Index: coverage.c
===
--- coverage.c (revision 190369)
+++ coverage.c (working copy)
@@ -261,6 +261,56 @@ str_eq (const void *p1, const void *p2)
return !strcmp (s1, s2);
}
+/* Command line option descriptor. */
+
+struct opt_
The following patch will be applied to google branches.
2011-08-11 David Li
* l-ipo.c (process_module_scope_static_func): Promote
aliases of static symbol in LIPO mode.
Index: l-ipo.c
===
--- l-ipo.c
modules -- we
saw
cases where funcdef_no is greater than 128k in a module leading to overflow and
conflict.
2011-05-21 David Li
* testsuite/gcc.dg/matrix/transpose-1.c: Do not inline mem_init.
* testsuite/gcc.dg/matrix/transpose-2.c: Ditto.
* testsuite/gcc.dg/matrix
modules. Those functions
will be eliminated later (they are not called from global dtor/ctor) --
there existence is important to make sure address taken (etc) attributes
for called dtor/ctors are consistent between profile-gen and use.
2011-05-30 David Li
* cgraphunit.c (cgraph_optimize
2011-06-07 David Li
Backport trunk r174536,174537,174762,174698
Index: doc/invoke.texi
===
--- doc/invoke.texi (revision 174725)
+++ doc/invoke.texi (working copy)
@@ -5217,11 +5217,12 @@ appended with a
Fix lipo test cases with new pass names.
2011-06-07 David Li
* testsuite/gcc.dg/tree-prof/lipo/indir-call-prof_0.c: Pass name fix.
* testsuite/gcc.dg/tree-prof/lipo/indir-call-prof_1.c: Pass name fix.
* testsuite/gcc.dg/tree-prof/lipo/val-prof-1_0.c: Pass name fix
This is trivial change to allow module id to be displayed in LIPO mode
in function header.
2011-06-07 David Li
* tree-pretty-print.c (revision 174779)
(dump_function_header): dump module id.
Index: tree-pretty-print.c
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
The patch will be committed to google/main to fix a problem in LIPO model
that leads to 'reference to discarded comdat section' ld warning. The problem
is caused by inconsistent comdat groups between primary and aux modules because
thunks were skipped in aux module.
2011-06-14
The patch is committed to google/main to fix lipo test regressions after trunk
merge.
2011-07-26 David Li
* value-prof.c (gimple_value_profile_transformations): Remove redundant
code.
* cgraphunit.c (cgraph_mark_functions_to_output): Fix assertion in lipo
mode
The following trivial patch will be applied to google branches.
Bootstrap and tessted on x86-64/linux
2011-07-28 David Li
* coverage.c (coverage_init): Remove checking of profile-use
flags.
Index: coverage.c
ssion test, and SPEC06 LIPO build.
Will commit to google/main.
Thanks,
David
2011-04-28 David Li
* testsuite/gcc.dg/tree-prof/lipo/inliner-1.c (revision 0): New test.
* testsuite/gcc.dg/tree-prof/lipo/gdb_cmd (revision 0): Ditto.
* testsuite/gcc.dg/tree-pr
2011-05-01 David Li
* testsuite/lib/profopt.exp (proc): Multiple source file support.
* testsuite/gcc.dg/tree-prof/lipo/lipo.exp (load_lib):
Multiple source file support.
* testsuite/g++.dg/tree-prof/lipo/lipo.exp: Ditto.
* testsuite/gcc.dg/tree-prof/lipo
Backport patches r17342, r173177, and r173147 from trunk to google/main + some
minor cleanups for problems found in test.
Tested: bootstrap + regression test + SPEC06 LIPO testing.
Backport r173177
2011-05-03 David Li
* tree-profile.c (init_ic_make_global_vars): Set
tls
-05-04 David Li
* coverage.c (incompatible_cl_args): Better handling
of option mismatch.
Index: coverage.c
===
--- coverage.c (revision 173353)
+++ coverage.c (working copy)
@@ -213,6 +213,27 @@ is_last_module
The following patch reverted r173158 from google/main -- -fstrict-enums
provides a better
implementation. The test cases are kept with slight modification.
