Re: Zero/Sign extension elimination using value ranges
On Tue, May 20, 2014 at 12:27:31PM +1000, Kugan wrote: 1. Handling NOP_EXPR or CONVERT_EXPR that are in the IL because they are required for type correctness. We have two cases here: A) Mode is smaller than word_mode. This is usually from where the zero/sign extensions are showing up in final assembly. For example : int = (int) short which usually expands to (set (reg:SI ) (sext:SI (subreg:HI (reg:SI We can expand this (set (reg:SI ) (((reg:SI If following is true: 1. Value stored in RHS and LHS are of the same signedness 2. Type can hold the value. i.e., In cases like char = (char) short, we check that the value in short is representable char type. (i.e. look at the value range in RHS SSA_NAME and see if that can be represented in types of LHS without overflowing) Subreg here is not a paradoxical subreg. We are removing the subreg and zero/sign extend here. I am assuming here that QI/HI registers are represented in SImode (basically word_mode) with zero/sign extend is used as in (zero_extend:SI (subreg:HI (reg:SI 117)). Wouldn't it be better to just set proper flags on the SUBREG based on value range info (SUBREG_PROMOTED_VAR_P and SUBREG_PROMOTED_UNSIGNED_P)? Then not only the optimizers could eliminate in zext/sext when possible, but all other optimizations could benefit from that. Jakub
Re: RFC: Doc update for attribute
After thinking about this some more, I believe I have some better text. Previously I used the word discouraged to describe this practice. The existing docs use the term avoid. I believe what you want is something more like the attached. Direct and clear, just like docs should be. If you are ok with this, I'll send it to gcc-patches. dw +While it +is discouraged, it is possible to write your own prologue/epilogue code +using asm and use ``C'' code in the middle. I wouldn't remove the last sentence since IMO it's not the intent of the feature to ever support that and the compiler doesn't guarantee it and may result in wrong code given that `naked' is a fragile low-level feature. I'm assuming you meant would remove. I wasn't comfortable including that sentence, but I was following the existing docs. Since they said you could only use basic asm, following that with a warning to avoid locals/if/etc was really confusing without this text. Also, as ugly as this is, apparently some people really do this (comment 6): https://gcc.gnu.org/bugzilla/show_bug.cgi?id=43404#c6 We don't have to doc every crazy thing people try to do with gcc. But since it's out there, maybe we should this time? If only to discourage it. I'm *slightly* more in favor of keeping it. But if you still feel it should go, it's gone. Index: extend.texi === --- extend.texi (revision 210624) +++ extend.texi (working copy) @@ -3332,16 +3332,15 @@ @item naked @cindex function without a prologue/epilogue code -Use this attribute on the ARM, AVR, MCORE, MSP430, NDS32, RL78, RX and SPU -ports to indicate that the specified function does not need prologue/epilogue -sequences generated by the compiler. -It is up to the programmer to provide these sequences. The -only statements that can be safely included in naked functions are -@code{asm} statements that do not have operands. All other statements, -including declarations of local variables, @code{if} statements, and so -forth, should be avoided. Naked functions should be used to implement the -body of an assembly function, while allowing the compiler to construct -the requisite function declaration for the assembler. +This attribute is available on the ARM, AVR, MCORE, MSP430, NDS32, +RL78, RX and SPU ports. It allows the compiler to construct the +requisite function declaration, while allowing the body of the +function to be assembly code. The specified function will not have +prologue/epilogue sequences generated by the compiler. Only Basic +@code{asm} statements can safely be included in naked functions +(@pxref{Basic Asm}). While using Extended @code{asm} or a mixture of +Basic @code{asm} and ``C'' code may appear to work, they cannot be +depended upon to work reliably and are not supported. @item near @cindex functions that do not handle memory bank switching on 68HC11/68HC12 @@ -6269,6 +6268,8 @@ efficient code, and in most cases it is a better solution. When writing inline assembly language outside of C functions, however, you must use Basic @code{asm}. Extended @code{asm} statements have to be inside a C function. +Functions declared with the @code{naked} attribute also require Basic +@code{asm} (@pxref{Function Attributes}). Under certain circumstances, GCC may duplicate (or remove duplicates of) your assembly code when optimizing. This can lead to unexpected duplicate @@ -6388,6 +6389,8 @@ Note that Extended @code{asm} statements must be inside a function. Only Basic @code{asm} may be outside functions (@pxref{Basic Asm}). +Functions declared with the @code{naked} attribute also require Basic +@code{asm} (@pxref{Function Attributes}). While the uses of @code{asm} are many and varied, it may help to think of an @code{asm} statement as a series of low-level instructions that convert input
Re: RFC: Doc update for attribute
Am 05/16/2014 07:16 PM, schrieb Carlos O'Donell: On 05/12/2014 11:13 PM, David Wohlferd wrote: After updating gcc's docs about inline asm, I'm trying to improve some of the related sections. One that I feel has problems with clarity is __attribute__ naked. I have attached my proposed update. Comments/corrections are welcome. In a related question: To better understand how this attribute is used, I looked at the Linux kernel. While the existing docs say only ... asm statements that do not have operands can safely be used, Linux routinely uses asm WITH operands. That's a bug. Period. You must not use naked with an asm that has operands. Any kind of operand might inadvertently cause the compiler to generate code and that would violate the requirements of the attribute and potentially generate an ICE. There is target hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS that is intended to cater that case. For example, the documentation indicates it only works with optimization turned off. But I don't know how reliable it is in general. For avr target it works as expected. https://gcc.gnu.org/onlinedocs/gccint/Misc.html#index-TARGET_005fALLOCATE_005fSTACK_005fSLOTS_005fFOR_005fARGS-4969 Johann
Roadmap for 4.9.1, 4.10.0 and onwards?
Hi, I've been tracking the latest releases of gcc since 4.7 or so (variously interested in C++1y support, cilk and openmp). One thing I've found hard to locate is information about planned inclusions for future releases. As much relies on unpredictable community contributions I don't expect there to be a concrete or reliable plan. However, equally I'm sure the steering committee have some ideas over what ought to be upcoming releases. Is this published anywhere? For example if I look at: https://gcc.gnu.org/projects/cxx1y.html There are 3 items marked no under C++14 support. Which if any are tabled for 4.10.0?More generally what targets (obviously subject to change) are there for 4.10.0? or 4.9.1? Regards, Bruce.
Supported targets
Hi, Slightly related to my previous question about the roadmap. I have two quite old targets based on (so far as I know) standard linux distributions. Should they still be supported? RHEL4 (kernel 2.6.9-55.ELsmp): I was able to compile 4.8.1 successfully when it was released. 4.9.0 fails as below. RHEL4 is end of life (but not extended life). My feeling is this ought to work and is probably a regression I should report? SUSE LINUX Enterprise Server 9 (i586) (kernal 2.6.5-7.111-smp) I was able to compile gcc 4.7.0 successfully when it was released. I had less luck with 4.8.0. 4.9.0 fails as below. However, this machine/distribution is so old it is not unreasonable to say it should be scrapped. My main targets are RHEL5 and RHEL6 which work perfectly. I also tried bootstrapping using 4.8.1 to build 4.9.0 on RHEL4 and 4.7.0 to build 4.9.0 on the Suse box rather than the ancient system installed versions (RHEL4 = gcc 3.4.6, Suse 9 = 3.3.3) but without success. Regards, Bruce. RHEL4 (kernel 2.6.9-55.ELsmp): [snip] ../../../../gcc-4.9.0/libsanitizer/include/system/linux/aio_abi.h:2:32: fatal error: linux/aio_abi.h: No such file or director y #include_next linux/aio_abi.h ^ compilation terminated. make[3]: *** [sanitizer_platform_limits_linux.lo] Error 1 make[3]: Leaving directory `/development/brucea/gcc/build/build/x86_64-unknown-linux-gnu/libsanitizer/sanitizer_common' make[2]: *** [install-recursive] Error 1 [snip] SUSE LINUX Enterprise Server 9 (i586) (kernal 2.6.5-7.111-smp) [snip] /development/dev1/brucea/gcc4.7/bin/../lib/gcc/i686-pc-linux-gnu/4.7.0/../../../../i686-pc-linux-gnu/bin/ld: /home/brucea/gcc4 .9/lib/libmpfr.so: undefined reference to symbol '___tls_get_addr@@GLIBC_2.3' /development/dev1/brucea/gcc4.7/bin/../lib/gcc/i686-pc-linux-gnu/4.7.0/../../../../i686-pc-linux-gnu/bin/ld: note: '___tls_get _addr@@GLIBC_2.3' is defined in DSO /lib/ld-linux.so.2 so try adding it to the linker command line /lib/ld-linux.so.2: could not read symbols: Invalid operation collect2: error: ld returned 1 exit status make[3]: *** [cc1] Error 1 [snip] Requires a later version of glibc?
Re: Supported targets
[snip] /development/dev1/brucea/gcc4.7/bin/../lib/gcc/i686-pc-linux-gnu/4.7.0/../.. /../../i686-pc-linux-gnu/bin/ld: /home/brucea/gcc4 .9/lib/libmpfr.so: undefined reference to symbol '___tls_get_addr@@GLIBC_2.3' /development/dev1/brucea/gcc4.7/bin/../lib/gcc/i686-pc-linux-gnu/4.7.0/../. ./../../i686-pc-linux-gnu/bin/ld: note: '___tls_get _addr@@GLIBC_2.3' is defined in DSO /lib/ld-linux.so.2 so try adding it to the linker command line /lib/ld-linux.so.2: could not read symbols: Invalid operation collect2: error: ld returned 1 exit status make[3]: *** [cc1] Error 1 [snip] Requires a later version of glibc? Yes, glibc 2.4 is required for GCC 4.9 because of this. -- Eric Botcazou
Re: Supported targets
On 20 May 2014 11:26, Bruce Adams wrote: RHEL4 (kernel 2.6.9-55.ELsmp): I was able to compile 4.8.1 successfully when it was released. 4.9.0 fails as below. RHEL4 is end of life (but not extended life). My feeling is this ought to work and is probably a regression I should report? Yes, I think it should be reported if it isn't in Bugzilla yet. You can use --disable-libsanitizer to build GCC without the failing library.
Re: Supported targets
On 20 May 2014 11:55, Eric Botcazou wrote: [snip] /development/dev1/brucea/gcc4.7/bin/../lib/gcc/i686-pc-linux-gnu/4.7.0/../.. /../../i686-pc-linux-gnu/bin/ld: /home/brucea/gcc4 .9/lib/libmpfr.so: undefined reference to symbol '___tls_get_addr@@GLIBC_2.3' /development/dev1/brucea/gcc4.7/bin/../lib/gcc/i686-pc-linux-gnu/4.7.0/../. ./../../i686-pc-linux-gnu/bin/ld: note: '___tls_get _addr@@GLIBC_2.3' is defined in DSO /lib/ld-linux.so.2 so try adding it to the linker command line /lib/ld-linux.so.2: could not read symbols: Invalid operation collect2: error: ld returned 1 exit status make[3]: *** [cc1] Error 1 [snip] Requires a later version of glibc? Yes, glibc 2.4 is required for GCC 4.9 because of this. Should that be noted at https://gcc.gnu.org/install/specific.html#x-x-linux-gnu ?
Re: Supported targets
Yes, glibc 2.4 is required for GCC 4.9 because of this. Should that be noted at https://gcc.gnu.org/install/specific.html#x-x-linux-gnu ? Probably, unless someone knows how to work around it. We traced it to the missing AS_NEEDED in /usr/lib/libc.so: /* GNU ld script Use the shared library, but some functions are only in the static library, so try that secondarily. */ OUTPUT_FORMAT(elf32-i386) GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a AS_NEEDED ( /lib/ld- linux.so.2 ) ) -- Eric Botcazou
Re: Supported targets
On Tue, May 20, 2014 at 01:14:24PM +0200, Eric Botcazou wrote: Yes, glibc 2.4 is required for GCC 4.9 because of this. Should that be noted at https://gcc.gnu.org/install/specific.html#x-x-linux-gnu ? Probably, unless someone knows how to work around it. We traced it to the missing AS_NEEDED in /usr/lib/libc.so: /* GNU ld script Use the shared library, but some functions are only in the static library, so try that secondarily. */ OUTPUT_FORMAT(elf32-i386) GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a AS_NEEDED ( /lib/ld- linux.so.2 ) ) But that should be generally needed only when linking with -Wl,-z,defs , without it the linker shouldn't care. Jakub
Re: Supported targets
But that should be generally needed only when linking with -Wl,-z,defs , without it the linker shouldn't care. Yet using a local libc.so with the missing AS_NEEDED is a (poor) workaround. -- Eric Botcazou
Re: [GSoC] writing test-case
On Mon, May 19, 2014 at 5:51 PM, Michael Matz m...@suse.de wrote: Hi, On Thu, 15 May 2014, Richard Biener wrote: To me predicate (and capture without expression or predicate) differs from expression in that predicate is clearly a leaf of the expression tree while we have to recurse into expression operands. Now, if we want to support applying predicates to the midst of an expression, like (plus predicate(minus @0 @1) @2) (...) then this would no longer be true. At the moment you'd write (plus (minus@3 @0 @1) @2) if (predicate (@3)) (...) which makes it clearer IMHO (with the decision tree building you'd apply the predicates after matching the expression tree anyway I suppose, so code generation would be equivalent). Syntaxwise I had this idea for adding generic predicates to expressions: (plus (minus @0 @1):predicate @2) (...) So you'd write (plus @0 :integer_zerop) instead of (plus @0 integer_zerop) ? If prefix or suffix doesn't matter much, but using a different syntax to separate expression from predicate seems to make things clearer. Optionally adding things like and/or for predicates might also make sense: (plus (minus @0 @1):positive_p(@0) || positive_p(@1) @2) (...) negation whould be more useful I guess. You open up a can of worms with ordering though: (plus (minus @0 @1) @2:operand_equal_p (@1, @2, 0)) which might be declared invalid or is equivalent to (plus (minus @0 @1) @2):operand_equal_p (@1, @2, 0) ? Note that your predicate placement doesn't match placement of captures for non-innermost expressions. capturing the outer plus would be (plus@3 (minus @0 @1) @2) not (plus (minus @0 @1) @2)@3 so maybe apply predicates there as well: (plus:operand_equal_p (@1, @2, 0) (minus @0 @1) @2) But I still think that doing all predicates within a if-expr makes the pattern less convoluted. Enabling/disabling a whole set of patterns with a common condition might still be a worthwhile addition. Richard. Ciao, Michael.
