Re: [google][RFA] add extra text to stack frame warnings (issue4479046)

[resending plain text. Sorry for the noise.] [sorry, i got stuck doing a bunch of other things.] On Fri, May 6, 2011 at 10:05, Andrew Pinski wrote: > > On Fri, May 6, 2011 at 1:52 AM, Chris Demetriou wrote: > > In theory, a more general warning-text-addition mechanism could be useful. > > e.g.

Re: Remove SETJMP_VIA_SAVE_AREA support

> No can do. It had been there for too many releases (even one is > one too many), at least 13 years according to the ChangeLogs. > I even went looking for it for PR48542 having a vague > recollection of it. Committed as obvious. OK, thanks. -- Eric Botcazou

Re: [PATCH] [Bug c++/49118] fake template nesting for operator-> chain

Applied, thanks. Jason

Re: [pph] New script to reproduce failures from a .log file (issue4601050)

On Jun 9, 2011, Lawrence Crowl wrote: > On 6/9/11, Diego Novillo wrote: >> +args="$@" I'd keep args in "$@" and use "$@" instead of $args, so as to avoid quoting issues. >> +line=$(grep "^spawn .*$pattern" $logf | sed -e "s:^spawn ::") > line=$(sed -e "/^spawn .*$pattern/ ! d ; s/^spawn //"

C++ PATCHes to fix issues with running the testsuite in C++0x mode

Periodically I run the G++ testsuite with --target_board=unix/-std=gnu++0x to catch any regressions relative to C++98 mode. Here are the fixes from this round: overflow.patch -- we were forgetting about overflow in some cases, causing us to treat an expression as a constant expression when it

v3 testsuite PATCH to avoid dg-excess-errors

dg-excess-errors is a very large hammer that is rarely what you want to use. To filter out certain messages, it is better to use dg-prune-output. So this patch removes most of the dg-excess-errors from the v3 testsuite, replacing some with dg-prune-output, but most need no replacement. Doin

Re: Remove SETJMP_VIA_SAVE_AREA support

On Thu, 9 Jun 2011, Eric Botcazou wrote: > > Poison it (in system.h)? > > Let's keep pretending that it never existed. :-) No can do. It had been there for too many releases (even one is one too many), at least 13 years according to the ChangeLogs. I even went looking for it for PR48542 having a

Re: FORBIDDEN_INC_DEC_CLASSES in ira-costs.c

On Thu, 9 Jun 2011, Vladimir Makarov wrote: > On 06/08/2011 12:37 PM, Hans-Peter Nilsson wrote: > > There's a lot of dead code inside the obsolete (removed in 2009) > > and nowhere else set #ifdef FORBIDDEN_INC_DEC_CLASSES. > > Remove and poison? > > > Yes, I believe so. > > I'd say borderline obvi

Re: [PATCH] [Bug c++/49118] fake template nesting for operator-> chain

On Jun 9, 2011, at 7:54 PM, Jason Merrill wrote: > Looks good, just need ChangeLog and testcase now. > > Jason The changelog is the .clog attachment to previous. I tried the testcase below but dejagnu seemed to hang with no compiler process running. I really don't know how to use dg, so perhap

Re: [PATCH] [Bug c++/49118] fake template nesting for operator-> chain

Looks good, just need ChangeLog and testcase now. Jason

Re: [pph] Rename pph_output__tree_header to respect naming convention (issue4528135)

On Thu, Jun 9, 2011 at 19:06, Diego Novillo wrote: >> +2011-06-09  Gabriel Charette   >> + >> +       * pph-streamer-out.c (pph_out_tree_header): Rename from >> +         pph_output_tree_header. Update all users. >> + > > OK. Committed at rev. 174879. Diego.

