Re: [Patch, libfortran] PR 52434/48878/38199 Improve floating point formatted writes
On Thu, Mar 15, 2012 at 01:52, Jerry DeLisle jvdeli...@charter.net wrote: I like the idea behind this patch. I confess, I have not studied the two test cases that you are modifying, but the changes seem to stick out with too many digits there. Is this really correct? When I get another moment, I will look closer. Maybe you can explain the need for this change a little more. Sorry if I am being to careful. Its been many months since I delved into the formatting code. The reason is that with the patch, the default rounding is whatever snprintf() gives us. At least with glibc, snprintf() rounds ties to even, but I'm not sure all implementations do this rather than rounding ties to away. So the testcase changes just make sure that we're not rounding a tie. E.g. 1.25 rounded to two significant digits is 1.2 if one rounds ties to even, 1.3 if ties are rounded away. So using 1.250001 ensures that the rounded value is 1.3 with both rounding modes. I suppose another option would be to assume that rounding is ties to even, and keep an eye out for regressions on other targets. -- Janne Blomqvist
[Ada] Implement Ada 2012 attributes First_Valid and Last_Valid
This patch implements the new attributes First_Valid and Last_Valid. These apply to static discrete types with at least one valid value. The static discrete type may have a static predicate (which is the case where these attributes are useful). They return the lowest and highest values for which valid values (that is values that satisfy any static predicate) exist. The following shows error detection in action (compiled with -gnat12 -gnatj60 -gnatld7) 1. procedure FLValidError (P : Integer) is 2.subtype R1 is integer range 1 .. 0; 3.subtype R2 is integer range 1 .. 10 4. with Dynamic_Predicate = R2 P; 5.subtype R3 is integer range 1 .. 10 6. with Static_Predicate = R3 12; 7.subtype R4 is integer range 1 .. P; 8. 9.Val : Integer; 10. 11. begin 12.Val := Float'First_Valid; -- Not discrete | prefix of First_Valid attribute must be discrete type 13.Val := Float'Last_Valid;-- Not discrete | prefix of Last_Valid attribute must be discrete type 14.Val := R1'First_Valid; -- No values | prefix of First_Valid attribute must be subtype with at least one value 15.Val := R1'Last_Valid; -- No values | prefix of Last_Valid attribute must be subtype with at least one value 16.Val := R2'First_Valid; -- Dynamic predicate | prefix of First_Valid attribute may not have dynamic predicate 17.Val := R2'Last_Valid; -- Dynamic_Predicate | prefix of Last_Valid attribute may not have dynamic predicate 18.Val := R3'First_Valid; -- No values | prefix of First_Valid attribute must be subtype with at least one value 19.Val := R3'Last_Valid; -- No values | prefix of Last_Valid attribute must be subtype with at least one value 20.Val := R4'First_Valid; -- Non-static subtype | prefix of First_Valid attribute must be a static subtype 21.Val := R4'Last_Valid; -- Non-static subtype | prefix of Last_Valid attribute must be a static subtype 22. end FLValidError; The following compiles and executes quietly 1. procedure FLRange is 2.subtype R1 is Integer range 1 .. 10; 3.subtype R2 is Integer range 1 .. 10 4. with Static_Predicate = R2 2 or R2 9; 5.subtype R3 is Integer range 1 .. 10 6. with Static_Predicate = R3 3 or R3 8; 7.subtype R4 is Integer range 1 .. 10 8. with Static_Predicate = R4 = 3 and R4 = 9; 9. 10.procedure Fail is 11.begin 12. raise Program_Error; 13.end Fail; 14. 15. begin 16.if R1'First_Valid /= 1 or else R1'Last_Valid /= 10 then 17. Fail; 18.end if; 19. 20.if R2'First_Valid /= 1 or else R2'Last_Valid /= 10 then 21. Fail; 22.end if; 23. 24.if R3'First_Valid /= 1 or else R3'Last_Valid /= 10 then 25. Fail; 26.end if; 27. 28.if R4'First_Valid /= 3 or else R4'Last_Valid /= 9 then 29. Fail; 30.end if; 31. end FLRange; Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Robert Dewar de...@adacore.com * exp_attr.adb (Expand_N_Attribute_Reference): Add handling of First_Valid/Last_Valid. * sem_attr.adb (Check_First_Last_Valid): New procedure (Analyze_Attribute): Add handling of First_Valid and Last_Valid (Eval_Attribute): ditto. * snames.ads-tmpl: Add entries for First_Valid and Last_Valid. Index: exp_attr.adb === --- exp_attr.adb(revision 185390) +++ exp_attr.adb(working copy) @@ -5701,10 +5701,12 @@ Attribute_Enabled | Attribute_Epsilon | Attribute_Fast_Math| + Attribute_First_Valid | Attribute_Has_Access_Values| Attribute_Has_Discriminants| Attribute_Has_Tagged_Values| Attribute_Large| + Attribute_Last_Valid | Attribute_Machine_Emax | Attribute_Machine_Emin | Attribute_Machine_Mantissa | Index: sem_attr.adb === --- sem_attr.adb(revision 185390) +++ sem_attr.adb(working copy) @@ -217,10 +217,14 @@ -- allowed with a type that
[Ada] Cleaning up of quantified expression analysis and expansion.