Bootstrap and tested with related test cases.
Ok for google/main?
2011-05-06 David Li
Revert r173158.
Index: tree-vrp.c
k ? (There is a little LIPO specific change that can be removed).
David
2011-05-18 David Li
* final.c (rest_of_clean_state): Call function header dumper.
* opts-global.c (handle_common_deferred_options): Handle new options.
* tree-cfg.c (gimple_dump_cfg): Call functio
Do you need to guard the jump function access with check if
(ipa_node_params_vector.exists ())?
Ideally, useful_cold_callee should be folded into the inline hints
estimation. Question about the heuristic: why filtering out
PASS_THROUGH parameter cases completely? Passing 'this' parameter in
many
URCE_FILE (current_function_decl),
- DECL_SOURCE_LINE (current_function_decl));
+ DECL_SOURCE_LINE (current_function_decl),
+ gimple_decl_printable_name (current_function_decl, 1));
}
}
Index: ChangeLog
==
Can you add missing documentation on functions like ...:get_count_info
-- documenting return value etc. Also it might be better to avoid
using 'set' as the local variable name. Change it to something more
specific.
thanks,
David
On Thu, Aug 22, 2013 at 3:56 PM, Dehao Chen wrote:
> This patch h
Ok.
David
On Tue, Aug 27, 2013 at 7:36 AM, Dehao Chen wrote:
> Patch updated.
>
> Thanks,
> Dehao
>
> On Mon, Aug 26, 2013 at 4:11 PM, Xinliang David Li wrote:
>> Can you add missing documentation on functions like ...:get_count_info
>> -- documenting return value
+ Honza
On Tue, Aug 27, 2013 at 10:56 AM, Teresa Johnson wrote:
> Ping #3.
>
> Thanks,
> Teresa
>
> On Mon, Aug 19, 2013 at 11:33 AM, Teresa Johnson wrote:
>> Ping.
>> Thanks,
>> Teresa
>>
>> On Mon, Aug 12, 2013 at 6:54 AM, Teresa Johnson wrote:
>>> On Tue, Aug 6, 2013 at 10:23 PM, Teresa John
Does this one look ok?
thanks,
David
On Thu, Aug 22, 2013 at 4:20 PM, Xinliang David Li wrote:
> Hi, In this patch, loop alignment peeling and loop versioning
> transformation will be reported via -fopt-info by default. This will
> help vectorizer size tuning.
>
> It also enhanc
eeds to be fixed to emit the function name
> before each block of messages, a la inform().
>
> Thanks,
> Teresa
>
> On Tue, Aug 27, 2013 at 11:22 AM, Xinliang David Li
> wrote:
>> Does this one look ok?
>>
>> thanks,
>>
>> David
>>
>
If this is the convention, we should probably have another patch to
fix all the existing opt-info messages.
thanks,
David
On Tue, Aug 27, 2013 at 1:23 PM, Mike Stump wrote:
> On Aug 27, 2013, at 11:22 AM, Xinliang David Li wrote:
>> Does this one look ok?
>
> We don't c
Richard, I have some comments about the patch.
> -ftree-vectorizer-verbose=This switch is deprecated. Use
> -fopt-info instead.
>
> ftree-slp-vectorize
> ! Common Report Var(flag_tree_slp_vectorize) Optimization
> Enable basic block vectorization (SLP) on trees
The code dealing with th
On Wed, Aug 28, 2013 at 7:09 AM, Teresa Johnson wrote:
> On Wed, Aug 28, 2013 at 4:01 AM, Richard Biener
> wrote:
>> On Wed, Aug 7, 2013 at 7:23 AM, Teresa Johnson wrote:
>>> On Tue, Aug 6, 2013 at 9:29 AM, Teresa Johnson wrote:
On Tue, Aug 6, 2013 at 9:01 AM, Martin Jambor wrote:
> H
On Wed, Aug 28, 2013 at 12:59 AM, Richard Biener wrote:
> On Tue, 27 Aug 2013, Xinliang David Li wrote:
>
>> Richard, I have some comments about the patch.
