Javier Miranda [Mon, 1 Jul 2019 13:34:45 +0000 (13:34 +0000)]
[Ada] Disable expansion of 'Min/'Max of floating point types
2019-07-01 Javier Miranda <miranda@adacore.com>
gcc/ada/
* exp_attr.adb (Expand_Min_Max_Attribute): Disable expansion of
'Min/'Max on integer, enumeration, fixed point and floating
point types since the CCG backend now provides in file
standard.h routines to support it.
package VS is new Membership_Set
(Element_Type => Vertex_Id,
"=" => "=",
Hash => Hash_Vertex);
-----------------------
-- Local subprograms --
-----------------------
procedure Check_Belongs_To_Component
(R : String;
G : Instance;
V : Vertex_Id;
Exp_Comp : Component_Id);
-- Verify that vertex V of graph G belongs to component Exp_Comp. R is the
-- calling routine.
procedure Check_Belongs_To_Some_Component
(R : String;
G : Instance;
V : Vertex_Id);
-- Verify that vertex V of graph G belongs to some component. R is the
-- calling routine.
procedure Check_Destination_Vertex
(R : String;
G : Instance;
E : Edge_Id;
Exp_V : Vertex_Id);
-- Vertify that the destination vertex of edge E of grah G is Exp_V. R is
-- the calling routine.
procedure Check_Distinct_Components
(R : String;
Comp_1 : Component_Id;
Comp_2 : Component_Id);
-- Verify that components Comp_1 and Comp_2 are distinct (not the same)
procedure Check_Has_Component
(R : String;
G : Instance;
G_Name : String;
Comp : Component_Id);
-- Verify that graph G with name G_Name contains component Comp. R is the
-- calling routine.
procedure Check_Has_Edge
(R : String;
G : Instance;
E : Edge_Id);
-- Verify that graph G contains edge E. R is the calling routine.
procedure Check_Has_Vertex
(R : String;
G : Instance;
V : Vertex_Id);
-- Verify that graph G contains vertex V. R is the calling routine.
procedure Check_No_Component
(R : String;
G : Instance;
V : Vertex_Id);
-- Verify that vertex V does not belong to some component. R is the calling
-- routine.
procedure Check_No_Component
(R : String;
G : Instance;
G_Name : String;
Comp : Component_Id);
-- Verify that graph G with name G_Name does not contain component Comp. R
-- is the calling routine.
procedure Check_No_Edge
(R : String;
G : Instance;
E : Edge_Id);
-- Verify that graph G does not contain edge E. R is the calling routine.
procedure Check_No_Vertex
(R : String;
G : Instance;
V : Vertex_Id);
-- Verify that graph G does not contain vertex V. R is the calling routine.
procedure Check_Number_Of_Components
(R : String;
G : Instance;
Exp_Num : Natural);
-- Verify that graph G has exactly Exp_Num components. R is the calling
-- routine.
procedure Check_Number_Of_Edges
(R : String;
G : Instance;
Exp_Num : Natural);
-- Verify that graph G has exactly Exp_Num edges. R is the calling routine.
procedure Check_Number_Of_Vertices
(R : String;
G : Instance;
Exp_Num : Natural);
-- Verify that graph G has exactly Exp_Num vertices. R is the calling
-- routine.
procedure Check_Outgoing_Edge_Iterator
(R : String;
G : Instance;
V : Vertex_Id;
Set : ES.Instance);
-- Verify that all outgoing edges of vertex V of graph G can be iterated
-- and appear in set Set. R is the calling routine.
procedure Check_Source_Vertex
(R : String;
G : Instance;
E : Edge_Id;
Exp_V : Vertex_Id);
-- Vertify that the source vertex of edge E of grah G is Exp_V. R is the
-- calling routine.
procedure Check_Vertex_Iterator
(R : String;
G : Instance;
Comp : Component_Id;
Set : VS.Instance);
-- Verify that all vertices of component Comp of graph G can be iterated
-- and appear in set Set. R is the calling routine.
function Create_And_Populate return Instance;
-- Create a brand new graph (see body for the shape of the graph)
procedure Error (R : String; Msg : String);
-- Output an error message with text Msg within the context of routine R
procedure Test_Add_Edge;
-- Verify the semantics of routine Add_Edge
procedure Test_Add_Vertex;
-- Verify the semantics of routine Add_Vertex
procedure Test_All_Edge_Iterator;
-- Verify the semantics of All_Edge_Iterator
procedure Test_All_Vertex_Iterator;
-- Verify the semantics of All_Vertex_Iterator
procedure Test_Component;
-- Verify the semantics of routine Component
procedure Test_Component_Iterator;
-- Verify the semantics of Component_Iterator
procedure Test_Contains_Component;
-- Verify the semantics of routine Contains_Component
procedure Test_Contains_Edge;
-- Verify the semantics of routine Contains_Edge
procedure Test_Contains_Vertex;
-- Verify the semantics of routine Contains_Vertex
procedure Test_Delete_Edge;
-- Verify the semantics of routine Delete_Edge
procedure Test_Destination_Vertex;
-- Verify the semantics of routine Destination_Vertex
procedure Test_Find_Components;
-- Verify the semantics of routine Find_Components
procedure Test_Is_Empty;
-- Verify the semantics of routine Is_Empty
procedure Test_Number_Of_Components;
-- Verify the semantics of routine Number_Of_Components
procedure Test_Number_Of_Edges;
-- Verify the semantics of routine Number_Of_Edges
procedure Test_Number_Of_Vertices;
-- Verify the semantics of routine Number_Of_Vertices
procedure Test_Outgoing_Edge_Iterator;
-- Verify the semantics of Outgoing_Edge_Iterator
procedure Test_Present;
-- Verify the semantics of routine Present
procedure Test_Source_Vertex;
-- Verify the semantics of routine Source_Vertex
procedure Test_Vertex_Iterator;
-- Verify the semantics of Vertex_Iterator;
procedure Unexpected_Exception (R : String);
-- Output an error message concerning an unexpected exception within
-- routine R.