Re: [pph] Rename pph_output__tree_header to respect naming convention (issue4528135)

On Thu, Jun 9, 2011 at 18:58, Gabriel Charette wrote: > This was probably introduced in sync with my renaming patch. Just noticed and > fixed it. > > There isn't the equivalent pph_input function, is that normal? Yeah, the bits written by the header output routine are read and used by the alloc_

[pph] Rename pph_output__tree_header to respect naming convention (issue4528135)

This was probably introduced in sync with my renaming patch. Just noticed and fixed it. There isn't the equivalent pph_input function, is that normal? Index: pph-streamer.c === --- pph-streamer.c (revision 174853) +++ pph-strea

Re: [PATCH 0/4] Docs: extend.texi

On Thu, Apr 28, 2011 at 01:20, Michael Witten wrote: > See the following emails for a few inlined patches > to /trunk/gcc/doc/extend.texi (revision 172911): > > [1] Docs: extend.texi: Add missing semicolon for consistency > [2] Docs: extend.texi: Remove trailing blanks from lines > [3] Docs: e

Re: Fix PR49154-related SEGV in IRA. And ping.

On Thu, 9 Jun 2011, H.J. Lu wrote: > >        PR rtl-optimization/49154 > >        * ira-costs.c (setup_regno_cost_classes_by_mode): If there > >        already is a matching slot in the hashtable, assign it to > >        classes_ptr. > > > > This caused: > > http://gcc.gnu.org/bugzilla/show_bug.cg

Re: Fix PR49154-related SEGV in IRA. And ping.

On Wed, Jun 8, 2011 at 7:27 PM, Hans-Peter Nilsson wrote: > First, a ping for > ; > updated regclass documentation.  Ok? > > Second, I updated the CRIS port to fit the proposed > documentation update (adding a class as the patch you sent, bu

Cgraph alias reorg 2/14 (introduction of alias walkers)

Hi, this patch adds the functions to walk aliases and updates cgraph.c infrastructure to use it. It also fixes the existing missed optimization where ipa-pure-const and friends did not update decls of the same body aliases. Bootstrapped/regtested x86_64-linux, comitted. Honza * cgraph.

CRIS regclass update, fixing build breakage

Updating register classes so that the narrowest class for the SRP register is now a singleton class, instead of having other register members without means to move to/from SRP cheaper than memory, matching the documentation update. Also fixing the buglet of not fixing the condition-code-register,

Re: [testsuite] skip ARM tests with conflicting options

On 06/08/2011 03:39 AM, Richard Earnshaw wrote: > On 08/06/11 03:14, Janis Johnson wrote: >> On 06/07/2011 06:25 PM, Mike Stump wrote: >>> On Jun 7, 2011, at 4:24 PM, Janis Johnson wrote: On 06/07/2011 02:07 PM, Joseph S. Myers wrote: > On Tue, 7 Jun 2011, Janis Johnson wrote: > >>

Re: [google] pessimize stack accounting during inlining

On Wed, Jun 8, 2011 at 1:54 AM, Richard Guenther wrote: > Well, it's still not a hard limit as we can't tell how many spill slots > or extra call argument or return value slots we need. Agreed.  It's not perfect.  But I've found this does a reasonable job of preventing the inliner from pushing th

Re: [testsuite] skip ARM neon-fp16 tests for other -mcpu values

On 06/08/2011 03:17 AM, Joseph S. Myers wrote: > On Tue, 7 Jun 2011, Janis Johnson wrote: > >> These tests fail when multilib options use -mfpu= and override the >> -mfpu=neon-fp16 used for the test: >> >> g++.dg/ext/arm-fp16/arm-fp16-ops-5.C >> g++.dg/ext/arm-fp16/arm-fp16-ops-6.C >> gc

Cgraph alias reorg 1/14

Hi, there is bug in record_eh_tables that fires in one of C++ testcases where personality function is present in the same unit and GCC happily optimizes it out because it fails to set address_taken flag. Bootstrapped/regtested x86_64-linux, will commit it shortly. * cgraphbuild.c (record_

Re: [PATCH] [Bug c++/49118] fake template nesting for operator-> chain

On Jun 6, 2011, at 1:28 PM, Jason Merrill wrote: > On 06/02/2011 03:25 PM, David Krauss wrote: >> Optimally the re-opened context would be the preceding operator-> function >> itself, to create the illusion of nested calls. However, the result of >> build_new_op may be a target_expr or a call_e