>>
>> > -ftree-vectorizer-verbose=This switch is deprecated. Use
>> > -fopt-info inst
asic3.c:8:foo: note: loop with 7 iterations completely unrolled
basic3.c:5:foo: note: loop with 7 iterations completely unrolled
Is this version ok after testing?
thanks,
David
On Wed, Aug 28, 2013 at 2:45 AM, Richard Biener
wrote:
> On Tue, Aug 27, 2013 at 10:30 PM, Xinliang David Li
02088)
+++ ChangeLog (working copy)
@@ -1,5 +1,17 @@
2013-08-29 Xinliang David Li
+ * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
+ Delay base decl alignment adjustment.
+ * tree-vectorizer.c (ensure_base_alignment): New function.
+ (vectorize_loops): Add dbg_cnt su
20, 2013 at 9:35 PM, Xinliang David Li wrote:
>> Do you need to guard the jump function access with check if
>> (ipa_node_params_vector.exists ())?
> I believe it is not necessary since, for example, ipa_analyze_node
> calls ipa_check_create_node_params that calls create. But I se
Except that in this form, the dump will be extremely large and not
suitable for very large applications. Besides, we might also want to
use the same machinery (dump_printf_loc etc) for dump file dumping.
The current behavior of using '-details' to turn on opt-info-all
messages for dump files are no
On Fri, Aug 30, 2013 at 12:51 PM, Teresa Johnson wrote:
> On Fri, Aug 30, 2013 at 9:27 AM, Xinliang David Li wrote:
>> Except that in this form, the dump will be extremely large and not
>> suitable for very large applications.
>
> Yes. I did some measurements for both a fai
On Fri, Aug 30, 2013 at 1:23 AM, Richard Biener
wrote:
> On Fri, Aug 30, 2013 at 1:28 AM, Xinliang David Li wrote:
>> I was debugging a runtime failure of SPEC06 xalancbmk built with LIPO.
>> Using -fdisable- option pinpoints the problem in slp vectorize
>> pass on a partic
Missing test cases?
Have you tested the optimization with SPEC2k and SPEC06? There are a
couple of benchmarks benefit greatly from devirtualization, such as
eon, povray etc. I believe astar will probably improve with this
optimization at O2 (it has hot virtual functions that are not
overridden a
>From Joseph:
"The
conversion is not safe for sqrt if the two types are double and long
double and long double is x86 extended, for example."
This is not reflected in the patch.
David
On Tue, Sep 3, 2013 at 2:27 PM, Joseph S. Myers wrote:
> On Tue, 3 Sep 2013, Cong Hou wrote:
>
>> + CASE_MAT
On Wed, Sep 4, 2013 at 1:53 PM, Cong Hou wrote:
> I have made a new patch according to your comments. I found several
> references saying that the precision 2p+2 is OK for the sqrt
> conversion (one here:
> http://www.cs.berkeley.edu/~fateman/generic/algorithms.pdf). The new
> patch is pasted as b
On Wed, Sep 4, 2013 at 1:59 PM, Joseph S. Myers wrote:
> On Wed, 4 Sep 2013, Cong Hou wrote:
>
>> I have made a new patch according to your comments. I found several
>> references saying that the precision 2p+2 is OK for the sqrt
>> conversion (one here:
>> http://www.cs.berkeley.edu/~fateman/gene
+
>/* Be careful about integer to fp conversions.
> These may overflow still. */
>if (FLOAT_TYPE_P (TREE_TYPE (arg0))
> Index: gcc/testsuite/gcc.c-torture/execute/20030125-1.c
> =======
> --- gcc/t
looks fine to me.
In the long run, I wonder if the machinery in diagnostic messages can
be reused for opt-info dumping -- i.e., support different streams. It
has many nice features including %qD specifier for printing tree
decls.
David
On Mon, Sep 9, 2013 at 12:01 PM, Teresa Johnson wrote:
> I'
The following patch enables GCC to record system include paths in
module infos. This allows more precise parsing context to be
established when compiling auxiliary modules -- FE behaves slightly
differently when parsing system headers. There are also a couple of
other cleanups in the patch.
