 11:17, Andrew Pinski wrote: > Fails on stage3 compiling libgcc > > testcase: > > #define TEMP (!0?4:8) > #if TEMP > 4 > int temp1 = TEMP; > #else > int temp2 = TEMP; > #endif > > > Should output temp2, not temp1. > Andrew, This appears fixed now. Are you still getting bootstrap failures on darwin? Thanks. Diego. ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-06-12 15:22 ` Diego Novillo @ 2003-06-12 15:42 ` Andrew Pinski 2003-06-12 16:01 ` Diego Novillo 0 siblings, 1 reply; 9+ messages in thread From: Andrew Pinski @ 2003-06-12 15:42 UTC (permalink / raw) To: Diego Novillo; +Cc: Andrew Pinski, gcc On Thursday, Jun 12, 2003, at 11:17 US/Eastern, Diego Novillo wrote: > On Tue, 2003-05-20 at 11:17, Andrew Pinski wrote: >> Fails on stage3 compiling libgcc >> >> testcase: >> >> #define TEMP (!0?4:8) >> #if TEMP > 4 >> int temp1 = TEMP; >> #else >> int temp2 = TEMP; >> #endif >> >> >> Should output temp2, not temp1. >> > Andrew, > > This appears fixed now. Are you still getting bootstrap failures on > darwin? Yes, the same error unless something change between [zhivago2:pinskia/src/gccssa] pinskia% more src/LAST_UPDATED Thu Jun 12 09:05:00 EDT 2003 Thu Jun 12 13:05:00 UTC 2003 and now. I will look into it a little more to see where the problem is because I said the secondary cause of effect of the problem not the first effect. Thanks, Andrew Pinski ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-06-12 15:42 ` Andrew Pinski @ 2003-06-12 16:01 ` Diego Novillo 2003-06-14 21:45 ` Andrew Pinski 0 siblings, 1 reply; 9+ messages in thread From: Diego Novillo @ 2003-06-12 16:01 UTC (permalink / raw) To: Andrew Pinski; +Cc: gcc On Thu, 2003-06-12 at 11:36, Andrew Pinski wrote: > On Thursday, Jun 12, 2003, at 11:17 US/Eastern, Diego Novillo wrote: > > > This appears fixed now. Are you still getting bootstrap failures on > > darwin? > > Yes, the same error unless something change between > Bummer. Wish I had access to a darwin box. > I will look into it a little more to see where > the problem is because I said the secondary cause of effect of the > problem > not the first effect. > Thanks. That'd be great Diego. ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 2003-06-12 16:01 ` Diego Novillo @ 2003-06-14 21:45 ` Andrew Pinski 0 siblings, 0 replies; 9+ messages in thread From: Andrew Pinski @ 2003-06-14 21:45 UTC (permalink / raw) To: Diego Novillo; +Cc: Andrew Pinski, gcc On Thursday, Jun 12, 2003, at 11:48 US/Eastern, Diego Novillo wrote: > On Thu, 2003-06-12 at 11:36, Andrew Pinski wrote: >> On Thursday, Jun 12, 2003, at 11:17 US/Eastern, Diego Novillo wrote: >> >>> This appears fixed now. Are you still getting bootstrap failures on >>> darwin? >> >> Yes, the same error unless something change between >> > Bummer. Wish I had access to a darwin box. > >> I will look into it a little more to see where >> the problem is because I said the secondary cause of effect of the >> problem >> not the first effect. >> > Thanks. That'd be great It looks to be aliasing problem because if I bootstrap with BOOTCFAGS='-fno-strict-aliasing -O2 -g', it works!!! Now to report another libmudflap problem because it looks like it is not including the right headers. Andrew Pinski ^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2003-06-14 20:34 UTC | newest] Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2003-05-20 16:17 [tree-ssa] bootstrap failure on ppc-apple-darwin6.6 Andrew Pinski 2003-05-20 20:59 ` Andrew Pinski 2003-05-20 21:16 ` Andrew Pinski 2003-05-21 16:07 ` Andrew Pinski 2003-05-21 16:17 ` Andrew Pinski 2003-06-12 15:22 ` Diego Novillo 2003-06-12 15:42 ` Andrew Pinski 2003-06-12 16:01 ` Diego Novillo 2003-06-14 21:45 ` Andrew Pinski
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).