Hello,
>> Tested and bootstrap on all languages except java (PR50879) on
>> ppc64-redhat-linux, enabling SMS on loops with SC 1.
>>
>> OK for mainline?
>>
>
> OK, seems reasonable.
>
> Please fix typo (in original comment):
> - scheduling passes doesn't touch it. */
> + sche
On 20/10/2011, at 10:32 PM, Jan Hubicka wrote:
> Hi,
>> Jan,
>>
>> The following patch started as a one-liner for ipa-inline-analysis.c:
>> account_size_time() to merge predicates when we are adding data to entry[0]
>> (i.e., when space for 32 size_time entries is exhausted):
>>
>> @@ -537,6 +
'addxc' is like 'addx', but it uses the 64-bit carry condition
as input instead of the 32-bit one.
This allows us to do ne/ltu/gtu setcc operations comparing 64-bit
integers without temporaries.
Unfortunately, VIS3 only has addxc (not subxc) and only the 3-register
opcode is provided (no immedia
Am 28.10.2011 01:27, schrieb Steve Kargl:
The attached patch tightens the checking of the arguments
of the dshiftl and dshiftr argument. It also does a kind
type conversion when either I or J is a BOZ to the kind
type of the non-BOZ I or J.
OK for trunk?
OK with a changelog entry for the NEAR
Maxim Kuvyrkov writes:
> The following patch adds 9 tests code quality tests for inlining and
> devirtualization optimizations. These tests were originally developed by
> Matt.
I had a test case for one C (not C++) dynamic devirtualization issue I
ran into in PR45631. It's a bit harder for th
On 2011.10.28 at 00:41 +, Joseph S. Myers wrote:
> On Fri, 28 Oct 2011, Markus Trippelsdorf wrote:
>
> > By popular demand, I've prepared a patch that updates the in-tree
> > libtool to version 2.4.2. It is needed for lto-bootstrap with
> > -fno-fat-lto-objects and FreeBSD10.x versions.
> > I
On 2011.10.27 at 17:29 -0700, Andi Kleen wrote:
> Markus Trippelsdorf writes:
>
> > By popular demand, I've prepared a patch that updates the in-tree
> > libtool to version 2.4.2. It is needed for lto-bootstrap with
> > -fno-fat-lto-objects and FreeBSD10.x versions.
> > It's a pretty big update
Noticed this while combing through a sparc 64-bit testsuite run.
Committed to trunk.
gcc/
* config/sparc/sparc.md (64-bit vector moves): Use 'e' not 'f'
constraint.
---
gcc/ChangeLog |3 +++
gcc/config/sparc/sparc.md |4 ++--
2 files changed, 5 insertions(+)
As discussed in the thread starting at:
http://gcc.gnu.org/ml/gcc/2011-10/msg00478.html
We need to make the tests done in copyprop_hardreg_forward_1()
match those performed in copy_value(), and in particular make
sure that if we are narrowing the mode on big-endian we do
not perform the substitu
---
gcc/builtins.c | 39 +---
gcc/expr.h |4 -
gcc/optabs.c | 267 ---
gcc/optabs.h |4 +
4 files changed, 126 insertions(+), 188 deletions(-)
diff --git a/gcc/builtins.c b/gcc/builtins.c
index cb9da83..756070f 100644
---
Cc: ja...@redhat.com
Cc: ubiz...@gmail.com
---
gcc/config/i386/i386.md |5 +-
gcc/config/i386/sync.md | 306 +--
2 files changed, 167 insertions(+), 144 deletions(-)
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index a11a71b..7ce5
Cc: ja...@redhat.com
---
gcc/omp-low.c | 55 +++
gcc/optabs.c | 134 +++--
gcc/optabs.h | 13 ++
3 files changed, 84 insertions(+), 118 deletions(-)
diff --git a/gcc/omp-low.c b/gcc/omp-low.c
index 5faa084..972cb6d
---
gcc/ChangeLog.mm | 58 +
gcc/c-family/ChangeLog.mm |4 +++
2 files changed, 62 insertions(+), 0 deletions(-)
diff --git a/gcc/ChangeLog.mm b/gcc/ChangeLog.mm
index 498cc15..758b510 100644
--- a/gcc/ChangeLog.mm
+++ b/gcc/ChangeLog.mm
---
gcc/builtins.c |7 +++
gcc/omp-low.c |3 +--
gcc/optabs.c | 27 +++
gcc/optabs.h |3 +++
