Re: Speedup configure and build with system.h

On Fri, Jan 22, 2016 at 12:09:43PM -0800, H.J. Lu wrote: > > * system.h (string, algorithm): Include only conditionally. > > (new): Include always under C++. > > * bb-reorder.c (toplevel): Define USES_ALGORITHM. > > * final.c (toplevel): Ditto. > > *

[PATCH][PR tree-optimization/69347] Speedup DOM slightly

So as noted in BZ69347 we have regressed a bit in the amount of time spent in DOM for included testcase. My previous patch helped a little, but there's still some low hanging fruit. One of the things that was added during this cycle was the ability to do a bit of secondary equivalence

C++ PATCH for c++/69392 (ICE with 'this' init-capture)

We're already handling capture of 'this' specially, but we need to do that in the case of init-capture as well. Tested x86_64-pc-linux-gnu, applying to trunk. commit 3d101d4f78d0940b54838695df35874d18f827d0 Author: Jason Merrill Date: Thu Jan 21 16:45:42 2016 -0500 PR

Re: C++ PATCH for c++/69379 (ICE with PTRMEM_CST wrapped in NOP_EXPR)

On 01/22/2016 12:20 PM, Marek Polacek wrote: On Thu, Jan 21, 2016 at 01:49:14PM -0500, Jason Merrill wrote: On 01/21/2016 01:25 PM, Marek Polacek wrote: The problem in this PR is that we have a PTRMEM_CST wrapped in NOP_EXPR and fold_convert can't digest that. Why didn't we fold away the

Re: Speedup configure and build with system.h

On Thu, Jan 21, 2016 at 8:57 AM, Michael Matz wrote: > Hi, > > this has bothered me for some time. The gcc configure with stage1 feels > like taking forever because some of the decl availability tests (checking > for C function) include system.h, and that, since a while,

[gomp4] fix atomic tests

These two tests presumed a particular ordering of atomicc operation execution, which is kind of anethema to why you'd want atomic ops and parallelizing loops. I've removed the more obviously incorrect assumptions, but I have a suspicion the fortran one at least still contains undefined

Re: [PATCH] c++/58109 - alignas() fails to compile with constant expression

On 01/20/2016 06:04 PM, Martin Sebor wrote: Right. The problem is this code in is_late_template_attribute: /* If the first attribute argument is an identifier, only consider second and following arguments. Attributes like mode, format, cleanup and several target

Re: [PATCH] Fix the remaining PR c++/24666 blockers (arrays decay to pointers too early)

On 01/22/2016 11:17 AM, Patrick Palka wrote: On Thu, 21 Jan 2016, Patrick Palka wrote: On Thu, 21 Jan 2016, Jason Merrill wrote: On 01/19/2016 10:30 PM, Patrick Palka wrote: * g++.dg/template/unify17.C: XFAIL. Hmm, I'm not comfortable with deliberately regressing this testcase.

[patch] libstdc++/69116 Constrain std::valarray functions and operators

The example in the PR is a sneaky little problem. When is included the following overload is declared: template _Expr operator<<(const _Tp& __t, const valarray<_Tp>& __v); This is a candidate function for any "a << b" expression with namespace std as an associated namespace. In order to do

[gomp4] Fix handling of subarrays with update directive

Hi, The attached patch fixes a defect reported with gcc 5.2 (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69414). It is also the case, the issue is present in the gomp4 branch. The patch also adds additional testing. Committed to gomp4 after bootstrap and regtesting. Thanks, Jim Index:

[PATCH, COMMITTED] Fix a ChangeLog entry

This fixes my name in gcc/ChangeLog. Index: ChangeLog === --- ChangeLog (revision 232742) +++ ChangeLog (working copy) @@ -144,7 +144,7 @@ * lra-coalesce.c (lra_coalesce): Invalidate value for the result pseudo

[gomp4] fix some tests

I discovered these tests were relying on implicitly using vector partitioning, rather than specifying it explicitly. Fixed thusly. nathan Index: gcc/testsuite/c-c++-common/goacc/reduction-1.c === ---

Re: [patch] Restore cross-language inlining into Ada

> > Why do you say so? There are C->Ada calls as there are Ada->C calls in > > plenty of existing software. > > But what percentage of the C->Ada ones are performance critical? Note that, > unlike the Ada->C or Ada/C++ ones, these have never been inlined and I can I think we was inlining them