This patch corrects the previous messy and erroneous analysis of quantified expression. Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Vincent Pucci pu...@adacore.com * exp_ch4.adb (Expand_N_Quantified_Expression): Expand the original quantified expression node. * sem_ch4.adb (Analyze_Quantified_Expression): Properly analyze the quantified expression and preserve the original non-analyzed quantified expression when an expansion is needed. * sem_ch5.adb (Analyze_Iteration_Scheme): Special treatment for quantified expressions. (Analyze_Iterator_Specification): Special treatment for quantified expressions. Index: sem_ch5.adb === --- sem_ch5.adb (revision 185390) +++ sem_ch5.adb (working copy) @@ -2087,8 +2087,18 @@ Check_Controlled_Array_Attribute (DS); - Make_Index (DS, LP, In_Iter_Schm = True); + -- The index is not processed during the analysis of a + -- quantified expression but delayed to its expansion where the + -- quantified expression is transformed into an expression with + -- actions. + if Nkind (Parent (N)) /= N_Quantified_Expression + or else Operating_Mode = Check_Semantics + or else Alfa_Mode + then + Make_Index (DS, LP, In_Iter_Schm = True); + end if; + Set_Ekind (Id, E_Loop_Parameter); -- If the loop is part of a predicate or precondition, it may @@ -2097,14 +2107,7 @@ -- because the second one may be created in a different scope, -- e.g. a precondition procedure, leading to a crash in GIGI. - -- Note that if the parent node is a quantified expression, - -- this preservation is delayed until the expansion of the - -- quantified expression where the node is rewritten as an - -- expression with actions. - - if (No (Etype (Id)) or else Etype (Id) = Any_Type) - and then Nkind (Parent (N)) /= N_Quantified_Expression - then + if No (Etype (Id)) or else Etype (Id) = Any_Type then Set_Etype (Id, Etype (DS)); end if; @@ -2241,14 +2244,14 @@ -- If domain of iteration is an expression, create a declaration for -- it, so that finalization actions are introduced outside of the loop. -- The declaration must be a renaming because the body of the loop may - -- assign to elements. + -- assign to elements. In case of a quantified expression, this + -- declaration is delayed to its expansion where the node is rewritten + -- as an expression with actions. - -- Note that if the parent node is a quantified expression, this - -- declaration is created during the expansion of the quantified - -- expression where the node is rewritten as an expression with actions. - if not Is_Entity_Name (Iter_Name) -and then Nkind (Parent (Parent (N))) /= N_Quantified_Expression +and then (Nkind (Parent (Parent (N))) /= N_Quantified_Expression + or else Operating_Mode = Check_Semantics + or else Alfa_Mode) then declare Id : constant Entity_Id := Make_Temporary (Loc, 'R', Iter_Name); Index: exp_ch4.adb === --- exp_ch4.adb (revision 185390) +++ exp_ch4.adb (working copy) @@ -7891,9 +7891,22 @@ Cond : Node_Id; Decl : Node_Id; I_Scheme : Node_Id; + Original_N : Node_Id; Test : Node_Id; begin + -- Retrieve the original quantified expression (non analyzed) + + if Present (Loop_Parameter_Specification (N)) then + Original_N := Parent (Parent (Loop_Parameter_Specification (N))); + else + Original_N := Parent (Parent (Iterator_Specification (N))); + end if; + + -- Rewrite N with the original quantified expression + + Rewrite (N, Original_N); + Decl := Make_Object_Declaration (Loc, Defining_Identifier = Tnn, @@ -7904,13 +7917,6 @@ Cond := Relocate_Node (Condition (N)); - -- Reset flag analyzed in the condition to force its analysis. Required - -- since the previous analysis was done with expansion disabled (see - -- Resolve_Quantified_Expression) and hence checks were not inserted - -- and record comparisons have not been expanded. - - Reset_Analyzed_Flags (Cond); - if Is_Universal then Cond := Make_Op_Not (Loc, Cond); end if; @@ -7926,9 +7932,14 @@ Make_Exit_Statement (Loc))); if Present (Loop_Parameter_Specification (N)) then -
[Ada] Misuse of constant class-wide objects
An object declaration of a class-wide object with a tag-indeterminate initial value is rewritten as a renaming of a dereference. The renaming must preserve the kind of the object (constant or variable). Previous to this patch, the compiler failed to reject a call to a primitive operation with an in-out controlling formal, when the actual was a constant class-wide object. Compiling main.adb in -gnat05 mode must yield: main.adb:16:04: actual for I must be a variable --- with Element; use Element; procedure Main is Object : constant Element.I_Interface'Class := Element.T_Class'(null record); begin Object.Add; end Main; --- package Element is type I_Interface is interface; procedure Add (I : in out I_Interface) is abstract; function Create return I_Interface'Class; type T_Class is new I_Interface with null record; overriding procedure Add (I : in out T_Class) is null; end Element; Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Ed Schonberg schonb...@adacore.com * exp_ch3.adb (Expand_N_Object_Declaration): When rewriting the declaration of a class-wide object, retain the Ekind to prevent subsequent misuse of constants. Index: exp_ch3.adb === --- exp_ch3.adb (revision 185390) +++ exp_ch3.adb (working copy) @@ -4829,10 +4829,12 @@ -- object renaming declaration ---because these identifiers -- were previously added by Enter_Name to the current scope. -- We must preserve the homonym chain of the source entity - -- as well. + -- as well. We must also preserve the kind of the entity, + -- which may be a constant. Set_Chars (Defining_Identifier (N), Chars (Def_Id)); Set_Homonym (Defining_Identifier (N), Homonym (Def_Id)); + Set_Ekind (Defining_Identifier (N), Ekind (Def_Id)); Exchange_Entities (Defining_Identifier (N), Def_Id); end; end if;
[Ada] Missing finalization call of interface class-wide object