procedure Check_Belongs_To_Some_Component
(R : String;
G : Instance;
V : Vertex_Id)
is
begin
if not Present (Component (G, V)) then
Error (R, "vertex " & V'Img & " does not belong to a component");
end if;
end Check_Belongs_To_Some_Component;
procedure Check_Distinct_Components
(R : String;
Comp_1 : Component_Id;
Comp_2 : Component_Id)
is
begin
if Comp_1 = Comp_2 then
Error (R, "components are not distinct");
end if;
end Check_Distinct_Components;
procedure Check_Has_Component
(R : String;
G : Instance;
G_Name : String;
Comp : Component_Id)
is
begin
if not Contains_Component (G, Comp) then
Error (R, "graph " & G_Name & " lacks component");
end if;
end Check_Has_Component;
procedure Check_Has_Edge
(R : String;
G : Instance;
E : Edge_Id)
is
begin
if not Contains_Edge (G, E) then
Error (R, "graph lacks edge " & E'Img);
end if;
end Check_Has_Edge;
procedure Check_Has_Vertex
(R : String;
G : Instance;
V : Vertex_Id)
is
begin
if not Contains_Vertex (G, V) then
Error (R, "graph lacks vertex " & V'Img);
end if;
end Check_Has_Vertex;
procedure Check_No_Component
(R : String;
G : Instance;
V : Vertex_Id)
is
begin
if Present (Component (G, V)) then
Error (R, "vertex " & V'Img & " belongs to a component");
end if;
end Check_No_Component;
procedure Check_No_Component
(R : String;
G : Instance;
G_Name : String;
Comp : Component_Id)
is
begin
if Contains_Component (G, Comp) then
Error (R, "graph " & G_Name & " contains component");
end if;
end Check_No_Component;
procedure Check_No_Edge
(R : String;
G : Instance;
E : Edge_Id)
is
begin
if Contains_Edge (G, E) then
Error (R, "graph contains edge " & E'Img);
end if;
end Check_No_Edge;
procedure Check_No_Vertex
(R : String;
G : Instance;
V : Vertex_Id)
is
begin
if Contains_Vertex (G, V) then
Error (R, "graph contains vertex " & V'Img);
end if;
end Check_No_Vertex;
procedure Check_Number_Of_Components
(R : String;
G : Instance;
Exp_Num : Natural)
is
Act_Num : constant Natural := Number_Of_Components (G);
begin
if Act_Num /= Exp_Num then
Error (R, "inconsistent number of components");
Error (R, " expected: " & Exp_Num'Img);
Error (R, " got : " & Act_Num'Img);
end if;
end Check_Number_Of_Components;
procedure Check_Number_Of_Edges
(R : String;
G : Instance;
Exp_Num : Natural)
is
Act_Num : constant Natural := Number_Of_Edges (G);
begin
if Act_Num /= Exp_Num then
Error (R, "inconsistent number of edges");
Error (R, " expected: " & Exp_Num'Img);
Error (R, " got : " & Act_Num'Img);
end if;
end Check_Number_Of_Edges;
procedure Check_Number_Of_Vertices
(R : String;
G : Instance;
Exp_Num : Natural)
is
Act_Num : constant Natural := Number_Of_Vertices (G);
begin
if Act_Num /= Exp_Num then
Error (R, "inconsistent number of vertices");
Error (R, " expected: " & Exp_Num'Img);
Error (R, " got : " & Act_Num'Img);
end if;
end Check_Number_Of_Vertices;
-- Verify that all edges in the range E1 .. E99 exist
for Curr_E in E1 .. E99 loop
Check_Has_Edge (R, G, Curr_E);
end loop;
-- Delete each edge that corresponds to an even position in Edge_Id
for Curr_E in E1 .. E99 loop
if Edge_Id'Pos (Curr_E) mod 2 = 0 then
Delete_Edge (G, Curr_E);
end if;
end loop;
-- Verify that all "even" edges are missing, and all "odd" edges are
-- present.
for Curr_E in E1 .. E99 loop
if Edge_Id'Pos (Curr_E) mod 2 = 0 then
Check_No_Edge (R, G, Curr_E);
else
Check_Has_Edge (R, G, Curr_E);
end if;
end loop;
* impunit.adb: Add GNAT.Graphs to list Non_Imp_File_Names_95.
* Makefile.rtl, gcc-interface/Make-lang.in: Register unit
GNAT.Graphs.
* libgnat/g-dynhta.adb: Various minor cleanups (use Present
rather than direct comparisons).
(Delete): Reimplement to use Delete_Node.
(Delete_Node): New routine.
(Destroy_Bucket): Invoke the provided destructor.
(Present): New routines.
* libgnat/g-dynhta.ads: Add new generic formal Destroy_Value.
Use better names for the components of iterators.
* libgnat/g-graphs.adb, libgnat/g-graphs.ads: New unit.
* libgnat/g-lists.adb: Various minor cleanups (use Present
rather than direct comparisons).
(Delete_Node): Invoke the provided destructor.
(Present): New routine.
* libgnat/g-lists.ads: Add new generic formal Destroy_Element.
Use better names for the components of iterators.
(Present): New routine.
* libgnat/g-sets.adb, libgnat/g-sets.ads (Destroy, Preset,
Reset): New routines.