David
Can you add a test case to the regression suite?
When the type of arguments are unsigned short/unsigned int, GCC does
not vectorize the loop anymore -- this is worth a separate bug to
track. punpcklwd instruction can be used to do zero extension of the
short type.
David
On Wed, Sep 11, 2013 at 6
On Thu, Sep 12, 2013 at 1:06 PM, Teresa Johnson wrote:
> After porting r198033 from google/4_7 to google/4_8 a test case failed
> with an assert when trying to take the strlen of profile_data_prefix.
>
> In most cases this is either set from the directory specified to
> -fprofile-generate=, or to
{
loop->force_vect = true;
Index: ChangeLog
===
--- ChangeLog (revision 202481)
+++ ChangeLog (working copy)
@@ -1,3 +1,24 @@
+2013-09-12 Xinliang David Li
+
+ * tree-if-conv.c (main_tree_if_conversio
:
> On Thu, Sep 12, 2013 at 1:20 PM, Xinliang David Li wrote:
>> On Thu, Sep 12, 2013 at 1:06 PM, Teresa Johnson wrote:
>>> After porting r198033 from google/4_7 to google/4_8 a test case failed
>>> with an assert when trying to take the strlen of profile_data_prefix.
&g
Ok.
David
On Fri, Sep 13, 2013 at 7:21 AM, Teresa Johnson wrote:
> Testing passes, is the below patch ok for google/4_8?
> Thanks, Teresa
>
> On Thu, Sep 12, 2013 at 10:18 PM, Teresa Johnson wrote:
>> On Thu, Sep 12, 2013 at 2:32 PM, Xinliang David Li
>> wrote:
&
On Fri, Sep 13, 2013 at 1:30 AM, Richard Biener
wrote:
> On Thu, Sep 12, 2013 at 10:31 PM, Xinliang David Li
> wrote:
>> Currently -ftree-vectorize turns on both loop and slp vectorizations,
>> but there is no simple way to turn on loop vectorization alone. The
>> logi
, Xinliang David Li wrote:
> On Fri, Sep 13, 2013 at 1:30 AM, Richard Biener
> wrote:
>> On Thu, Sep 12, 2013 at 10:31 PM, Xinliang David Li
>> wrote:
>>> Currently -ftree-vectorize turns on both loop and slp vectorizations,
>>> but there is no simple way to t
Ok -- then my updated patch is wrong then. The implementation in the
first version matches the requirement.
thanks,
David
On Fri, Sep 13, 2013 at 9:45 AM, Joseph S. Myers
wrote:
> On Fri, 13 Sep 2013, Richard Biener wrote:
>
>> @@ -1691,6 +1695,12 @@ common_handle_option (struct gcc_options
>>
Updated patch implementing the logic that more specific option wins.
Ok for trunk?
thanks,
David
On Fri, Sep 13, 2013 at 9:48 AM, Xinliang David Li wrote:
> Ok -- then my updated patch is wrong then. The implementation in the
> first version matches the requirement.
>
> thank
revision 202229)
+++ ChangeLog (working copy)
@@ -1,3 +1,12 @@
+2013-09-13 Xinliang David Li
+
+ * omp-low.c (expand_omp_simd): Build optimization cl
+ node if necessary.
+ * opts.c (finish_option): Enable/Disable tree if-cvt
+ if not explicitly set.
+
I noticed there are a couple of dump_printf_loc instances in
coverage.c not ended with newline. They should be fixed.
David
On Tue, Sep 10, 2013 at 6:32 AM, Teresa Johnson wrote:
> On Mon, Sep 9, 2013 at 9:55 PM, Xinliang David Li wrote:
>> looks fine to me.