Re: [GSoC] first phase
On Mon, May 19, 2014 at 7:30 PM, Prathamesh Kulkarni bilbotheelffri...@gmail.com wrote: Hi, Unfortunately I shall need to take this week off, due to university exams, which are up-to 27th May. I will start working from 28th on pattern matching with decision tree, and try to cover up for the first week. I am extremely sorry about this. I thought I would be able to do both during exam week, but the exam load has become too much -:( Ok. In the first phase (up-to 23rd June), I hope to get genmatch ready: a) pattern matching with decision tree. b) Add patterns to test genmatch. c) Depending upon the patterns, extending the meta-description d) Other fixes: * capturing outermost expressions. For example this pattern does not get simplified (match_and_simplify (plus@2 (negate @0) @1) if (!TYPE_SATURATING (TREE_TYPE (@2))) (minus @1 @0)) I guess this happens because in write_nary_simplifiers: if (s-match-type != OP_EXPR) continue; Yeah. Maybe this is not correct way to fix this, should we also pass lhs to generated gimple_match_and_simplify ? I guess that would be the capture for outermost expression. Unfortunately it is not available for all API entries. The type of the expression is, though. I lean towards rejecting the capture at parsing time and providing a special capture (for example @@, or just @0, or @T to denote it's a type, or just refer magically to 'type'). That is, (match_and_simplify (plus (negate @0) @1) if (!TYPE_SATURATING (type)) (minus @1 @0)) works for me. For above pattern, I guess @2 represents lhs. So for this test-case: int foo (int x, int y) { int t1 = -x; int t2 = t1 + y; return t2; } t2 would be @2, t1 would be @0 and y would be @1. Is that correct ? This would create issues when lhs is NULL, for example, in call to built-in functions ? Yeah, or if the machinery is called via gimple_build () where there is no existing lhs. * avoid using statement expressions for code gen of expression * rewriting code-generator using visitor classes, and other refactoring (using std::string for example), etc. I have a very rough time-line in mind, for completing tasks: 28th may - 31st may a) Have test-case for each pattern present (except COND_EXPR) in match.pd I guess most of it is already done, a few patterns are remaining. Good. b) Small fixes (for example, those mentioned above). Good. c) Have an initial idea/prototype for implementing decision tree 1st June - 15th June a) Implementing decision tree b) Adding patterns in match.pd to test the decision tree in match.pd, and accompanying test-cases in tree-ssa/match-*.c 16th June - 23rd June a) Support for GENERIC code generation. b) Refactoring and backup time for backlog. GENERIC code generation: I am a bit confused about this. Currently, pattern matching is implemented for GENERIC. However I believe simplification is done on GIMPLE. For example: (match_and_simplify (plus (negate @0) @1) (minus @0 @1)) If given input is GENERIC , it would do matching on GENERIC, but shall transform (minus @0 @1) to it's GIMPLE equivalent. Is that correct ? Correct. Err, not sure what it will do - I implemented it only to support the weird cases where GENERIC is nested inside GIMPLE, like for a_2 = b_3 0 ? c_4 : d_5; thus the comment in match.pd: /* Due to COND_EXPRs weirdness in GIMPLE the following won't work without some hacks in the code generator. */ (match_and_simplify (cond (bit_not @0) @1 @2) (cond @0 @2 @1)) the code generator would need to know that COND_EXPR has a GENERIC op0 ... same applies to REALPART_EXPR, but there the hacks are already in place ;) * Should we have a separate GENERIC match-and-simplify API like for gimple instead of having GENERIC matching in gimple_match_and_simplify ? Yes. The GENERIC API follows the API of fold_{unary,binary,ternary}. I suppose we simply provide a slightly different name for them (but use the original API for recursing and call ourselves from the original API). * Do we add another pattern type, something like generic_match_and_simplify that will do the transform on GENERIC for example: (generic_match_and_simplify (plus (negate @0) @1) (minus @0 @1)) would produce GENERIC equivalent of (minus @0 @1). or maybe keep match_and_simplify, and tell the transform operand to produce GENERIC. Something like: (match_and_simplify (plus (negate @0) @1) GENERIC: (minus @0 @1)) we simply process each pattern twice, once we generate the GIMPLE match-and-simplify routine and once we generate the GENERIC match-and-simplify routine. The patterns are supposed to be the same for both and always apply to both. Another thing I would like to do in first phase is figure out dependencies of tree-ssa-forwprop on GENERIC folding (for instance fold_comparison patterns). Yeah. Having patterns for comparison simplification is important for other parts of the compiler as well, thus
Re: [GSoC] writing test-case
Hi, On Tue, 20 May 2014, Richard Biener wrote: Syntaxwise I had this idea for adding generic predicates to expressions: (plus (minus @0 @1):predicate @2) (...) So you'd write (plus @0 :integer_zerop) instead of (plus @0 integer_zerop) ? plus is binary, where is your @1? If you want to not capture the second operand but still have it tested for a predicates, then yes, the first form it would be. If prefix or suffix doesn't matter much, but using a different syntax to separate expression from predicate seems to make things clearer. Optionally adding things like and/or for predicates might also make sense: (plus (minus @0 @1):positive_p(@0) || positive_p(@1) @2) (...) negation whould be more useful I guess. You open up a can of worms with ordering though: (plus (minus @0 @1) @2:operand_equal_p (@1, @2, 0)) which might be declared invalid or is equivalent to It wouldn't necessarily be invalid, the predicate would apply to @2; but check operands 1 and 0 as well, which might be surprising. In this case it might indeed be equivalent to : (plus (minus @0 @1) @2):operand_equal_p (@1, @2, 0) Note that your predicate placement doesn't match placement of captures for non-innermost expressions. capturing the outer plus would be (plus@3 (minus @0 @1) @2) You're right, I'd allow placing the predicate directly behind the capture, i.e.: (plus@3:predicate (minus @0 @1) @2) But I still think that doing all predicates within a if-expr makes the pattern less convoluted. I think it simply depends on the scope of the predicate. If it's a predicate applying to multiple operands from different nested level an if-expr is clearer (IMHO). If it applies to one operand it seems more natural to place it directly next to that operand. I.e.: (minus @0 @1:non_negative) // better vs. (minus @0 @1) (if (non_negative (@1)) But: (plus@3 (minus @0 @1) @2) // better (if (operand_equal_p (@1, @2, 0)) vs: (plus@3:operand_equal_p (@1, @2, 0) (minus @0 @1) @2) That is we could require that predicates that are applied with ':' need to be unary and apply to the one expression to which they are bound. Enabling/disabling a whole set of patterns with a common condition might still be a worthwhile addition. Right, but that seems orthogonal to the above? Ciao, Michael.
Re: [GSoC] first phase
On Tue, May 20, 2014 at 5:46 PM, Richard Biener richard.guent...@gmail.com wrote: On Mon, May 19, 2014 at 7:30 PM, Prathamesh Kulkarni bilbotheelffri...@gmail.com wrote: Hi, Unfortunately I shall need to take this week off, due to university exams, which are up-to 27th May. I will start working from 28th on pattern matching with decision tree, and try to cover up for the first week. I am extremely sorry about this. I thought I would be able to do both during exam week, but the exam load has become too much -:( Ok. In the first phase (up-to 23rd June), I hope to get genmatch ready: a) pattern matching with decision tree. b) Add patterns to test genmatch. c) Depending upon the patterns, extending the meta-description d) Other fixes: * capturing outermost expressions. For example this pattern does not get simplified (match_and_simplify (plus@2 (negate @0) @1) if (!TYPE_SATURATING (TREE_TYPE (@2))) (minus @1 @0)) I guess this happens because in write_nary_simplifiers: if (s-match-type != OP_EXPR) continue; Yeah. Maybe this is not correct way to fix this, should we also pass lhs to generated gimple_match_and_simplify ? I guess that would be the capture for outermost expression. Unfortunately it is not available for all API entries. The type of the expression is, though. I lean towards rejecting the capture at parsing time and providing a special capture (for example @@, or just @0, or @T to denote it's a type, or just refer magically to 'type'). That is, (match_and_simplify (plus (negate @0) @1) if (!TYPE_SATURATING (type)) (minus @1 @0)) works for me. For above pattern, I guess @2 represents lhs. So for this test-case: int foo (int x, int y) { int t1 = -x; int t2 = t1 + y; return t2; } t2 would be @2, t1 would be @0 and y would be @1. Is that correct ? This would create issues when lhs is NULL, for example, in call to built-in functions ? Yeah, or if the machinery is called via gimple_build () where there is no existing lhs. * avoid using statement expressions for code gen of expression * rewriting code-generator using visitor classes, and other refactoring (using std::string for example), etc. I have a very rough time-line in mind, for completing tasks: 28th may - 31st may a) Have test-case for each pattern present (except COND_EXPR) in match.pd I guess most of it is already done, a few patterns are remaining. Good. b) Small fixes (for example, those mentioned above). Good. c) Have an initial idea/prototype for implementing decision tree 1st June - 15th June a) Implementing decision tree b) Adding patterns in match.pd to test the decision tree in match.pd, and accompanying test-cases in tree-ssa/match-*.c 16th June - 23rd June a) Support for GENERIC code generation. b) Refactoring and backup time for backlog. GENERIC code generation: I am a bit confused about this. Currently, pattern matching is implemented for GENERIC. However I believe simplification is done on GIMPLE. For example: (match_and_simplify (plus (negate @0) @1) (minus @0 @1)) If given input is GENERIC , it would do matching on GENERIC, but shall transform (minus @0 @1) to it's GIMPLE equivalent. Is that correct ? Correct. Err, not sure what it will do - I implemented it only to support the weird cases where GENERIC is nested inside GIMPLE, like for a_2 = b_3 0 ? c_4 : d_5; thus the comment in match.pd: /* Due to COND_EXPRs weirdness in GIMPLE the following won't work without some hacks in the code generator. */ (match_and_simplify (cond (bit_not @0) @1 @2) (cond @0 @2 @1)) the code generator would need to know that COND_EXPR has a GENERIC op0 ... same applies to REALPART_EXPR, but there the hacks are already in place ;) * Should we have a separate GENERIC match-and-simplify API like for gimple instead of having GENERIC matching in gimple_match_and_simplify ? Yes. The GENERIC API follows the API of fold_{unary,binary,ternary}. I suppose we simply provide a slightly different name for them (but use the original API for recursing and call ourselves from the original API). * Do we add another pattern type, something like generic_match_and_simplify that will do the transform on GENERIC for example: (generic_match_and_simplify (plus (negate @0) @1) (minus @0 @1)) would produce GENERIC equivalent of (minus @0 @1). or maybe keep match_and_simplify, and tell the transform operand to produce GENERIC. Something like: (match_and_simplify (plus (negate @0) @1) GENERIC: (minus @0 @1)) we simply process each pattern twice, once we generate the GIMPLE match-and-simplify routine and once we generate the GENERIC match-and-simplify routine. The patterns are supposed to be the same for both and always apply to both. Yes, for patterns where transform operand is not c-code. I guess this wouldn't work for transforms written as c-code that uses IR-specific API ? As long
Re: [GSoC] first phase
On Tue, May 20, 2014 at 2:59 PM, Prathamesh Kulkarni bilbotheelffri...@gmail.com wrote: On Tue, May 20, 2014 at 5:46 PM, Richard Biener richard.guent...@gmail.com wrote: On Mon, May 19, 2014 at 7:30 PM, Prathamesh Kulkarni bilbotheelffri...@gmail.com wrote: Hi, Unfortunately I shall need to take this week off, due to university exams, which are up-to 27th May. I will start working from 28th on pattern matching with decision tree, and try to cover up for the first week. I am extremely sorry about this. I thought I would be able to do both during exam week, but the exam load has become too much -:( Ok. In the first phase (up-to 23rd June), I hope to get genmatch ready: a) pattern matching with decision tree. b) Add patterns to test genmatch. c) Depending upon the patterns, extending the meta-description d) Other fixes: * capturing outermost expressions. For example this pattern does not get simplified (match_and_simplify (plus@2 (negate @0) @1) if (!TYPE_SATURATING (TREE_TYPE (@2))) (minus @1 @0)) I guess this happens because in write_nary_simplifiers: if (s-match-type != OP_EXPR) continue; Yeah. Maybe this is not correct way to fix this, should we also pass lhs to generated gimple_match_and_simplify ? I guess that would be the capture for outermost expression. Unfortunately it is not available for all API entries. The type of the expression is, though. I lean towards rejecting the capture at parsing time and providing a special capture (for example @@, or just @0, or @T to denote it's a type, or just refer magically to 'type'). That is, (match_and_simplify (plus (negate @0) @1) if (!TYPE_SATURATING (type)) (minus @1 @0)) works for me. For above pattern, I guess @2 represents lhs. So for this test-case: int foo (int x, int y) { int t1 = -x; int t2 = t1 + y; return t2; } t2 would be @2, t1 would be @0 and y would be @1. Is that correct ? This would create issues when lhs is NULL, for example, in call to built-in functions ? Yeah, or if the machinery is called via gimple_build () where there is no existing lhs. * avoid using statement expressions for code gen of expression * rewriting code-generator using visitor classes, and other refactoring (using std::string for example), etc. I have a very rough time-line in mind, for completing tasks: 28th may - 31st may a) Have test-case for each pattern present (except COND_EXPR) in match.pd I guess most of it is already done, a few patterns are remaining. Good. b) Small fixes (for example, those mentioned above). Good. c) Have an initial idea/prototype for implementing decision tree 1st June - 15th June a) Implementing decision tree b) Adding patterns in match.pd to test the decision tree in match.pd, and accompanying test-cases in tree-ssa/match-*.c 16th June - 23rd June a) Support for GENERIC code generation. b) Refactoring and backup time for backlog. GENERIC code generation: I am a bit confused about this. Currently, pattern matching is implemented for GENERIC. However I believe simplification is done on GIMPLE. For example: (match_and_simplify (plus (negate @0) @1) (minus @0 @1)) If given input is GENERIC , it would do matching on GENERIC, but shall transform (minus @0 @1) to it's GIMPLE equivalent. Is that correct ? Correct. Err, not sure what it will do - I implemented it only to support the weird cases where GENERIC is nested inside GIMPLE, like for a_2 = b_3 0 ? c_4 : d_5; thus the comment in match.pd: /* Due to COND_EXPRs weirdness in GIMPLE the following won't work without some hacks in the code generator. */ (match_and_simplify (cond (bit_not @0) @1 @2) (cond @0 @2 @1)) the code generator would need to know that COND_EXPR has a GENERIC op0 ... same applies to REALPART_EXPR, but there the hacks are already in place ;) * Should we have a separate GENERIC match-and-simplify API like for gimple instead of having GENERIC matching in gimple_match_and_simplify ? Yes. The GENERIC API follows the API of fold_{unary,binary,ternary}. I suppose we simply provide a slightly different name for them (but use the original API for recursing and call ourselves from the original API). * Do we add another pattern type, something like generic_match_and_simplify that will do the transform on GENERIC for example: (generic_match_and_simplify (plus (negate @0) @1) (minus @0 @1)) would produce GENERIC equivalent of (minus @0 @1). or maybe keep match_and_simplify, and tell the transform operand to produce GENERIC. Something like: (match_and_simplify (plus (negate @0) @1) GENERIC: (minus @0 @1)) we simply process each pattern twice, once we generate the GIMPLE match-and-simplify routine and once we generate the GENERIC match-and-simplify routine. The patterns are supposed to be the same for both and always apply to both. Yes, for patterns where transform operand is not c-code. I
Re: [GSoC] writing test-case
On Tue, May 20, 2014 at 2:20 PM, Michael Matz m...@suse.de wrote: Hi, On Tue, 20 May 2014, Richard Biener wrote: Syntaxwise I had this idea for adding generic predicates to expressions: (plus (minus @0 @1):predicate @2) (...) So you'd write (plus @0 :integer_zerop) instead of (plus @0 integer_zerop) ? plus is binary, where is your @1? I know it's zero so I don't need it captured. (match_and_simplify (plus @0 integer_zerop) @0) mind that all predicates apply to leafs only at the moment. If you want to not capture the second operand but still have it tested for a predicates, then yes, the first form it would be. Ok. If prefix or suffix doesn't matter much, but using a different syntax to separate expression from predicate seems to make things clearer. Optionally adding things like and/or for predicates might also make sense: (plus (minus @0 @1):positive_p(@0) || positive_p(@1) @2) (...) negation whould be more useful I guess. You open up a can of worms with ordering though: (plus (minus @0 @1) @2:operand_equal_p (@1, @2, 0)) which might be declared invalid or is equivalent to It wouldn't necessarily be invalid, the predicate would apply to @2; but check operands 1 and 0 as well, which might be surprising. In this case it might indeed be equivalent to : (plus (minus @0 @1) @2):operand_equal_p (@1, @2, 0) Note that your predicate placement doesn't match placement of captures for non-innermost expressions. capturing the outer plus would be (plus@3 (minus @0 @1) @2) You're right, I'd allow placing the predicate directly behind the capture, i.e.: (plus@3:predicate (minus @0 @1) @2) But I still think that doing all predicates within a if-expr makes the pattern less convoluted. I think it simply depends on the scope of the predicate. If it's a predicate applying to multiple operands from different nested level an if-expr is clearer (IMHO). If it applies to one operand it seems more natural to place it directly next to that operand. I.e.: (minus @0 @1:non_negative) // better vs. (minus @0 @1) (if (non_negative (@1)) But: (plus@3 (minus @0 @1) @2) // better (if (operand_equal_p (@1, @2, 0)) vs: (plus@3:operand_equal_p (@1, @2, 0) (minus @0 @1) @2) That is we could require that predicates that are applied with ':' need to be unary and apply to the one expression to which they are bound. Your example is to leafs which we already support as (minus @0 non_negative@1) any good example that is not convoluted where that applies to a non-leaf position? Enabling/disabling a whole set of patterns with a common condition might still be a worthwhile addition. Right, but that seems orthogonal to the above? Right. Richard. Ciao, Michael.