Ed Schonberg [Mon, 1 Jul 2019 13:34:30 +0000 (13:34 +0000)]
[Ada] Compiler abort on use of Invalid_Value on numeric positive subtype
Invalid_Value in most cases uses a predefined numeric value from a
built-in table, but if the type does not include zero in its range, the
literal 0 is used instead. In that case the value (produced by a call to
Get_Simple_Init_Val) must be resolved for proper type information.
The following must compile quietly:
gnatmake -q main
----
with Problems; use Problems;
with Text_IO; use Text_IO;
function P1 return Integer;
function P2 return Long_Integer;
-- Max. number of prime factors a number can have is log_2 N
-- For N = 600851475143, this is ~ 40
-- type P3_Factors is array (1 .. 40) of Long_Integer;
function P3 return Long_Integer;
type P4_Palindrome is range 100*100 .. 999*999;
function P4 return P4_Palindrome;
end Problems;
----
package body Problems is
function P1 return Integer is separate;
function P2 return Long_Integer is separate;
function P3 return Long_Integer is separate;
function P4 return P4_Palindrome is separate;
end Problems;
----
separate(Problems)
function P1 return Integer is
Sum : Integer range 0 .. 500_500 := 0;
begin
for I in Integer range 1 .. 1000 - 1 loop
if I mod 3 = 0 or I mod 5 = 0 then
Sum := Sum + I;
end if;
end loop;
return Sum;
end P1;
--
separate(Problems)
function P2 return Long_Integer is
subtype Total is Long_Integer range 0 .. 8_000002e6 ;
subtype Elem is Total range 0 .. 4e7 ;
Sum : Total := 0;
a, b, c : Elem;
begin
a := 1;
b := 2;
loop
if b mod 2 = 0 then
Sum := Sum + b;
end if;
c := b;
b := a + b;
a := c;
exit when b >= 4e6;
end loop;
return Sum;
end P2;
--
with Text_IO; use Text_IO;
with Ada.Numerics.Elementary_Functions; use Ada.Numerics.Elementary_Functions;
separate(Problems)
function P3 return Long_Integer is
while Dividend > 1 loop
Quotient := Dividend / Factor;
if Dividend mod Factor = 0 then
GPF := Factor;
Dividend := Quotient;
else
if Factor >= Quotient then
GPF := Dividend;
exit;
end if;
Factor := Factor + 1;
end if;
end loop;
return GPF;
end P3;
----
with Text_IO; use Text_IO;
separate(Problems)
function P4 return P4_Palindrome is
type TripleDigit is range 100 .. 999;
a, b: TripleDigit := TripleDigit'First;
function Is_Palindrome (X : in P4_Palindrome) return Boolean is
type Int_Digit is range 0 .. 9;
type Int_Digits is array (1 .. 6) of Int_Digit;
type Digit_Extractor is range 0 .. P4_Palindrome'Last;
Y : Digit_Extractor := Digit_Extractor (X);
X_Digits : Int_Digits;
begin
for I in reverse X_Digits'Range loop
X_Digits (I) := Int_Digit (Y mod 10);
Y := Y / 10;
end loop;
return
(X_Digits (1) = X_Digits (6) and X_Digits (2) = X_Digits (5) and
X_Digits (3) = X_Digits (4)) or
(X_Digits (2) = X_Digits (6) and X_Digits (3) = X_Digits (5) and
X_Digits(1) = 0);
end Is_Palindrome;
begin
for a in TripleDigit'Range loop
for b in TripleDigit'Range loop
c := P4_Palindrome (a * b);
if Is_Palindrome (c) then
if Max_Palindrome'Valid or else c > Max_Palindrome then
Max_Palindrome := c;
end if;
end if;
end loop;
end loop;
return Max_Palindrome;
end;
2019-07-01 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* exp_attr.adb (Expand_Attribute_Reference, case Invalid_Value):
Resolve result of call to Get_Simple_Init_Val, which may be a
conversion of a literal.
[Ada] Crash due to missing freeze nodes in transient scope
The following patch updates the freezing of expressions to insert the
generated freeze nodes prior to the expression that produced them when
the context is a transient scope within a type initialization procedure.
This ensures that the nodes are properly interleaved with respect to the
constructs that generated them.
* freeze.adb (Freeze_Expression): Remove the horrible useless
name hiding of N. Insert the freeze nodes generated by the
expression prior to the expression when the nearest enclosing
scope is transient.
gcc/testsuite/
* gnat.dg/freezing1.adb, gnat.dg/freezing1.ads,
gnat.dg/freezing1_pack.adb, gnat.dg/freezing1_pack.ads: New
testcase.
make_early_clobber_and_input_conflicts records allocno conflicts
between inputs and earlyclobber outputs. It (rightly) avoids
doing this for inputs that are explicitly allowed to match the
output due to matching constraints.
The problem is that whether this matching is allowed varies
between alternatives. At the moment the code avoids adding
a clobber if *any* enabled alternative allows the match,
even if some other operand makes that alternative impossible.
The specific instance of this for SVE is that some alternatives
allow matched earlyclobbers when a third operand X is constant zero.
We should avoid adding conflicts when X really is constant zero,
but should ignore the match if X is nonzero or nonconstant.
ira_setup_alts can already filter these alternatives out for us,
so all we need to do is use it in process_bb_node_lives. The
preferred_alternatives variable is only used for this earlyclobber
detection, so no other check should be affected.
With the previous patch to check the reject weight in ira_setup_alts,
this has the effect of ignoring expensive alternatives if we have
other valid alternatives with zero cost. It seems reasonable to base
the heuristic on only the alternatives that we'd actually like to use,
but if this ends up being too aggressive, we could instead make the new
reject behaviour conditional and only use it for add_insn_allocno_copies.