Re: -fdump-passes -fenable-xxx=func_name_list

Patch is temporally rolled back. Richard, looks like deeper pass manager cleanup is needed -- I would like to delay that. For now, this leaves us two choices -- 1) do cfunc push/pop, or 2) do pass dump while executing. None of them is ideal, but safe enough. Thanks, David On Thu, Jun 9, 2011 at

Go patch committed: Use backend interface for zero initialization

This patch to the Go frontend uses the backend interface for zero initialization. This basically removes quite a bit of unnecessary code, which was more flexible than it ever really needed to be. Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu. Committed to mainline. Ian 2011-06

Re: -fdump-passes -fenable-xxx=func_name_list

Though I can not reproduce it, it might be related to what Richard mentioned in the review -- The TODO's are executed though the pass is not. This opened up a can of worm -- I will revert the patches for now. David On Thu, Jun 9, 2011 at 3:05 PM, H.J. Lu wrote: > On Tue, Jun 7, 2011 at 11:54 AM

[patch committed] Fix PR target/49307

Hi, The attached patch is to fix PR target/49307. In the faulty case, protector inserts PIC codes after the result register R0 is set and IRA combines a few of them into a load from a memory which needs R0 to spill. The patch use a unspec and an insn_and_split to hide the above PIC codes from IR

Re: -fdump-passes -fenable-xxx=func_name_list

Can you send me a trace? I can not reproduce the problem. David On Thu, Jun 9, 2011 at 3:05 PM, H.J. Lu wrote: > On Tue, Jun 7, 2011 at 11:54 AM, Xinliang David Li wrote: >> Please review the attached two patches. >> >> In the first patch, gate functions are cleaned up. All the per >> function

Re: -fdump-passes -fenable-xxx=func_name_list

It also breaks arm backend. ../trunk/configure '--build=x86_64-build_pc-linux-gnu' '--host=x86_64-build_pc-linux-gnu' '--target=arm-unknown-linux-gnueabi' '--with-sysroot=/home/carrot/x-tools/arm-unknown-linux-gnueabi/arm-unknown-linux-gnueabi/sys-root' '--disable-multilib' '--with-float=soft' '--

Re: -fdump-passes -fenable-xxx=func_name_list

On Tue, Jun 7, 2011 at 11:54 AM, Xinliang David Li wrote: > Please review the attached two patches. > > In the first patch, gate functions are cleaned up. All the per > function legality checks are moved into the executor and the > optimization heuristic checks (optimize for size) remain in the >

Re: [pph] New script to reproduce failures from a .log file (issue4601050)

On Thu, Jun 9, 2011 at 13:48, Lawrence Crowl wrote: > > line=$(sed -e "/^spawn .*$pattern/ ! d ; s/^spawn //" $logf) > > Has one fewer process and one fewer pipe.  Sed is your friend. sed and I have a complicated relationship. But, sure :) > I think this code fails when multiple lines match th

[patch libgcc committed]: Fix missing include of windows.h for mingw targets

Hello, by recent changes about HAVE_ENABLE_EXECUTE_STACK and moving code into libgcc2.c the necessary header include of windows.h was removed for config/i386/mingw32.h for LIBGCC2 case. This header is necessary for trampoline code in libgcc2.c, as for mingw target platform API is used. ChangeLog

Re: [pph] New script to reproduce failures from a .log file (issue4601050)

On 6/9/11, Diego Novillo wrote: > This small script searches for the spawn line corresponding to a given > grep pattern inside a dejagnu log file. If it finds the spawn line, > it executes it with any other arguments provided on the command line. > > It's generally useful for cutting and pasting

[google] cprop pass's gate function cleanup (issue4572055)

The patch fixed the breakage due to the r174848 backport to google/main -- the gate function for cprop is in a different file from trunk. 2011-06-09 David Li * gcse.c (gate_rtl_cprop): Gate cleanup (execute_rtl_cprop): Gate cleanup Index: gcse.c ===

Re: [PATCH] Remove set_sizetype

> 2011-06-09 Richard Guenther > > * stor-layout.c (initialize_sizetypes): Give names to all > sizetype kinds. Thanks! -- Eric Botcazou