Re: [PATCH] Improve TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS callback

On 01/22/2016 10:34 AM, Wilco Dijkstra wrote: Improve TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS target hook. It turns out there is another case where the register allocator uses the union of register classes without checking that the cost of the resulting register class is lower than both (see

Re: Patch RFA: Add option -fcollectible-pointers, use it in ivopts

On 01/22/2016 08:03 PM, Ian Lance Taylor wrote: Updated patch. I've verified that I'm changing the only relevant place in tree-ssa-loop-ivopts.c that creates a POINTER_PLUS_EXPR, so I do think that this is the only changed to fix the problem for ivopts. I don't think so. One of the problems

Re: [gomp4] fix some tests

These two libgomp tests were likewise presuming vector partitioning, without specifying it. nathan 2016-01-22 Nathan Sidwell * testsuite/libgomp.oacc-c-c++-common/routine-1.c: Specify vector. * testsuite/libgomp.oacc-c-c++-common/routine-2.c: Likewise. Index:

Re: [PATCH 1/2][AArch64] Implement AAPCS64 updates for alignment attribute

> Ok, rebased onto a more recent build, and bootstrapping with Ada posed no > problems. Sorry for the noise. Great, no problem, and thanks for double checking. -- Eric Botcazou

[PATCH] PR c++/69399: Add HAVE_WORKING_CXX_BUILTIN_CONSTANT_P

Without the fix for PR 65656, g++ miscompiles __builtin_constant_p in wi::lrshift in wide-int.h. Add a check with PR 65656 testcase to verify that C++ __builtin_constant_p works properly. Tested on x86-64 with working GCC: gcc/auto-host.h:/* #undef HAVE_WORKING_CXX_BUILTIN_CONSTANT_P */

Re: Patch RFA: Add option -fcollectible-pointers, use it in ivopts

Updated patch. I've verified that I'm changing the only relevant place in tree-ssa-loop-ivopts.c that creates a POINTER_PLUS_EXPR, so I do think that this is the only changed to fix the problem for ivopts. OK for mainline? Ian gcc/ChangeLog: 2016-01-22 Ian Lance Taylor *

[PATCH][ARM][1/4] PR target/65932: Add testcase

Hi all, This patch adds an execute testcase that is fixed by the arm.h hunk of Jim Wilson's patch that he posted at: https://gcc.gnu.org/ml/gcc-patches/2015-06/msg02132.html I propose this testsuite change instead of the testsuite changes in that patch since the gcc.target/arm/wmul-[123].c

[PATCH][ARM][2/4] Fix operand costing logic for SMUL[TB][TB]

Hi all, As part of investigating the codegen effects of a fix for PR 65932 I found we assign too high a cost for the sign-extending multiply instruction SMULBB. This is because we add the cost of a multiply-extend but then also recurse into the SIGN_EXTEND sub-expressions rather than the

[PATCH][cse][3/4] Don't overwrite original rtx when folding source of set

Hi all, This patch is a consequence of the thread I started at https://gcc.gnu.org/ml/gcc/2016-01/msg00100.html The problem is that the fold_rtx call in cse_insn may overwrite its argument if the insn argument is non-NULL. This leads to CSE not considering the original form of the RTX when

[PATCH][ARM][0/4] Fixing PR target/65932

Hi all, PR target/65932 is a wrong-code bug affecting arm and has manifested itself when compiling the Linux kernel, so it's something that we really ought to fix. The problem stems from the fact that PROMOTE_MODE and TARGET_PROMOTE_FUNCTION_MODE don't match up on arm. PROMOTE_MODE also marks

Re: [PATCH][ARM] (cleanup) remove arn_builtin_macro_redefined overwrite

On 21/01/16 10:22, Christian Bruel wrote: A tiny patch to cleanup the cpp_opts->warn_builtin_macro_redefined setting in arm_pragma_target_parse. Removes this hack obsoleted by the use of NODE_CONDITIONAL for pr target/69180. regtested for arm-linux-gnueabi -mfpu=vfp,

Re: [PATCH] Fix up reduction-1{1,2} testcases (PR middle-end/68221)

On Thu, Jan 21, 2016 at 07:17:59AM +0100, Thomas Schwinge wrote: > Ping. I'd prefer to keep the code as is, it is closer to what could result from the user trying to do a similar thing, and thus has a better chance of keeping being supported. Jakub

Re: [PATCH] Fix use of declare'd vars by routine procedures.