4 files changed, 30 insertions(+), 10 deletions(-)
diff --git a/gcc/builtins.c b/gcc/builtins.c
index 3367b50..cb9da83 100644
--- a/gcc/builtins.c
+++ b/gcc/
---
gcc/c-family/c-cppbuiltin.c | 50 ++-
1 files changed, 35 insertions(+), 15 deletions(-)
diff --git a/gcc/c-family/c-cppbuiltin.c b/gcc/c-family/c-cppbuiltin.c
index bb9893a..bf83c26 100644
--- a/gcc/c-family/c-cppbuiltin.c
+++ b/gcc/c-family/c-cppbui
---
gcc/genopinit.c |6 ++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/gcc/genopinit.c b/gcc/genopinit.c
index e658b44..1954783 100644
--- a/gcc/genopinit.c
+++ b/gcc/genopinit.c
@@ -259,6 +259,12 @@ static const char * const optabs[] =
"set_direct_optab_handler (atom
---
gcc/optabs.c |8
1 files changed, 8 insertions(+), 0 deletions(-)
diff --git a/gcc/optabs.c b/gcc/optabs.c
index b021042..307101b 100644
--- a/gcc/optabs.c
+++ b/gcc/optabs.c
@@ -7697,6 +7697,14 @@ maybe_gen_insn (enum insn_code icode, unsigned int nops,
case 6:
retur
---
gcc/builtins.c |7 ---
1 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/gcc/builtins.c b/gcc/builtins.c
index f47c3b1..3367b50 100644
--- a/gcc/builtins.c
+++ b/gcc/builtins.c
@@ -5523,11 +5523,12 @@ expand_builtin_atomic_is_lock_free (tree exp)
void
expand_builtin_mem_
This exposed a wealth of problems in code that has heretofore never
been tested. The fourth patch makes certain that all expansions of
compare-and-swap go through a single routine.
I've tested the whole series with and without the last patch. So
that I've tested both the sync_ and atomic_ paths.
On Thu, 2011-10-27 at 20:43 -0500, Peter Bergner wrote:
> Index: gcc/config/rs6000/476.opt
[snip]
> +Target Var(rs6000_link_stack) Init(1) Save
Oops, this should actually be Init(-1). The hunk above was just my
way of testing the modified code more by enabling it by default.
Sorry about that.
Pe
On 20/10/2011, at 10:11 PM, Jan Hubicka wrote:
> static clause_t
> -evaluate_conditions_for_edge (struct cgraph_edge *e, bool inline_p)
> +evaluate_conditions_vals_binfos_for_edge (struct cgraph_edge *e,
> + bool inline_p,
> +
Jason Merrill writes:
| Recent discussion on the committee reflector pointed out issues with
| mutable and constexpr; this patch implements what I think is a
| reasonable solution.
Yes, I agree. Thank you!
-- Gaby
I made a few formatting tweaks as follows, and checked in the patch.
Thanks!
Jason
commit 3e01e07381cdb16727fe6011202372ebcdfc
Author: Jason Merrill
Date: Thu Oct 27 17:44:39 2011 -0400
cleanup
diff --git a/gcc/cp/decl2.c b/gcc/cp/decl2.c
index 1e514af..be9044b 100644
--- a/gcc/cp/d
Recent discussion on the committee reflector pointed out issues with
mutable and constexpr; this patch implements what I think is a
reasonable solution.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit 4b41346ab344632f6c22adb3d947a093cdb770ac
Author: Jason Merrill
Date: Thu Oct 27 14:56
This patch provides a flag that allows atomics operation in-lining to be
turned off. The default is on, which is what we have today.
by specifying -fno-inline-atomics, none of the new __atomic routines
will be inlined, and will be left as external function calls. The
__sync routines are not
On 27/10/2011, at 12:22 PM, Martin Jambor wrote:
> Hi,
>
> I've been asked by Maxim Kuvyrkov to revive the following patch which
> has not made it to 4.6. Currently, when type based devirtualization
> detects a potential type change, it simply gives up on gathering any
> information on the objec
The following patch adds 9 tests code quality tests for inlining and
devirtualization optimizations. These tests were originally developed by Matt.