Re: [SPARC] Fix another thinko

> 2011-06-05 Eric Botcazou > > * config/sparc/sparc.c (output_return): Fix thinko in the output of an > EH return when delayed branches are disabled. Of course the length attribute needs to be adjusted as well. Tested on SPARC/Solaris 8, applied on the mainline and 4.6/4.5/4.4 bran

Re: [Patch, AVR]: Fix PR46779

Georg-Johann Lay schrieb: Denis Chertykov schrieb: 2011/6/9 Georg-Johann Lay : This is a tentative patch to fix PR46779 and hopefully also related issues like PR45291. - /* Disallow QImode in stack pointer regs. */ - if ((regno == REG_SP || regno == (REG_SP + 1)) && mode == QImode) + /

Re: [Patch, AVR]: Fix PR46779

2011/6/9 Georg-Johann Lay : > Denis Chertykov schrieb: >> 2011/6/9 Georg-Johann Lay : >>> This is a tentative patch to fix PR46779 and hopefully also related >>> issues like PR45291. >>> >> -  /* Disallow QImode in stack pointer regs.  */ >> -  if ((regno == REG_SP || regno == (REG_SP + 1)) && mode

Re: RFA (fold): PATCH for c++/49290 (folding *(T*)(ar+10))

On 06/07/2011 10:24 AM, Jason Merrill wrote: On 06/07/2011 10:05 AM, Richard Guenther wrote: In that case you could do what Jakub suggested - but only for rvalues of course. Right, and I need to handle lvalues as well. Can't you instead adjust the type you feed to fold_indirect_ref_1 in the

Re: [Patch, AVR]: Fix PR46779

Denis Chertykov schrieb: > 2011/6/9 Georg-Johann Lay : >> This is a tentative patch to fix PR46779 and hopefully also related >> issues like PR45291. >> > - /* Disallow QImode in stack pointer regs. */ > - if ((regno == REG_SP || regno == (REG_SP + 1)) && mode == QImode) > + /* Don't allocate d

Re: [PATCH] Un-obsolete Interix

Unneeded and removed. Also found some trailing spaces in i386-interix.h, now removed. On 6/9/2011 12:02 PM, Joseph S. Myers wrote: Why are you adding an include of to i386-interix.h? This header (and most common system headers) should generally be included via system.h on the host, or tsystem.

Re: First Patch

Daniel Carrera wrote: I think I understand the problem. We have: if (count == 0 || (count == 1 && images[0] == caf_this_image)) *stat = 0; Well, you could have kept the return: if (count == 0 || (count == 1&& images[0] == caf_this_image)) -return 0; +*stat = 0; Tobias

Re: [PATCH] Un-obsolete Interix

Why are you adding an include of to i386-interix.h? This header (and most common system headers) should generally be included via system.h on the host, or tsystem.h on the target, not directly. -- Joseph S. Myers jos...@codesourcery.com

Re: [Patch, AVR]: Fix thinko in avr_function_arg_advance

2011/6/9 Georg-Johann Lay : > This patch fixes a thinko in avr_function_arg_advance. > > Without the patch, following code will report false positive for > passing args in fixed regs like: > > error: Register r0 is needed to pass a parameter but is fixed > error: Register r1 is needed to pass a par

[Patch, AVR]: Fix thinko in avr_function_arg_advance

This patch fixes a thinko in avr_function_arg_advance. Without the patch, following code will report false positive for passing args in fixed regs like: error: Register r0 is needed to pass a parameter but is fixed error: Register r1 is needed to pass a parameter but is fixed This is the code:

[pph] Do not emit PCH if generating PPH (issue4591061)

We were trying to generate PCH and PPH information at the same time. We never noticed because PPH is generated after PCH, so we were just clobbering over the previous dump. Found it by accident while debugging a GC ICE. This should make testing slightly faster. Committed to the branch.

Re: [lto, testsuite] Don't use visibility on targets that don't support it (PR lto/47334)

On Jun 9, 2011, at 12:14 AM, Rainer Orth wrote: > Mike Stump writes: >> On Apr 5, 2011, at 1:56 AM, Rainer Orth wrote: >>> * lto.c (promote_var): Only set VISIBILITY_HIDDEN if >>> HAVE_GAS_HIDDEN. >> >> This looks wrong, there are more things that have visibility than those >> things that us

Re: [Patch, AVR]: Fix PR46779

2011/6/9 Georg-Johann Lay : > This is a tentative patch to fix PR46779 and hopefully also related > issues like PR45291. > -  /* Disallow QImode in stack pointer regs.  */ -  if ((regno == REG_SP || regno == (REG_SP + 1)) && mode == QImode) +  /* Don't allocate data to non-GENERAL_REGS registers.  