On Sun, Jan 17, 2016 at 05:01:24PM -0600, James Norris wrote: > The attached patch addresses the failure of declare-4 in the libgomp > testsuite. > The primary failure was the use a variable with a link clause for an OpenACC > routine function. The patch changes the test to use a create clause.

Re: TR29124 C++ Special Maths - Make pull functions into global namespace.

On 21/01/16 19:07 -0500, Ed Smith-Rowland wrote: On 01/21/2016 07:29 AM, Jonathan Wakely wrote: On 20/01/16 20:30 -0500, Ed Smith-Rowland wrote: Now that libstdc++ installs a proper math.h we can piggyback on that to put in the last bit of TR29124. This patch adds the math special functions

Re: [patch] Restore cross-language inlining into Ada

> I only updated > - /* Don't inline if the callee can throw non-call exceptions but the > - caller cannot. > - FIXME: this is obviously wrong for LTO where STRUCT_FUNCTION is > missing. - Move the flag into cgraph node or mirror it in the inline > summary. */ - else if (callee_fun

Re: [patch] Restore cross-language inlining into Ada

> I am fine with it being relaxed and permitting inlining !non_call_exceptions > to non_call_exceptions functions.. It would be also cool to have a > testcases. Thanks, patch installed with check_match changed to check_maybe_up, I'll work towards adding an Ada+C LTO test but this will require

Re: [gomp4] Un-parallelized OpenACC kernels constructs with nvptx offloading: "avoid offloading"

On Fri, Jan 22, 2016 at 08:40:26AM +0100, Thomas Schwinge wrote: > On Thu, 21 Jan 2016 22:54:26 +0100, I wrote: > > On Mon, 18 Jan 2016 18:26:49 +0100, Tom de Vries > > wrote: > > > [...] [OpenACC] kernels region [...] > > > that parloops does not manage to parallelize: >

Re: [gomp4] Un-parallelized OpenACC kernels constructs with nvptx offloading: "avoid offloading"

Hi Jakub! On Fri, 22 Jan 2016 09:36:25 +0100, Jakub Jelinek wrote: > On Fri, Jan 22, 2016 at 08:40:26AM +0100, Thomas Schwinge wrote: > > On Thu, 21 Jan 2016 22:54:26 +0100, I wrote: > > > On Mon, 18 Jan 2016 18:26:49 +0100, Tom de Vries > > > wrote: >

Re: [PATCH 1/3, libgomp] Resolve libgomp plugin deadlock on exit, libgomp proper parts

On Tue, Jan 19, 2016 at 03:02:48PM +0900, Chung-Lin Tang wrote: > Ping x 2. I like the general idea of moving the gomp_fatal stuff up, out of the plugins, but we now have another plugin, so it can't apply as is. And in the intelmic plugin, adding return true; after abort (); looks wrong. So, can

Re: Suspicious code in fold-const.c

On Fri, Jan 22, 2016 at 12:06 AM, Andrew MacLeod wrote: > I was trying the ttype prototype for static type checking on fold-const.c to > see how long it would take me to convert such a large file, and it choked on > this snippet of code in fold_unary_loc, around lines

Re: [patch] Document restriction of scalar_storage_order

> This patch is OK. Thanks. > I wish somebody could fix the existing documentation for this attribute > to use the present tense instead of the future to describe GCC's current > behavior, though :-S I think it's a common idiom: if you do this, then the compiler will do that. I can see it

Re: [PATCH][ARM] Fix PR target/69403: Bug in thumb2_ior_scc_strict_it pattern

Hi Han, On 21/01/16 22:57, Han Shen wrote: Hi Kyrill, the patched gcc generates correct asm for me for the test case. (I'll then build the whole system to see if other it-block related bugs are gone too.) One short question, the newly generated RTL for x = x |(a) will be orr %0,

PR 69400: Invalid 128-bit modulus result

As described in the PR, wi::divmod_internal was sign- rather than zero-extending a modulus result in cases where the result has fewer HWIs than the precision and the upper bit of the upper HWI was set. This patch tries to make things more robust by getting wi_pack to handle the canonicalisation

Re: [hsa merge 07/10] IPA-HSA pass

On Wed, Jan 20, 2016 at 09:53:30PM +0300, Ilya Verbin wrote: > If you're OK with this, I'll install this patch: > > > libgomp/ > * target.c (gomp_get_target_fn_addr): Allow host fallback if target > function wasn't mapped to the device with non-shared memory. Ok, thanks. > diff