Re: RFC: Doc update for attribute
On 05/20/2014 03:02 AM, David Wohlferd wrote: After thinking about this some more, I believe I have some better text. Previously I used the word discouraged to describe this practice. The existing docs use the term avoid. I believe what you want is something more like the attached. Direct and clear, just like docs should be. David, Thanks for the new patch. If you are ok with this, I'll send it to gcc-patches. Looks good to me. Cheers, Carlos. Index: extend.texi === --- extend.texi (revision 210624) +++ extend.texi (working copy) @@ -3332,16 +3332,15 @@ @item naked @cindex function without a prologue/epilogue code -Use this attribute on the ARM, AVR, MCORE, MSP430, NDS32, RL78, RX and SPU -ports to indicate that the specified function does not need prologue/epilogue -sequences generated by the compiler. -It is up to the programmer to provide these sequences. The -only statements that can be safely included in naked functions are -@code{asm} statements that do not have operands. All other statements, -including declarations of local variables, @code{if} statements, and so -forth, should be avoided. Naked functions should be used to implement the -body of an assembly function, while allowing the compiler to construct -the requisite function declaration for the assembler. +This attribute is available on the ARM, AVR, MCORE, MSP430, NDS32, +RL78, RX and SPU ports. It allows the compiler to construct the +requisite function declaration, while allowing the body of the +function to be assembly code. The specified function will not have +prologue/epilogue sequences generated by the compiler. Only Basic +@code{asm} statements can safely be included in naked functions +(@pxref{Basic Asm}). While using Extended @code{asm} or a mixture of +Basic @code{asm} and ``C'' code may appear to work, they cannot be +depended upon to work reliably and are not supported. @item near @cindex functions that do not handle memory bank switching on 68HC11/68HC12 @@ -6269,6 +6268,8 @@ efficient code, and in most cases it is a better solution. When writing inline assembly language outside of C functions, however, you must use Basic @code{asm}. Extended @code{asm} statements have to be inside a C function. +Functions declared with the @code{naked} attribute also require Basic +@code{asm} (@pxref{Function Attributes}). Under certain circumstances, GCC may duplicate (or remove duplicates of) your assembly code when optimizing. This can lead to unexpected duplicate @@ -6388,6 +6389,8 @@ Note that Extended @code{asm} statements must be inside a function. Only Basic @code{asm} may be outside functions (@pxref{Basic Asm}). +Functions declared with the @code{naked} attribute also require Basic +@code{asm} (@pxref{Function Attributes}). While the uses of @code{asm} are many and varied, it may help to think of an @code{asm} statement as a series of low-level instructions that convert input
Re: RFC: Doc update for attribute
On 05/20/2014 03:59 AM, Georg-Johann Lay wrote: Am 05/16/2014 07:16 PM, schrieb Carlos O'Donell: On 05/12/2014 11:13 PM, David Wohlferd wrote: After updating gcc's docs about inline asm, I'm trying to improve some of the related sections. One that I feel has problems with clarity is __attribute__ naked. I have attached my proposed update. Comments/corrections are welcome. In a related question: To better understand how this attribute is used, I looked at the Linux kernel. While the existing docs say only ... asm statements that do not have operands can safely be used, Linux routinely uses asm WITH operands. That's a bug. Period. You must not use naked with an asm that has operands. Any kind of operand might inadvertently cause the compiler to generate code and that would violate the requirements of the attribute and potentially generate an ICE. There is target hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS that is intended to cater that case. For example, the documentation indicates it only works with optimization turned off. But I don't know how reliable it is in general. For avr target it works as expected. https://gcc.gnu.org/onlinedocs/gccint/Misc.html#index-TARGET_005fALLOCATE_005fSTACK_005fSLOTS_005fFOR_005fARGS-4969 It's still a bug for now. That hook is there because we've allowed bad code to exist for so long that at this point we must for legacy reasons allow some type of input arguments in the asm. However, that doesn't mean we should actively promote this feature or let users use it (until we fix it). Ideally you do want to use the named input arguments as r types to avoid needing to know the exact registers used in the call sequence. Referencing the variables by name and letting gcc emit the right register is useful, but only if it works consistently and today it doesn't. Features that fail to work depending on the optimization level should not be promoted in the documentation. We should document what works and file bugs or fix what doesn't work. Cheers, Carlos.
Weird startup issue with -fsplit-stack
Hello, I'm trying to support -fsplit-stack in GNU Emacs. The most important problem is that GC uses conservative scanning of a C stack, so I need to iterate over stack segments. I'm doing this by using __splitstack_find, as described in libgcc/generic-morestack.c; but now I'm facing the weird issue with startup: Core was generated by `./temacs --batch --load loadup bootstrap'. Program terminated with signal SIGSEGV, Segmentation fault. #0 __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:486 486 pushq %rax (gdb) bt 10 #0 __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:486 #1 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #2 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #3 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #4 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #5 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #6 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #7 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #8 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #9 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 (More stack frames follow...) (gdb) bt -10 #87310 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87311 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87312 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87313 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87314 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87315 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87316 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87317 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #87318 0x003791a21d65 in __libc_start_main (main=0x4d111d main, argc=5, argv=0x7fffacc868d8, init=optimized out, fini=optimized out, rtld_fini=optimized out, stack_end=0x7fffacc868c8) at libc-start.c:285 #87319 0x00405f69 in _start () (gdb) Unfortunately I was unable to reproduce this issue with small test programs, so there is no simple and easy-to-use recipe. Anyway, if someone would like to try: bzr branch bzr://bzr.savannah.gnu.org/emacs/trunk cd trunk cat /path/to/emacs_split_stack.patch | patch -p0 # 'configure' options for 'smallest possible' configuration CPPFLAGS='-DSPLIT_STACK=1' CFLAGS='-O0 -g3 -fsplit-stack' ./configure --prefix=/some/dir --without-all --without-x --disable-acl make I'm using (homebrew) GCC 4.9.0 and (stock) gold 2.24 on a Fedora 20 system. Dmitry === modified file 'src/alloc.c' --- src/alloc.c 2014-05-19 19:19:05 + +++ src/alloc.c 2014-05-20 14:01:56 + @@ -4932,11 +4932,28 @@ #endif /* not GC_SAVE_REGISTERS_ON_STACK */ #endif /* not HAVE___BUILTIN_UNWIND_INIT */ - /* This assumes that the stack is a contiguous region in memory. If - that's not the case, something has to be done here to iterate - over the stack segments. */ +#ifdef SPLIT_STACK + + /* This assumes gcc = 4.6.0 with -fsplit-stack + and corresponding support in libgcc. */ + { +size_t stack_size; +extern void * __splitstack_find (void *, void *, size_t *, + void **, void **, void **); +void *next_segment = NULL, *next_sp = NULL, *initial_sp = NULL, *stack; + +while ((stack = __splitstack_find (next_segment, next_sp, stack_size, + next_segment, next_sp, initial_sp))) + mark_memory (stack, (char *) stack + stack_size); + } + +#else /* not SPLIT_STACK */ + + /* This assumes that the stack is a contiguous region in memory. */ mark_memory (stack_base, end); +#endif /* SPLIT_STACK */ + /* Allow for marking a secondary stack, like the register stack on the ia64. */ #ifdef GC_MARK_SECONDARY_STACK
Re: soft-fp functions support without using libgcc
If you have a working compiler that is missing some functions provided by libgcc, that should be sufficient to build libgcc. Meaning that even if i am unable build libgcc to my new architecture, I should be able to able to provide soft-fp support to the architecture? Btw i get the following error when i build gcc: configure:2627: error: in `/target-arch/target-arch-gcc/builddir/target-arch/libgcc': configure:2630: error: cannot compute suffix of object files: cannot compile And regarding soft-fp, I get the following error when i use soft-fp functions in a test program: : In function `test': (.text+0x0): undefined reference to `__floatsisf' In function `test': : In function `test': (.text+0x2c): undefined reference to `__mulsf3' : In function `test': (.text+0x2e): undefined reference to `__fixsfsi' Is this due to libgcc build fail or it just linking error? In other words, if you want soft-fp for IEEE float, the job should be very simple because that has already been done. If you want soft-fp for CDC 6000 float, you have to do a full implementation of that. Actually i want soft-fp for standard IEEE 754 Sheheryar On Fri, May 16, 2014 at 6:34 PM, paul_kon...@dell.com wrote: On May 16, 2014, at 12:25 PM, Ian Bolton ian.bol...@arm.com wrote: On Fri, May 16, 2014 at 6:34 AM, Sheheryar Zahoor Qazi sheheryar.zahoor.q...@gmail.com wrote: I am trying to provide soft-fp support to a an 18-bit soft-core processor architecture at my university. But the problem is that libgcc has not been cross-compiled for my target architecture and some functions are missing so i cannot build libgcc.I believe soft-fp is compiled in libgcc so i am usable to invoke soft-fp functions from libgcc. It is possible for me to provide soft-fp support without using libgcc. How should i proceed in defining the functions? Any idea? And does any archoitecture provide floating point support withoput using libgcc? I'm sorry, I don't understand the premise of your question. It is not necessary to build libgcc before building libgcc. That would not make sense. If you have a working compiler that is missing some functions provided by libgcc, that should be sufficient to build libgcc. If you replace cross-compiled with ported, I think it makes senses. Can one provide soft-fp support without porting libgcc for their architecture? By definition, in soft-fp you have to implement the FP operations in software. That’s not quite the same as porting libgcc to the target architecture. It should translate to porting libgcc (the FP emulation part) to the floating point format being used. In other words, if you want soft-fp for IEEE float, the job should be very simple because that has already been done. If you want soft-fp for CDC 6000 float, you have to do a full implementation of that. paul
Re: negative latencies
On 05/19/2014 02:13 AM, shmeel gutl wrote: Are there hooks in gcc to deal with negative latencies? In other words, an architecture that permits an instruction to use a result from an instruction that will be issued later. Could you explain more on *an example* what are you trying to achieve with the negative latency. Scheduler is based on a critical path algorithm. Generally speaking latency time can be negative for this algorithm. But I guess that is not what you are asking. At first glance it seems that it will will break a few things. 1) The definition of dependencies cannot come from the simple ordering of rtl. 2) The scheduling problem starts to look like get off the train 3 stops before me. 3) The definition of live ranges needs to use actual instruction timing information, not just instruction sequencing. The hooks in the scheduler seem to be enough to stop damage but not enough to take advantage of this feature.
Re: Roadmap for 4.9.1, 4.10.0 and onwards?
On Tue, 2014-05-20 at 11:09 +0100, Bruce Adams wrote: Hi, I've been tracking the latest releases of gcc since 4.7 or so (variously interested in C++1y support, cilk and openmp). One thing I've found hard to locate is information about planned inclusions for future releases. As much relies on unpredictable community contributions I don't expect there to be a concrete or reliable plan. However, equally I'm sure the steering committee have some ideas over what ought to be upcoming releases. As a whole, the steering committee does not have any idea, because GCC development is based upon volunteer contributions. However, some members of the steering committee might work in large organization having a team of GCC contributors. That team might have its own (private) agenda. But every patch has to be approved by someone else. So I don't think that the steering committee knows a lot more than you and me. Regards. -- Basile STARYNKEVITCH http://starynkevitch.net/Basile/ email: basileatstarynkevitchdotnet mobile: +33 6 8501 2359 8, rue de la Faiencerie, 92340 Bourg La Reine, France *** opinions {are only mine, sont seulement les miennes} ***
RE: Roadmap for 4.9.1, 4.10.0 and onwards?
-Original Message- From: gcc-ow...@gcc.gnu.org [mailto:gcc-ow...@gcc.gnu.org] On Behalf Of Basile Starynkevitch Sent: 20 May 2014 16:29 To: Bruce Adams Cc: gcc@gcc.gnu.org Subject: Re: Roadmap for 4.9.1, 4.10.0 and onwards? On Tue, 2014-05-20 at 11:09 +0100, Bruce Adams wrote: Hi, I've been tracking the latest releases of gcc since 4.7 or so (variously interested in C++1y support, cilk and openmp). One thing I've found hard to locate is information about planned inclusions for future releases. As much relies on unpredictable community contributions I don't expect there to be a concrete or reliable plan. However, equally I'm sure the steering committee have some ideas over what ought to be upcoming releases. As a whole, the steering committee does not have any idea, because GCC development is based upon volunteer contributions. I understand the argument but I am not sure it's the way to go. Even if the project is based on volunteer contributions it would be interesting to have a tentative roadmap. This, I would think, would also help possible beginner volunteers know where to start if they wanted to contribute to the project. So the roadmap could be a list of features (big or small) of bug fixes that we would like fixed for a particular version. Even if we don't want to name it roadmap it would still be interesting to have a list of things that are being worked on or on the process of being merged into mainline and therefore will make it to the next major version. That being said I know it's hard to set sometime apart to write this kind of thing given most of us prefer to be hacking on GCC. From a newcomer point of view, however, not having things like a roadmap makes it look like the project is heading nowhere.
Re: soft-fp functions support without using libgcc
On Tue, May 20, 2014 at 7:37 AM, Sheheryar Zahoor Qazi sheheryar.zahoor.q...@gmail.com wrote: If you have a working compiler that is missing some functions provided by libgcc, that should be sufficient to build libgcc. Meaning that even if i am unable build libgcc to my new architecture, I should be able to able to provide soft-fp support to the architecture? You need to build soft-fp as part of libgcc. What I am saying is that you don't need soft-fp support in order to build libgcc. Btw i get the following error when i build gcc: configure:2627: error: in `/target-arch/target-arch-gcc/builddir/target-arch/libgcc': configure:2630: error: cannot compute suffix of object files: cannot compile You need to look in target-arch/libgcc/config.log to see what the problem is. And regarding soft-fp, I get the following error when i use soft-fp functions in a test program: : In function `test': (.text+0x0): undefined reference to `__floatsisf' In function `test': : In function `test': (.text+0x2c): undefined reference to `__mulsf3' : In function `test': (.text+0x2e): undefined reference to `__fixsfsi' Is this due to libgcc build fail or it just linking error? It's because libgcc was not built. Ian
Re: Roadmap for 4.9.1, 4.10.0 and onwards?