[pph] New script to reproduce failures from a .log file (issue4601050)

This small script searches for the spawn line corresponding to a given grep pattern inside a dejagnu log file. If it finds the spawn line, it executes it with any other arguments provided on the command line. It's generally useful for cutting and pasting failed test cases. I did not find anythin

Re: Typed DWARF stack and convert to untyped

On Thu, Jun 09, 2011 at 04:53:19PM +0200, Jakub Jelinek wrote: > I'd like to propose convert to untyped operation, e.g. > DW_OP_GNU_convert <0> could do it (and maybe DW_OP_GNU_reinterpret <0>), > these would convert to an integral value of the same size as DWARF > address and make it untyped. As

[PATCH] Fix PR lto/49302 (corrected)

Richard, thanks for the comments. Here's the revised patch for expanding cabs in gimple. Given the direction this pass is taking, should it eventually be renamed from cse_sincos to something more general? Thanks, Bill 2011-06-09 Bill Schmidt PR lto/49302 * tree-ssa-math-opt

[google] backport 174848, 174849 to google/main

Backported -fdump-passes option impl. David

[PATCH] Un-obsolete Interix

This patch restores interix v3 and above to working condition, combines the multitude of config/*interix*.h files into one file, removes unused bits, and fixes PR 47096. Comments welcome. --Douglas B Rupp 2011-06-09 Douglas B Rupp * /config/mh-interix: Remove. * /configur

[Patch, AVR]: Fix PR46779

This is a tentative patch to fix PR46779 and hopefully also related issues like PR45291. In the present version of avr_hard_regno_mode_ok, QImode is forbidden in r29/r28. If a 16-bit value or composite is allocated to r28, this can lead to odd subregs like (set (subreg:QI (reg:HI r28) 0) ...) Th

[testsuite]: Skip tests for targets with int < 32 bits

This patch fixes testsuite failures because the testcases assume sizeof(int) >= 4. * gcc.c-torture/compile/pr49029.c: Add dg-require-effective-target int32plus * gcc.c-torture/compile/pr49163.c: Ditto. Index: gcc.c-torture/compile/pr49029.c =

[PATCH, i386]: Merge two alternatives in *movdi_internal_rex64

Hello! --- and reoder alternatives a bit. 2011-06-09 Uros Bizjak * config/i386/i386.md (*movdi_internal_rex64): Merge alternatives 6 and 8. Tested on x86_64-pc-linux-gnu, committed to mainline. Uros. Index: i386.md

Re: Fix PR49154-related SEGV in IRA. And ping.

On 06/08/2011 10:27 PM, Hans-Peter Nilsson wrote: First, a ping for ; updated regclass documentation. Ok? Sorry for the delay. I actually thought about a better formulation for some time and forgot about this because I had to work on

Re: FORBIDDEN_INC_DEC_CLASSES in ira-costs.c

On 06/08/2011 12:37 PM, Hans-Peter Nilsson wrote: There's a lot of dead code inside the obsolete (removed in 2009) and nowhere else set #ifdef FORBIDDEN_INC_DEC_CLASSES. Remove and poison? Yes, I believe so. I'd say borderline obvious, but maybe there's instead reason to reinstate it, if that

Re: [lra] A new branch

On 06/07/2011 09:46 AM, Joseph S. Myers wrote: On Mon, 30 May 2011, Vladimir Makarov wrote: * doc/tm.texi.in (TARGET_REGISTER_BANK): New hook. Please include the text of documentation for new hooks in target.def instead of tm.texi.in where possible (so tm.texi.in just has the @hook lin

[lra] patch to speed up LRA

The following patch removes coalescing when it is not necessary, decrease number of insns to process for constraints, fix a typo, and implements Joseph Myers proposals. The patch was successfully bootstraped (with LRA which is default) on x86-64, itanium, and ppc64. 2011-06-09 Vladimir Maka

