Your message dated Fri, 21 Aug 2015 13:43:08 +0000 with message-id <[email protected]> and subject line Bug#796274: Removed package(s) from unstable has caused the Debian Bug report #75773, regarding [PR optimization/3507, optimization/3996] Non-optimal code to be marked as done.
This means that you claim that the problem has been dealt with. If this is not the case it is now your responsibility to reopen the Bug report if necessary, and/or fix the problem forthwith. (NB: If you are a system administrator and have no idea what this message is talking about, this may indicate a serious mail system misconfiguration somewhere. Please contact [email protected] immediately.) -- 75773: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=75773 Debian Bug Tracking System Contact [email protected] with problems
--- Begin Message ---Package: gcc-3.0 Version: 3.0.1-0pre010727 Severity: wishlist Hi, this simple code snippet generates a bit suboptimal code when compiled with gcc-3.0. For some unknown reason simple '--b->cnt' is correctly assembled as 'decl (somememory)', but when this expression is tested for zero (which is available on i386 in Z flag after simple decrement), it loads value into register, decrements, then even again tests for zero, and finally stores it back to memory ... This is visible in both freeblock and freeblockC. Second problem is that when decremented field in structure is not integer (for example char) and function is chained into another one with same arguments, some of arguments are for unknown reason written back to stack, although their value did not change (I even tried to mark argument as 'struct blockChar* const b', but no change on generated code). This is visible in freeblockC only. These are all acceptable because of gcc-2.95.x generated even worse code. But what's not acceptable happens in readAsyncCallback. For some strange reason it allocates 16bytes of stack, for unknown reason (note that it was compiled with -mpreferred-stack-boundary=2, so it is not any stack aligning). It is smallest example of problem I'm suffering of - I have function which reserves 364 bytes on stack with gcc-3.0, but only 120 bytes when compiled with gcc-2.95, and code generated by gcc-3.0 does not even touch couple of on-stack variables which were optimized, but then somehow left in stack frame. Thanks, Petr Vandrovec [email protected] struct blockInt { unsigned int cnt; }; void __freeblock(struct blockInt*); void freeblock(struct blockInt* b) { if (--b->cnt) __freeblock(b); } void simpleDec(struct blockInt* b) { --b->cnt; } struct blockChar { unsigned char cnt; void* address; }; void __freeblockC(struct blockChar*); void freeblockC(struct blockChar* b) { if (--b->cnt) __freeblockC(b); } void simpleDecC(struct blockChar* b) { --b->cnt; } struct semaphore { void* queue; int value; }; struct LB_async { struct semaphore* sema; }; void LBL_sSignal(struct semaphore*); static inline void __LBL_sSignal(struct semaphore* sem) { if (!--sem->value && sem->queue) { LBL_sSignal(sem); } } void readAsyncCallback(struct LB_async* req) { __LBL_sSignal(req->sema); } ---------- generated asm ---------- .file "x.c" # GNU C version 3.0.1 20010728 (Debian prerelease) (i386-linux) # compiled by GNU C version 3.0.1 20010728 (Debian prerelease). # options passed: -lang-c -D__GNUC__=3 -D__GNUC_MINOR__=0 # -D__GNUC_PATCHLEVEL__=1 -D__ELF__ -Dunix -Dlinux -D__ELF__ -D__unix__ # -D__linux__ -D__unix -D__linux -Asystem=posix -D__OPTIMIZE__ # -D__STDC_HOSTED__=1 -W -Wall -Acpu=i386 -Amachine=i386 -Di386 -D__i386 # -D__i386__ -D__tune_i386__ -mpreferred-stack-boundary=2 -O2 -W -Wall # -fomit-frame-pointer -fverbose-asm # options enabled: -fdefer-pop -fomit-frame-pointer # -foptimize-sibling-calls -fcse-follow-jumps -fcse-skip-blocks # -fexpensive-optimizations -fthread-jumps -fstrength-reduce -fpeephole # -fforce-mem -ffunction-cse -finline -fkeep-static-consts -fcaller-saves # -fpcc-struct-return -fgcse -frerun-cse-after-loop -frerun-loop-opt # -fdelete-null-pointer-checks -fschedule-insns2 -fsched-interblock # -fsched-spec -fbranch-count-reg -freorder-blocks -fcommon -fverbose-asm # -fgnu-linker -fregmove -foptimize-register-move -fargument-alias # -fstrict-aliasing -fident -fpeephole2 -fguess-branch-probability # -fmath-errno -m80387 -mhard-float -mno-soft-float -mieee-fp # -mfp-ret-in-387 -mpreferred-stack-boundary=2 .text .align 4 .globl freeblock .type freeblock,@function freeblock: movl 4(%esp), %edx movl (%edx), %eax decl %eax testl %eax, %eax movl %eax, (%edx) je .L2 jmp __freeblock .p2align 2 .L2: ret .Lfe1: .size freeblock,.Lfe1-freeblock .align 4 .globl simpleDec .type simpleDec,@function simpleDec: movl 4(%esp), %eax decl (%eax) ret .Lfe2: .size simpleDec,.Lfe2-simpleDec .align 4 .globl freeblockC .type freeblockC,@function freeblockC: movl 4(%esp), %edx movb (%edx), %al decl %eax testb %al, %al movb %al, (%edx) je .L5 movl %edx, 4(%esp) jmp __freeblockC .p2align 2 .L5: ret .Lfe3: .size freeblockC,.Lfe3-freeblockC .align 4 .globl simpleDecC .type simpleDecC,@function simpleDecC: movl 4(%esp), %eax decb (%eax) ret .Lfe4: .size simpleDecC,.Lfe4-simpleDecC .align 4 .globl readAsyncCallback .type readAsyncCallback,@function readAsyncCallback: subl $16, %esp movl 20(%esp), %eax movl (%eax), %edx movl 4(%edx), %eax decl %eax testl %eax, %eax movl %eax, 4(%edx) je .L11 .L10: addl $16, %esp ret .p2align 2 .L11: movl (%edx), %ecx testl %ecx, %ecx je .L10 pushl %edx call LBL_sSignal popl %eax jmp .L10 .Lfe5: .size readAsyncCallback,.Lfe5-readAsyncCallback .ident "GCC: (GNU) 3.0.1 20010728 (Debian prerelease)"
--- End Message ---
--- Begin Message ---Version: 4.6.4-7+rm Dear submitter, as the package gcc-4.6 has just been removed from the Debian archive unstable we hereby close the associated bug reports. We are sorry that we couldn't deal with your issue properly. For details on the removal, please see https://bugs.debian.org/796274 The version of this package that was in Debian prior to this removal can still be found using http://snapshot.debian.org/. This message was generated automatically; if you believe that there is a problem with it please contact the archive administrators by mailing [email protected]. Debian distribution maintenance software pp. Scott Kitterman (the ftpmaster behind the curtain)
--- End Message ---