- Original Message - From: Paulo Matos pma...@broadcom.com To: Basile Starynkevitch bas...@starynkevitch.net; Bruce Adams tortoise...@yahoo.co.uk Cc: gcc@gcc.gnu.org gcc@gcc.gnu.org Sent: Tuesday, May 20, 2014 5:04 PM Subject: RE: Roadmap for 4.9.1, 4.10.0 and onwards? -Original Message- From: gcc-ow...@gcc.gnu.org [mailto:gcc-ow...@gcc.gnu.org] On Behalf Of Basile Starynkevitch Sent: 20 May 2014 16:29 To: Bruce Adams Cc: gcc@gcc.gnu.org Subject: Re: Roadmap for 4.9.1, 4.10.0 and onwards? On Tue, 2014-05-20 at 11:09 +0100, Bruce Adams wrote: Hi, I've been tracking the latest releases of gcc since 4.7 or so (variously interested in C++1y support, cilk and openmp). One thing I've found hard to locate is information about planned inclusions for future releases. As much relies on unpredictable community contributions I don't expect there to be a concrete or reliable plan. However, equally I'm sure the steering committee have some ideas over what ought to be upcoming releases. As a whole, the steering committee does not have any idea, because GCC development is based upon volunteer contributions. I understand the argument but I am not sure it's the way to go. Even if the project is based on volunteer contributions it would be interesting to have a tentative roadmap. This, I would think, would also help possible beginner volunteers know where to start if they wanted to contribute to the project. So the roadmap could be a list of features (big or small) of bug fixes that we would like fixed for a particular version. Even if we don't want to name it roadmap it would still be interesting to have a list of things that are being worked on or on the process of being merged into mainline and therefore will make it to the next major version. That being said I know it's hard to set sometime apart to write this kind of thing given most of us prefer to be hacking on GCC. From a newcomer point of view, however, not having things like a roadmap makes it look like the project is heading nowhere. If you think of gcc as a large distributed agile project the road map may be buried somewhere in the bug database. Perhaps its a matter of mining the relevant details or encouraging practices that make them mineable? The bugzilla has fields for assignee, priority and target milestone that could be used as hints. The trouble is its very low level. The intent is buried in the communities subjective interpretation of priority. I don't know how well that mirrors the actual values in the priority fields. I wouldn't expect it to without a conscious effort. If I search for ALL cilk 4.9 or ALL cilk it is still not obvious that the cilk branch was merged into main prior to release 4.9.0. Though that could be down to my unfamiliarity with more complex queries in bugzilla. Regards, Bruce.
Re: Roadmap for 4.9.1, 4.10.0 and onwards?
On 05/20/14 04:09, Bruce Adams wrote: Hi, I've been tracking the latest releases of gcc since 4.7 or so (variously interested in C++1y support, cilk and openmp). One thing I've found hard to locate is information about planned inclusions for future releases. As much relies on unpredictable community contributions I don't expect there to be a concrete or reliable plan. However, equally I'm sure the steering committee have some ideas over what ought to be upcoming releases. Is this published anywhere? The steering committee doesn't get involved in that aspect of development. It's just not in the committee's charter. There is no single roadmap for the GCC project and that's a direct result of the decentralized development. Looking forward to the next major GCC release (4.10 or 5.0): At a high level, wrapping up the C++11 ABI transition is high on the list for the next major GCC release. As is the ongoing efforts to clean up the polymorphism in gimple (and maybe RTL). Those aren't really user visible features, but they're a ton of work. I'm hoping the Intel team can push the last remaining Cilk+ feature through (Cilk_for). Jakub is working on Fortran support for OpenMP4. Others are working on OpenACC support. Richi's work on folding looks promising, but I'm not sure of its relative priority. There's work to bring AArch64 and Power 8 to first class support... Honza's work on IPA, etc etc. C++14 support will continue to land as bits are written. I'm certainly missing lots of important stuff... WRT to gcc-4.9.1, like most (all?) point releases, it's primarily meant to address bugs in the prior release. I wouldn't expect significant features to be appearing in 4.9.x releases. Jeff
Re: Weird startup issue with -fsplit-stack
On Tue, May 20, 2014 at 7:18 AM, Dmitry Antipov dmanti...@yandex.ru wrote: I'm trying to support -fsplit-stack in GNU Emacs. The most important problem is that GC uses conservative scanning of a C stack, so I need to iterate over stack segments. I'm doing this by using __splitstack_find, as described in libgcc/generic-morestack.c; but now I'm facing the weird issue with startup: Core was generated by `./temacs --batch --load loadup bootstrap'. Program terminated with signal SIGSEGV, Segmentation fault. #0 __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:486 486 pushq %rax (gdb) bt 10 #0 __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:486 #1 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #2 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 #3 0x005f15df in __morestack () at ../../../gcc-4.9.0/libgcc/config/i386/morestack.S:502 This is the call to __morestack_block_signals in morestack.S. It should only be possible if __morestack_block_signals or something it calls directly has a split stack. __morestack_block_signals has the no_split_stack attribute, meaning that it should never call __morestack. __morestack_block_signals only calls pthread_sigmark or sigprocmask, neither of which should be compiled with -fsplit-stack. So something has gone wrong, but I don't know what. I would recommend tracing the code instruction by instruction to see why __morestack_block_signals calls back into __morestack. Or, if that analysis is wrong, see what else is happening. I can advise but I don't have time to look at this in detail. Sorry. Ian
Re: Roadmap for 4.9.1, 4.10.0 and onwards?
If I search for ALL cilk 4.9 or ALL cilk it is still not obvious that the cilk branch was merged into main prior to release 4.9.0. Though that could be down to my unfamiliarity with more complex queries in bugzilla. Our bugzilla is usually used for tracking bugs, not merging of feature branches. https://gcc.gnu.org/gcc-4.9/changes.html#c-family announces the addition of Cilk Plus. Merges of major new features should probably also be announced on https://gcc.gnu.org/news.html
Re: negative latencies
On 20-May-14 06:13 PM, Vladimir Makarov wrote: On 05/19/2014 02:13 AM, shmeel gutl wrote: Are there hooks in gcc to deal with negative latencies? In other words, an architecture that permits an instruction to use a result from an instruction that will be issued later. Could you explain more on *an example* what are you trying to achieve with the negative latency. Scheduler is based on a critical path algorithm. Generally speaking latency time can be negative for this algorithm. But I guess that is not what you are asking. The architecture has an exposed pipeline where instructions read registers during the required cycle. So if one instruction produces its results in the third pipeline stage and a second instruction reads the register in the sixth pipeline stage, the second instruction can read the results of the first instruction even if it is issued three cycles earlier. The problem that I see is that the haifa scheduler schedules one cycle at a time, in a forward order, by picking from a list of instructions that can be scheduled without delays. So, in the above example, if instruction one is scheduled during cycle 3, it can't schedule instruction two during cycle 0, 1, or 2 because its producer dependency (instruction one) hasn't been scheduled yet. It won't be able to schedule it until cycle 3. So I am asking if there is an existing mechanism to back schedule instruction two once instruction one is issued. Thanks, Shmeel At first glance it seems that it will will break a few things. 1) The definition of dependencies cannot come from the simple ordering of rtl. 2) The scheduling problem starts to look like get off the train 3 stops before me. 3) The definition of live ranges needs to use actual instruction timing information, not just instruction sequencing. The hooks in the scheduler seem to be enough to stop damage but not enough to take advantage of this feature.
Re: Zero/Sign extension elimination using value ranges
On 20/05/14 16:52, Jakub Jelinek wrote: On Tue, May 20, 2014 at 12:27:31PM +1000, Kugan wrote: 1. Handling NOP_EXPR or CONVERT_EXPR that are in the IL because they are required for type correctness. We have two cases here: A) Mode is smaller than word_mode. This is usually from where the zero/sign extensions are showing up in final assembly. For example : int = (int) short which usually expands to (set (reg:SI ) (sext:SI (subreg:HI (reg:SI We can expand this (set (reg:SI ) (((reg:SI If following is true: 1. Value stored in RHS and LHS are of the same signedness 2. Type can hold the value. i.e., In cases like char = (char) short, we check that the value in short is representable char type. (i.e. look at the value range in RHS SSA_NAME and see if that can be represented in types of LHS without overflowing) Subreg here is not a paradoxical subreg. We are removing the subreg and zero/sign extend here. I am assuming here that QI/HI registers are represented in SImode (basically word_mode) with zero/sign extend is used as in (zero_extend:SI (subreg:HI (reg:SI 117)). Wouldn't it be better to just set proper flags on the SUBREG based on value range info (SUBREG_PROMOTED_VAR_P and SUBREG_PROMOTED_UNSIGNED_P)? Then not only the optimizers could eliminate in zext/sext when possible, but all other optimizations could benefit from that. Thanks for the comments. Here is an attempt (attached) that sets SUBREG_PROMOTED_VAR_P based on value range into. Is this the good place to do this ? Thanks, Kugan diff --git a/gcc/cfgexpand.c b/gcc/cfgexpand.c index b7f6360..d23ae76 100644 --- a/gcc/cfgexpand.c +++ b/gcc/cfgexpand.c @@ -3120,6 +3120,60 @@ expand_return (tree retval) } } + +static bool +is_assign_promotion_redundant (struct separate_ops *ops) +{ + double_int type_min, type_max; + double_int min, max; + bool uns = TYPE_UNSIGNED (ops-type); + double_int msb; + + /* We remove extension for integral stmts. */ + if (!INTEGRAL_TYPE_P (ops-type)) +return false; + + if (TREE_CODE_CLASS (ops-code) == tcc_unary) +{ + switch (ops-code) + { + case CONVERT_EXPR: + case NOP_EXPR: + + /* Get the value range. */ + if (TREE_CODE (ops-op0) != SSA_NAME + || POINTER_TYPE_P (TREE_TYPE (ops-op0)) + || get_range_info (ops-op0, min, max) != VR_RANGE) + return false; + + msb = double_int_one.rshift (TYPE_PRECISION (TREE_TYPE (ops-op0))); + if (!uns min.cmp (msb, uns) == 1 + max.cmp (msb, uns) == 1) + { + min = min.sext (TYPE_PRECISION (TREE_TYPE (ops-op0))); + max = max.sext (TYPE_PRECISION (TREE_TYPE (ops-op0))); + } + + /* Signedness of LHS and RHS should match or value range of RHS +should be all positive values to make zero/sign extension redundant. */ + if ((uns != TYPE_UNSIGNED (TREE_TYPE (ops-op0))) + (min.cmp (double_int_zero, TYPE_UNSIGNED (TREE_TYPE (ops-op0))) == -1)) + return false; + + type_max = tree_to_double_int (TYPE_MAX_VALUE (ops-type)); + type_min = tree_to_double_int (TYPE_MIN_VALUE (ops-type)); + + /* If rhs value range fits lhs type, zero/sign extension is + redundant. */ + if (max.cmp (type_max, uns) != 1 + (type_min.cmp (min, uns)) != 1) + return true; + } +} + + return false; +} + /* A subroutine of expand_gimple_stmt, expanding one gimple statement STMT that doesn't require special handling for outgoing edges. That is no tailcalls and no GIMPLE_COND. */ @@ -3240,6 +3294,12 @@ expand_gimple_stmt_1 (gimple stmt) } ops.location = gimple_location (stmt); + if (promoted is_assign_promotion_redundant (ops)) + { + promoted = false; + SUBREG_PROMOTED_VAR_P (target) = 0; + } + /* If we want to use a nontemporal store, force the value to register first. If we store into a promoted register, don't directly expand to target. */
Re: Roadmap for 4.9.1, 4.10.0 and onwards?
On 05/20/14 04:09, Bruce Adams wrote: Hi, I've been tracking the latest releases of gcc since 4.7 or so (variously interested in C++1y support, cilk and openmp). One thing I've found hard to locate is information about planned inclusions for future releases. As much relies on unpredictable community contributions I don't expect there to be a concrete or reliable plan. However, equally I'm sure the steering committee have some ideas over what ought to be upcoming releases. Is this published anywhere? The steering committee doesn't get involved in that aspect of development. It's just not in the committee's charter. There is no single roadmap for the GCC project and that's a direct result of the decentralized development. Looking forward to the next major GCC release (4.10 or 5.0): At a high level, wrapping up the C++11 ABI transition is high on the list for the next major GCC release. As is the ongoing efforts to clean up the polymorphism in gimple (and maybe RTL). Those aren't really user visible features, but they're a ton of work. I'm hoping the Intel team can push the last remaining Cilk+ feature through (Cilk_for). Jakub is working on Fortran support for OpenMP4. Others are working on OpenACC support. Richi's work on folding looks promising, but I'm not sure of its relative priority. There's work to bring AArch64 and Power 8 to first class support... Honza's work on IPA, etc etc. For IPA/FDO I think we are on track to merge some of more interesting Google's changes (autoFDO, perhaps LIPO and other FDO improvements) and Martin's pass for merging identical code. I am personally trying to focus on two things - first is to cleanup APIs of symbol table and IPA infrastructure after the C++ conversion and try to get things working well for LTO of large binaries - this is important change for optimizers, since we go from units consisting of hundred functions to units consiting of million of functions and heuristics needs to retune. And I also hope we will continue pushing bits making LTO more transparent and reliable (command line arguments, debug info etc.) Honza C++14 support will continue to land as bits are written. I'm certainly missing lots of important stuff... WRT to gcc-4.9.1, like most (all?) point releases, it's primarily meant to address bugs in the prior release. I wouldn't expect significant features to be appearing in 4.9.x releases. Jeff
Reducing Register Pressure through Live range Shrinking through Loops!!