2019-07-01 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* ira-lives.c (process_bb_node_lives): Use ira_setup_alts.
ira_get_dup_out_num punted on operands that are matched to
earlyclobber outputs:
/* It is better ignore an alternative with early clobber. */
else if (*str == '&')
goto fail;
But I'm not sure why this is the right thing to do. At this stage
we've established that *all* alternatives of interest require the
input to match the output, so
(a) the earlyclobber can only affect other operands and
(b) not tying the registers is bound to introduce a move
The code was part of the initial commit and so isn't obviously
related to a specific testcase. Also, I can imagine LRA makes
a much better job of this situation than reload did. (Certainly
SVE uses matched earlyclobbers extensively and I haven't seen any
problems.)
In case this turns out to regress something important: the main
case that matters for SVE is the one in which all alternatives
are earlyclobber.
2019-07-01 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* ira.c (ira_get_dup_out_num): Don't punt for earlyclobbers.
Use recog_data to test for an output operand.
SVE has a prefix instruction (MOVPRFX) that acts as a move but is
designed to be easily fusible with the following instruction. The SVE
port therefore has lots of patterns with constraints of the form:
A: operand 0: =w,?w
...
operand n: 0, w
where the first alternative is a single instruction and the second
alternative uses MOVPRFX.
Ideally we want operand n to be allocated to the same register as
operand 0 in this case.
add_insn_allocno_copies is the main IRA routine that deals with tied
operands. It is (rightly) very conservative, and only handles cases in
which we're confident about saving a full move. So for a pattern like:
B: operand 0: =w,w
...
operand n: 0,w
we don't (and shouldn't) assume that tying operands 0 and n would
save the cost of a move.
But in A, the second alternative has a ? marker, which makes it more
expensive than the first alternative by a full reload. So I think for
copy elision we should ignore the untied operand n in the second
alternative of A.
One approach would be to add '*' markers to each pattern and make
ira_get_dup_out_num honour them. But I think the rule applies on
first principles, so marking with '*' shouldn't be necessary.
This patch instead makes ira_get_dup_out_num ignore expensive
alternatives if there are other alternatives that match exactly.
The cheapest way of doing that seemed to be to take expensive
alternatives out of consideration in ira_setup_alts, which provides
a bitmask of alternatives and has all the information available.
add_insn_allocno_copies is the only current user of ira_setup_alts,
so no other code should be affected.
If all available alternatives are disparaged or need a reload,
there's not much we can do to cut them down at this stage,
since it's hard to predict which operands will be reloaded and
which registers will need to be spilled.
An interesting case is patterns like this msp430 one:
Here alternative 3 is significantly more expensive then alternative 0
(reject costs 0 and 606 respectively). But if operand 1 is an integer
constant, we'll still use alternative 3 if operand 2 is an allocated
register. On the other hand, if operand 1 is an integer constant but
operand 2 is spilled to memory, we'll move the constant into a register
and use the first alternative.
So in this case, if operand 1 is a register, we should consider
only the first two alternatives and thus try to tie operand 1
to operand 0 (which we didn't do previously). If operand 1 is a
constant integer, we should consider at least alternatives 0, 1 and 3.
We could exclude alternative 2, but I don't have any evidence that
that's useful.
2019-07-01 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* ira.c (ira_setup_alts): If any valid alternatives have zero cost,
exclude any others that are disparaged or that are bound to need
a reload or spill.
(ira_get_dup_out_num): Expand comment.
ira_setup_alts has its own code to calculate the start of the
constraint string for each operand/alternative combination,
but preprocess_constraints now provides that information in (almost)
constant time for non-asm instructions. Using it here should speed
up the common case at the cost of potentially slowing down the handling
of asm statements.
The real reason for doing this is that a later patch wants to use
more of the operand_alternative information.
2019-07-01 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* ira.c (ira_setup_alts): Use preprocess_constraints to get the
constraint string for each operand/alternative combo. Only handle
'%' at the start of constraint strings, and look for it outside
the main loop.
add_insn_allocno_copies and its subroutines used HARD_REG_SET to
represent a bitmask of alternatives. There's not really any connection
between the number of registers and the maximum number of alternatives,
so this patch uses alternative_mask instead (which wasn't around when
this code was added).
This is just a minor clean-up making way for later patches.
2019-07-01 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* ira-int.h (ira_setup_alts, ira_get_dup_out_num): Use
alternative_mask instead of HARD_REG_SET to represent a
bitmask of alternatives.
* ira.c (ira_setup_alts, ira_get_dup_out_num): Likewise.
* ira-conflicts.c (add_insn_allocno_copies): Likewise.
Martin Liska [Mon, 1 Jul 2019 08:08:29 +0000 (10:08 +0200)]
Fix 2 clang warnings.
2019-07-01 Martin Liska <mliska@suse.cz>
* edit-context.c (test_applying_fixits_unreadable_file): Do not
use () for a constructor call.
(test_applying_fixits_line_out_of_range): Likewise.
* ggc-page.c (alloc_page): Use (void *) for %p printf format
argument.
(free_page): Likewise.
* gdbhooks.py (GdbPrettyPrinters.add_printer_for_types): Reorder
parameter names to match usage (no functional change).
(GdbPrettyPrinters.add_printer_for_regex): Ditto.
Uros Bizjak [Sun, 30 Jun 2019 21:12:07 +0000 (23:12 +0200)]
sse.md (ssse3_abs<mode>2): Rename from abs<mode>2.
* config/i386/sse.md (ssse3_abs<mode>2): Rename from abs<mode>2.
(abs<mode>2): New expander.
* config/i386/i386-builtin.def (__builtin_ia32_pabsb):
Use CODE_FOR_ssse3_absv8qi2.
(__builtin_ia32_pabsw): Use CODE_FOR_ssse3_absv4hi2.
(__builtin_ia32_pabsd): Use CODE_FOR_ssse3_absv2si2.