[PATCH][ARM][4/4] Adjust gcc.target/arm/wmul-[123].c tests

Hi all, In this final patch I adjust the troublesome gcc.target/arm/wmul-[123].c tests to make them more helpful. gcc.target/arm/wmul-[12].c may now generate either sign-extending multiplies (+accumulate) or normal 32-bit multiplies since the arguments to the multiplies are already sign-extended

Re: PR 69400: Invalid 128-bit modulus result

On Fri, Jan 22, 2016 at 09:43:52AM +, Richard Sandiford wrote: > gcc/ > PR tree-optimization/69400 > * wide-int.cc (wi_pack): Take the precision as argument and > perform canonicalization here rather than in the callers. > Use the main loop to handle all full-width

Re: gomp_target_fini

On Thu, Jan 21, 2016 at 04:24:46PM +0100, Bernd Schmidt wrote: > Thomas, I've mentioned this issue before - there is sometimes just too much > irrelevant stuff to wade through in your patch submissions, and it > discourages review. The discussion of the actual problem begins more than > halfway

Re: [PATCH][ARM] Fix PR target/69403: Bug in thumb2_ior_scc_strict_it pattern

On Fri, Jan 22, 2016 at 9:32 AM, Kyrill Tkachov wrote: > Hi Han, > > On 21/01/16 22:57, Han Shen wrote: >> >> Hi Kyrill, the patched gcc generates correct asm for me for the test >> case. (I'll then build the whole system to see if other it-block >> related bugs are

Re: [PATCH][Testsuite] Fix PR66877

On Fri, Jan 22, 2016 at 11:51 AM, Alan Lawrence wrote: > This is a scan-tree-dump failure in vect-over-widen-3-big-array.c, that occurs > only on ARM - the only platform to have vect_widen_shift. > > Tested on arm-none-eabi (armv8-crypto-neon-fp, plus a non-neon variant),

[PATCH][Testsuite] Fix PR66877

This is a scan-tree-dump failure in vect-over-widen-3-big-array.c, that occurs only on ARM - the only platform to have vect_widen_shift. Tested on arm-none-eabi (armv8-crypto-neon-fp, plus a non-neon variant), also aarch64 (token platform without vect_widen_shift). gcc/testsuite/ChangeLog:

Re: Fix PR 67665: ICE when passing two empty files directly to cc1 with -g

On Fri, Jan 22, 2016 at 2:17 AM, Andrew Pinski wrote: > On Wed, Jan 13, 2016 at 4:36 AM, Richard Biener > wrote: >> On Wed, Jan 13, 2016 at 9:27 AM, Andrew Pinski wrote: >>> Hi, >>> The support -combine was removed a while back

Re: C++ PATCH for c++/69379 (ICE with PTRMEM_CST wrapped in NOP_EXPR)

On Fri, Jan 22, 2016 at 03:38:26PM -0500, Jason Merrill wrote: > If we have a NOP_EXPR to the same type, we should strip it here. This helps for the unreduced testcases in the PR, but not for the reduced one, because for the reduced one, the types are not the same. One type is struct { void

Re: [PATCH] Fix the remaining PR c++/24666 blockers (arrays decay to pointers too early)

On Fri, 22 Jan 2016, Jason Merrill wrote: On 01/22/2016 11:17 AM, Patrick Palka wrote: On Thu, 21 Jan 2016, Patrick Palka wrote: On Thu, 21 Jan 2016, Jason Merrill wrote: On 01/19/2016 10:30 PM, Patrick Palka wrote: * g++.dg/template/unify17.C: XFAIL. Hmm, I'm not comfortable with

Re: [PATCH] Fix -minline-stringops-dynamically (PR target/69432)

> Hi! > > With this option we generate a conditional library call. When expanding > such a call, do_pending_stack_adjust () is performed from expand_call > and if there are any needed sp adjustments, they are emitted. The > problem is that this happens only in conditionally executed code, so on

[gomp4] gang partitioning

I've committed this patch to gomp4 branch. It changes the auto partitioning logic to allocate the outermost loop to the outermost available partitioning. For instance, gang partitioning will be used for the outermost loop of a parallel region. Innermost loops remain partitioned at the