Hello All: Simpson does the Live range shrinking and reduction of register pressure by using the computation that are not load and store but the arithmetic computation. The computation where the operands and registers are live at the entry and exit of the basic block but not touched inside the block then the computation is moved at the end of the block the reducing the register pressure inside the block by one. Extension of the Simpson work by extending the computation not being touched inside the basic block to the spanning of the Loops. If the Live ranges spans the Loops and live at the entry and exit of the Loop but the computation is not being touched inside the Loops then the computation is moved after the exit of the Loop. REDUCTION OF REGISTER PRESSURE THROUGH LIVE RANGE SHRINKING INSIDE THE LOOPS for each Loop starting from inner to outer do the following begin RELIEFIN(i) = null if i is the entry of the cfg. Else For all predecessors j RELIEFOUT(j) RELIEFOUT(i) = RELIEFIN(i) exposed union relief INSERT(I,j) = RELIEFOUT(i) RELIEFIN(i) Intersection Live(i) end The Simpson approach does takes the nesting depth into consideration of placing the computation and the relieve of the register pressure. Simpson approach doesn't takes into consideration the computation which spans throughout the loop and the operands and results are live at the entry of the Loop and exit of the Loop but not touched inside the Loops can be useful in reduction of register pressure inside the Loops. This approach will be useful in Region Based Register Allocator for Live Range Splitting at the Region Boundaries. Extension of the Simpson approach is to consider the data flow analysis with respect to the given Loop rather than having it for entire control flow graph. This data flow analysis starts from the inner loop and extends it to the outer loop. If the reference is not through the nested depth or with some depth then the computation can be placed accordingly. For register allocator by Graph coloring the live ranges that are with respect to operands and results of the computation are taken into consideration and for the above approach put into the stack during simplification phase of Graph Coloring so that there is a chance of getting such Live ranges colorable and thus reduces the register pressure. This is extended to splitting approach based on containment of Live ranges OPTIMAL PLACEMENT OF THE COMPUTATION FOR SINGLE ENTRY AND MULTIPLE EXIT LOOPS The placement of the computation to reduce the register pressure for Single Entry and Multiple exit by Simpson approach lead to unoptimal solution. The unoptimal Solution is because of the exit node of the loop does not post dominates all the basic block inside the Loops. Due to this the placement of the computation just after the tail block of the Loop will lead to incorrect results. In order to perform the Optimal Solution of the placement of the computation, the computation needs to be placed the block just after all the exit points of the Loop reconverge and which will post dominates all the blocks of the Loops. This will take care of reducing the register pressure for the Loops that are single Entry and Multiple Exit. For irreducible Loops the optimization to convert to reducible is done before the register allocation that reduces the register pressure and will be applicable to structured control flow and thus reduces the register pressure. The Live range shrinkage reducing register pressure takes load and store into consideration but not computation as proposed by Simpson. I am proposing to extend in GCC for the computation to reduce register pressure and for the Loop as given above for both Single Entry and Single Exit and Single Entry and Multiple Exit Loops. Please let me know what do you think. Thanks Regards Ajit
Re: Weird startup issue with -fsplit-stack
On 05/20/2014 10:16 PM, Ian Lance Taylor wrote: This is the call to __morestack_block_signals in morestack.S. It should only be possible if __morestack_block_signals or something it calls directly has a split stack. __morestack_block_signals has the no_split_stack attribute, meaning that it should never call __morestack. __morestack_block_signals only calls pthread_sigmark or sigprocmask, neither of which should be compiled with -fsplit-stack. So something has gone wrong, but I don't know what. Thanks - that was an application's own copy of pthread_sigmask (compiled with -fsplit-stack) linked into the binary due to a subtle configuration issue. The next major problem is that -fsplit-stack code randomly crashes with the useless gdb backtrace, usually pointing to the very beginning of the function (plus occasional Cannot access memory at... messages), e.g.: (gdb) bt 1 #0 0x005a615b in mark_object (arg=0) at ../../trunk/src/alloc.c:6039 6037 void 6038 mark_object (Lisp_Object arg) == 6039 { IIUC this usually (with traditional stack) happens due to stack overflow. But what may be the case with -fsplit-stack? I do not receive any error messages from libgcc, and there are a lot of free heap memory. If that matters, mark_object is recursive, and recursion depth may be very high, up to a few tens of thousands calls. Dmitry
[Bug middle-end/61243] New: [4.10 Regression] verify_flow_info failed: No region crossing jump at section boundary in bb 65
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61243 Bug ID: 61243 Summary: [4.10 Regression] verify_flow_info failed: No region crossing jump at section boundary in bb 65 Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: middle-end Assignee: unassigned at gcc dot gnu.org Reporter: Joost.VandeVondele at mat dot ethz.ch Recent regression in the one day between good: r210596 bad: r210629 Unfortunately, only happens with -fprofile-use on our large application (CP2K). Any suggestion on how to get this into a testcase ? I can provide unreduced .F, .mod and .gcda files. /data/vjoost/gnu/cp2k/cp2k/makefiles/../src/qs_integrate_potential_low.F:526:0: error: No region crossing jump at section boundary in bb 65 END SUBROUTINE xyz_to_vab ^ /data/vjoost/gnu/cp2k/cp2k/makefiles/../src/qs_integrate_potential_low.F:526:0: internal compiler error: verify_flow_info failed 0x6edead verify_flow_info() ../../gcc/gcc/cfghooks.c:260 0xf60a88 try_optimize_cfg ../../gcc/gcc/cfgcleanup.c:2860 0xf60a88 cleanup_cfg(int) ../../gcc/gcc/cfgcleanup.c:3025 0xf5363b execute ../../gcc/gcc/bb-reorder.c:2353 Please submit a full bug report, with preprocessed source if appropriate. Please include the complete backtrace with any bug report. See http://gcc.gnu.org/bugs.html for instructions. make[2]: *** [qs_integrate_potential_low.o] Error 1
[Bug libstdc++/60758] Infinite backtrace in __cxa_end_cleanup
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60758 --- Comment #9 from Alexey Merzlyakov alexey.merzlyakov at samsung dot com --- The following PR has been opened for Thumb1 problem: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61223
[Bug middle-end/61243] [4.10 Regression] verify_flow_info failed: No region crossing jump at section boundary in bb 65
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61243 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added CC||rsandifo at gcc dot gnu.org --- Comment #1 from Uroš Bizjak ubizjak at gmail dot com --- The same failure is triggered with profiledbootstrap build. First recorded bootstrap failure is at r210603 [1]. Last good bootstrap is at r210599 [2]. It looks that r210603 is problematic [3]. Author CC'd. [1] https://gcc.gnu.org/ml/gcc-regression/2014-05/msg00224.html [2] https://gcc.gnu.org/ml/gcc-testresults/2014-05/msg01581.html [3] https://gcc.gnu.org/ml/gcc-cvs/2014-05/msg00646.html
[Bug middle-end/61243] [4.10 Regression] verify_flow_info failed: No region crossing jump at section boundary in bb 65
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61243 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added Status|UNCONFIRMED |NEW Last reconfirmed||2014-05-20 Target Milestone|--- |4.10.0 Ever confirmed|0 |1 --- Comment #2 from Uroš Bizjak ubizjak at gmail dot com --- Confirmed.
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 --- Comment #9 from Mukund Sivaraman muks at banu dot com --- Hi Jakub, Markus We discussed this during our daily standup call today, and there are two points we'd like to make: 1. The qsort() defintion in C99 doesn't explicitly state that base must not be NULL, though it seems you are deducing that from the initial element of which is pointed to by base. The POSIX definition of qsort() adds this: If the nel argument has the value zero, the comparison function pointed to by compar shall not be called and no rearrangement shall take place. 2. From our perpective as users of GCC, this kind of agressive optimization seems counter-intuitive. We'd like code to compile to correct object code first before performance. When the compiler knows at that point that base (=x) is NULL as an argument to qsort(), why isn't it warning when the attribute expects it to be non-NULL, esp. as it is using this inferred decision to optimize code down below? The compiler knows x is NULL at this point in this codepath regardless of what qsort()'s attributes say. Why is it using the attribute then? qsort() also does not assert (at runtime) that base is non-NULL. There is no way to detect this for code which used to run correctly before, but doesn't anymore (without it _hopefully_ crashing somewhere). Other similar functions such as memcpy(), etc. also have this annotation in glibc, whereas there is no definition of n=0 case in C99. This example of qsort() is in libc, but imagine a case where a program uses a 3rd party system installed utility shared library. If the library, in a new version, adds a nonnull annotation for a function, but the library function itself continues to work for NULL input, see what happens to the program: The library is not affected, but the pointer in the calling program is affected if the compiler infers that the pointer is non-NULL due to the attribute. The calling program is now buggy due to a change in the library. How do we discover it? It makes sense to just avoid the qsort() in our case and we will update our code to do so, but please consider the arguments above.
[Bug go/61244] New: gccgo: ICE in write_specific_type_functions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61244 Bug ID: 61244 Summary: gccgo: ICE in write_specific_type_functions Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: go Assignee: ian at airs dot com Reporter: dvyukov at google dot com Created attachment 32824 -- https://gcc.gnu.org/bugzilla/attachment.cgi?id=32824action=edit reproducer $ gccgo -v Using built-in specs. COLLECT_GCC=gccgo COLLECT_LTO_WRAPPER=libexec/gcc/x86_64-unknown-linux-gnu/4.10.0/lto-wrapper Target: x86_64-unknown-linux-gnu Configured with: ../configure --enable-languages=c,c++,go --disable-bootstrap --enable-checking=yes --disable-multilib --prefix=gcc_trunk/install Thread model: posix gcc version 4.10.0 20140516 (experimental) (GCC) The reproducer program is attached. $ go build -compiler=gccgo 0.go go1: internal compiler error: in write_specific_type_functions, at go/gofrontend/types.cc:1819 0x628f2a Type::write_specific_type_functions(Gogo*, Named_type*, std::basic_stringchar, std::char_traitschar, std::allocatorchar const, Function_type*, std::basic_stringchar, std::char_traitschar, std::allocatorchar const, Function_type*) ../../gcc/go/gofrontend/types.cc:1819 0x6276bf Type::specific_type_functions(Gogo*, Named_type*, Function_type*, Function_type*, Named_object**, Named_object**) ../../gcc/go/gofrontend/types.cc:1798 0x627c86 Type::type_functions(Gogo*, Named_type*, Function_type*, Function_type*, Named_object**, Named_object**) ../../gcc/go/gofrontend/types.cc:1675 0x630194 Type::type_descriptor_constructor(Gogo*, int, Named_type*, Methods const*, bool) ../../gcc/go/gofrontend/types.cc:2016 0x6321f3 Array_type::array_type_descriptor(Gogo*, Named_type*) ../../gcc/go/gofrontend/types.cc:6121 0x62a5db Type::make_type_descriptor_var(Gogo*) ../../gcc/go/gofrontend/types.cc:1249 0x62a796 Type::type_descriptor_pointer(Gogo*, Location) ../../gcc/go/gofrontend/types.cc:1206 0x5bf412 Type_descriptor_expression::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/expressions.cc:13863 0x5bb6eb Call_expression::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/expressions.cc:9425 0x5a23f6 Unary_expression::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/expressions.cc:4039 0x607491 If_statement::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/statements.cc:3200 0x5db097 Block::get_backend(Translate_context*) ../../gcc/go/gofrontend/gogo.cc:5454 0x60741c Block_statement::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/statements.cc:1811 0x5db097 Block::get_backend(Translate_context*) ../../gcc/go/gofrontend/gogo.cc:5454 0x60741c Block_statement::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/statements.cc:1811 0x5db097 Block::get_backend(Translate_context*) ../../gcc/go/gofrontend/gogo.cc:5454 0x60741c Block_statement::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/statements.cc:1811 0x5db097 Block::get_backend(Translate_context*) ../../gcc/go/gofrontend/gogo.cc:5454 0x60741c Block_statement::do_get_backend(Translate_context*) ../../gcc/go/gofrontend/statements.cc:1811 0x5db097 Block::get_backend(Translate_context*) ../../gcc/go/gofrontend/gogo.cc:5454
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 Jakub Jelinek jakub at gcc dot gnu.org changed: What|Removed |Added Status|WAITING |RESOLVED Resolution|--- |INVALID --- Comment #10 from Jakub Jelinek jakub at gcc dot gnu.org --- If you believe the nonnull attribute on qsort is incorrect, then you should report that as glibc bug, not gcc bug, the prototype is provided by glibc. The more aggressive GCC optimization is documented e.g. in https://gcc.gnu.org/gcc-4.9/porting_to.html plus we hope to add -fsanitize=undefined instrumentation for this in the upcoming GCC version, so you find it out more easily. When the compiler knows at that point that base (=x) is NULL as an argument to qsort(), why isn't it warning when the attribute expects it to be non-NULL, esp. as it is using this inferred decision to optimize code down below? But the compiler doesn't know there that x is NULL. The compiler sees a call to a function which must not be called with NULL, and from that derives the value range of x to be anything but NULL. Instead of qsort consider here some less controversial function, e.g. memcpy, where the standard is very clear that memcpy (NULL, , 0); or memcpy (, NULL, 0); is invalid despite the length 0.
[Bug tree-optimization/61245] New: ICE at in expand_ANNOTATE, at internal-fn.c:127 called from cfgexpand.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61245 Bug ID: 61245 Summary: ICE at in expand_ANNOTATE, at internal-fn.c:127 called from cfgexpand.c Product: gcc Version: 4.9.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: tree-optimization Assignee: unassigned at gcc dot gnu.org Reporter: vincenzo.innocente at cern dot ch apologize for not reducing (trivial reduction (bar below) works) given cat NaiveDod.cc #includearray #includevector #includeutility unsigned int N; float * a, *b, *c; void bar() { #pragma GCC ivdep for (auto i=0U; iN; ++i) a[i] = b[i]*c[i]; } templateint N struct SoA { using s_t = unsigned int; using Ind = unsigned int; auto size() const { return m_n;} float operator()(Ind i, Ind j) { return data[j][i];} float const operator()(Ind i, Ind j) const { return data[j][i];} std::arraystd::vectorfloat,N data; s_t m_n=0; }; templateint N void doT(SoAN soa) { #pragma GCC ivdep for (auto i=0U; isoa.size(); ++i) soa(i,0) = soa(i,1)*soa(i,2); } void doIt(SoA3 soa) { doT(soa); } produces c++ -std=c++1y -Ofast -Wall -fopt-info-vec -fno-tree-slp-vectorize -march=nehalem -S NaiveDod.cc NaiveDod.cc:10:17: note: loop vectorized NaiveDod.cc:10:17: note: loop peeled for vectorization to enhance alignment NaiveDod.cc: In function 'void doIt(SoA3)': NaiveDod.cc:34:17: internal compiler error: in expand_ANNOTATE, at internal-fn.c:127 for (auto i=0U; isoa.size(); ++i) ^ 0x9e9a97 expand_ANNOTATE ../../gcc-trunk/gcc/internal-fn.c:127 0x820a7a expand_call_stmt ../../gcc-trunk/gcc/cfgexpand.c:2236 0x820a7a expand_gimple_stmt_1 ../../gcc-trunk/gcc/cfgexpand.c:3202 0x820a7a expand_gimple_stmt ../../gcc-trunk/gcc/cfgexpand.c:3354 0x821aee expand_gimple_basic_block ../../gcc-trunk/gcc/cfgexpand.c:5194 0x823746 execute ../../gcc-trunk/gcc/cfgexpand.c:5803 Please submit a full bug report, with preprocessed source if appropriate. Please include the complete backtrace with any bug report. See http://gcc.gnu.org/bugs.html for instructions. c++ -v Using built-in specs. COLLECT_GCC=c++ COLLECT_LTO_WRAPPER=/afs/cern.ch/user/i/innocent/w4/libexec/gcc/x86_64-unknown-linux-gnu/4.10.0/lto-wrapper Target: x86_64-unknown-linux-gnu Configured with: ../gcc-trunk//configure --prefix=/afs/cern.ch/user/i/innocent/w4 --enable-languages=c,c++,lto,fortran -enable-gold=yes --enable-lto --with-gmp-lib=/usr/local/lib64 --with-mpfr-lib=/usr/local/lib64 -with-mpc-lib=/usr/local/lib64 --enable-cloog-backend=isl --with-cloog=/usr/local --with-ppl-lib=/usr/local/lib64 -enable-libitm -disable-multilib Thread model: posix gcc version 4.10.0 20140520 (experimental) [trunk revision 210630] (GCC)
[Bug middle-end/61225] [4.10 Regression] Several new failures after r210458 on x86_64-*-* with -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61225 --- Comment #3 from zhenqiang.chen at linaro dot org --- I can not reproduce gcc.dg/guality/pr43051-1.c fail with options -fno-diagnostics-show-caret -fdiagnostics-color=never -O3 -fomit-frame-pointer -funroll-all-loops -m32 -mtune=core2 What are your final options to build the test case?