Current mainline can optimize only test 5. Martin's patch in
http://gcc.gnu.org/ml/gcc-patches/2011-10/msg02506.html will enable
optimization of
On Thu, 2011-10-13 at 10:03 -0700, Richard Henderson wrote:
> On 10/13/2011 08:49 AM, Peter Bergner wrote:
> > + if (TARGET_LINK_STACK)
> > + asm_fprintf (file, "\tbl 1f\n\tb 2f\n1:\n\tblr\n2:\n");
> > + else
> > + asm_fprintf (file, "\tbcl 20,31,1f\n1:\n");
>
> Wouldn't it be
On Fri, 28 Oct 2011, Markus Trippelsdorf wrote:
> By popular demand, I've prepared a patch that updates the in-tree
> libtool to version 2.4.2. It is needed for lto-bootstrap with
> -fno-fat-lto-objects and FreeBSD10.x versions.
> It's a pretty big update as you can see by the following diffstat.
Markus Trippelsdorf writes:
> By popular demand, I've prepared a patch that updates the in-tree
> libtool to version 2.4.2. It is needed for lto-bootstrap with
> -fno-fat-lto-objects and FreeBSD10.x versions.
> It's a pretty big update as you can see by the following diffstat. I
> cannot attach
.. maybe one final piece of information for today.
I'm trying to figure out where the very different args argument is
coming from.
Both in 4.5 and in mainline we have the same final chain of
tusbst_copy_build, preceded by tsubst_copy_and_build, tusbst_expr,
tsubt, tsubst_template_arg, the ar
On 10/27/2011 05:17 AM, Sergey Ostanevich wrote:
> + /* MOVCC semantics implies that source is always read which is wrong
> + for devices I/O that are defined using volatile in C. PR47698 */
> +
> + if (MEM_P (operands[2]) && MEM_VOLATILE_P (operands[2]))
> +return false;
This looks to b
.. thus now I'm investigating, comparing 4.5 to mainline, where the
former gets the TYPE_BINFO of the TREE_OPERAND (member, 0). And the
answer is "simple": at the beginning of tsubst_copy_and_build case
COMPONENT_REF itself! Where:
member = TREE_OPERAND (t, 1);
if (BASELINK_P (member))
On Thu, Oct 27, 2011 at 12:24:46AM +1030, Alan Modra wrote:
> more code than duplicating epilogues. From what I've seen, the
> duplicate tails are generally very small. I guess I should dump out
> some info so we can get a better idea.
There were 545 occurrences of shrink-wrap in the gcc/ dir fo
On 10/27/2011 03:47 PM, Torvald Riegel wrote:
> Fix tm_region association of blocks with the "over" label.
>
> * trans-mem.c (struct tm_region): Extended comment.
> (tm_region_init): Fix tm_region association of blocks with the "over"
> label used for transcation abort.
>
By popular demand, I've prepared a patch that updates the in-tree
libtool to version 2.4.2. It is needed for lto-bootstrap with
-fno-fat-lto-objects and FreeBSD10.x versions.
It's a pretty big update as you can see by the following diffstat. I
cannot attach the patch even as a gzip file, because o
This is the final patch.
Context diff also provided as gfc_conv_intrinsic_arith has many blank changes.
The changes in gfc_conv_intrinsic_arith involve the new initialization for
parent_se and ploop, a few additional conditions for scalar-specific stuff,
and a specific handling for scalar masks in
In the non-scalar case the gfc_se structs are initialized using the se pointer
given as argument to the function. To accomodate this, this patch changes
initialization to NULL with initialization to a pointer (initially set to
NULL for now).
The patch explains it better.
OK?
2011-10-19 Mikael Mori
In the non-scalar case the scalarizer needs to take code out of the outer loop,
which means that the inner one (the "sum" loop) has already been handled
before entering gfc_conv_intrinsic_arith, which means loop shall be a pointer
to that loop instead of the address of the local loop in that case.
This uses named argument to make it easier to follow.
OK?