Re: Speedup configure and build with system.h

On Fri, Jan 22, 2016 at 09:23:48PM +0100, Jakub Jelinek wrote: > On Fri, Jan 22, 2016 at 12:09:43PM -0800, H.J. Lu wrote: > > > * system.h (string, algorithm): Include only conditionally. > > > (new): Include always under C++. > > > * bb-reorder.c (toplevel): Define

[PATCH] Avoid unnecessary creation of VEC_COND_EXPR in the vectorizer

Hi! I've noticed we create a VEC_COND_EXPR tree just to grab the arguments from it to construct a ternary gimple assign. The following patch fixes that by creating the ternary gimple assign directly. Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? 2016-01-22 Jakub Jelinek

[PATCH] Fix -minline-stringops-dynamically (PR target/69432)

Hi! With this option we generate a conditional library call. When expanding such a call, do_pending_stack_adjust () is performed from expand_call and if there are any needed sp adjustments, they are emitted. The problem is that this happens only in conditionally executed code, so on some path

Re: TR29124 C++ Special Maths - Make pull functions into global namespace.

On 01/22/2016 05:39 AM, Jonathan Wakely wrote: On 21/01/16 19:07 -0500, Ed Smith-Rowland wrote: On 01/21/2016 07:29 AM, Jonathan Wakely wrote: On 20/01/16 20:30 -0500, Ed Smith-Rowland wrote: Now that libstdc++ installs a proper math.h we can piggyback on that to put in the last bit of

Re: [patch] Document restriction of scalar_storage_order

> Isn't this kind of implied by the already documented restriction of > taking the address of a union field? Still, you can add this or any kind > of similar language if you like. It's stronger I think since you can do type punning without taking an address. -- Eric Botcazou

Re: Remove redundant unshare_expr from ipa-prop

On Fri, Jan 22, 2016 at 05:36:14PM +1100, Kugan wrote: > There is a redundant unshare_expr in ipa-prop. Attached patch removes > it. Bootstrapped and regression tested on x86_64-pc-linux-gnu with no > new regressions. > > Is this OK for trunk? > > Thanks, > Kugan > > gcc/ChangeLog: > >

Re: [PATCH][cse][3/4] Don't overwrite original rtx when folding source of set

On Fri, Jan 22, 2016 at 10:52 AM, Kyrill Tkachov wrote: > Hi all, > > This patch is a consequence of the thread I started at > https://gcc.gnu.org/ml/gcc/2016-01/msg00100.html > The problem is that the fold_rtx call in cse_insn may overwrite its argument > if the insn

Re: [patch] Restore cross-language inlining into Ada

On Fri, Jan 22, 2016 at 12:06 PM, Eric Botcazou wrote: >> I am fine with it being relaxed and permitting inlining !non_call_exceptions >> to non_call_exceptions functions.. It would be also cool to have a >> testcases. > > Thanks, patch installed with check_match changed

RE: [PATCH] [ARC] Add basic support for double load and store instructions

Thank u for the feedback. I hope this new patch solves the outstanding issues. Please find it attached. //Claudiu 0001-ARC-Add-basic-support-for-double-load-and-store-inst.patch Description: 0001-ARC-Add-basic-support-for-double-load-and-store-inst.patch

Re: [PATCH][ARM,AARCH64] target/PR68674: relayout vector_types in expand_expr

On Fri, Jan 22, 2016 at 12:41 PM, Christian Bruel wrote: > > > On 01/19/2016 04:18 PM, Richard Biener wrote: >> >> maybe just if (currently_expanding_to_rtl)? >> >> But yes, this looks like a safe variant of the fix. >> >> Richard. >> > thanks, currently_expanding_to_rtl

Re: [patch] Restore cross-language inlining into Ada

> > I only updated > > - /* Don't inline if the callee can throw non-call exceptions but the > > - caller cannot. > > - FIXME: this is obviously wrong for LTO where STRUCT_FUNCTION is > > missing. - Move the flag into cgraph node or mirror it in the inline > > summary. */ - else if

Re: Speedup configure and build with system.h

On Thu, 2016-01-21 at 18:10 +0100, Richard Biener wrote: > On Thu, Jan 21, 2016 at 5:57 PM, Michael Matz wrote: > > Hi, > > > > this has bothered me for some time. The gcc configure with stage1 > > feels > > like taking forever because some of the decl availability tests > >