Iain Sandoe [Sun, 30 Jun 2019 14:58:45 +0000 (14:58 +0000)]
[PATCH, Ada] Push -shared-libgcc where needed.
Gnatlink has code that checks for duplicate '-shared-libgcc’ switches (but not
duplicate ‘static-libgcc’) and also pushes ’static-libgcc' onto the link line for
targets that default to static linking, provided '-shared-libgcc' is not present.
For targets that should use a shared libgcc we need the same process to be
applied (in inverse), in the event that they do not default to providing the
shared flag implicitly.
So this adds the complementary set of tests for the shared case and pushes
the shared flag as needed. As a minor tidy-up there’s no need push duplicates
of the libgcc switch onto the link line when one has already been seen (given by
the user).
The patch does not alter any of the platform defaults for static/shared libgcc,
but it ensures that the intent of the link is explicit.
gcc/ada/
2019-06-30 Iain Sandoe <iain@sandoe.co.uk>
* gnatlink.adb (Link_Step): Remove duplicate -static-libgcc switches.
Push -shared-libgcc explicitly, when it is the target default (unless
overidden by the static flag).
When the user has put an instance of shared/static-libgcc do not push
a duplicate of this.
Eric Botcazou [Sat, 29 Jun 2019 09:01:27 +0000 (09:01 +0000)]
decl.c (gnat_to_gnu_entity): Beep up comment on SAVED...
* gcc-interface/decl.c (gnat_to_gnu_entity): Beep up comment on SAVED,
and tweak comment on the assertion about the scopes of Itypes. Do not
skip the regular processing for Itypes that are E_Record_Subtype with
a Cloned_Subtype. Get the Cloned_Subtype for every E_Record_Subtype
if the type is dummy and hasn't got its own freeze node.
<E_Record_Subtype>: Save again the DECL of the Cloned_Subtype, if any.
<E_Access_Subtype>: Save again the DECL of the equivalent type.
(Gigi_Equivalent_Type) <E_Access_Subtype>: New case.
Eric Botcazou [Sat, 29 Jun 2019 08:22:35 +0000 (08:22 +0000)]
utils.c (unchecked_convert): Tweak comment.
* gcc-interface/utils.c (unchecked_convert): Tweak comment. Only skip
dereferences when padding to have the same size on both sides. Do it
for destination types with self-referential size too.
Eric Botcazou [Sat, 29 Jun 2019 08:10:20 +0000 (08:10 +0000)]
decl.c (gnat_to_gnu_entity): If the type requires strict alignment, then set the RM size to the type size.
* gcc-interface/decl.c (gnat_to_gnu_entity) <E_Record_Type>: If the
type requires strict alignment, then set the RM size to the type size.
Rework handling of alignment and sizes of tagged types in ASIS mode.
(validate_size): Rename local variable and remove special handling for
strict-alignment types.
* gcc-interface/utils.c (finish_record_type): Constify local variables
and use properly typed constants.
Eric Botcazou [Sat, 29 Jun 2019 07:53:27 +0000 (07:53 +0000)]
decl.c (gnat_to_gnu_field): Rework error messages for fields requiring strict alignment...
* gcc-interface/decl.c (gnat_to_gnu_field): Rework error messages for
fields requiring strict alignment, add explicit test on Storage_Unit
for position and size, and mention type alignment for position.
Eric Botcazou [Sat, 29 Jun 2019 07:30:22 +0000 (07:30 +0000)]
decl.c (set_nonaliased_component_on_array_type): Add missing guard for the presence of TYPE_CANONICAL.
* gcc-interface/decl.c (set_nonaliased_component_on_array_type): Add
missing guard for the presence of TYPE_CANONICAL.
(set_reverse_storage_order_on_array_type): Likewise.
Eric Botcazou [Sat, 29 Jun 2019 07:16:37 +0000 (07:16 +0000)]
expr.c (expand_expr_real_1): Apply the big-endian adjustment for bit-fields to all aggregate types.
* expr.c (expand_expr_real_1) <BIT_FIELD_REF>: Apply the big-endian
adjustment for bit-fields to all aggregate types.
ada/
* gcc-interface/gigi.h (make_packable_type): Remove default value.
(value_factor_p): Tweak prototype.
* gcc-interface/decl.c (gnat_to_gnu_entity): Add comment.
(gnat_to_gnu_component_type): Likewise.
(gnat_to_gnu_field): Likewise. Fetch the position of the field earlier
and simplify the condition under which the type is packed. Declare
local variable is_bitfield. Pass 1 as max_align to make_packable_type
if it is set to true.
(copy_and_substitute_in_layout): Pass 0 to make_packable_type.
* gcc-interface/utils.c (make_packable_array_type): New function.
(make_packable_type): Use it to rewrite the type of array field.
(maybe_pad_type): Pass align parameter to make_packable_type.
(create_field_decl): Minor tweaks.
(value_factor_p): Assert that FACTOR is a power of 2 and replace the
modulo computation by a masking operation.
Michael Meissner [Fri, 28 Jun 2019 20:19:54 +0000 (20:19 +0000)]
Update pc-relative support.
2019-06-28 Michael Meissner <meissner@linux.ibm.com>
* config/rs6000/predicates.md (pcrel_address): Use
SYMBOL_REF_LOCAL_P to determine if a label is local.
(pcrel_external_address): New predicate.
(non_prefixed_mem_operand): Delete, predicate not used.
* config/rs6000/rs6000.h (SYMBOL_FLAG_PCREL_P): Delete, we now use
SYMBOL_REF_LOCAL_P to determine if we can use pc-relative
addressing.
(SYMBOL_REF_PCREL_P): Likewise.
A recent RTX cost commit has changed the costs for ARC700 leading to errors in slsr-13.c test.