Re: [PATCH] Remove set_sizetype

On Thu, 9 Jun 2011, Richard Guenther wrote: > On Thu, 9 Jun 2011, Eric Botcazou wrote: > > > > Now that there is a single place left to call set_sizetype we can > > > remove it and initialize sizetypes properly from the start > > > (in initialize_sizetypes). > > > > Can you give a meaningful TYP

Typed DWARF stack and convert to untyped

Hi! On Fri, Mar 25, 2011 at 12:32:37PM +0100, Jakub Jelinek wrote: > This patch on top of > http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01224.html > and > http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01723.html > implements parts of Cary's typed DWARF stack proposal: > http://www.dwarfstd.org/doc

Re: [PATCH] Print location for conflicting global regs in warning

Andi, PR49344 occured between 174833 and 174836 which correspond to your commits for this patch and the two other ones. Could you please have a look? BTW what is the way to CC you in bugzilla? TIA Dominique

Re: [PATCH] c-pragma: adding a data field to pragma_handler

Thanks for this patch. You need to adjust the formatting of the comments to match the existing style (in particular, indentation of second and subsequent lines of comments, and watch out for spelling errors (exemple, abstact). For a struct, the explanations of individual fields should be in t

Re: [C++ Patch] PR 29003

OK. Jason

Re: [C++ Patch] PR 29003

On 06/09/2011 04:06 PM, Jason Merrill wrote: On 06/09/2011 06:07 AM, Paolo Carlini wrote: I tested on x86_64-linux the below patchlet for a long standing accepts-invalid. Is it ok for mainline? Or do we want a different error message? A somehow tighter check? The error message needs to say som

Re: [PATCH][C family] Fix pp_c_type_specifier IDENTIFIER_NODE handling

On Thu, 9 Jun 2011, Richard Guenther wrote: > 2011-06-09 Richard Guenther > > c-family/ > * c-pretty-print.c (pp_c_type_specifier): Use pp_c_identifier > to print a IDENTIFIER_NODE. OK. -- Joseph S. Myers jos...@codesourcery.com

Re: [C++ Patch] PR 29003

On 06/09/2011 06:07 AM, Paolo Carlini wrote: I tested on x86_64-linux the below patchlet for a long standing accepts-invalid. Is it ok for mainline? Or do we want a different error message? A somehow tighter check? The error message needs to say something about typedef being the problem. Mayb

[PATCH][C family] Fix pp_c_type_specifier IDENTIFIER_NODE handling

The current IDENTIFIER_NODE handling in pp_c_type_specifier can't work because the first thing pp_c_tree_decl_identifier does is assert that the passed tree is a DECL. Instead the following patch arranges us to use pp_c_identifier to dump an IDENTIFIER_NODE. This (obvious?) patch is required to

Re: [PATCH] Fix noreturn predictor to also work with no regular exits

On Thu, 9 Jun 2011, Richard Guenther wrote: > > This is a complementary fix for the leslie3d problem. For functions > without a regular exit (for example one that exits with exit (0)) > we do not apply noreturn predictors. The following patch notices > that edges to noreturn calls are still un

Re: [PATCH] better warning for section conflict

On Wed, Jun 8, 2011 at 3:07 AM, Andi Kleen wrote: > When erroring on a section type conflict print both locations. > This makes it a lot easier to track them down, especially in LTO. > > This is still not quite good before it prints only one > conflict pair currently instead of all, but at least a

Re: [PATCH] Remove set_sizetype

On Thu, 9 Jun 2011, Eric Botcazou wrote: > > Now that there is a single place left to call set_sizetype we can > > remove it and initialize sizetypes properly from the start > > (in initialize_sizetypes). > > Can you give a meaningful TYPE_NAME to sizetype? Ada dump files now read: > > _GLOBAL.