This patch corrects the code which detects whether an interface class-wide object has been initialized by a controlled function call. -- Source -- -- element.ads with Ada.Containers.Indefinite_Doubly_Linked_Lists; with Ada.Containers.Indefinite_Holders; package Element is type I_Interface is interface; procedure Add (I : in out I_Interface) is abstract; function = (Left, Right : I_Interface) return Boolean is abstract; procedure Clear (Self : in out I_Interface'Class); package Interface_Holder is new Ada.Containers.Indefinite_Holders (Element_Type = I_Interface'Class, = = =); function Create return I_Interface'Class; type T_Abstract_Element is abstract tagged null record; function = (Left, Right : T_Abstract_Element) return Boolean; type T_Concrete_Element is new T_Abstract_Element with null record; package Element_Collection is new Ada.Containers.Indefinite_Doubly_Linked_Lists (Element_Type = T_Abstract_Element'Class); type T_Class is new I_Interface with record Attributs : Element_Collection.List; end record; overriding procedure Add (I : in out T_Class); overriding function = (Left, Right : T_Class) return Boolean is (True); end Element; -- element.adb package body Element is function Create return I_Interface'Class is begin return T_Class'(Attributs = Element_Collection.Empty_List); end Create; overriding procedure Add (I : in out T_Class) is begin I.Attributs.Append (T_Concrete_Element'(null record)); end Add; function = (Left, Right : T_Abstract_Element) return Boolean is begin return False; end =; procedure Clear (Self : in out I_Interface'Class) is Elmt : T_Class := T_Class (Self); begin Elmt.Attributs.Clear; end Clear; end Element; -- main.adb with Element; use type Element.I_Interface; procedure Main is Holder : Element.Interface_Holder.Holder := Element.Interface_Holder.To_Holder (Element.Create); begin for I in 1 .. 100 loop declare Object : Element.I_Interface'Class := Holder.Element; begin Object.Add; Holder.Replace_Element (Object); end; end loop; end Main; - -- Compilation and expected output -- - $ gnatmake -q -gnat12 main.adb -largs -lgmem $ ./main $ gnatmem ./main $ Global information $ -- $Total number of allocations:30203 $Total number of deallocations :30203 $Final Water Mark (non freed mem) : 0 Bytes $High Water Mark: 13.98 Kilobytes Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Hristian Kirtchev kirtc...@adacore.com * exp_util.adb (Initialized_By_Ctrl_Function): Add code to process the case when a function call appears in object.operation format. Index: exp_util.adb === --- exp_util.adb(revision 185390) +++ exp_util.adb(working copy) @@ -3960,11 +3960,28 @@ -- function Initialized_By_Ctrl_Function (N : Node_Id) return Boolean is - Expr : constant Node_Id := Original_Node (Expression (N)); + Expr : Node_Id := Original_Node (Expression (N)); + begin + if Nkind (Expr) = N_Function_Call then +Expr := Name (Expr); + end if; + + -- The function call may appear in object.operation format. Strip + -- all prefixes and retrieve the function name. + + loop +if Nkind (Expr) = N_Selected_Component then + Expr := Selector_Name (Expr); +else + exit; +end if; + end loop; + return -Nkind (Expr) = N_Function_Call - and then Needs_Finalization (Etype (Expr)); + Nkind_In (Expr, N_Expanded_Name, N_Identifier) + and then Ekind (Entity (Expr)) = E_Function + and then Needs_Finalization (Etype (Entity (Expr))); end Initialized_By_Ctrl_Function; --
[Ada] Ada 2012 AI05-288: conformance for formal access-to-subprogram types
AI05-288 specifies that subtype conformance is required for actual types for generic formal access-to-subprogram types, rather than just mode conformance. This is a binding interpretation. Compiling pack2.adb must yield: pack2.adb:4:37: not subtype conformant with declaration at pack1.ads:2 pack2.adb:4:37: type of N does not match generic type ACC is access procedure (N : Natural); package Pack1 is procedure Test1 (A : out ACC); end Pack1; --- package body Pack1 is procedure Test1 (A : out ACC) is begin null; end Test1; end Pack1; --- package Pack2 is procedure Cause_A_Problem; end Pack2; --- with Pack1; package body Pack2 is type Actual_Acc is access procedure (N : Integer); package New_Pack1 is new Pack1 (Actual_Acc); procedure Cause_A_Problem is X : Actual_Acc; begin New_Pack1.Test1 (X); X (-1); -- PROBLEM end Cause_A_Problem; end Pack2; --- Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Ed Schonberg schonb...@adacore.com * sem_ch6.ads, sem_ch6.adb (Check_Subtype_Conformant): add Get_Inst formal, so that conformance within an instantiation follows renamings of formals. This is similar to what is already done in Check_Mode_conformant. * sem_ch12.adb (Vailidate_Access_Subprogram_Instance): check that formal and actual are subtype conformant. Previously only mode conformance was required. Index: sem_ch12.adb === --- sem_ch12.adb(revision 185390) +++ sem_ch12.adb(working copy) @@ -10433,7 +10433,11 @@ Abandon_Instantiation (Actual); end if; - Check_Mode_Conformant + -- In Ada 2012, actuals for access_to_subprograms must be subtype + -- conformant with the generic formal. Previous to AI05-288 only mode + -- conformance was required. + + Check_Subtype_Conformant (Designated_Type (Act_T), Designated_Type (A_Gen_T), Actual, Index: sem_ch6.adb === --- sem_ch6.adb (revision 185390) +++ sem_ch6.adb (working copy) @@ -7251,14 +7251,16 @@ (New_Id : Entity_Id; Old_Id : Entity_Id; Err_Loc : Node_Id := Empty; - Skip_Controlling_Formals : Boolean := False) + Skip_Controlling_Formals : Boolean := False; + Get_Inst : Boolean := False) is Result : Boolean; pragma Warnings (Off, Result); begin Check_Conformance (New_Id, Old_Id, Subtype_Conformant, True, Result, Err_Loc, - Skip_Controlling_Formals = Skip_Controlling_Formals); + Skip_Controlling_Formals = Skip_Controlling_Formals, + Get_Inst = Get_Inst); end Check_Subtype_Conformant; --- Index: sem_ch6.ads === --- sem_ch6.ads (revision 185390) +++ sem_ch6.ads (working copy) @@ -28,7 +28,7 @@ type Conformance_Type is (Type_Conformant, Mode_Conformant, Subtype_Conformant, Fully_Conformant); - pragma Ordered (Conformance_Type); + -- pragma Ordered (Conformance_Type); -- Conformance type used in conformance checks between specs and bodies, -- and for overriding. The literals match the RM definitions of the -- corresponding terms. This is an ordered type, since each conformance @@ -141,7 +141,8 @@ (New_Id : Entity_Id; Old_Id : Entity_Id; Err_Loc : Node_Id := Empty; - Skip_Controlling_Formals : Boolean := False); + Skip_Controlling_Formals : Boolean := False; + Get_Inst : Boolean := False); -- Check that two callable entities (subprograms, entries, literals) -- are subtype conformant, post error message if not (RM 6.3.1(16)), -- the flag being placed on the Err_Loc node if it is specified, and
Re: [dwarf]:
Thanks, committed. (And sorry for the truncated subject). Tristan. On Mar 14, 2012, at 6:50 PM, Jason Merrill wrote: OK. Jason
Re: [C] Handle #pragma before a parameter
Thanks, committed. Tristan. On Mar 14, 2012, at 5:33 PM, Joseph S. Myers wrote: On Wed, 14 Mar 2012, Tristan Gingold wrote: Hi, it happens that some system headers on VMS have #pragma between parameters. This is spotted by building the Ada runtime. This patch simply handles them. Manually tested by building for ia64-hp-openvms. No regressions for c on x86_64-darwin. Ok for trunk ? OK. -- Joseph S. Myers jos...@codesourcery.com
[Ada] New pragma and aspect Contract_Case to refine contracts
A new pragma and aspect Contract_Case is defined, which allows defining
fine-grain specifications that can complement or replace the contract given
by a precondition and a postcondition. Additionally, the Contract_Case pragma
or aspect can be used by testing and formal verification tools. The compiler
checks its validity and, depending on the assertion policy at the point of
declaration of the pragma or aspect, it may insert a check in the executable.