This commit fixes this issue by reverting the cost computation for short instructions.
Jan Beulich [Fri, 28 Jun 2019 08:46:56 +0000 (08:46 +0000)]
x86: fix CVT{,T}PD2PI insns
With just an "m" constraint misaligned memory operands won't be forced
into a register, and hence cause #GP. So far this was guaranteed only
in the case that CVT{,T}PD2DQ were chosen (which looks to be the case on
x86-64 only).
Switch the second alternative to Bm and also replace
nonimmediate_operand by vector_operand.
Dennis Zhang [Fri, 28 Jun 2019 08:42:09 +0000 (08:42 +0000)]
[Arm] Remove constraint strings from define_expand constructs in the back end
A number of Arm define_expand patterns have specified constraints for
their operands. But the constraint strings are ignored at expand time
and are therefore redundant/useless. We now avoid specifying constraints
in new define_expands, but we should clean up the existing define_expand
definitions.
2019-06-28 Dennis Zhang <dennis.zhang@arm.com>
* config/arm/arm.md: Remove redundant constraints from
define_expand but leave reload_inm and reload_outm patterns
untouched since they need special constraints to work.
* config/arm/arm-fixed.md: Remove redundant constraints from
define_expand.
* config/arm/iwmmxt.md: Likewise.
* config/arm/neon.md: Likewise.
* config/arm/sync.md: Likewise.
* config/arm/thumb1.md: Likewise.
* config/arm/vec-common.md: Likewise.
This patch adds libgcc configuration option to disable TM clone
registry. This option helps to reduce code size for embedded targets
which do not need transactional memory support.
Jason Merrill [Thu, 27 Jun 2019 21:29:19 +0000 (17:29 -0400)]
PR c++/55442 - memory-hog with highly recursive constexpr.
This testcase in the PR is extremely recursive, and therefore uses a huge
amount of memory on caching the results of individual calls. We no longer
need to track all calls to catch infinite recursion, as we have other limits
on maximum depth and operations count. So let's only cache a few calls at
the top level: 8 seems to be a reasonable compromise.
Jakub Jelinek [Thu, 27 Jun 2019 21:23:09 +0000 (23:23 +0200)]
re PR tree-optimization/91010 (ICE: Segmentation fault (in location_wrapper_p))
PR tree-optimization/91010
* tree-vect-stmts.c (scan_operand_equal_p): If offset1 == offset2,
return true. Otherwise, don't call operand_equal_p if offset1 or
offset2 is NULL and just return false.
Iain Sandoe [Thu, 27 Jun 2019 19:08:16 +0000 (19:08 +0000)]
[Darwin, PPC] Allow the user to override the use of hard float in kexts.
The default for the kernel is soft-float, however a user writing a kernel
extension might want to make use of hard float. This change makes
" -mkernel -mhard-float " work as expected.
Iain Sandoe [Thu, 27 Jun 2019 18:56:53 +0000 (18:56 +0000)]
[Darwin, PPC] Do not use longcall for 64b code.
The linker [ld64] that supports 64Bit does not need the JBSR longcall
optimisation, and will not work with the most generic case (where the
symbol is undefined external, but there is no symbl stub). So switch
the longcall option off. ld64 will generate branch islands as needed.
2019-06-27 Iain Sandoe <iain@sandoe.co.uk>
* config/rs6000/rs6000.c (darwin_rs6000_override_options): Do not
use longcall for 64b code.
Bill Schmidt [Thu, 27 Jun 2019 13:59:20 +0000 (13:59 +0000)]
rs6000.c (rs6000_option_override_internal): Enable -fvariable-expansion-in-unroller by default.
2019-06-27 Bill Schmidt <wschmidt@linux.ibm.com>
* config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
-fvariable-expansion-in-unroller by default.
* doc/invoke.texi (-fvariable-expansion-in-unroller): Document new
default for Power.
Jan Hubicka [Thu, 27 Jun 2019 12:08:12 +0000 (12:08 +0000)]
lto-common.c: tree-pretty-print.h
* lto-common.c: tree-pretty-print.h
(type_streaming_finished): New static var.
(gimple_register_canonical_type_1): Return updated hash; handle ODR
types.
(iterative_hash_canonical_type): Update use of
gimple_register_canonical_type_1.
Jan Hubicka [Thu, 27 Jun 2019 12:07:43 +0000 (14:07 +0200)]
class.c (layout_class_type): Set TYPE_CXX_ODR_P for as-base type copy.
* class.c (layout_class_type): Set TYPE_CXX_ODR_P for as-base
type copy.
* ipa-devirt.c (odr_type_d): Add tbaa_enabled flag.
(add_type_duplicate): When odr hash is not allocated, to nothing.
(odr_based_tbaa_p): New function.
(set_type_canonical_for_odr_type): New function.
* ipa-utils.h (enable_odr_based_tbaa, odr_based_tbaa_p,
set_type_canonical_for_odr_type): New.
* tree.c (gimple_canonical_types_compatible_p): ODR types with
ODR based TBAA are not equivalent to non-ODR types.
* lto-common.c: Include demangle.h and tree-pretty-print.h
(type_streaming_finished): New static var.
(gimple_register_canonical_type_1): Return updated hash; handle ODR
types.
(iterative_hash_canonical_type): Update use of
gimple_register_canonical_type_1.
* g++.dg/lto/alias-2_0.C: New testcase.
* g++.dg/lto/alias-2_1.C: New testcase.
Jakub Jelinek [Thu, 27 Jun 2019 11:13:10 +0000 (13:13 +0200)]
re PR target/90991 (_mm_loadu_ps instrinsic translates to vmovaps in combination with _mm512_insertf32x4)
PR target/90991
* config/i386/sse.md (avx_vec_concat<mode>): Use nonimmediate_operand
instead of register_operand for operands[1], add m to its constraints
if operands[2] uses "C" constraint. Ensure in condition that if
operands[2] is not 0, then operands[1] is not a MEM. For last two
alternatives, use unaligned loads instead of aligned if operands[1] is
misaligned_operand.