>>
>> In the
ize without a profile.
thanks,
David
On Mon, Sep 16, 2013 at 11:06 AM, Teresa Johnson wrote:
> On Mon, Sep 16, 2013 at 10:57 AM, Xinliang David Li
> wrote:
>> I noticed there are a couple of dump_printf_loc instances in
>> coverage.c not ended with newline. They should be
I incorporated all the comments and committed the change (also fixed a
test failure with --help=optimizers).
thanks,
David
On Mon, Sep 16, 2013 at 3:07 AM, Richard Biener
wrote:
> On Fri, Sep 13, 2013 at 6:56 PM, Xinliang David Li wrote:
>> Updated patch implementing the logic
On Mon, Sep 16, 2013 at 3:13 AM, Richard Biener
wrote:
> On Fri, Sep 13, 2013 at 5:16 PM, Xinliang David Li wrote:
>> On Fri, Sep 13, 2013 at 1:30 AM, Richard Biener
>> wrote:
>>> On Thu, Sep 12, 2013 at 10:31 PM, Xinliang David Li
>>> wrote:
>>>>
Ok -- abandon the patch.
On Mon, Sep 16, 2013 at 2:04 AM, Richard Biener
wrote:
> On Sat, Sep 14, 2013 at 8:10 AM, Xinliang David Li wrote:
>> tree if conversion is an enabler pass for vectorization, so by
>> default, it is only turned on when vectorization is on, but may also
&
On Tue, Sep 17, 2013 at 1:20 AM, Richard Biener
wrote:
> On Mon, Sep 16, 2013 at 10:24 PM, Xinliang David Li
> wrote:
>> On Mon, Sep 16, 2013 at 3:13 AM, Richard Biener
>> wrote:
>>> On Fri, Sep 13, 2013 at 5:16 PM, Xinliang David Li
>>> wrote:
>>&
On Tue, Sep 17, 2013 at 8:45 AM, Jakub Jelinek wrote:
> On Tue, Sep 17, 2013 at 08:37:57AM -0700, Xinliang David Li wrote:
>> >> char a[1];
>> >>
>> >> void foo(int n)
>> >> {
>> >> int* b = (int*)(a+n);
>
ok.
David
On Tue, Sep 17, 2013 at 4:53 PM, Dehao Chen wrote:
> This patch makes AutoFDO honor system paths stored in the profile.
>
> Bootstrapped and passed regression tests.
>
> OK for google-4_8 branch?
>
> Thanks,
> Dehao
>
> Index: gcc/auto-profile.c
> ==
Ok.
David
On Wed, Sep 18, 2013 at 10:21 AM, Dehao Chen wrote:
> This patch disables SLP for AutoFDO.
>
> Bootstrapped and passed unittests.
>
> OK for google-4_8?
>
> Thanks,
> Dehao
>
> Index: gcc/opts.c
> ===
> --- gcc/opts.c (rev
On Tue, Sep 17, 2013 at 1:20 AM, Richard Biener
wrote:
> On Mon, Sep 16, 2013 at 10:24 PM, Xinliang David Li
> wrote:
>> On Mon, Sep 16, 2013 at 3:13 AM, Richard Biener
>> wrote:
>>> On Fri, Sep 13, 2013 at 5:16 PM, Xinliang David Li
>>> wrote:
>>&
On Wed, Sep 18, 2013 at 4:51 PM, Dehao Chen wrote:
> This patch fixup the call graph edge targets during AutoFDO pass, so
> that when rebuilding call graph edges, it can find the correct callee.
>
> Bootstrapped and passed regression test. Benchmark tests on-going.
>
> Ok for google-4_8 branch?
>
/* Local pure-const may imply need to fixup the cfg. */
> + if (execute_fixup_cfg () & TODO_cleanup_cfg)
> + cleanup_tree_cfg ();
> +
>current_function_decl = NULL;
>pop_cfun ();
> }
>
> - cgraph_pre_profiling_inlining_done = true;
> -
Looks good.
David
On Thu, Sep 19, 2013 at 1:15 PM, Dehao Chen wrote:
> This patch sets cgraph_node count during AutoFDO annotation, otherwise
> execute_fixup_cfg will clear all the BB counts.
>
> bootstrapped and passed regression test.
>
> OK for google-4_8 branch?