When compiled with assertions enabled (-gnata) and run, the following code
raises an exception:
$ ./main
$ raised SYSTEM.ASSERTIONS.ASSERT_FAILURE : failed contract case from p.ads:3
--- main.adb
1. with P; use P;
2.
3. procedure Main is
4.X : Integer;
5. begin
6.X := 10;
7.Incr (X);
8.X := -10;
9.Incr (X);
10. end Main;
--- p.adb
1. package body P is
2.procedure Incr (X : in out Integer) is
3.begin
4. if X /= -10 then
5. X := X + 1;
6. end if;
7.end Incr;
8. end P;
--- p.ads
1. package P is
2.procedure Incr (X : in out Integer) with
3. Contract_Case = (Name = ,
4.Mode = Nominal,
5.Requires = X 0,
6.Ensures = X = X'Old + 1);
7. end P;
---
Tested on x86_64-pc-linux-gnu, committed on trunk
2012-03-15 Yannick Moy m...@adacore.com
* aspects.adb, aspects.ads (Aspect_Id): New GNAT aspect
Aspect_Contract_Case.
* gnat_rm.texi Document the new pragma/aspect
Contract_Case. Correct the documentation of the existing
pragma/aspect Test_Case with the new semantics.
* sem_attr.adb (Analyze_Attribute): Allow use of 'Result in the
Ensures component of a Contract_Case pragma.
* sem_ch13.adb (Analyze_Aspect_Specifications): Check new aspect
and translate it into a pragma.
(Check_Aspect_At_Freeze_Point): Take into account the new aspect.
* sem_ch3.adb, sinfo.adb, sinfo.ads Renaming of TC (for test case)
into CTC (for contract and test case).
* sem_ch6.adb (Process_PPCs): Generate Check pragmas from
Contract_Case pragmas, similarly to what is done already for
postconditions.
* sem_prag.adb, sem_prag.ads (Check_Contract_Or_Test_Case):
Renaming of Check_Test_Case.
(Analyze_Pragma, Sig_Flags): Take into account the new pragma.
* sem_util.adb, sem_util.ads Renaming to take into account the
new pragma, so that functions which applied only to Test_Case
now apply to both Test_Case and Contract_Case.
* par-prag.adb, sem_warn.adb, snames.ads-tmpl Take into account
the new pragma.
Index: sem_ch3.adb
===
--- sem_ch3.adb (revision 185390)
+++ sem_ch3.adb (working copy)
@@ -2204,9 +2204,9 @@
Check_Subprogram_Contract (Sent);
- Prag := Spec_TC_List (Contract (Sent));
+ Prag := Spec_CTC_List (Contract (Sent));
while Present (Prag) loop
- Analyze_TC_In_Decl_Part (Prag, Sent);
+ Analyze_CTC_In_Decl_Part (Prag, Sent);
Prag := Next_Pragma (Prag);
end loop;
end if;
Index: gnat_rm.texi
===
--- gnat_rm.texi(revision 185410)
+++ gnat_rm.texi(working copy)
@@ -120,6 +120,7 @@
* Pragma Complete_Representation::
* Pragma Complex_Representation::
* Pragma Component_Alignment::
+* Pragma Contract_Case::
* Pragma Convention_Identifier::
* Pragma CPP_Class::
* Pragma CPP_Constructor::
@@ -855,6 +856,7 @@
* Pragma Complete_Representation::
* Pragma Complex_Representation::
* Pragma Component_Alignment::
+* Pragma Contract_Case::
* Pragma Convention_Identifier::
* Pragma CPP_Class::
* Pragma CPP_Constructor::
@@ -1704,6 +1706,108 @@
pragma @code{Pack}, pragma @code{Component_Alignment}, or a record rep
clause), the GNAT uses the default alignment as described previously.
+@node Pragma Contract_Case
+@unnumberedsec Pragma Contract_Case
+@cindex Contract cases
+@findex Contract_Case
+@noindent
+Syntax:
+
+@smallexample @c ada
+pragma Contract_Case (
+ [Name =] static_string_Expression
+ ,[Mode =] (Nominal | Robustness)
+ [, Requires = Boolean_Expression]
+ [, Ensures = Boolean_Expression]);
+@end smallexample
+
+@noindent
+The @code{Contract_Case} pragma allows defining fine-grain specifications
+that can complement or replace the contract given by a precondition and a
+postcondition. Additionally, the @code{Contract_Case} pragma can be used
+by testing and formal verification tools. The compiler checks its validity and,
+depending on the assertion policy at the point of declaration of the pragma,
+it may insert a check in the executable. For
[Ada] Issue warning on suspicious contract cases when -gnatw.t is set
The warning option -gnatw.t already issued warnings on suspicious
postconditions. This extends it to contract cases, which is a GNAT
pragma/aspect allowing to express fine-grain contracts. GNAT now detects
these cases on the following code:
$ gcc -c -gnatc -gnatw.t -gnat12 p.ads
1. package P is
2.function A_Is_Positive (X : Integer) return Boolean with
3. Contract_Case = (Name= normal case,
|
warning: Ensures component refers only to pre-state
warning: contract cases do not mention result
4.Mode= Nominal,
5.Ensures = X = 0);
6.procedure A_Incr (X : in Integer; Y : out Integer) with
7. Contract_Case = (Name= normal case,
|
warning: Ensures component refers only to pre-state
8.Mode= Nominal,
9.Ensures = X = X + 1);
10. end P;
Tested on x86_64-pc-linux-gnu, committed on trunk
2012-03-15 Yannick Moy m...@adacore.com
* gnat_ugn.texi Document the extension of option -gnatw.t.
* sem_ch3.adb (Analyze_Declaration): Check for suspicious
contracts only after contract cases have been semantically
analyzed.
* sem_ch6.adb (Check_Subprogram_Contract): Consider also Ensures
components of contract cases for detecting suspicious contracts.
Index: sem_ch3.adb
===
--- sem_ch3.adb (revision 185415)
+++ sem_ch3.adb (working copy)
@@ -2196,19 +2196,26 @@
Spec := Specification (Original_Node (Decl));
Sent := Defining_Unit_Name (Spec);
+ -- Analyze preconditions and postconditions
+
Prag := Spec_PPC_List (Contract (Sent));
while Present (Prag) loop
Analyze_PPC_In_Decl_Part (Prag, Sent);
Prag := Next_Pragma (Prag);
end loop;
- Check_Subprogram_Contract (Sent);
+ -- Analyze contract-cases and test-cases
Prag := Spec_CTC_List (Contract (Sent));
while Present (Prag) loop
Analyze_CTC_In_Decl_Part (Prag, Sent);
Prag := Next_Pragma (Prag);
end loop;
+
+ -- At this point, entities have been attached to identifiers.