* gcc.target/i386/avx2-pr90991-1.c: New test.
* gcc.target/i386/avx512dq-pr90991-2.c: New test.
When I refactored the filesystem_error code I changed it to only use the
constructor parameter in the what() string, instead of the string
returned by system_error::what(). That meant it no longer included the
description of the error_code that system_error adds. This restores the
previous behaivour, as encouraged by the standard ("Implementations
should include the system_error::what() string and the pathnames of
path1 and path2 in the native format in the returned string").
PR libstdc++/91012
* src/c++17/fs_path.cc (filesystem_error::_Impl): Use a string_view
for the what_arg parameters.
(filesystem_error::filesystem_error): Pass system_error::what() to
the _Impl constructor.
* testsuite/27_io/filesystem/filesystem_error/cons.cc: Ensure that
filesystem_error::what() contains system_error::what().
Jeff Law [Wed, 26 Jun 2019 21:36:27 +0000 (15:36 -0600)]
re PR tree-optimization/90883 (Generated code is worse if returned struct is unnamed)
PR tree-optimization/90883
* tree-ssa-alias.c (stmt_kills_ref_p): Handle BUILT_IN_CALLOC.
* tree-ssa-dse.c: Update various comments to distinguish between
dead and redundant stores.
(initialize_ao_ref_for_dse): Handle BUILT_IN_CALLOC.
(dse_optimize_redundant_stores): New function.
(delete_dead_or_redundant_call): Renamed from delete_dead_call.
Distinguish between dead and redundant calls in dump output. All
callers updated.
(delete_dead_or_redundant_assignment): Similarly for assignments.
(dse_optimize_stmt): Handle _CHK variants. For statements which
store 0 into multiple memory locations, try to prove a subsequent
store is redundant.
PR tree-optimization/90883
* g++.dg/tree-ssa/pr90883.C: New test.
* gcc.dg/tree-ssa/ssa-dse-36.c: New test.
Iain Sandoe [Wed, 26 Jun 2019 19:04:50 +0000 (19:04 +0000)]
[PATCH, PPC 2/2] Fix Darwin bootstrap after split of rs6000.c.
To fix this we need to make the branch islands (or code) visible between
both files. This keeps the generation side in rs6000.c and moves the
output routine to rs6000-logue.c, placing a reference to the islands
vector in rs6000-internal.h.
2019-06-26 Iain Sandoe <iain@sandoe.co.uk>
* config/rs6000/rs6000-internal.h (branch_island): New typedef.
(branch_islands): New extern.
* config/rs6000/rs6000-logue.c (macho_branch_islands): Moved from
* config/rs6000/rs6000.c: .. here.
Iain Sandoe [Wed, 26 Jun 2019 19:00:16 +0000 (19:00 +0000)]
[PATCH, PPC 1/2] Make sure the gt- files are built for all sub-targets.
The new gt-rs6000-logue.h is common to all sub-targets in the port, so
it needs to be added for them.
It seems better to place the common target_gtfiles in the powerpc*-*-*
section, rather than duplicating them in sub-targets. This would make it
less likely that a sub-target would be overlooked in any future file
introductions.
2019-06-26 Iain Sandoe <iain@sandoe.co.uk>
* config.gcc (powerpc*-*-linux*): Move target_gtfiles from here..
(powerpc*-*-*) ... to here.
Jonathan Wakely [Wed, 26 Jun 2019 14:38:23 +0000 (15:38 +0100)]
Add new helper traits for signed/unsigned integer types
Reuse the __is_one_of alias in additional places, and define traits to
check for signed/unsigned integer types so we don't have to duplicate
those checks elsewhere.
The additional overloads for std::byte in <bit> were reviewed by LEWG
and considered undesirable, so this patch removes them.
* include/bits/fs_path.h (path::__is_encoded_char): Use __is_one_of.
* include/std/bit (_If_is_unsigned_integer_type): Remove.
(_If_is_unsigned_integer): Use __is_unsigned_integer.
(rotl(byte, unsigned), rotr(byte, unsigned), countl_zero(byte))
(countl_one(byte), countr_zero(byte), countr_one(byte))
(popcount(byte), ispow2(byte), ceil2(byte), floor2(byte))
(log2p1(byte)): Remove.
* include/std/charconv (__detail::__is_one_of): Move to <type_traits>.
(__detail::__is_int_to_chars_type): Remove.
(__detail::__integer_to_chars_result_type): Use __is_signed_integer
and __is_unsigned_integer.
* include/std/type_traits (__is_one_of): Move here from <charconv>.
(__is_signed_integer, __is_unsigned_integer): New helpers.
* testsuite/26_numerics/bit/bit.pow.two/ceil2.cc: Remove test for
std::byte overload.
* testsuite/26_numerics/bit/bit.pow.two/floor2.cc: Likewise.
* testsuite/26_numerics/bit/bit.pow.two/ispow2.cc: Likewise.
* testsuite/26_numerics/bit/bit.pow.two/log2p1.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countl_one.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countl_zero.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countr_one.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countr_zero.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/popcount.cc: Likewise.
* testsuite/26_numerics/bit/bitops.rot/rotl.cc: Likewise.
* testsuite/26_numerics/bit/bitops.rot/rotr.cc: Likewise.
Nathan Sidwell [Wed, 26 Jun 2019 12:58:39 +0000 (12:58 +0000)]
[PR preprocessor/90927] Fixe dependency output
https://gcc.gnu.org/ml/gcc-patches/2019-06/msg01664.html
libcpp/
PR preprocessor/90927
* mkdeps.c (mkdeps::vec::operator[]): Add non-const variant.