>
> Thanks,
> Dehao
>
> Index
e_profiling_inlining_done = true;
> - cgraph_process_module_scope_statics ();
> - /* Now perform link to allow cross module inlining. */
> - cgraph_do_link ();
> - varpool_do_link ();
> - cgraph_unify_type_alias_sets ();
> -
>return TODO_rebuild_cgraph_edges;
> }
&
cleanup are moved after
> annotation. So setting of cgraph node's count is still needed, right?
>
> Thanks,
> Dehao
>
> On Thu, Sep 19, 2013 at 9:28 PM, Xinliang David Li wrote:
>> I did not catch this in the last review. The cleanup CFG should be
>> done be
On Sat, Sep 21, 2013 at 12:54 PM, Jan Hubicka wrote:
> Hi,
> this is upated version of patch discussed at
> http://gcc.gnu.org/ml/gcc-patches/2012-12/msg00841.html
>
> It makes CORE tuning to more follow the optimization guidelines.
> In particular it removes some tuning flags for features I imple
On Sat, Sep 21, 2013 at 3:51 PM, Xinliang David Li wrote:
> On Sat, Sep 21, 2013 at 12:54 PM, Jan Hubicka wrote:
>> Hi,
>> this is upated version of patch discussed at
>> http://gcc.gnu.org/ml/gcc-patches/2012-12/msg00841.html
>>
>> It makes CORE tuning to more f
Thanks. I modified the patch so that the max allowed peel iterations
can be specified via a parameter. Testing on going. Ok for trunk ?
David
On Mon, Sep 23, 2013 at 4:31 AM, Richard Biener
wrote:
> On Wed, Sep 18, 2013 at 10:21 PM, Xinliang David Li
> wrote:
>> On Tue, Sep 17,
Basically GCC reports that the loop is vectorized, but the vectorized
loop is never executed because of the bogus alias check ' a + 16 < a'
generated. In trunk, the vectorized version is eliminated, but it
remains as a dead code with gcc 48.
David
On Mon, Sep 23, 2013 at 5:26 PM, Cong Hou wrote:
I wish there is better heuristic in the future. For now it is ok.
David
On Wed, Sep 25, 2013 at 2:48 PM, Dehao Chen wrote:
> This patch disables aggressive loop peeling when profile is available.
> This prevents extensive code bloat which leads to increased i-cache
> misses.
>
> Bootstrapped and
I took the liberty to pick up Richard's original fvect-cost-model
patch and made some modification.
What has not changed:
1) option -ftree-vect-loop-version is removed;
2) three cost models are introduced: cheap, dynamic, and unlimited;
3) unless explicitly specified, cheap model is the default at
Looks ok.
David
On Thu, Sep 26, 2013 at 9:00 AM, Dehao Chen wrote:
> This patch fix the bug when setting max-lipo-group in AutoFDO:
>
> Bootstrapped and passed regression test.
>
> OK for google branches?
>
> Thanks,
> Dehao
>
> Index: gcc/auto-profile.c
> ===
yes.
David
On Thu, Sep 26, 2013 at 9:26 AM, Dehao Chen wrote:
> This fixes an ICE when lipo_cmp_type handles NULL_PTR_TYPE.
>
> Bootstrapped and regression test on going?
>
> OK for google branches?
>
> Thanks,
> Dehao
>
> Index: gcc/l-ipo.c
>
On Thu, Sep 26, 2013 at 7:37 AM, Richard Biener
wrote:
> On Thu, Sep 26, 2013 at 1:10 AM, Xinliang David Li wrote:
>> I took the liberty to pick up Richard's original fvect-cost-model
>> patch and made some modification.
>>
>> What has not changed:
>>
This patch improves linux kernel performance with a large workload, so
it is good to first submit this to trunk and backport it.
thanks,
David
On Thu, Sep 26, 2013 at 3:27 PM, Jan Hubicka wrote:
>> Hi,
>>
>> Current default probably for builtin_expect is 0.9996.