[Bug go/61246] New: gccgo: ICE in do_determine_types [GoSmith]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61246 Bug ID: 61246 Summary: gccgo: ICE in do_determine_types [GoSmith] Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: go Assignee: ian at airs dot com Reporter: dvyukov at google dot com $ gccgo -v Using built-in specs. COLLECT_GCC=gccgo COLLECT_LTO_WRAPPER=libexec/gcc/x86_64-unknown-linux-gnu/4.10.0/lto-wrapper Target: x86_64-unknown-linux-gnu Configured with: ../configure --enable-languages=c,c++,go --disable-bootstrap --enable-checking=yes --disable-multilib --prefix=gcc_trunk/install Thread model: posix gcc version 4.10.0 20140516 (experimental) (GCC) The reproducer is attached. $ go build -compiler=gccgo 0.go go1: internal compiler error: in do_determine_types, at go/gofrontend/statements.cc:400 0x6078ed Temporary_statement::do_determine_types() ../../gcc/go/gofrontend/statements.cc:400 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x6073fa If_statement::do_determine_types() ../../gcc/go/gofrontend/statements.cc:3166 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x6073fa If_statement::do_determine_types() ../../gcc/go/gofrontend/statements.cc:3166 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x6073fa If_statement::do_determine_types() ../../gcc/go/gofrontend/statements.cc:3166 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x5d2d4b Block::determine_types() ../../gcc/go/gofrontend/gogo.cc:5410 0x5d5ad8 Function::determine_types() ../../gcc/go/gofrontend/gogo.cc:4562 0x5d5ad8 Gogo::determine_types() ../../gcc/go/gofrontend/gogo.cc:2755 0x5cea3c go_parse_input_files(char const**, unsigned int, bool, bool) ../../gcc/go/gofrontend/go.cc:107
[Bug go/61246] gccgo: ICE in do_determine_types [GoSmith]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61246 --- Comment #1 from Dmitry Vyukov dvyukov at google dot com --- Created attachment 32825 -- https://gcc.gnu.org/bugzilla/attachment.cgi?id=32825action=edit reproducer
[Bug tree-optimization/61247] New: vectorization fails if conversion from unsigned int to signed int is involved
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61247 Bug ID: 61247 Summary: vectorization fails if conversion from unsigned int to signed int is involved Product: gcc Version: 4.9.0 Status: UNCONFIRMED Severity: minor Priority: P3 Component: tree-optimization Assignee: unassigned at gcc dot gnu.org Reporter: vincenzo.innocente at cern dot ch in the following example cat uintLoop.cc unsigned int N; float * a, *b, *c; using Ind = /*unsigned*/ int; inline float val(float * x, Ind i) { return x[i];} inline float const val(float const * x, Ind i) { return x[i];} void foo() { #pragma GCC ivdep for (auto i=0U; iN; ++i) val(a,i) = val(b,i)*val(c,i); } using Ind = /*unsigned*/ int; does not vectorize with c++ -std=c++1y -Ofast -Wall -fopt-info-vec-missed -fno-tree-slp-vectorize -march=nehalem -S uintLoop.cc uintLoop.cc:12:17: note: not vectorized: not suitable for gather load _8 = *_17; uintLoop.cc:12:17: note: bad data references. using Ind = unsigned int; vectorize fine minor, just annoying
[Bug tree-optimization/61247] vectorization fails if conversion from unsigned int to signed int is involved
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61247 Andrew Pinski pinskia at gcc dot gnu.org changed: What|Removed |Added Target||LP64 Status|UNCONFIRMED |NEW Last reconfirmed||2014-05-20 Ever confirmed|0 |1 --- Comment #1 from Andrew Pinski pinskia at gcc dot gnu.org --- This is a LP64 issue. In that on x32 or x86 (or ILP32 with AARCH64), we can vectorize this.
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 --- Comment #11 from Andrew Pinski pinskia at gcc dot gnu.org --- (In reply to Jakub Jelinek from comment #10) If you believe the nonnull attribute on qsort is incorrect, then you should report that as glibc bug, not gcc bug, the prototype is provided by glibc. The more aggressive GCC optimization is documented e.g. in https://gcc.gnu.org/gcc-4.9/porting_to.html plus we hope to add -fsanitize=undefined instrumentation for this in the upcoming GCC version, so you find it out more easily. It is not incorrect as the C standard says this about qsort: nmemb can have the value zero on a call to that function; the comparison function is not called, a search finds no matching element, and sorting performs no rearrangement. Pointer arguments on such a call shall still have valid values, as described in 7.1.4. POSIX 2008 defers to the C standard now so this is neither a glibc or a GCC bug in the end.
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 --- Comment #12 from Andrew Pinski pinskia at gcc dot gnu.org --- (In reply to Andrew Pinski from comment #11) It is not incorrect as the C standard says this about qsort: nmemb can have the value zero on a call to that function; the comparison function is not called, a search finds no matching element, and sorting performs no rearrangement. Pointer arguments on such a call shall still have valid values, as described in 7.1.4. 7.1.4 says this: Each of the following statements applies unless explicitly stated otherwise in the detailed descriptions that follow: If an argument to a function has an invalid value (such as a value outside the domain of the function, or a pointer outside the address space of the program, or a null pointer, or a pointer to non-modifiable storage when the corresponding parameter is not const-qualified) or a type (after promotion) not expected by a function with variable number of arguments, the behavior is undefined. So there is not need to say it was detected to be non-null as the null pointer case is mentioned in 7.1.4.
[Bug ipa/61144] [4.9/4.10 Regression] Invalid optimizations for extern vars with local weak definitions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61144 --- Comment #19 from Rich Felker bugdal at aerifal dot cx --- Here is the commit that seems to have introduced the bug: https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=df8d3e8981a99e264b49876f0f5064bdb30ac981 In the function ctor_for_folding, the following erroneous logic appears: + /* Non-readonly alias of readonly variable is also de-facto readonly, + because the variable itself is in readonly section. + We also honnor READONLY flag on alias assuming that user knows + what he is doing. */ + if (!TREE_READONLY (decl) !TREE_READONLY (real_decl)) +return error_mark_node; This treats the value of an alias as a compile-time constant if either the alias itself or the alias target has TREE_READONLY being true. Replacing the with || seems to make the problem go away in my test case, and makes a bit more sense (perhaps that was the original intent?), but it's only sufficient if TREE_READONLY is always false for weak aliases (since they can always be overridden by a strong symbol from another translation unit, even if the alias is const-qualified). I'm not sure where the value of TREE_READONLY is set for aliases yet but I'll keep looking...
[Bug tree-optimization/61221] [4.10 Regression] ICE on valid code at -O1 and above on x86_64-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61221 --- Comment #5 from Richard Biener rguenth at gcc dot gnu.org --- Author: rguenth Date: Tue May 20 08:16:13 2014 New Revision: 210633 URL: http://gcc.gnu.org/viewcvs?rev=210633root=gccview=rev Log: 2014-05-20 Richard Biener rguent...@suse.de PR tree-optimization/61221 * tree-ssa-pre.c (el_to_update): Remove. (eliminate_dom_walker::before_dom_children): Handle released VDEFs by value-numbering them to the associated VUSE. Update stmt immediately for substituted call address. (eliminate): Remove delayed stmt updating code. * tree-ssa-sccvn.c (vuse_ssa_val): New function valueizing possibly late re-numbered vuses. (vn_reference_lookup_2): Adjust. (vn_reference_lookup_pieces): Likewise. (vn_reference_lookup): Likewise. Modified: trunk/gcc/ChangeLog trunk/gcc/tree-ssa-pre.c trunk/gcc/tree-ssa-sccvn.c
[Bug middle-end/61245] ICE at in expand_ANNOTATE, at internal-fn.c:127 called from cfgexpand.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61245 Richard Biener rguenth at gcc dot gnu.org changed: What|Removed |Added Status|UNCONFIRMED |NEW Last reconfirmed||2014-05-20 CC||burnus at gcc dot gnu.org, ||rguenth at gcc dot gnu.org Component|tree-optimization |middle-end Ever confirmed|0 |1 Known to fail||4.10.0, 4.9.0 --- Comment #1 from Richard Biener rguenth at gcc dot gnu.org --- Confirmed. An ANNOTATE survives CFG building: ;; 2 loops found ;; ;; Loop 0 ;; header 0, latch 1 ;; depth 0, outer -1 ;; nodes: 0 1 2 3 4 5 ;; ;; Loop 1 ;; header 4, latch 3 ;; depth 1, outer 0 ;; nodes: 4 3 void doT(SoAN) [with int N = 3] (struct SoA soa) { ... bb 3: D.30093 = SoA3::operator() (soa, i, 0); D.30094 = SoA3::operator() (soa, i, 1); D.30095 = *D.30094; D.30096 = SoA3::operator() (soa, i, 2); D.30097 = *D.30096; D.30098 = D.30095 * D.30097; *D.30093 = D.30098; i = i + 1; bb 4: D.30101 = SoA3::size (soa); D.30102 = D.30101 i; D.30100 = ANNOTATE (D.30102, 0); retval.4 = D.30100; if (retval.4 != 0) goto bb 3; else goto bb 5; bb 5:
[Bug go/61248] New: gccgo: spurious error: too many arguments [GoSmith]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61248 Bug ID: 61248 Summary: gccgo: spurious error: too many arguments [GoSmith] Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: go Assignee: ian at airs dot com Reporter: dvyukov at google dot com gcc version 4.10.0 20140516 (experimental) (GCC) Program: package main func main() { var f func(int, interface{}) go f(0, recover()) } $ go build -compiler=gccgo args.go args.go:5:10: error: too many arguments go f(0, recover())
[Bug c++/61245] ICE at in expand_ANNOTATE, at internal-fn.c:127 called from cfgexpand.c
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61245 Richard Biener rguenth at gcc dot gnu.org changed: What|Removed |Added Component|middle-end |c++ --- Comment #2 from Richard Biener rguenth at gcc dot gnu.org --- For some reason we create a temporary, probably because the very specific ANNOTATE is inside a cleanup_point: if (cleanup_point ANNOTATE_EXPR SoA3::size ((struct SoA *) soa) i, ivdep) goto D.28269; else goto D.28267; that is, the wrapped expressions has side-effects. We should build the ANNOTATE_EXPR outside of that cleanup. C++ FE support issue.
[Bug c++/61242] [4.9/4.10 Regression] Bogus no matching function for call to ‘Foo::Create(brace-enclosed initializer list)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61242 Richard Biener rguenth at gcc dot gnu.org changed: What|Removed |Added Target Milestone|--- |4.9.1
[Bug ipa/61144] [4.9/4.10 Regression] Invalid optimizations for extern vars with local weak definitions
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61144 --- Comment #20 from Rich Felker bugdal at aerifal dot cx --- On further investigation, it looks like the code I cited deals with strong aliases as well as weak ones, and in the strong alias case, the strong folding behavior might be desirable. A better fix seems to be adding an explicit check for weak aliases (DECL_WEAK(decl)) when an alias is found and returning error_mark_node in that case. Note that prior to the above-mentioned commit, the !TREE_READONLY(decl) case was always treated as non-foldable, so there seems to have been no subtlety to avoiding errors with weak aliases. But the new code performs much more aggressive constant folding and thus needs to avoid stepping on weak aliases.
[Bug rtl-optimization/61241] built-in memset makes the caller function slower
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61241 ktkachov at gcc dot gnu.org changed: What|Removed |Added CC||ktkachov at gcc dot gnu.org --- Comment #3 from ktkachov at gcc dot gnu.org --- Can you please send the patch to gcc-patc...@gcc.gnu.org including a ChangeLog
[Bug c/61240] [4.8/4.9/4.10 Regression] Incorrect warning integer overflow in expression on pointer-pointer subtraction
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61240 Richard Biener rguenth at gcc dot gnu.org changed: What|Removed |Added Status|UNCONFIRMED |NEW Last reconfirmed||2014-05-20 Target Milestone|--- |4.8.3 Summary|Incorrect warning integer |[4.8/4.9/4.10 Regression] |overflow in expression on |Incorrect warning integer |pointer-pointer subtraction |overflow in expression on ||pointer-pointer subtraction Ever confirmed|0 |1 --- Comment #1 from Richard Biener rguenth at gcc dot gnu.org --- We warn for integer_cst 0x76d54e28 type integer_type 0x76c407e0 long int constant public overflow 1 via c-common.c:overflow_warning called from #1 0x0065b80e in parser_build_binary_op (location=5653, code=MINUS_EXPR, arg1=..., arg2=...) at /space/rguenther/src/svn/trunk/gcc/c/c-typeck.c:3411 #2 0x0068f5ed in c_parser_binary_expression (parser=0x76d67000, after=0x0, omp_atomic_lhs=tree 0x0) at /space/rguenther/src/svn/trunk/gcc/c/c-parser.c:6282 #3 0x0068dfee in c_parser_conditional_expression ( parser=0x76d67000, after=0x0, omp_atomic_lhs=tree 0x0) at /space/rguenther/src/svn/trunk/gcc/c/c-parser.c:5934 #4 0x0068dd75 in c_parser_expr_no_commas (parser=0x76d67000, after=0x0, omp_atomic_lhs=tree 0x0) at /space/rguenther/src/svn/trunk/gcc/c/c-parser.c:5852 when building p - (p + -1U) which gets simplified to - -1U - 1U (with overflow set - as it's sizetype arithmetic). pointer_diff is guilty here which calls /* First do the subtraction as integers; then drop through to build the divide operator. Do not do default conversions on the minus operator in case restype is a short type. */ op0 = build_binary_op (loc, MINUS_EXPR, convert (inttype, op0), convert (inttype, op1), 0); doing 0 - -1U, converting them to inttype (long int) first. I suggest to do that conversion and strip overflow bits in the POINTER_PLUS_EXPR decomposition part.
[Bug target/60991] [avr] Stack corruption when using 24-bit integers __int24 or __memx pointers in large stack frame
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60991 --- Comment #7 from Georg-Johann Lay gjl at gcc dot gnu.org --- Author: gjl Date: Tue May 20 08:37:50 2014 New Revision: 210635 URL: http://gcc.gnu.org/viewcvs?rev=210635root=gccview=rev Log: gcc/ 2014-05-20 Senthil Kumar Selvaraj senthil_kumar.selva...@atmel.com Backport from mainline r210325 2014-05-12 Senthil Kumar Selvaraj senthil_kumar.selva...@atmel.com PR target/60991 * config/avr/avr.c (avr_out_store_psi): Use correct constant to restore Y. gcc/testsuite/ 2014-05-20 Senthil Kumar Selvaraj senthil_kumar.selva...@atmel.com Backport from mainline r210325 2014-05-12 Senthil Kumar Selvaraj senthil_kumar.selva...@atmel.com PR target/60991 * gcc.target/avr/pr60991.c: New testcase. Added: branches/gcc-4_7-branch/gcc/testsuite/gcc.target/avr/pr60991.c Modified: branches/gcc-4_7-branch/gcc/ChangeLog branches/gcc-4_7-branch/gcc/config/avr/avr.c branches/gcc-4_7-branch/gcc/testsuite/ChangeLog
[Bug rtl-optimization/61239] [4.10 Regression]: ICE in decompose, at rtl.h when compiling vshuf-v16hi.c using -mavx2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61239 Richard Biener rguenth at gcc dot gnu.org changed: What|Removed |Added Target Milestone|--- |4.10.0
[Bug debug/61237] gcc puts line number 11 ahead of line number 10 in function call in debug info
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61237 --- Comment #1 from Richard Biener rguenth at gcc dot gnu.org --- Well, '-foo (2, 3)' _has_ to be executed before the call to foo. line-numbers are a bad representation for C sequence point rules. I would say this is not a bug.