2011-10-19 Mikael Morin
* trans-intrinsic.c (gfc_conv_intrinsic_arith): Small argument handling
cleanup.
diff --git a/trans-intrinsic.c b/trans-intrinsic.c
index 342d2cb..b701502 100644
--- a/trans-intrinsic.c
+++ b/trans
This makes the convention that all conditions are based on arguments presence
and their rank. In the hope that it makes code easier to read.
OK?
2011-10-19 Mikael Morin
* trans-intrinsic.c (gfc_conv_intrinsic_arith): Update conditions.
diff --git a/trans-intrinsic.c b/trans-intrinsic.c
This is the purpose of all the serie of patches: inline sum and product.
Patches 62..65, possibly don't deserve a patch of their own, they just make
the changes in gfc_conv_intrinsic_arith easier to read.
The combined diff (context, ignoring indent changes) also attached here.
Patch 62: Change co
Not exactly a regression, but frontend optimizations can make things worse
by forcing temporary generation.
This disables common function elimination and binary operator optimization
for inline intrinsics.
I hope I didn't miss anything else.
OK?
2011-10-19 Mikael Morin
* frontend-passes
This makes the necessary changes for the error markers to be generated at the
right place in the testcase.
that is for:
foo(&
sum(bar,1) &
)
if sum(bar,1) requires a temporary, the warning is emitted on the second line
instead of the first one, so that one can distinguish between ar
Patches 58 and 59 fix the {min,max}loc and {min,max}val intrinsics which use
multiple loops. See the comments in the patches for details.
Patch 57 avoids duplicated offset calculation in the code generated.
OK?
2011-10-19 Mikael Morin
* trans-intrinsic.c (gfc_conv_intrinsic_minmaxloc):
gfc_conv_elemental_dependency had some ad-hoc fixes to replace one array by its
temporary in the scalarizer.
It was using a fake gfc_loopinfo struct so that gfc_trans_create_temp_array
doesn't overwrite the real one. It was using fake array references and custom
delta calculation, so that scalariz
This adds dependency checking for inline functions.
OK?
2011-10-19 Mikael Morin
* trans.h (gfc_inline_intrinsic_function_p): Move prototype...
* gfortran.h (gfc_inline_intrinsic_function_p): ... here.
* dependency.c (gfc_check_argument_var_dependency): Check dependencies
This is a collection of changes not directly related, but made necessary by
the emergence of regressions in the testsuite without them.
Combined patch not attached, as the separate patches being largely independant,
it is the concatenation of them.
Patch 54: Enable dependency checking for inli
This updates gfc_trans_preloop_setup to handle multiple loops.
The preliminary patches 1..6 have made it quite painless:
- the condition "this is the the outermost dimension" is changed from
`dim == loop->dimen - 1' to `dim == loop->dimen - 1 && loop->parent == NULL'
- to retrieve information
This uses the just added parent loop pointer to create an outermost_loop
function and use it throughout the scalarizer.
OK?
2011-10-19 Mikael Morin
* trans-array.c (outermost_loop): New function.
(gfc_trans_array_constructor, gfc_set_vector_loop_bounds,
gfc_add_loop_ss_c
This introduces a convenience pointer parent for the gfc_loopinfo struct
(patch 49).
Patches 50 and 51 (with context diff) use it to update
gfc_trans_array_constructor and constant_array_constructor_loop_size
respectively.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_loopinfo): New
This introduces the nested_loop list of nested loops inside a gfc_loopinfo
struct (patch 47).
Patch 48 adds to the scalarizer functions self-recursive calls on the nested
loops.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_loopinfo): New fields nested and next.
* trans-arra
This updates gfc_trans_create_temp_array with looking at parent (outer) ss.
Context diff also provided.
OK?
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_create_temp_array): Loop over the parents.
diff --git a/trans-array.c b/trans-array.c
index d918fa8..1a86ae6 100644
--- a/trans-
gfc_trans_create_temp_arrays has been using loop bounds to set array
temporary bounds; this is a sensible thing to do as gfc_conv_loop_setup
tries to set loop bounds according to the most information (especially w.r.t.
constantness) it can find.
For plain arrays like `a(:,:)' or `b(:,1,:)', it's
As there will be more than one gfc_ss struct pointing to a single gfc_ss_info,
it needs to be reference counted. This introduces reference counting.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss_info): New field refcount.