Re: [patch] Restore cross-language inlining into Ada

On Fri, Jan 22, 2016 at 1:00 PM, Jan Hubicka wrote: >> > I only updated >> > - /* Don't inline if the callee can throw non-call exceptions but the >> > - caller cannot. >> > - FIXME: this is obviously wrong for LTO where STRUCT_FUNCTION is >> > missing. - Move the

Re: [patch] Restore cross-language inlining into Ada

> Why do you say so? There are C->Ada calls as there are Ada->C calls in > plenty of existing software. But what percentage of the C->Ada ones are performance critical? Note that, unlike the Ada->C or Ada/C++ ones, these have never been inlined and I can imagine the kind of trouble this would

Re: [patch] Restore cross-language inlining into Ada

> Hmm, I see now. I wonder if we can also inline > can_thorw_non_call_exceptions to !can_throw_non_call_exceptions provied > that we set the flag in ipa-inline-transform. That way we can inline Ada to > C and the observation about no EH regions should still hold. I'd say you're the only one

Re: [patch] Restore cross-language inlining into Ada

> > Hmm, I see now. I wonder if we can also inline > > can_thorw_non_call_exceptions to !can_throw_non_call_exceptions > > provied > > that we set the flag in ipa-inline-transform. That way we can inline Ada to > > C and the observation about no EH regions should still hold. > > I'd say you're

Re: Patch RFA: Add option -fcollectible-pointers, use it in ivopts

On Fri, Jan 22, 2016 at 11:25 AM, Bernd Schmidt wrote: > On 01/22/2016 08:03 PM, Ian Lance Taylor wrote: >> >> Updated patch. >> >> I've verified that I'm changing the only relevant place in >> tree-ssa-loop-ivopts.c that creates a POINTER_PLUS_EXPR, so I do think >> that

Re: [PATCH] Avoid unnecessary creation of VEC_COND_EXPR in the vectorizer

On January 22, 2016 11:09:06 PM GMT+01:00, Jakub Jelinek wrote: >Hi! > >I've noticed we create a VEC_COND_EXPR tree just to grab the arguments >from >it to construct a ternary gimple assign. > >The following patch fixes that by creating the ternary gimple assign >directly.

Re: Speedup configure and build with system.h

On January 22, 2016 11:15:38 PM GMT+01:00, Jakub Jelinek wrote: >On Fri, Jan 22, 2016 at 09:23:48PM +0100, Jakub Jelinek wrote: >> On Fri, Jan 22, 2016 at 12:09:43PM -0800, H.J. Lu wrote: >> > > * system.h (string, algorithm): Include only >conditionally. >> > >

Re: [PATCH] libitm: Fix HTM fastpath.

On Fri, 2016-01-08 at 12:07 +0100, Torvald Riegel wrote: > This patch fixes a thinko in the HTM fastpath implementation. In a > nutshell, we also need to monitor the HTM fastpath control (ie, > htm_fastpath) variable from within a HW transaction on the HTM fastpath, > so that such a HW

[PATCH] Fix aarch64 bootstrap (pr69416)

The bare CONST_INT inside the CCmode IF_THEN_ELSE is causing combine to make incorrect simplifications. At this stage it feels safer to wrap the CONST_INT inside of an UNSPEC than make more generic changes to combine. But we should definitely investigate combine's CCmode issues for gcc7.

Re: [PATCH] ARM PR68620 (ICE with FP16 on armeb)

On 20/01/16 21:10, Christophe Lyon wrote: On 19 January 2016 at 15:51, Alan Lawrence wrote: On 19/01/16 11:15, Christophe Lyon wrote: For neon_vdupn, I chose to implement neon_vdup_nv4hf and neon_vdup_nv8hf instead of updating the VX iterator because I thought it

Re: [PATCH] Fix aarch64 bootstrap (pr69416)

On 22/01/16 17:07, Richard Henderson wrote: > The bare CONST_INT inside the CCmode IF_THEN_ELSE is causing combine to > make incorrect simplifications. At this stage it feels safer to wrap > the CONST_INT inside of an UNSPEC than make more generic changes to > combine. > > But we should

Re: [PATCH 1/2][AArch64] Implement AAPCS64 updates for alignment attribute

On 21/01/16 17:23, Alan Lawrence wrote: > On 18/01/16 17:10, Eric Botcazou wrote: >> >> Could you post the list of files that differ? How do they differ exactly? > > Hmmm. Well, I definitely had this failing to bootstrap once. I repeated that, > to > try to identify exactly what the differences