Re: Remove SETJMP_VIA_SAVE_AREA support

> Poison it (in system.h)? Let's keep pretending that it never existed. :-) -- Eric Botcazou

Re: [PATCH] Remove set_sizetype

> Now that there is a single place left to call set_sizetype we can > remove it and initialize sizetypes properly from the start > (in initialize_sizetypes). Can you give a meaningful TYPE_NAME to sizetype? Ada dump files now read: _GLOBAL.SZ0_discr31 (integer p0, integer p1) { return p1 <= p0

Re: [PATCH] Fix leslie3d performance regression with -flto/-fwhole-program

On Thu, 9 Jun 2011, Tobias Burnus wrote: > Richard Guenther wrote: > > leslie3d terminates itself with STOP which translates to a noreturn > > function. This causes our edge hot/cold prediction machinery to not > > apply any predictions based on paths to noreturn functions > > > Bootstrapped and

Re: [PATCH] Fix leslie3d performance regression with -flto/-fwhole-program

Richard Guenther wrote: > leslie3d terminates itself with STOP which translates to a noreturn > function. This causes our edge hot/cold prediction machinery to not > apply any predictions based on paths to noreturn functions > Bootstrapped and tested on x86_64-unknown-linux-gnu. > Ok for trunk?

[PATCH] Fix noreturn predictor to also work with no regular exits

This is a complementary fix for the leslie3d problem. For functions without a regular exit (for example one that exits with exit (0)) we do not apply noreturn predictors. The following patch notices that edges to noreturn calls are still unlikely if they have exactly one predecessor and is not

Re: [build] Move ENABLE_EXECUTE_STACK to toplevel libgcc

On 06/09/2011 09:42 AM, Rainer Orth wrote: > Indeed, that did the trick. The following revised patch has been > bootstrapped without regressions on i386-pc-solaris2.10. I've already > got build, Darwin and Mingw approval and can commit the osf and solaris > parts on my own. If the C parts are ok

[PATCH] Fix leslie3d performance regression with -flto/-fwhole-program

leslie3d terminates itself with STOP which translates to a noreturn function. This causes our edge hot/cold prediction machinery to not apply any predictions based on paths to noreturn functions (because there is no path that does not lead to a noreturn function). This in turn makes us predict t

[C++ Patch] PR 29003

Hi, I tested on x86_64-linux the below patchlet for a long standing accepts-invalid. Is it ok for mainline? Or do we want a different error message? A somehow tighter check? Thanks, Paolo. // /cp 2011-06-09 Paolo Carlini PR c++/29003 * decl.c (grokdeclarat

[patch, testsuite] Fix vectorizer testsuite failures on ARM

Hi, This patch fixes several vectorizer testsuite failures on ARM: - vect-16.c checks that the vectorization fails without -ffast-math, but -ffast-math is a default flag for vector tests on ARM. I renamed the test to no-fast-math-vect-16.c to avoid the use of the flag for it. - vect-peel-3.c and v

Re: Merge score7 to score.c and remove forwarding functions.

You didn't test this, did you? Since it reverts my fixes to score_handle_option to avoid global state and it's hardly going to build cleanly with the old prototype for a hook that has changed prototype. -- Joseph S. Myers jos...@codesourcery.com

Re: [lto, testsuite] Don't use visibility on targets that don't support it (PR lto/47334)

> On Thu, Jun 9, 2011 at 9:14 AM, Rainer Orth > wrote: > > Mike Stump writes: > > > >> On Apr 5, 2011, at 1:56 AM, Rainer Orth wrote: > >>>  * lto.c (promote_var): Only set VISIBILITY_HIDDEN if > >>>      HAVE_GAS_HIDDEN. > >> > >> This looks wrong, there are more things that have visibility tha

Re: [build] Move ENABLE_EXECUTE_STACK to toplevel libgcc

On Thu, 9 Jun 2011, Rainer Orth wrote: > Indeed, that did the trick. The following revised patch has been > bootstrapped without regressions on i386-pc-solaris2.10. I've already > got build, Darwin and Mingw approval and can commit the osf and solaris > parts on my own. If the C parts are ok no

Re: [lto, testsuite] Don't use visibility on targets that don't support it (PR lto/47334)