+ -- This is required to be able to detect suspicious contracts.
+
+ Check_Subprogram_Contract (Sent);
end if;
Next (Decl);
Index: gnat_ugn.texi
===
--- gnat_ugn.texi (revision 185410)
+++ gnat_ugn.texi (working copy)
@@ -5696,9 +5696,11 @@
@emph{Activate warnings on suspicious contracts.}
@cindex @option{-gnatw.t} (@command{gcc})
This switch activates warnings on suspicious postconditions (whether a
-pragma @code{Postcondition} or a @code{Post} aspect in Ada 2012). A
-function postcondition is suspicious when it does not mention the result
-of the function. A procedure postcondition is suspicious when it only
+pragma @code{Postcondition} or a @code{Post} aspect in Ada 2012)
+and suspicious contract cases (pragma @code{Contract_Case}). A
+function postcondition or contract case is suspicious when no postcondition
+or contract case for this function mentions the result of the function.
+A procedure postcondition or contract case is suspicious when it only
refers to the pre-state of the procedure, because in that case it should
rather be expressed as a precondition. The default is that such warnings
are not generated. This warning can also be turned on using @option{-gnatwa}.
Index: sem_ch6.adb
===
--- sem_ch6.adb (revision 185415)
+++ sem_ch6.adb (working copy)
@@ -6953,6 +6953,9 @@
-- Last postcondition on the subprogram, or else Empty if either no
-- postcondition or only inherited postconditions.
+ Last_Contract_Case : Node_Id := Empty;
+ -- Last contract-case on the subprogram, or else Empty
+
Attribute_Result_Mentioned : Boolean := False;
-- Whether attribute 'Result is mentioned in a postcondition
@@ -6971,9 +6974,14 @@
-- reference to attribute 'Old, in order to ignore its prefix, which
-- is precisely evaluated in the pre-state. Otherwise return OK.
+ procedure Process_Contract_Cases (Spec : Node_Id);
+ -- This processes the Spec_CTC_List from Spec, processing any contract
+ -- case from the list. The caller has checked that Spec_CTC_List is
+ -- non-Empty.
+
procedure Process_Post_Conditions (Spec : Node_Id; Class : Boolean);
-- This processes the Spec_PPC_List from Spec, processing any
- -- postconditions from the list. If Class is True, then only
+ -- postcondition from the list. If Class is
[Ada] Switch to control maximum number of instantiations
This patch adds the switch (-gnateinn, MAX_INSTANTIATIONS=nn in VMS) to control the maximum number of instantiations. This may be used to increase the limit from the default of 8000 in the very rare case where a single unit legitimately has more than 8000 instantiations. The following program: 1. procedure MaxInst is 2.generic package P is 3. X : Integer; 4.end; 5. 6.package P1 is new P; 7.package P2 is new P; 8.package P4 is new P; 9.package P5 is new P; 10.package P6 is new P; 11.package P7 is new P; 12.package P8 is new P; 13.package P9 is new P; 14. begin 15.null; 16. end; If compiled with -gnatei4 -gnatj64, yields maxinst.adb:12:04: too many instantiations, exceeds max of 4, limit can be changed using -gnateinn switch compilation abandoned Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Robert Dewar de...@adacore.com * errout.ads: Add entry for translating -gnateinn to /MAX_INSTANTIATIONS for VMS. * hostparm.ads (Max_Instantiations): Moved to Opt. * opt.ads (Maximum_Instantiations): Moved from Hostparm, and renamed. * sem_ch12.adb (Maximum_Instantiations): New name of Max_Instantiations (Analyze_Package_Instantiation): Change error msg for too many instantiations (mention -gnateinn switch). * switch-c.adb (Scan_Front_End_Switches): Implement -gnateinn switch. * switch.ads: Minor comment update. * usage.adb (Usage): Output line for -maxeinn switch. * vms_data.ads: Add entry for MAX_INSTANTIATIONS (-gnateinn). Index: switch-c.adb === --- switch-c.adb(revision 185390) +++ switch-c.adb(working copy) @@ -6,7 +6,7 @@ -- -- -- B o d y -- -- -- --- Copyright (C) 2001-2011, Free Software Foundation, Inc. -- +-- Copyright (C) 2001-2012, Free Software Foundation, Inc. -- -- -- -- GNAT is free software; you can redistribute it and/or modify it under -- -- terms of the GNU General Public License as published by the Free Soft- -- @@ -482,6 +482,13 @@ Generate_Processed_File := True; Ptr := Ptr + 1; + -- -gnatei (max number of instantiations) + + when 'i' = + Ptr := Ptr + 1; + Scan_Pos + (Switch_Chars, Max, Ptr, Maximum_Instantiations, C); + -- -gnateI (index of unit in multi-unit source) when 'I' = Index: usage.adb === --- usage.adb (revision 185390) +++ usage.adb (working copy) @@ -6,7 +6,7 @@ -- -- --B o d y -- -- -- --- Copyright (C) 1992-2011, Free Software Foundation, Inc. -- +-- Copyright (C) 1992-2012, Free Software Foundation, Inc. -- -- -- -- GNAT is free software; you can redistribute it and/or modify it under -- -- terms of the GNU General Public License as published by the Free Soft- -- @@ -197,6 +197,11 @@ Write_Switch_Char (eG); Write_Line (Generate preprocessed source); + -- Line for -gnatei switch + + Write_Switch_Char (einn); + Write_Line (Set maximumum number of instantiations to nn); + -- Line for -gnateI switch Write_Switch_Char (eInn); Index: sem_ch12.adb === --- sem_ch12.adb(revision 185417) +++ sem_ch12.adb(working copy) @@ -34,7 +34,6 @@ with Fname;use Fname; with Fname.UF; use Fname.UF; with Freeze; use Freeze; -with Hostparm; with Itypes; use Itypes; with Lib; use Lib; with Lib.Load; use Lib.Load; @@ -3784,8 +3783,10 @@ -- Here is a defence against a ludicrous number of instantiations -- caused by a circular set of instantiation attempts. -if Pending_Instantiations.Last Hostparm.Max_Instantiations then - Error_Msg_N (too many instantiations, N); +if Pending_Instantiations.Last Maximum_Instantiations then + Error_Msg_Uint_1 := UI_From_Int (Maximum_Instantiations); + Error_Msg_N (too many instantiations, exceeds max of^, N); +
[Ada] Add unit file name for subunits in Alfa section of ALI file
The unit file name is needed when processing Alfa references from subunits,
in the formal verification backend of GNAT. Thus, add the unit file name
information for subunits in the line of the Alfa section that gives the
subunit file name.