Re: C++ PATCH for c++/69379 (ICE with PTRMEM_CST wrapped in NOP_EXPR)

On Thu, Jan 21, 2016 at 01:49:14PM -0500, Jason Merrill wrote: > On 01/21/2016 01:25 PM, Marek Polacek wrote: > >The problem in this PR is that we have a PTRMEM_CST wrapped in NOP_EXPR > >and fold_convert can't digest that. > > Why didn't we fold away the NOP_EXPR before calling fold_convert? I

Re: Suspicious code in fold-const.c

On 01/22/2016 02:37 PM, Andrew MacLeod wrote: /* If the initializer is non-void, then it's a normal expression that will be assigned to the slot. */ (*) if (!VOID_TYPE_P (t)) (*) return RECURSE (t); I suspect this should also be if

Re: [gomp4] Un-parallelized OpenACC kernels constructs with nvptx offloading: "avoid offloading"

On Fri, Jan 22, 2016 at 02:18:38PM +0100, Bernd Schmidt wrote: > On 01/22/2016 09:36 AM, Jakub Jelinek wrote: > > > >I think it is a bad idea to go against what the user wrote. Warning that > >some code might not be efficient? Perhaps (if properly guarded with some > >warning option one can turn

Re: [gomp4] Un-parallelized OpenACC kernels constructs with nvptx offloading: "avoid offloading"

On 01/22/2016 02:25 PM, Jakub Jelinek wrote: What about #pragma oacc parallel? That would never do that? It shouldn't, no (IMO). Bernd

Re: [PATCH][AArch64] Replace insn to zero up DF register

On 12/16/2015 03:30 PM, Evandro Menezes wrote: > >    On 10/30/2015 05:24 AM, Marcus Shawcroft wrote: > >    On 20 October 2015 at 00:40, Evandro Menezes >wrote: > >    In the existing targets, it seems that it's always faster to zero >up a DF > >   

Re: [PATCH][ARM] Fix PR target/69245 Rewrite arm_set_current_function

Hi Christian, On 22/01/16 14:07, Christian Bruel wrote: Hi Kyrill, On 01/21/2016 01:22 PM, Kyrill Tkachov wrote: Hi Christian, On 21/01/16 10:36, Christian Bruel wrote: The current arm_set_current_function was both awkward and buggy. For instance using partially set TARGET_OPTION set from

Re: Suspicious code in fold-const.c

On 01/22/2016 06:03 AM, Richard Biener wrote: On Fri, Jan 22, 2016 at 12:06 AM, Andrew MacLeod wrote: I was trying the ttype prototype for static type checking on fold-const.c to see how long it would take me to convert such a large file, and it choked on this snippet of

Re: [gomp4] Un-parallelized OpenACC kernels constructs with nvptx offloading: "avoid offloading"

On 01/22/2016 09:36 AM, Jakub Jelinek wrote: I think it is a bad idea to go against what the user wrote. Warning that some code might not be efficient? Perhaps (if properly guarded with some warning option one can turn off, either on a per-source file or using pragmas even more fine grained).

Re: Speedup configure and build with system.h

Hi, On Fri, 22 Jan 2016, Oleg Endo wrote: > and have been put into system.h because there have > been problems with malloc poisoning and C++ stdlib implementation other > than libstdc++, which sometimes pull other headers which then cause > trouble. The fix for this set of errors was to

[PATCH][Testsuite] Fix scan-tree-dump failures with vect_multiple_sizes

Since r230292, these tests in gcc.dg/vect have been failing on ARM, AArch64, and x86_64 with -march=haswell (among others - when prefer_avx128 is true): vect-outer-1-big-array.c scan-tree-dump-times vect "grouped access in outer loop" 2 vect-outer-1.c scan-tree-dump-times vect "grouped access

Re: [PATCH PR68542]

Richard, I fixed all remarks pointed by you in vectorizer part of patch. Could you take a look on modified patch. Uros, Could you please review i386 part of patch related to support of conditional branches with vector comparison. Bootstrap and regression testing did not show any new failures.

Re: [PATCH 4/5] Don't mark targets of unconditional jumps with side effects as FALLTHRU.