>> This makes the freq of unlike
it might worth extend __builtin_expect to take more parameters:
1) argument to specify actual probability: __builtin_expect (x, 10, 0.6)
2) a more general way of doing this is to allow specifying multiple
values, and the probability is determined by # of occurances:
__builtin_expect (x, 10, 10, 20)
Nice extension. Test cases would be great to have.
thanks,
David
On Fri, Sep 27, 2013 at 7:50 AM, Jan Hubicka wrote:
> Hi,
> this patch makes it possible to access value range info from setmem/movstr
> that
> I plan to use in i386 memcpy/memset expansion code. It is all quite
> straighforward
turned on at -O2. To override this, use
option -fvect-cost-model=[cheap|dynamic|unlimited].
New Languages and Language specific improvements
thanks,
David
On Thu, Sep 26, 2013 at 11:09 AM, Xinliang David Li wrote:
> On Thu, Sep 26, 2013 at 7:37 AM, Richard Biener
> wrote:
>>
On Fri, Sep 27, 2013 at 11:50 AM, Teresa Johnson wrote:
> David and Rong,
>
> The following patch will disable -g/-gmlt when instrumenting for LIPO
> since they will affect the recorded ggc_memory used in the module
> grouping decision. Added -fripa-allow-debug to override this behavior.
>
> Passe
ok.
David
On Fri, Sep 27, 2013 at 1:03 PM, Teresa Johnson wrote:
> On Fri, Sep 27, 2013 at 12:01 PM, Xinliang David Li
> wrote:
>> On Fri, Sep 27, 2013 at 11:50 AM, Teresa Johnson
>> wrote:
>>> David and Rong,
>>>
>>> The following patch will
The key is that grouping results should not dependent on the presence
of -g flags. The downside of the patch is that it may slightly
underestimate the memory pressure at profile-use time, but that should
not be a big problem.
David
On Fri, Sep 27, 2013 at 1:18 PM, Rong Xu wrote:
> I don't quite
d.growth -= (info->size
* (100 - PARAM_VALUE (PARAM_COMDAT_SHARING_PROBABILITY))
+ 50) / 100;
What is the purpose of '50' here?
The patch is fine for Google branch.
Other tunings to think about -- I think the sharing probability should
not be a fixed value -- but depending on the functio
You can also add a test case of this form:
int foo( int t, int n, int *dst)
{
int j = 0;
int s = 1;
t++;
for (j = 0; j < n; j++)
{
dst[j] = t;
s *= t;
}
return s;
}
where without the fix the loop vectorization is missed.
David
On Fri, Sep 27, 2013 at
On Sat, Sep 28, 2013 at 3:05 PM, Jan Hubicka wrote:
>> > Nice extension. Test cases would be great to have.
>> Fore those you need i386 changes to actually use the info. I will post that
>> after some cleanup and additional testing.
>
> Hi,
> since I already caught your attention, here is the tar
Yes, that will do. Can you do it for me? I can't do testing easily
on arm myself.
thanks,
David
On Mon, Sep 30, 2013 at 3:29 AM, Kyrill Tkachov wrote:
> Hi Richard, David,
>
>> In principle yes. Note that it changes the behavior of -O2
>> -ftree-vectorize
>> as -ftree-vectorize does not i
On Tue, Oct 1, 2013 at 10:31 AM, Cong Hou wrote:
> The current uniform_vector_p() function only returns non-NULL when the
> vector is directly a uniform vector. For example, for the following
> gimple code:
>
> vect_cst_.15_91 = {_9, _9, _9, _9, _9, _9, _9, _9};
>
>
> The current implementation ca
On Tue, Oct 1, 2013 at 2:37 PM, Xinliang David Li wrote:
> On Tue, Oct 1, 2013 at 10:31 AM, Cong Hou wrote:
>> The current uniform_vector_p() function only returns non-NULL when the
>> vector is directly a uniform vector. For example, for the following
>> gimple code:
>&g
ok (for google-47 and google/main)
thanks,
David
On Fri, Sep 28, 2012 at 10:22 AM, Rong Xu wrote:
> Comments are inlined.
> Attached is the new patch.