* trans-array.c (free_ss_info): Decrement refcount
To be able, from one outer gfc_ss struct, to descend into the corresponding
inner gfc_ss struct in the nested loop, this introduces a nested_ss field.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss): New field nested_ss.
* trans-array.c (gfc_mark_ss_chain_used): Mark recur
We are going to loop over the parent array infos.
This introduces a new parent field to permit this loop (patch 40).
Then it will be used to loop over multiple parent ss:
- gfc_set_loop_bounds_from_array_spec (patch 41 with context diff)
- gfc_trans_array_constructor (patch 42)
- set_vector_loop
To avoid confusion between multiple loops, a new loop field is added to
the gfc_ss (patch 35).
Then we can remove superfluous loop argument in function interfaces:
- set_vector_loop_bounds (patch 36)
- gfc_trans_array_constructor (patch 37)
- gfc_trans_create_temp_array (patch 39)
Patch 38 prepa
gfc_trans_create_temp_array has code like this:
for (n = 0; n < loop->dimen; n++)
{
if (size == NULL)
{
/* blah */
continue;
}
/* bleh */
}
We are going to update this to handle more than one loop.
However the two branches will get different trea
This part of the serie takes care of updating the scalarizer code once its
core structures have changed.
There are basically two kinds of changes:
- for functions working on loops mostly one needs to take care of more than
one loop. This is done by attaching a list of nested loops to every
g
The code clearing specloop in gfc_conv_loop_setup would need some explanation.
This patch moves it to gfc_trans_create_temp_array (where the reason for
it are more clear) with a proper comment.
OK?
2011-10-19 Mikael Morin
* trans-array.c (gfc_conv_loop_setup, gfc_trans_create_temp_array
Loop-centered functions will be called recursively to handle nested loops.
This is the case of gfc_conv_loop_setup. However, gfc_conv_loop_setup has
a tempoary handling part in the middle which doesn't need to be called
recursively. This patch moves the beginning and ending of gfc_conv_loop_setup
i
This moves expr field from gfc_ss to gfc_ss_info.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_ss_info): Move field expr from
the former struct to the latter.
* trans-array.c
(gfc_get_array_ss, gfc_get_scalar_ss,
gfc_trans_constant_arr
This moves where field from gfc_ss to gfc_ss_info.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_ss_info): Move field
gfc_ss::where into gfc_ss_info.
* trans-array.c (gfc_add_loop_ss_code):
Update reference chains.
* trans-stmt.c (gfc_t
This moves data::info field from gfc_ss to gfc_ss_info.
The name is changed to array, as it is for all the non-scalar and non-temp
cases, thus all the array cases.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_ss_info): Move field
gfc_ss::data::info into gfc_s
This moves useflags field from gfc_ss to gfc_ss_info.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_ss_info): Move field
gfc_ss::useflags into gfc_ss_info.
* trans-array.c (gfc_mark_ss_chain_used, gfc_trans_preloop_setup,
gfc_trans_scalarizing_
This moves data::temp field from gfc_ss to gfc_ss_info.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_ss_info): Move member struct
gfc_ss::data::temp into gfc_ss_info::data.
* trans-array.c (gfc_get_temp_ss, gfc_conv_loop_setup): Update reference
This moves data::scalar field from gfc_ss to gfc_ss_info.
The expr subfield is renamed to value, as it is not the expression really,
it is a reference to a variable containing the pre-calculated value.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_ss_info): Move membe
This moves string_length field from gfc_ss to gfc_ss_info.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_ss_info): Move field
string_length from the former struct to the latter.
* trans-array.c
(gfc_get_temp_ss, gfc_trans_array_constructor, gfc
This moves type field from gfc_ss to a new gfc_ss_info struct.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss_info): New struct.
(gfc_get_ss_info): New macro.
(struct gfc_ss): Move type field to struct gfc_ss_info.
Add an info field of type gfc_ss_info.
This moves shape field from gfc_ss to gfc_array_info.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss, struct gfc_array_info): Move shape field
from the former struct to the latter.
* trans-array.c (gfc_conv_ss_startstride, gfc_conv_loop_setup):
Update field
This moves dim and dimen fields from gfc_array_info to gfc_ss.
The gfc_ss::data::temp::dimen field is also removed as it is now redundant.