On 22/01/16 08:44, Andreas Krebbel wrote: On 01/22/2016 12:10 AM, Jeff Law wrote: On 01/21/2016 03:05 AM, Andreas Krebbel wrote: On 01/02/2016 08:16 PM, Marcin Kościelnicki wrote: When an unconditional jump with side effects targets an immediately following label, rtl_tidy_fallthru_edge is

Re: Speedup configure and build with system.h

Hi, On Thu, 21 Jan 2016, Richard Biener wrote: > I'm inclined to say #define INCLUDE_ALGORITHM is a better name, I've done this. On a different (slower) machine than the one from the initial mail: without patch, -j31 bootstrap all,ada: real35m2.655s user395m28.135s sys 12m10.814s

Re: [PATCH] Fix the remaining PR c++/24666 blockers (arrays decay to pointers too early)

On Thu, 21 Jan 2016, Patrick Palka wrote: On Thu, 21 Jan 2016, Jason Merrill wrote: On 01/19/2016 10:30 PM, Patrick Palka wrote: * g++.dg/template/unify17.C: XFAIL. Hmm, I'm not comfortable with deliberately regressing this testcase. template -void bar (void (T[5])); // {

Re: [PATCH 4/4][AArch64] Cost CCMP instruction sequences to choose better expand order

This breaks bootstrap on aarch64 by miscompiling the stage2 compiler. ../../../libgomp/priority_queue.h:422:11: internal compiler error: RTL flag check: MEM_VOLATILE_P used with unexpected rtx code 'mem' in set_mem_attributes_minus_bitpos, at emit-rtl.c:1833 if (list->tasks == node)

Re: [PATCH PR68542]

On Fri, Jan 22, 2016 at 6:29 AM, Yuri Rumyantsev wrote: > Richard, > > I fixed all remarks pointed by you in vectorizer part of patch. Could > you take a look on modified patch. > > Uros, > > Could you please review i386 part of patch related to support of > conditional

Re: [PATCH][ARM][4/4] Adjust gcc.target/arm/wmul-[123].c tests

Hi Bernd, On 22/01/16 14:53, Bernd Schmidt wrote: On 01/22/2016 10:52 AM, Kyrill Tkachov wrote: AFAICT the new sequence is better than the old one even for -mtune=cortex-a9 since it contains two fewer instructions. Just curious (I think this patch series is good but will leave it to the arm

[Patch Obvious] gcc.dg/vect/bb-slp-pr68892.c requires vectorization of doubles

Hi, As title. This testcase fails on arm-none-linux-gnueabihf, because we don't have vectorization of doubles there. Committed as obvious as revision 232731. Thanks, James --- 2016-01-22 James Greenhalgh * gcc.dg/vect/bb-slp-pr68892.c: Require

Re: [PATCH, rs6000] Fix PR63354

On Fri, Jan 22, 2016 at 12:42 AM, Bill Schmidt wrote: > Hi, > > On Thu, 2016-01-21 at 21:21 -0600, Bill Schmidt wrote: >> The testcase will need a slight adjustment, as currently it fails on >> powerpc64 with -m32 testing. Working on a fix. >> >> Bill >> > > This

Re: [PATCH][ARM][4/4] Adjust gcc.target/arm/wmul-[123].c tests

On 01/22/2016 10:52 AM, Kyrill Tkachov wrote: AFAICT the new sequence is better than the old one even for -mtune=cortex-a9 since it contains two fewer instructions. Just curious (I think this patch series is good but will leave it to the arm folks) - are these instructions equally expensive?

Re: [PATCH, rs6000] Fix PR63354

OK, thanks, Joseph! I'll make that adjustment later today. Bill On Fri, 2016-01-22 at 15:51 +, Joseph Myers wrote: > On Thu, 21 Jan 2016, Bill Schmidt wrote: > > > +/* { dg-do compile { target { powerpc64*-linux-* } } } */ > > That's suboptimal; you should allow powerpc*-*-linux* targets

[PATCH] Improve TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS callback

Improve TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS target hook. It turns out there is another case where the register allocator uses the union of register classes without checking that the cost of the resulting register class is lower than both (see

Re: [PATCH, rs6000] Fix PR63354

On Thu, 21 Jan 2016, Bill Schmidt wrote: > +/* { dg-do compile { target { powerpc64*-linux-* } } } */ That's suboptimal; you should allow powerpc*-*-linux* targets so that the test is also run for --enable-targets=all powerpc-linux builds when testing a -m64 multilib. -- Joseph S. Myers