[Bug target/60991] [avr] Stack corruption when using 24-bit integers __int24 or __memx pointers in large stack frame
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60991 Georg-Johann Lay gjl at gcc dot gnu.org changed: What|Removed |Added Priority|P3 |P4 Status|UNCONFIRMED |RESOLVED Known to work||4.7.4, 4.8.3, 4.9.1 Keywords||wrong-code Resolution|--- |FIXED Target Milestone|--- |4.9.1 Known to fail|4.7.2, 4.8.1|4.7.3, 4.8.2, 4.9.0 Severity|critical|normal --- Comment #8 from Georg-Johann Lay gjl at gcc dot gnu.org --- Fixed in 4.7.4, 4.8.3, 4.9.1
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 --- Comment #13 from Jakub Jelinek jakub at gcc dot gnu.org --- (In reply to Andrew Pinski from comment #11) It is not incorrect as the C standard says this about qsort: nmemb can have the value zero on a call to that function; the comparison function is not called, a search finds no matching element, and sorting performs no rearrangement. Pointer arguments on such a call shall still have valid values, as described in 7.1.4. POSIX 2008 defers to the C standard now so this is neither a glibc or a GCC bug in the end. I've missed the Pointer arguments on such a call shall still have valid values, as described in 7.1.4. sentence in C99 7.20.5 (was looking for that in 7.20.5.2), with that it is exactly the same thing in this regard as memcpy etc.
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 --- Comment #14 from Mukund Sivaraman muks at banu dot com --- (In reply to Jakub Jelinek from comment #10) But the compiler doesn't know there that x is NULL. The compiler sees a See comment #3. It generates 2 codepaths, one where (nalloc == 0) and another where (nalloc != 0). For the former, it deletes the if statement and isc_mem_put() call at the free_rdatas label completely: free_rdatas: if (x != NULL) isc_mem_put(mctx, x, nalloc * sizeof(struct xrdata)); return (result); } and instead reduces free_rdata's definition to: free_rdatas: return (result); } How does the compiler do that if it has not inferred that x is NULL there? OTOH, you're the compiler developers, so if you say it doesn't know that x is NULL, then that is that. :) Maybe the part of compiler code that does this doesn't know it. Note that despite all this discussion of correctness, this optimization is counter intuitive and will bite developers. There should at least be warnings where they could be generated. The point about correctness with C standards is taken and agreed. See what is happening from a programmer's point of view: an explicit NULL check is deleted. There are no warnings about qsort() used with NULL arguments where it seems the compiler could warn (see above). Also consider the use of notnull as an API annotation change by 3rd party libraries, which can make caller code buggy without any way to notice it. At the very least, if it is possible to detect that the pointer is NULL by static analysis and it is being passed to a function that has the notnull attribute, please warn mentioning inferences being made.
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 --- Comment #15 from Andrew Pinski pinskia at gcc dot gnu.org --- (In reply to Mukund Sivaraman from comment #14) (In reply to Jakub Jelinek from comment #10) But the compiler doesn't know there that x is NULL. The compiler sees a See comment #3. It generates 2 codepaths, one where (nalloc == 0) and another where (nalloc != 0). For the former, it deletes the if statement and isc_mem_put() call at the free_rdatas label completely: free_rdatas: if (x != NULL) isc_mem_put(mctx, x, nalloc * sizeof(struct xrdata)); return (result); } and instead reduces free_rdata's definition to: free_rdatas: return (result); } How does the compiler do that if it has not inferred that x is NULL there? OTOH, you're the compiler developers, so if you say it doesn't know that x is NULL, then that is that. :) Maybe the part of compiler code that does this doesn't know it. Note that despite all this discussion of correctness, this optimization is counter intuitive and will bite developers. There should at least be warnings where they could be generated. The point about correctness with C standards is taken and agreed. See what is happening from a programmer's point of view: an explicit NULL check is deleted. There are no warnings about qsort() used with NULL arguments where it seems the compiler could warn (see above). Also consider the use of notnull as an API annotation change by 3rd party libraries, which can make caller code buggy without any way to notice it. At the very least, if it is possible to detect that the pointer is NULL by static analysis and it is being passed to a function that has the notnull attribute, please warn mentioning inferences being made. The warning did not make it into gcc 4.9 due to the patches to do the warning were not ready. Gcc 4.10 should warn about it. If it does not then that is a bug.
[Bug target/61249] New: _mm_frcz_ss, _mm_frcz_sd: __builtin_ia32_vfrczss, __builtin_ia32_vfrczsd require 2 arguments
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61249 Bug ID: 61249 Summary: _mm_frcz_ss, _mm_frcz_sd: __builtin_ia32_vfrczss, __builtin_ia32_vfrczsd require 2 arguments Product: gcc Version: 4.8.3 Status: UNCONFIRMED Severity: normal Priority: P3 Component: target Assignee: unassigned at gcc dot gnu.org Reporter: mt at debian dot org Looking at https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html on the one hand and AMD's AMD64 Architecture Programmer’s Manual Volume 6: 128-Bit and 256-Bit XOP and FMA4 Instructions on the other hand, vfrczss/vfrczsd require a second argument to specify the destination. Yet r205495 changed _mm_frcz_ss/_mm_frcz_sd so that only a single argument is passed to the __builtin_ia32_vfrczss/vfrczsd calls. This was detected at language level (inconsistent types), I can only speculate that this may cause invalid code to be generated (or null operands). Best, Michael PS.: The problem persists in the 4.9 branch as xopintrin.h hasn't been touched since 4.8.
[Bug fortran/61126] [4.10 Regression] gfortran does not enable -Wununused-parameter with -Wextra
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61126 --- Comment #23 from Dominique d'Humieres dominiq at lps dot ens.fr --- Adding -Wall to the dg-options let the test succeed (in line with the gfortran manual, see comment 21): --- ../_clean/gcc/testsuite/gfortran.dg/wextra_1.f2012-10-21 13:06:18.0 +0200 +++ gcc/testsuite/gfortran.dg/wextra_1.f2014-05-15 12:05:56.0 +0200 @@ -1,5 +1,5 @@ ! { dg-do compile } -! { dg-options -Wextra } +! { dg-options -Wall -Wextra } program main integer, parameter :: x=3 ! { dg-warning Unused parameter } real :: a Indeed the questions asked in comment 22 should be answered.
[Bug c/61236] GCC 4.9 generates incorrect object code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61236 --- Comment #16 from Mukund Sivaraman muks at banu dot com --- (In reply to Andrew Pinski from comment #15) At the very least, if it is possible to detect that the pointer is NULL by static analysis and it is being passed to a function that has the notnull attribute, please warn mentioning inferences being made. The warning did not make it into gcc 4.9 due to the patches to do the warning were not ready. Gcc 4.10 should warn about it. If it does not then that is a bug. Thank you for this. :) It should detect at least some cases in that case. If qsort() can cause this sort of disruption to a caller if NULL is passed, I guess a change in glibc to add an assert(base != NULL) or similar abort is also in order given that the caller code becomes buggy otherwise. Do you agree? This would catch remaining cases before any inferred decisions are executed after the qsort().
[Bug pch/61250] New: Random pch failures on x86_64-apple-darwin13.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61250 Bug ID: 61250 Summary: Random pch failures on x86_64-apple-darwin13. Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: pch Assignee: unassigned at gcc dot gnu.org Reporter: dominiq at lps dot ens.fr CC: iains at gcc dot gnu.org Host: x86_64-apple-darwin13 Target: x86_64-apple-darwin13 Build: x86_64-apple-darwin13 When running the test suite on x86_64-apple-darwin13 with -j8 (4 cores, 8 threads), I see random failures for the tests g++.dg/pch/pch.C gcc.dg/pch/save-temps-1.c They are either an ICE: Internal compiler error: Error reporting routines re-entered. (may be related to pr59877), or a program timed out leading to an assembly comparison: WARNING: program timed out. FAIL: g++.dg/pch/pch.C -g -I. -Dwith_PCH (test for excess errors) FAIL: g++.dg/pch/pch.C -g assembly comparison AFAICT these failures occur for any set of options or -m32/-m64. The number of failures ranges from 0 to 4/5 and seems (weakly) correlated to the machine load (-j8+backup).
[Bug target/61249] _mm_frcz_ss, _mm_frcz_sd: __builtin_ia32_vfrczss, __builtin_ia32_vfrczsd require 2 arguments
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61249 --- Comment #1 from Uroš Bizjak ubizjak at gmail dot com --- (In reply to Michael Tautschnig from comment #0) Looking at https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html on the one hand and AMD's AMD64 Architecture Programmer’s Manual Volume 6: 128-Bit and 256-Bit XOP and FMA4 Instructions on the other hand, vfrczss/vfrczsd require a second argument to specify the destination. Yet r205495 changed _mm_frcz_ss/_mm_frcz_sd so that only a single argument is passed to the __builtin_ia32_vfrczss/vfrczsd calls. You should not use __builtin_* function directly. They are internal implementation details, published interface is in relevant *.h files. This was detected at language level (inconsistent types), I can only speculate that this may cause invalid code to be generated (or null operands). Did you see the runtime tests in the referred commit?
[Bug fortran/61126] [4.10 Regression] gfortran does not enable -Wununused-parameter with -Wextra
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61126 --- Comment #24 from Manuel López-Ibáñez manu at gcc dot gnu.org --- (In reply to Dominique d'Humieres from comment #23) Adding -Wall to the dg-options let the test succeed (in line with the gfortran manual, see comment 21): This is enough to make the test pass, but it is not enough to get the correct behavior and it will get broken by other reason. The patch in comment #10 is needed independently of what gfortran chooses.
[Bug target/61249] _mm_frcz_ss, _mm_frcz_sd: __builtin_ia32_vfrczss, __builtin_ia32_vfrczsd require 2 arguments
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61249 --- Comment #2 from Michael Tautschnig mt at debian dot org --- Thanks a lot for your quick reply. Yet r205495 changed _mm_frcz_ss/_mm_frcz_sd so that only a single argument is passed to the __builtin_ia32_vfrczss/vfrczsd calls. You should not use __builtin_* function directly. They are internal implementation details, published interface is in relevant *.h files. Yes, sure. It would still be nice if they were well typed. This was detected at language level (inconsistent types), I can only speculate that this may cause invalid code to be generated (or null operands). Did you see the runtime tests in the referred commit? Just re-ran those tests and looking at the generated assembly (should have done that before, my apologies) it seems that two operands are generated: vfrczsd %xmm1, %xmm0 and vfrczss %xmm1, %xmm1 So shall I read this as an imprecision in documentation? Best, Michael
[Bug target/61195] single precision fmov does not need to switch mode
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61195 chrbr at gcc dot gnu.org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |FIXED --- Comment #3 from chrbr at gcc dot gnu.org --- in 4.10
[Bug target/61231] [4.9/4.10 Regression] bootstrap comparision failure on powerpc64le-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61231 --- Comment #3 from Matthias Klose doko at gcc dot gnu.org --- is test/compile sufficient, or do you have to run it?
[Bug target/61231] [4.9/4.10 Regression] bootstrap comparision failure on powerpc64le-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61231 --- Comment #4 from Frank Ch. Eigler fche at redhat dot com --- is test/compile sufficient, or do you have to run it? Just compile.
[Bug fortran/61251] New: Hang in write from inside a function
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61251 Bug ID: 61251 Summary: Hang in write from inside a function Product: gcc Version: 4.8.2 Status: UNCONFIRMED Severity: normal Priority: P3 Component: fortran Assignee: unassigned at gcc dot gnu.org Reporter: peter.machon at arcor dot de Writing from a functions stops execution of programs. ** Simple code: program test write(*,*) f(3e0) contains function f(x) real :: f,x write(*,*)x f=2e0*x end function end program * The write function inside the function f(x) causes the hang. Sorry if it is supposed to be like this. But with e.g. the PGI compiler it works fine. I'm working with MacOSX 10.6.8 and a self-build gcc, downloaded from the gnu page. gfortran -v : Using built-in specs. COLLECT_GCC=gfortran COLLECT_LTO_WRAPPER=/Users/pm/bin/gcc_4.8.2/libexec/gcc/x86_64-apple-darwin10.8.0/4.8.2/lto-wrapper Target: x86_64-apple-darwin10.8.0 Configured with: ./configure --prefix=/Users/pm/bin/gcc_4.8.2/ Thread model: posix gcc version 4.8.2 (GCC)
[Bug c++/61252] New: Invalid code produced for omp simd reduction(min:var) where var is reference
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61252 Bug ID: 61252 Summary: Invalid code produced for omp simd reduction(min:var) where var is reference Product: gcc Version: 4.9.1 Status: UNCONFIRMED Severity: normal Priority: P3 Component: c++ Assignee: unassigned at gcc dot gnu.org Reporter: hazeman11 at gmail dot com Created attachment 32826 -- https://gcc.gnu.org/bugzilla/attachment.cgi?id=32826action=edit minimal code with reduce_bad Invalid code produced ( loop optimized out ) for case when reduction variable is a reference. I've looked into OpenMP specification but didn't find requirement for reduction variable to be local ( althought I didn't spend much time on it :) ). System: ubuntu 12.04 64bit GCC -v COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/local/libexec/gcc/x86_64-unknown-linux-gnu/4.9.1/lto-wrapper Target: x86_64-unknown-linux-gnu Configured with: ../configure --enable-checking=release --enable-languages=c,c++ --enable-multiarch --disable-multilib --enable-shared --enable-threads=posix --with-abi=m64 --program-suffix=-4.9 --with-gmp=/usr/lib --with-mpc=/usr/lib --with-mpfr=/usr/lib --without-included-gettext --with-system-zlib --with-tune=generic --prefix=/usr/local Thread model: posix gcc version 4.9.1 20140514 (prerelease) (GCC) Compilation options: g++ -fopenmp -O3 -g3 -ffast-math -Wall -ftree-vectorize -march=core-avx2 -c test.cpp -o test.o -- code -- void reduce_bad( int N, float* a0, float* a1, float maxstep ) { #pragma omp simd reduction(min:maxstep) for(int i=0;iN;i++) { maxstep = std::min(a0[i],a1[i]); } } --- and here is workaround for this bug -- code -- void reduce_good( int N, float* a0, float* a1, float ret ) { float maxstep = ret; #pragma omp simd reduction(min:maxstep) for(int i=0;iN;i++) { maxstep = std::min(a0[i],a1[i]); } ret = maxstep; } --
[Bug c++/61252] Invalid code produced for omp simd reduction(min:var) where var is reference
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61252 Jakub Jelinek jakub at gcc dot gnu.org changed: What|Removed |Added CC||jakub at gcc dot gnu.org --- Comment #1 from Jakub Jelinek jakub at gcc dot gnu.org --- I see a problem in the OpenMP lowering, that said, I hope you know that your testcase doesn't compute the minimum from all iterations, but just the last few ones (without -fopenmp just the last one, with -fopenmp from one to vectorization factor of the loop). So the value you get doesn't make any sense.