The only non-trivial change is the removal of the dim array initialization in
gfc_conv_loop_setup (when the temp_ss's type is changed from GFC_SS_TEMP
to GFC_SS
This renames gfc_ss_info to gfc_array_info.
OK?
2011-10-19 Mikael Morin
* trans.h (struct gfc_ss_info, struct gfc_array_info):
Rename the former to the latter.
* trans-array.c (gfc_get_array_ss, gfc_trans_allocate_array_storage,
get_array_ref_dim, gfc_trans_creat
Same as previous patch, dim_ok uses dimensions and needs a gfc_ss struct as
argument.
The name is changed to the more descriptive transposed_dims and the logic
is inverted (dim_ok <=> !transposed_dims).
OK?
2011-10-19 Mikael Morin
* trans-array.c (dim_ok, transposed_dims): Rename the fo
Same as previous patch, get_array_ref_dim uses dimensions and thus needs
a gfc_ss struct as argument.
OK?
2011-10-19 Mikael Morin
* trans-array.c (get_array_ref_dim): Change argument type and name.
Obtain previous argument from the new argument in the body.
(gfc_trans_cr
Same as previous patch, gfc_set_vector_loop_bounds uses dimensions, and thus
needs a gfc_ss struct as argument.
gfc_ prefix removed along the way.
OK?
2011-10-19 Mikael Morin
* trans-array.c (gfc_set_vector_loop_bounds, set_vector_loop_bounds):
Rename the former to the latter.
gfc_trans_create_temp_array uses dimensions heavily, and dimensions are to be
moved from gfc_ss_info to gfc_ss. To have them still available in
gfc_trans_create_temp_array, the gfc_ss_info argument should be a gfc_ss.
OK?
2011-10-19 Mikael Morin
* trans-array.h (gfc_trans_create_temp_ar
This is a follow-up to the previous patch. It symplifies name obtention
so that later we can change structs with less pain. :-)
OK?
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_array_bound_check): Use ss argument
to get name.
diff --git a/trans-array.c b/trans-array.c
inde
The next patch is going to need array expression (available through gf_ss)
instead of just the array descriptor in gfc_trans_array_bound_check. This makes
the gfc_ss struct available as argument.
This requires an update of gfc_conv_array_index_offset too.
Both functions are static, they lose their
loop.temp_dim is set by gfc_trans_create_temp_array; there should be no reason
why that value wouldn't suit gfc_trans_constant_array_constructor, which has
then no reason to set the value directly.
This patch remove that code.
Then, the loop argument is useless and can be removed too.
The function
This is another preliminary change, to update function interfaces requiring
it, so that afterwards structures can be changed internally without impacting
function interfaces.
The main reason for these changes is that gfc_ss_info's dim and dimen fields
are to be moved to struct gfc_ss. Thus function
We should not be writing to gfc_ss_terminator.
It is working without this patch because gfc_ss_terminator's next pointer is
NULL, so the loop stops just after it, and because we are writing zero to
gfc_ss_terminator, but it is already all zeros anyway.
OK?
2011-10-19 Mikael Morin
* tran
We can't use temporaries to guess loop dimensions, as temporaries' bounds are
calculated from loop dimensions.
In the union:
union
{
struct { ... } scalar;
struct { ... } temp;
struct gfc_ss_info info
}
data;
We
This is a followup to my recent coarray/scalarizer changes, especially that one
cleaning-up gfc_walk_variable_expr/gfc_walk_array_ref:
http://gcc.gnu.org/ml/fortran/2011-09/msg00072.html
Scalar coarrays are not handled as AR_ELEMENT in the scalarizer.
Thus, no need to handle coarrays here.
OK?
2011
2011-10-19 Mikael Morin
* trans-array.c (gfc_conv_ss_startstride): Access array bounds along
array dimensions instead of loop dimensions.
diff --git a/trans-array.c b/trans-array.c
index 8359af2..f4d8a85 100644
--- a/trans-array.c
+++ b/trans-array.c
@@ -3347,9 +3347,11 @@ done:
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_preloop_setup): Assertify one condition.
diff --git a/trans-array.c b/trans-array.c
index 5500ec4..8359af2 100644
--- a/trans-array.c
+++ b/trans-array.c
@@ -2885,8 +2885,7 @@ gfc_trans_preloop_setup (gfc_loopinfo * loop, int dim, int f
As the first line of context shows, if the first condition is false, the second
is false too. Thus, the first condition is useless.