Tested on x86_64-pc-linux-gnu, committed on trunk
2012-03-15 Yannick Moy m...@adacore.com
* alfa.ads Update the decription of ALI sections.
(Alfa_File_Record): Add a component Unit_File_Name to store the
unit file name for subunits.
* get_alfa.adb, put_alfa.adb Adapt to the possible presence of
a unit file name.
* lib-xref-alfa.adb (Add_Alfa_File): For subunits, retrieve the
file name of the unit.
Index: alfa.ads
===
--- alfa.ads(revision 185390)
+++ alfa.ads(working copy)
@@ -6,7 +6,7 @@
-- --
-- S p e c --
-- --
--- Copyright (C) 2011, Free Software Foundation, Inc. --
+-- Copyright (C) 2011-2012, Free Software Foundation, Inc. --
-- --
-- GNAT is free software; you can redistribute it and/or modify it under --
-- terms of the GNU General Public License as published by the Free Soft- --
@@ -70,7 +70,7 @@
-- subprogram declaration and body, when both present, define two different
-- scopes.
- --FD dependency-number filename
+ --FD dependency-number filename (- unit-filename)?
-- This header precedes scope information for the unit identified by
-- dependency number and file name. The dependency number is the index
@@ -89,6 +89,8 @@
-- reading of the Alfa information, and means that the Alfa information
-- can stand on its own without needing other parts of the ALI file.
+ -- The optional unit filename is given only for subunits.
+
--FS . scope line type col entity (- spec-file . spec-scope)?
-- (The ? mark stands for an optional entry in the syntax)
@@ -314,6 +316,10 @@
File_Name : String_Ptr;
-- Pointer to file name in ALI file
+ Unit_File_Name : String_Ptr;
+ -- Pointer to file name for unit in ALI file, when File_Name refers to a
+ -- subunit. Otherwise null.
+
File_Num : Nat;
-- Dependency number in ALI file
Index: put_alfa.adb
===
--- put_alfa.adb(revision 185390)
+++ put_alfa.adb(working copy)
@@ -6,7 +6,7 @@
-- --
-- B o d y --
-- --
--- Copyright (C) 2011, Free Software Foundation, Inc. --
+-- Copyright (C) 2011-2012, Free Software Foundation, Inc. --
-- --
-- GNAT is free software; you can redistribute it and/or modify it under --
-- terms of the GNU General Public License as published by the Free Soft- --
@@ -49,6 +49,18 @@
Write_Info_Char (F.File_Name (N));
end loop;
+ -- If file is a subunit, print the file name for the unit
+
+ if F.Unit_File_Name /= null then
+Write_Info_Char (' ');
+Write_Info_Char ('-');
+Write_Info_Char ('');
+Write_Info_Char (' ');
+for N in F.Unit_File_Name'Range loop
+ Write_Info_Char (F.Unit_File_Name (N));
+end loop;
+ end if;
+
Write_Info_Terminate;
-- Loop through scope entries for this file
Index: lib-xref-alfa.adb
===
--- lib-xref-alfa.adb (revision 185390)
+++ lib-xref-alfa.adb (working copy)
@@ -6,7 +6,7 @@
-- --
-- B o d y --
-- --
--- Copyright (C) 2011, Free Software Foundation, Inc. --
+-- Copyright (C) 2011-2012, Free Software Foundation, Inc. --
-- --
-- GNAT is free software; you can redistribute it and/or modify it under --
-- terms of the GNU General Public License as published by the Free Soft- --
@@ -214,6 +214,8 @@
S : constant Source_File_Index := Source_Index (U);
+ File_Name, Unit_File_Name : String_Ptr;
+
begin
-- Source file
[Ada] Remove redundant warning on function postcondition not mentioning 'Result
When all function postconditions and contract-cases get a warning for only referring to pre-state, there is no need to issue another warning for not mentioning 'Result. This is in particular the case when there is a single postcondition. Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Yannick Moy m...@adacore.com * sem_ch6.adb (Check_Subprogram_Contract): Do not issue warning on missing 'Result in postcondition if all postconditions and contract-cases already get a warning for only referring to pre-state. Index: sem_ch6.adb === --- sem_ch6.adb (revision 185421) +++ sem_ch6.adb (working copy) @@ -6937,6 +6937,10 @@ Attribute_Result_Mentioned : Boolean := False; -- Whether attribute 'Result is mentioned in a postcondition + No_Warning_On_Some_Postcondition : Boolean := False; + -- Whether there exists a postcondition or a contract-case without a + -- corresponding warning. + Post_State_Mentioned : Boolean := False; -- Whether some expression mentioned in a postcondition can have a -- different value in the post-state than in the pre-state. @@ -7081,7 +7085,9 @@ Post_State_Mentioned := False; Ignored := Find_Post_State (Arg); - if not Post_State_Mentioned then + if Post_State_Mentioned then + No_Warning_On_Some_Postcondition := True; + else Error_Msg_N (?`Ensures` component refers only to pre-state, Prag); end if; @@ -7133,7 +7139,9 @@ Post_State_Mentioned := False; Ignored := Find_Post_State (Arg); - if not Post_State_Mentioned then + if Post_State_Mentioned then + No_Warning_On_Some_Postcondition := True; + else Error_Msg_N (?postcondition refers only to pre-state, Prag); end if; @@ -7177,12 +7185,15 @@ end if; -- Issue warning for functions whose postcondition does not mention - -- 'Result after all postconditions have been processed. + -- 'Result after all postconditions have been processed, and provided + -- all postconditions do not already get a warning that they only refer + -- to pre-state. if Ekind_In (Spec_Id, E_Function, E_Generic_Function) and then (Present (Last_Postcondition) or else Present (Last_Contract_Case)) and then not Attribute_Result_Mentioned +and then No_Warning_On_Some_Postcondition then if Present (Last_Postcondition) then if Present (Last_Contract_Case) then
[Ada] Indirect calls in static elaboration model