(deps_add_target): Deal with out of order unquoted targets.
/* Return true if the OFFSET is valid for the quad address instructions that
use d-form (register + offset) addressing. */
diff --git a/gcc/config/rs6000/rs6000-logue.c b/gcc/config/rs6000/rs6000-logue.c
index 9df4b5a..adc137b 100644
--- a/gcc/config/rs6000/rs6000-logue.c
+++ b/gcc/config/rs6000/rs6000-logue.c
@@ -4025,8 +4025,8 @@ rs6000_output_function_prologue (FILE *file)
/* -mprofile-kernel code calls mcount before the function prolog,
so a profiled leaf function should stay a leaf function. */
-static bool
-rs6000_keep_leaf_when_profiled ()
+bool
+rs6000_keep_leaf_when_profiled (void)
{
return TARGET_PROFILE_KERNEL;
}
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 3fc4029..bcfc881 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -1338,7 +1338,6 @@ static bool rs6000_secondary_reload_move (enum rs6000_reg_type,
secondary_reload_info *,
bool);
rtl_opt_pass *make_pass_analyze_swaps (gcc::context*);
-static bool rs6000_keep_leaf_when_profiled () __attribute__ ((unused));
static tree rs6000_fold_builtin (tree, int, tree *, bool);
/* Hash table stuff for keeping track of TOC entries. */
--
1.8.3.1
Jakub Jelinek [Wed, 26 Jun 2019 08:26:18 +0000 (10:26 +0200)]
re PR target/90991 (_mm_loadu_ps instrinsic translates to vmovaps in combination with _mm512_insertf32x4)
PR target/90991
* config/i386/sse.md
(*<extract_type>_vinsert<shuffletype><extract_suf>_0): Use vmovupd,
vmovups, vmovdqu, vmovdqu32 or vmovdqu64 instead of the aligned
insns if operands[2] is misaligned_operand.
Li Jia He [Wed, 26 Jun 2019 08:23:06 +0000 (08:23 +0000)]
[RS6000] Change maddld match_operand from DI to GPR
From PowerPC ISA3.0, the description of `maddld RT, RA.RB, RC` is as follows:
64-bit RA and RB are multiplied and then the RC is signed extend to 128 bits,
and add them together.
We only apply it to 64-bit mode (DI) when implementing maddld. However, if we
can guarantee that the result of the maddld operation will be limited to 32-bit
mode (SI), we can still apply it to 32-bit mode (SI).
gcc/ChangeLog
2019-06-26 Li Jia He <helijia@linux.ibm.com>
* config/rs6000/rs6000.h (TARGET_MADDLD): Remove the restriction of
TARGET_POWERPC64.
* config/rs6000/rs6000.md (maddld): Change maddld match_operand from DI
to GPR.
gcc/testsuite/ChangeLog
2019-06-26 Li Jia He <helijia@linux.ibm.com>
Martin Liska [Wed, 26 Jun 2019 06:44:28 +0000 (08:44 +0200)]
Fix missing else keyword seen with clang-static-analyzer:
2019-06-26 Martin Liska <mliska@suse.cz>
* bb-reorder.c (connect_better_edge_p): Add missing else
statement in the middle of if-else statements.
/home/marxin/Programming/gcc/gcc/bb-reorder.c:1031:2: warning: Value stored to 'is_better_edge' is never read
is_better_edge = true;
^ ~~~~
/home/marxin/Programming/gcc/gcc/bb-reorder.c:1034:2: warning: Value stored to 'is_better_edge' is never read
is_better_edge = false;
^ ~~~~~
Jason Merrill [Wed, 26 Jun 2019 04:56:07 +0000 (00:56 -0400)]
PR c++/70462 - unnecessary base ctor variant with final.
As pointed out in the PR, we don't need base 'tor variants for a final
class, since it can never be a base. I tried also dropping complete
variants for abstract classes, but that runs into ABI compatibility problems
with older releases that refer to those symbols.
* optimize.c (populate_clone_array): Skip base variant if
CLASSTYPE_FINAL.
(maybe_clone_body): We don't need an alias if we are only defining
one clone.
Steven G. Kargl [Wed, 26 Jun 2019 04:31:14 +0000 (04:31 +0000)]
re PR fortran/90988 (Wrong error message with variables named "PUBLIC*")
2019-06-24 Steven G. Kargl <kargl@gcc.gnu.org>
PR Fortran/90988
* decl.c (access_attr_decl): Use temporary variable to reduce
unreadability of code. Normalize jumping to return.
(gfc_match_protected): Fix parsing error. Add comments to
explain code. Remove dead code.
(gfc_match_private): Use temporary variable to reduce unreadability
of code. Fix parsing error. Move code to test for blank PRIVATE.
Remove dead code.
(gfc_match_public): Move code to test for blank PUBLIC. Fix
parsing error. Remove dead code.
Ian Lance Taylor [Wed, 26 Jun 2019 00:17:32 +0000 (00:17 +0000)]
libgo/testsuite: ignore symbols with a leading dot in symtogo
On AIX, a function has two symbols, a text symbol (with a leading dot)
and a data one (without it).
As the tests must be run only once, only the data symbol can be used to
retrieve the final go symbol. Therefore, all symbols beginning with a dot
are ignored by symtogo.
Ian Lance Taylor [Wed, 26 Jun 2019 00:04:36 +0000 (00:04 +0000)]
cmd/go: silence ar with D flag failures
The first call of ar must not show its output in order to avoid useless
error messages about D flag.
The corresponding Go toolchain patch is CL 182077.
projects / thirdparty / gcc.git / log