[Bug c++/61252] Invalid code produced for omp simd reduction(min:var) where var is reference
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61252 --- Comment #2 from hazeman11 at gmail dot com --- Yep sorry for so stupid example. I've reduced it to bare minimum without looking whether it does make sense. Ofcourse something like maxstep = std::min(std::min(a0[i],a1[i]),maxstep); would make much more sense. Inside of my original loop ( where i've spotted the problem ) was much much larger.
[Bug go/61253] New: gccgo: spurious error: expected '-' or '=' [GoSmith]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61253 Bug ID: 61253 Summary: gccgo: spurious error: expected '-' or '=' [GoSmith] Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: go Assignee: ian at airs dot com Reporter: dvyukov at google dot com gcc version 4.10.0 20140516 (experimental) (GCC) The program is: package main func main() { c := make(chan int) v := new(int) b := new(bool) select { case (*v), (*b) = -c: } } $ go build -compiler=gccgo /tmp/switch.go src.go:8:11: error: expected '-' or '=' case (*v), (*b) = -c:
[Bug target/61223] [gcc-4.10 regression] libstdc++ build fail due to pop lr register
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61223 Richard Earnshaw rearnsha at gcc dot gnu.org changed: What|Removed |Added Status|UNCONFIRMED |NEW Last reconfirmed||2014-05-20 Summary|libstdc++ build fail due to |[gcc-4.10 regression] |pop lr register |libstdc++ build fail due to ||pop lr register Ever confirmed|0 |1
[Bug c++/61252] Invalid code produced for omp simd reduction(min:var) where var is reference
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61252 --- Comment #3 from hazeman11 at gmail dot com --- Created attachment 32827 -- https://gcc.gnu.org/bugzilla/attachment.cgi?id=32827action=edit corrected minimal example
[Bug c++/58664] [c++11] ICE initializing array of incomplete type within union
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58664 --- Comment #3 from paolo at gcc dot gnu.org paolo at gcc dot gnu.org --- Author: paolo Date: Tue May 20 13:30:40 2014 New Revision: 210642 URL: http://gcc.gnu.org/viewcvs?rev=210642root=gccview=rev Log: /cp 2014-05-20 Paolo Carlini paolo.carl...@oracle.com PR c++/58664 * typeck2.c (cxx_incomplete_type_inform): New. (cxx_incomplete_type_diagnostic): Use it. * decl.c (grokdeclarator): Check the element type of an incomplete array type; call the above. * cp-tree.h (cxx_incomplete_type_inform): Declare. /testsuite 2014-05-20 Paolo Carlini paolo.carl...@oracle.com PR c++/58664 * g++.dg/cpp0x/nsdmi-union6.C: New. * g++.dg/parse/pr58664.C: Likewise. * g++.dg/cpp0x/nsdmi6.C: Tweak. * g++.dg/parse/crash31.C: Likewise. * g++.dg/template/error2.C: Likewise. * g++.dg/template/inherit8.C: Likewise. * g++.dg/template/offsetof2.C: Likewise. Added: trunk/gcc/testsuite/g++.dg/cpp0x/nsdmi-union6.C trunk/gcc/testsuite/g++.dg/parse/pr58664.C Modified: trunk/gcc/cp/ChangeLog trunk/gcc/cp/cp-tree.h trunk/gcc/cp/decl.c trunk/gcc/cp/typeck2.c trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/g++.dg/cpp0x/nsdmi6.C trunk/gcc/testsuite/g++.dg/parse/crash31.C trunk/gcc/testsuite/g++.dg/template/error2.C trunk/gcc/testsuite/g++.dg/template/inherit8.C trunk/gcc/testsuite/g++.dg/template/offsetof2.C
[Bug c++/58664] [c++11] ICE initializing array of incomplete type within union
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58664 Paolo Carlini paolo.carlini at oracle dot com changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED Assignee|paolo.carlini at oracle dot com|unassigned at gcc dot gnu.org Target Milestone|--- |4.10.0 --- Comment #4 from Paolo Carlini paolo.carlini at oracle dot com --- Fixed for 4.10.0.
[Bug go/61254] New: gccgo: spurious error: slice end must be integer [GoSmith]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61254 Bug ID: 61254 Summary: gccgo: spurious error: slice end must be integer [GoSmith] Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: go Assignee: ian at airs dot com Reporter: dvyukov at google dot com gcc version 4.10.0 20140516 (experimental) (GCC) The program is: package main func main() { ((([][]int{})[:])[0])[0]++ } $ go build -compiler=gccgo /tmp/index.go src.go:4:15: error: slice end must be integer ((([][]int{})[:])[0])[0]++ gc compiles successfully.
[Bug go/61255] New: gccgo: spurious error: argument 2 has incompatible type [GoSmith]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61255 Bug ID: 61255 Summary: gccgo: spurious error: argument 2 has incompatible type [GoSmith] Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: go Assignee: ian at airs dot com Reporter: dvyukov at google dot com gcc version 4.10.0 20140516 (experimental) (GCC) The program is: package main func main() { append([]byte{}, 0)[0]++ } $ go build -compiler=gccgo src.go src.go:4:2: error: argument 2 has incompatible type append([]byte{}, 0)[0]++ gc compiles successfully.
[Bug libfortran/30617] Implement a run time diagnostic for invalid recursive I/O
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=30617 Dominique d'Humieres dominiq at lps dot ens.fr changed: What|Removed |Added CC||peter.machon at arcor dot de --- Comment #37 from Dominique d'Humieres dominiq at lps dot ens.fr --- *** Bug 61251 has been marked as a duplicate of this bug. ***
[Bug fortran/61251] Hang in write from inside a function
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61251 Dominique d'Humieres dominiq at lps dot ens.fr changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |DUPLICATE --- Comment #1 from Dominique d'Humieres dominiq at lps dot ens.fr --- The code is invalid: recursive I/O, see pr30617 for discussion (hanging is darwin specific). *** This bug has been marked as a duplicate of bug 30617 ***
[Bug c++/58704] [c++11] ICE initializing array member of template class
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58704 Paolo Carlini paolo.carlini at oracle dot com changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |paolo.carlini at oracle dot com --- Comment #3 from Paolo Carlini paolo.carlini at oracle dot com --- Mine. Related to c++/58753 and c++/58930.
[Bug rtl-optimization/61241] built-in memset makes the caller function slower
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61241 --- Comment #4 from ma.jiang at zte dot com.cn --- (In reply to ktkachov from comment #3) Can you please send the patch to gcc-patc...@gcc.gnu.org including a ChangeLog Done! Thanks.
[Bug bootstrap/61210] [4.10 regression] bootstrap failure with clang
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61210 --- Comment #8 from rsandifo at gcc dot gnu.org rsandifo at gcc dot gnu.org --- Author: rsandifo Date: Tue May 20 14:18:44 2014 New Revision: 210645 URL: http://gcc.gnu.org/viewcvs?rev=210645root=gccview=rev Log: gcc/cp/ PR bootstrap/61210 * pt.c (tsubst_copy, tsubst_omp_for_iterator, tsubst_expr) (tsubst_copy_and_build): Perform recursive substitutions in a deterministic order. Modified: trunk/gcc/cp/ChangeLog trunk/gcc/cp/pt.c
[Bug middle-end/61225] [4.10 Regression] Several new failures after r210458 on x86_64-*-* with -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61225 Dominique d'Humieres dominiq at lps dot ens.fr changed: What|Removed |Added CC||hjl at gcc dot gnu.org --- Comment #4 from Dominique d'Humieres dominiq at lps dot ens.fr --- I can not reproduce gcc.dg/guality/pr43051-1.c fail with options ... The patch at https://gcc.gnu.org/ml/gcc-patches/2014-05/msg01579.html fixes the issues I saw on x86_64-apple-darwin13 (not an approval). I don't see the other issues at https://gcc.gnu.org/ml/gcc-regression/2014-05/msg00155.html in particular the guality test (not run on darwin): CCing H.J. Lu. Personal opinion: the guality tests are just a mess that should be fixed or removed.
[Bug lto/61256] New: [4.10 regression] Building spec2000/252.eon with LTO got a compfail after r210522
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61256 Bug ID: 61256 Summary: [4.10 regression] Building spec2000/252.eon with LTO got a compfail after r210522 Product: gcc Version: 4.10.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: lto Assignee: unassigned at gcc dot gnu.org Reporter: izamyatin at gmail dot com Linking on x86 as follows g++ -m64 -Ofast -flto -funroll-loops -m64 -Ofast -flto -funroll-loops -DSPEC_CPU2000_LP64 ... gives lto1: internal compiler error: in gimple_get_virt_method_for_vtable, at gimple-fold.c:3276 0x730833 gimple_get_virt_method_for_vtable(long, tree_node*, unsigned long, bool*) ../../gcc/gimple-fold.c:3276 0x730a23 gimple_get_virt_method_for_binfo(long, tree_node*, bool*) ../../gcc/gimple-fold.c:3377 0x77a133 record_target_from_binfo ../../gcc/ipa-devirt.c:867 0x77a30f record_target_from_binfo ../../gcc/ipa-devirt.c:884 0x77a9bb possible_polymorphic_call_targets_1 ../../gcc/ipa-devirt.c:931 0x77e609 possible_polymorphic_call_targets(tree_node*, long, ipa_polymorphic_call_context, bool*, void**, int*) ../../gcc/ipa-devirt.c:1743 0x7a46f9 possible_polymorphic_call_targets ../../gcc/ipa-utils.h:121 0x7a46f9 walk_polymorphic_call_targets ../../gcc/ipa.c:177 0x7a46f9 symtab_remove_unreachable_nodes(bool, _IO_FILE*) ../../gcc/ipa.c:407 0x858ec7 execute_todo ../../gcc/passes.c:1843 Please submit a full bug report, with preprocessed source if appropriate. Please include the complete backtrace with any bug report. See http://gcc.gnu.org/bugs.html for instructions. lto-wrapper: g++ returned 1 exit status /usr/bin/ld: lto-wrapper failed collect2: error: ld returned 1 exit status specmake: *** [eon] Error 1 Also 471.omnetpp from spec2006 fails with the same error
[Bug target/44557] internal compiler error: in gen_thumb_movhi_clobber, at config/arm/arm.md:5811
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=44557 --- Comment #10 from Chung-Lin Tang cltang at gcc dot gnu.org --- The ICE still happens under -mno-lra (and using reload).
[Bug other/61257] New: configure should check if sys/sdt.h is usable, not just checking the existance of the header
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61257 Bug ID: 61257 Summary: configure should check if sys/sdt.h is usable, not just checking the existance of the header Product: gcc Version: 4.9.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: other Assignee: unassigned at gcc dot gnu.org Reporter: doko at gcc dot gnu.org splitting out to a new issue, this started in PR61231, and it came up before in https://gcc.gnu.org/ml/gcc-patches/2012-12/msg01122.html (Note that strictly speaking, systemtap per se doesn't need to support an architecture for the sys/sdt.h header file to work there. gdb is a fully independent client of sys/sdt.h markers.) Perhaps the way to go forward is to have the gcc configury test-compile some toy sys/sdt.h code [1], and activate the probes only if that works. [1] #include sys/sdt.h int main () { DTRACE_PROBE(foo,bar); return 0; }
[Bug target/61231] [4.9/4.10 Regression] bootstrap comparision failure on powerpc64le-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61231 Matthias Klose doko at gcc dot gnu.org changed: What|Removed |Added CC||vmakarov at gcc dot gnu.org --- Comment #5 from Matthias Klose doko at gcc dot gnu.org --- I was wrong about blaming the sys/sdt.h, header, and filed PR61257 for that. Now I tracked down (with enough coffee I hope) r210519 as the patch introducing the bootstrap failure. Reverting it lets the bootstrap succeed on powerpc64le-linux-gnu. The build is configured with: --enable-secureplt --with-cpu=power7 --with-tune=power8 --disable-multilib --enable-multiarch --disable-werror --with-long-double-128 --enable-checking=release --build=powerpc64le-linux-gnu --host=powerpc64le-linux-gnu --target=powerpc64le-linux-gnu
[Bug c++/60373] half warning: visibility attribute ignored because it
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60373 --- Comment #2 from paolo at gcc dot gnu.org paolo at gcc dot gnu.org --- Author: paolo Date: Tue May 20 15:16:48 2014 New Revision: 210646 URL: http://gcc.gnu.org/viewcvs?rev=210646root=gccview=rev Log: /cp 2014-05-20 Paolo Carlini paolo.carl...@oracle.com PR c++/60373 * decl.c (duplicate_decls): Replace pair of warning_at with warning_at + inform. (maybe_commonize_var): Likewise. /testsuite 2014-05-20 Paolo Carlini paolo.carl...@oracle.com PR c++/60373 * g++.dg/cpp0x/Wattributes1.C: New. * g++.dg/ext/visibility/redecl1.C: Adjust. * g++.dg/ext/visibility/visibility-7.C: Likewise. Added: trunk/gcc/testsuite/g++.dg/cpp0x/Wattributes1.C Modified: trunk/gcc/cp/ChangeLog trunk/gcc/cp/decl.c trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/g++.dg/ext/visibility/redecl1.C trunk/gcc/testsuite/g++.dg/ext/visibility/visibility-7.C
[Bug rtl-optimization/60969] [4.9/4.10 Regression] ICE in output_129 in MMXMOV of mode MODE_SF for march=pentium4
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60969 Matthias Klose doko at gcc dot gnu.org changed: What|Removed |Added CC||doko at gcc dot gnu.org --- Comment #26 from Matthias Klose doko at gcc dot gnu.org --- this fix causes PR61231 (bootstrap failure on powerpc64le-linux-gnu).
[Bug c++/60373] half warning: visibility attribute ignored because it
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60373 Paolo Carlini paolo.carlini at oracle dot com changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED Assignee|paolo.carlini at oracle dot com|unassigned at gcc dot gnu.org --- Comment #3 from Paolo Carlini paolo.carlini at oracle dot com --- Fixed for 4.10.0.
[Bug target/61231] [4.9/4.10 Regression] bootstrap comparision failure on powerpc64le-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61231 --- Comment #6 from Peter Bergner bergner at gcc dot gnu.org --- Created attachment 32828 -- https://gcc.gnu.org/bugzilla/attachment.cgi?id=32828action=edit Test case that errors out with invalid assembly on big-endian [bergner@makalu-lp1 BUGS]$ /home/bergner/gcc/build/gcc-fsf-mainline-r210518/gcc/xg++ -B/home/bergner/gcc/build/gcc-fsf-mainline-r210518/gcc/ -c -O2 -m64 pr61231.ii [bergner@makalu-lp1 BUGS]$ /home/bergner/gcc/build/gcc-fsf-mainline-r210520/gcc/xg++ -B/home/bergner/gcc/build/gcc-fsf-mainline-r210520/gcc -c -O2 -m64 pr61231.ii /tmp/ccNkPBLm.s: Assembler messages: /tmp/ccNkPBLm.s:65: Error: operand out of domain (2 is not a multiple of 4) The problematic assembly is: lwa 3,2(31) The problem is that the offset for the lwa instruction needs to be a multiple of 4. I'm not exactly sure that Vlad's patch is at fault here. It could be that it's just exposing a latent bug?