This patch makes the static elaboration model more conservative in the case of indirect calls, by treating Subp'Access as a call for elaboration purposes. The following test should print 3, even when compiled with the binder switch -p, which enables pessimistic (worst-case) elaboration order. gnatmake -f a4 -bargs -p Expected output: warning: use of -p switch questionable warning: since all units compiled with static elaboration model 3 package a1 is function f return Integer; end a1; with a2; package body a1 is function f return integer is begin return a2.f; end; end a1; package a2 is function f return Integer; end a2; package body a2 is function Ident (X : Integer) return Integer is begin return X; end; Var : Integer := Ident (3); function f return Integer is begin return Var; end f; end a2; with a1; package a3 is type P is access function return Integer; PP : P := a1.f'Access; R : Integer := PP.all; end a3; with a3; with Text_IO; use Text_IO; procedure a4 is begin Put_Line (a3.R'Img); end; Tested on x86_64-pc-linux-gnu, committed on trunk 2012-03-15 Bob Duff d...@adacore.com * debug.adb: Add new debug switch -gnatd.U, which disables the support added below, in case someone trips over a cycle, and needs to disable this. * sem_attr.adb (Analyze_Access_Attribute): Treat Subp'Access as a call for elaboration purposes. * sem_elab.ads, sem_elab.adb (Check_Elab_Call): Add support for Subp'Access. Index: debug.adb === --- debug.adb (revision 185390) +++ debug.adb (working copy) @@ -138,7 +138,7 @@ -- d.R -- d.S Force Optimize_Alignment (Space) -- d.T Force Optimize_Alignment (Time) - -- d.U + -- d.U Ignore indirect calls for static elaboration -- d.V -- d.W Print out debugging information for Walk_Library_Items -- d.X Use Expression_With_Actions @@ -642,6 +642,12 @@ -- d.T Force Optimize_Alignment (Time) mode as the default + -- d.U Ignore indirect calls for static elaboration. The static + -- elaboration model is conservative, especially regarding indirect + -- calls. If you say Proc'Access, it will assume you might call + -- Proc. This can cause elaboration cycles at bind time. This flag + -- reverts to the behavior of earlier compilers. + -- d.W Print out debugging information for Walk_Library_Items, including -- the order in which units are walked. This is primarily for use in -- debugging CodePeer mode. Index: sem_attr.adb === --- sem_attr.adb(revision 185420) +++ sem_attr.adb(working copy) @@ -28,6 +28,7 @@ with Atree;use Atree; with Casing; use Casing; with Checks; use Checks; +with Debug;use Debug; with Einfo;use Einfo; with Errout; use Errout; with Eval_Fat; @@ -54,6 +55,7 @@ with Sem_Ch10; use Sem_Ch10; with Sem_Dim; use Sem_Dim; with Sem_Dist; use Sem_Dist; +with Sem_Elab; use Sem_Elab; with Sem_Elim; use Sem_Elim; with Sem_Eval; use Sem_Eval; with Sem_Res; use Sem_Res; @@ -644,6 +646,13 @@ Kill_Current_Values; end if; +-- Treat as call for elaboration purposes and we are all +-- done. Suppress this treatment under debug flag. + +if not Debug_Flag_Dot_UU then + Check_Elab_Call (N); +end if; + return; -- Component is an operation of a protected type Index: sem_elab.adb === --- sem_elab.adb(revision 185390) +++ sem_elab.adb(working copy) @@ -6,7 +6,7 @@ -- -- -- B o d y -- -- -- --- Copyright (C) 1997-2011, Free Software Foundation, Inc. -- +-- Copyright (C) 1997-2012, Free Software Foundation, Inc. -- -- -- -- GNAT is free software; you can redistribute it and/or modify it under -- -- terms of the GNU General Public License as published by the Free Soft- -- @@ -180,7 +180,7 @@ Inter_Unit_Only : Boolean; Generate_Warnings : Boolean := True; In_Init_Proc : Boolean := False); - -- This is the internal recursive routine that is called to check for a + -- This is the internal recursive routine that is called to check for -- possible elaboration error. The argument N is a subprogram call or -- generic instantiation to be checked, and E is the entity of the called -- subprogram, or instantiated generic unit. The
Re: [Patch/cfgexpand]: also consider assembler_name to call expand_main_function
On Wed, 14 Mar 2012, Tristan Gingold wrote: On Mar 14, 2012, at 5:08 PM, Richard Guenther wrote: On Wed, 14 Mar 2012, Tristan Gingold wrote: Hi, the code to call expand_main_function currently only checks DECL_NAME. This leads to a hack in ada/gcc-interface/utils.c to handle the gnatbind generated file that could declare: package ada_main is … function my_main (argc : Integer; argv : System.Address; envp : System.Address) return Integer; pragma Export (C, my_main, main); … end ada_main; But expand_main_function is also called for function whose name is main but assembly name isn't. Eg: package pkg is procedure main; end pkg; So I think we should consider the assembler name is set, otherwise the decl name. Manually tested on ia64-hp-openvms (where this issue was discovered). No C regressions for x86_64-darwin. Ok for trunk ? There are more checks for MAIN_NAME_P, so this certainly isn't enough. And if it is a good idea then the whole check, whether a FUNCTION_DECL is considered 'main' should be put into a function in tree.[ch] and used everywhere. Note that what is 'main' is controlled by main_identifier_node, controlled by frontends. So - why is that not enough to control for Ada? Indeed, I think we could handle this issue in gigi for Ada. (I also think we don't want to handle crazy C code such as 'int my_main () asm (main)'. But, unless I missed something, doing this in gigi won't work with LTO. Well. To make this work in LTO the main function (thus, the program entry point) should be marked at cgraph level and all users of MAIN_NAME_P should instead check a flag on the cgraph node. Will write a predicate in tree.[ch]. Please instead transition main-ness to the cgraph. Thanks, Richard.
[Patch, Fortran] PR52585 - Fix ASSOCIATE with proc-pointer dummies
A rather obvious patch: With proc-pointer dummies, one compared the
address of the pointer and not of the pointer target.
Build and regtested on x86-64-linux.
OK for the trunk? (What's the sentiment regarding backporting to 4.7.1?)
Tobias
PS: The patch looks larger than it is: I converted some spaces into tabs.
2012-03-15 Tobias Burnus bur...@net-b.de
PR fortran/52585
* trans-intrinsic.c (gfc_conv_associated): Fix handling of
procpointer dummy arguments.