>
> Thanks,
>
> -Rong
>
> On Tue, Sep 25, 2012 at 2:25 PM, Xinliang David Li wrote:
>> On Mon, Sep 24, 2012 at 2
On Mon, Oct 1, 2012 at 2:37 PM, Michael Meissner
wrote:
> I tracked down some of the other code that previously used REPORT_DETAILS, and
> MSG_NOTE is the new way to do the same thing. This bootstraps and no
> unexpected errors occur during make check. Is it ok to install?
>
> 2012-10-01 Michae
sier to read/maintain.
>
I find it confusing to be honest. The redundant check serves no purpose.
David
> Sharad
> Sharad
>
>
> On Mon, Oct 1, 2012 at 3:45 PM, Xinliang David Li wrote:
>> On Mon, Oct 1, 2012 at 2:37 PM, Michael Meissner
>> wrote:
>>> I tracked
What is the status of switch expansion GIMPLE rewrite? If it is not
planned for 4.8, It will be desirable to include this fix into trunk.
It also helps set up a good base line to test against regression.
thanks,
David
On Tue, Oct 2, 2012 at 6:09 PM, Easwaran Raman wrote:
> Hi,
> This patch pr
thanks for the update!
David
On Wed, Oct 3, 2012 at 10:37 AM, Steven Bosscher wrote:
> On Wed, Oct 3, 2012 at 6:12 PM, Xinliang David Li wrote:
>> What is the status of switch expansion GIMPLE rewrite? If it is not
>> planned for 4.8, It will be desirable to include this fix in
Hi Jason, Sri has addressed the comments you had on FE part. Can you
take a look if it is ok? Stage-1 is going to be closed soon, and we
hope to get this major feature in 4.8.
thanks,
David
On Tue, Sep 18, 2012 at 9:29 AM, Sriraman Tallam wrote:
> Ping.
>
> On Fri, Aug 24, 2012 at 5:34 PM,
thanks. That will be helpful.
David
On Fri, Oct 5, 2012 at 2:09 PM, Dehao Chen wrote:
> Sure, I'll add a detailed documentation in a gcc wiki page.
>
> Dehao
>
> On Fri, Oct 5, 2012 at 2:01 PM, Xinliang David Li wrote:
>> Dehao, the file auto-profile.c has some
xray feature is not in trunk yet.
David
On Fri, Oct 5, 2012 at 3:53 PM, Diego Novillo wrote:
> Harshit, why didn't you propose this patch for trunk? Why should we
> make it a google-local patch?
>
>
> Diego.
>
> On Fri, Sep 28, 2012 at 5:24 AM, Harshit Chopra wrote:
>> commit fc3a55ccec9bc770c
Ok for google branches.
Please consider resend the original xray patch to trunk (gcc-4_8) You
need to make the runtime bits available publicly though.
thanks,
David
On Fri, Sep 28, 2012 at 2:24 AM, Harshit Chopra wrote:
> commit fc3a55ccec9bc770c79f8a221f5abd397befc8f6
> Author: Harshit Chopra
Is there an agreed way for file naming?
David
On Wed, Oct 10, 2012 at 1:48 PM, Steven Bosscher wrote:
> On Wed, Oct 10, 2012 at 10:20 PM, Diego Novillo wrote:
>> * tree-asan.c: New file.
>> * tree-asan.h: New file.
>
> Nit: do we still need the "tree-" prefix? IMHO not.
>
> Ciao!
On Wed, Oct 10, 2012 at 2:08 PM, Steven Bosscher wrote:
> On Wed, Oct 10, 2012 at 11:00 PM, Xinliang David Li wrote:
>> Is there an agreed way for file naming?
>
> It was not my intent to start a bike shed discussion. This was just
> something I've been wondering for some
In a different thread, I proposed the following alternative to 'try_xxx':
template T* symbol::cast_to(symbol* p) {
if (p->is())
return static_cast(p);
return 0;
}
cast:
template T& symbol:as(symbol* p) {
assert(p->is())
return static_cast(*p);
}
David
On Wed, Sep 19, 2012
1 - 100 of 1051 matches
Mail list logo