On Thu, Jun 9, 2011 at 9:14 AM, Rainer Orth wrote: > Mike Stump writes: > >> On Apr 5, 2011, at 1:56 AM, Rainer Orth wrote: >>>  * lto.c (promote_var): Only set VISIBILITY_HIDDEN if >>>      HAVE_GAS_HIDDEN. >> >> This looks wrong, there are more things that have visibility than those >> things

Re: Dump before flag

On Thu, Jun 9, 2011 at 12:31 AM, Xinliang David Li wrote: > this is the patch that just removes the TODO_dump flag and forces it > to dump. The original code cfun->last_verified = flags & > TODO_verify_all looks weird -- depending on TODO_dump is set or not, > the behavior of the update is differe

Re: [PATCH] c-pragma: adding a data field to pragma_handler

You are right, the new version is in the diff. The diff for the test hasn't changed and is in the previous mail. In the previous version of the file, the registered_pragmas was not better freed. I don't know if it is really important (it would need a callback at the end of the front-end passes)

Re: [PATCH] Fix PR lto/49302

On Wed, Jun 8, 2011 at 6:56 PM, William J. Schmidt wrote: > This patch adds code to the cse_sincos pass to expand builtin cabs (x) > into sqrt (r*r + i*i), where r = realpart(x) and i = imagpart(x).  This > is usually handled by fold_builtin_cabs; however, PR49302 shows a case > where compile flag

[v3] Fix two typos in

Hi, tested x86_64-linux, committed. Thanks, Paolo. 2011-06-09 Paolo Carlini * include/std/type_traits (__is_copy_assignable_impl, __is_nt_copy_assignable_impl): Fix typos. Index: include/std/type_traits ===

Re: C++ PATCH for c++/49107 (excessive instantiation due to noexcept)

On 06/08/2011 11:32 PM, Jason Merrill wrote: 49107 is an issue with the addition of noexcept to the standard library causing extra instantiations, to the point that code that worked before doesn't work now due to circular dependency. Someone suggested that to avoid this problem we could defer

Re: [PATCH] Fix ICE in reset_unmarked_insns_debug_uses (PR middle-end/49308)

On Tue, Jun 07, 2011 at 02:24:49PM +0200, Jakub Jelinek wrote: > Fixed thusly. Additionally I've renamed a variable that was shadowing > a variable of the same name, which confused me quite a bit when debugging > it - I was expecting insn to be a DEBUG_INSN, while it was something > completely dif

Re: [build] Move ENABLE_EXECUTE_STACK to toplevel libgcc

"Joseph S. Myers" writes: > On Tue, 7 Jun 2011, Rainer Orth wrote: > >> > So my view is that you should define __LIBGCC_TRAMPOLINE_SIZE__, only if >> > -fbuilding-libgcc. >> >> I can give it a try if I can figure out how to define -fbuilding-libgcc >> via the option handling machinery. I just

Re: toplevel libgcc md-unwind-support.h change breaks build

From: Rainer Orth Date: Thu, 09 Jun 2011 09:33:35 +0200 > Of course we do. I must have been dreaming when writing parts of this > patch. I've now double-checked all md_unwind_header filenames, found > just another typo, and installed the following as obvious. Thanks.

Re: toplevel libgcc md-unwind-support.h change breaks build

David Miller writes: > It's using "linux.h" instead of "linux-unwind.h" for whatever reason. > > So we need something like this? > > diff --git a/libgcc/config.host b/libgcc/config.host > index 24282e1..713b2a9 100644 > --- a/libgcc/config.host > +++ b/libgcc/config.host > @@ -600,7 +600,7 @@ spa

[Patch : H8300] Bug fix for bit insn and minor tweaks to insns

Hi, The following patch fixes an ICE that is generated when the compiler tries to perform bit manipulation for logical operations when the source and destination address does not match. The testcase is also included in the patch(gcc.dg). The additional condition in the insn takes care of the ICE

Re: [lto, testsuite] Don't use visibility on targets that don't support it (PR lto/47334)

Mike Stump writes: > On Apr 5, 2011, at 1:56 AM, Rainer Orth wrote: >> * lto.c (promote_var): Only set VISIBILITY_HIDDEN if >> HAVE_GAS_HIDDEN. > > This looks wrong, there are more things that have visibility than those > things that use GAS and have .hidden. Darwin I think is one of them