OK?
2011-10-19 Mikael Morin
* trans-array.c (get_array_ref_dim): Remove redundant condition.
diff --git a/trans-array.c b/trans-array.c
index 83fa7b6..5500
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_preloop_setup): Factor loop index
initialization.
diff --git a/trans-array.c b/trans-array.c
index 91359e9..e3134f5 100644
--- a/trans-array.c
+++ b/trans-array.c
@@ -2863,16 +2863,15 @@ gfc_trans_preloop_setup (gfc_loopinfo * l
This is a collection of independant and mostly trivial cleanups.
I don't attach the combined patch as it is the concatenation of the separate
patches.
Patch 07: Useless coarray code removal.
Patch 08: Remove redundant condition.
Patch 09: Assertify condition.
Patch 10: Use array dimensions instead
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_preloop_setup): Move common code...
(add_array_offset): ...into that new function.
diff --git a/trans-array.c b/trans-array.c
index 476978e..f615e4e 100644
--- a/trans-array.c
+++ b/trans-array.c
@@ -2830,6 +2830,34 @@ gfc_conv_
This is for consistency. As dim == dimen-1 means "we are in the outermost loop",
we should check against the loop dimension, not the array dimension.
An assertion is added to check that it is in fact the same.
OK?
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_preloop_setup): Use lo
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_preloop_setup): Remove redundant assertion.
Special case outermost loop.
diff --git a/trans-array.c b/trans-array.c
index e3134f5..f5e30ae 100644
--- a/trans-array.c
+++ b/trans-array.c
@@ -2867,7 +2867,10 @@ gfc_trans_preloop_s
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_preloop_setup): Move code earlier.
diff --git a/trans-array.c b/trans-array.c
index 4b21476..91359e9 100644
--- a/trans-array.c
+++ b/trans-array.c
@@ -2876,6 +2876,17 @@ gfc_trans_preloop_setup (gfc_loopinfo * loop, int dim, int flag,
This is a 6 steps program to update gfc_trans_preloop_setup.
gfc_trans_preloop_setup appeared completely rewritten. These step by step
patches show that it is not the case.
Combined patch attached here.
OK?
diff --git a/trans-array.c b/trans-array.c
index 3472804e4c6344e68d15af2feee627429ca61018..f
2011-10-19 Mikael Morin
* trans-array.c (gfc_trans_preloop_setup): Move array reference
initialisation earlier. Factor subsequent array references.
diff --git a/trans-array.c b/trans-array.c
index 3472804..4b21476 100644
--- a/trans-array.c
+++ b/trans-array.c
@@ -2842,6 +2842,7
The attached patch tightens the checking of the arguments
of the dshiftl and dshiftr argument. It also does a kind
type conversion when either I or J is a BOZ to the kind
type of the non-BOZ I or J.
OK for trunk?
2011-10-27 Steven G. Kargl
PR fortran/50573
* check.c (gfc_ch
Hi again,
On 10/27/2011 08:57 AM, Paolo Carlini wrote:
Ok, thanks, I'll work more on this later today. Then, if we fix that we don't
error out at all, right? Because this is a reject valid, AFAIK.
In that case, lookup_qualified_name shouldn't be returning error_mark_node.
A little more informa
Richard,
Just as Matt posted his findings about the effect of iterating early
optimizations, I've got the new patch ready. This patch is essentially a
complete rewrite and addresses the comments you made.
On 18/10/2011, at 9:56 PM, Richard Guenther wrote:
>>>
>>> If we'd want to iterate earl
This patch fixes the detection of transactions regions for transactions
that have __transaction_cancel statements. The block that has the "over"
label to which an aborted transaction jumps does not belong to this
transaction, but to the enclosing region. The included testcase would
fail with prior
> 2011-10-27 Chung-Lin Tang
>
> PR rtl-optimization/49720
> * simplify-rtx.c (simplify_relational_operation_1): Detect
> infinite recursion condition in "(eq/ne (plus x cst1) cst2)
> simplifies to (eq/ne x (cst2 - cst1))" case.
>
> testsuite/
> * g++.dg/tortur
1 - 100 of 177 matches
Mail list logo