2012-03-15 Tobias Burnus bur...@net-b.de
PR fortran/52585
* gfortran.dg/proc_ptr_36.f90: New.
diff --git a/gcc/fortran/trans-intrinsic.c b/gcc/fortran/trans-intrinsic.c
index ac9f507..2ec97c2 100644
--- a/gcc/fortran/trans-intrinsic.c
+++ b/gcc/fortran/trans-intrinsic.c
@@ -5761,10 +5787,14 @@ gfc_conv_associated (gfc_se *se, gfc_expr *expr)
/* No optional target. */
if (ss1 == gfc_ss_terminator)
{
- /* A pointer to a scalar. */
- arg1se.want_pointer = 1;
- gfc_conv_expr (arg1se, arg1-expr);
- tmp2 = arg1se.expr;
+ /* A pointer to a scalar. */
+ arg1se.want_pointer = 1;
+ gfc_conv_expr (arg1se, arg1-expr);
+ if (arg1-expr-symtree-n.sym-attr.proc_pointer
+ arg1-expr-symtree-n.sym-attr.dummy)
+ arg1se.expr = build_fold_indirect_ref_loc (input_location,
+ arg1se.expr);
+ tmp2 = arg1se.expr;
}
else
{
@@ -5794,12 +5824,21 @@ gfc_conv_associated (gfc_se *se, gfc_expr *expr)
if (ss1 == gfc_ss_terminator)
{
- /* A pointer to a scalar. */
- gcc_assert (ss2 == gfc_ss_terminator);
- arg1se.want_pointer = 1;
- gfc_conv_expr (arg1se, arg1-expr);
- arg2se.want_pointer = 1;
- gfc_conv_expr (arg2se, arg2-expr);
+ /* A pointer to a scalar. */
+ gcc_assert (ss2 == gfc_ss_terminator);
+ arg1se.want_pointer = 1;
+ gfc_conv_expr (arg1se, arg1-expr);
+ if (arg1-expr-symtree-n.sym-attr.proc_pointer
+ arg1-expr-symtree-n.sym-attr.dummy)
+ arg1se.expr = build_fold_indirect_ref_loc (input_location,
+ arg1se.expr);
+
+ arg2se.want_pointer = 1;
+ gfc_conv_expr (arg2se, arg2-expr);
+ if (arg2-expr-symtree-n.sym-attr.proc_pointer
+ arg2-expr-symtree-n.sym-attr.dummy)
+ arg2se.expr = build_fold_indirect_ref_loc (input_location,
+ arg2se.expr);
gfc_add_block_to_block (se-pre, arg1se.pre);
gfc_add_block_to_block (se-post, arg1se.post);
tmp = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node,
--- /dev/null 2012-03-15 07:05:00.651809558 +0100
+++ /home/tob/projects/gcc-git/gcc/gcc/testsuite/gfortran.dg/proc_ptr_36.f90 2012-03-15 11:34:46.0 +0100
@@ -0,0 +1,48 @@
+! { dg-do run }
+!
+! PR fortran/52585
+!
+! Test proc-pointer dummies with ASSOCIATE
+!
+! Contributed by Mat Cross of NAG
+!
+module m0
+ abstract interface
+subroutine sub
+end subroutine sub
+ end interface
+ interface
+subroutine s(ss, isassoc)
+ import sub
+ logical :: isassoc
+ procedure(sub), pointer, intent(in) :: ss
+end subroutine s
+ end interface
+end module m0
+
+use m0, only : sub, s
+procedure(sub) :: sub2, pp
+pointer :: pp
+pp = sub2
+if (.not. associated(pp)) call abort ()
+if (.not. associated(pp,sub2)) call abort ()
+call s(pp, .true.)
+pp = null()
+if (associated(pp)) call abort ()
+if (associated(pp,sub2)) call abort ()
+call s(pp, .false.)
+end
+
+subroutine s(ss, isassoc)
+ use m0, only : sub
+ logical :: isassoc
+ procedure(sub), pointer, intent(in) :: ss
+ procedure(sub) :: sub2
+ if (isassoc .neqv. associated(ss)) call abort ()
+ if (isassoc .neqv. associated(ss,sub2)) call abort ()
+end subroutine s
+
+subroutine sub2
+end subroutine sub2
+
+! { dg-final { cleanup-modules m0 } }
Re: [PATCH] gfortran testsuite: implicitly cleanup-modules
On Tue, Mar 13, 2012 at 01:30:29PM -0700, Mike Stump wrote:
On Mar 13, 2012, at 9:38 AM, Bernhard Reutner-Fischer wrote:
Could some of the testsuite maintainers please eyeball?
I've eyed it, the only thing that stood out was:
-foreach testcase [lsort [glob -nocomplain $srcdir/$subdir/*.F]] {
-if ![runtest_file_p $runtests $testcase] then {
- continue
-}
-fortran-torture $testcase
-}
-
-foreach testcase [lsort [glob -nocomplain $srcdir/$subdir/*.f90]] {
-if ![runtest_file_p $runtests $testcase] then {
- continue
-}
-fortran-torture $testcase
-}
which, I hope does what you want.
It replaces a manually unrolled loop with a loop over the
testcase-extensions and works as expected. Or did you mean something
else?
Assuming you like those parts, Ok.
committed as r185430.
Since it touches non-fortran, please watch for and respond to any problems it
might cause.
Of course, as always.
PS: I saw that the cleanup-tree-dump is also a bit redundant.
When looking at e.g. gcc/testsuite/gcc.dg/pr52578.c one would think that
scan-tree-dump-times might be able to automagically collect which
dumpfiles to cleanup as last step in dg-final.
One could go one step further and even implicitly add the corresponding
dg-options according to the expected scan-tree-dump files although this
might not be worth it iff there are passes that may produce dump-files
different to their pass name (did not look if this is possible). One
would have to strip the pass-number off the dump-file for passes that
are run more than once, too. Just a thought..
Thanks and cheers,
Re: [Patch/cfgexpand]: also consider assembler_name to call expand_main_function
On Mar 15, 2012, at 10:37 AM, Richard Guenther wrote: On Wed, 14 Mar 2012, Tristan Gingold wrote: On Mar 14, 2012, at 5:08 PM, Richard Guenther wrote: On Wed, 14 Mar 2012, Tristan Gingold wrote: Hi, the code to call expand_main_function currently only checks DECL_NAME. This leads to a hack in ada/gcc-interface/utils.c to handle the gnatbind generated file that could declare: package ada_main is … function my_main (argc : Integer; argv : System.Address; envp : System.Address) return Integer; pragma Export (C, my_main, main); … end ada_main; But expand_main_function is also called for function whose name is main but assembly name isn't. Eg: package pkg is procedure main; end pkg; So I think we should consider the assembler name is set, otherwise the decl name. Manually tested on ia64-hp-openvms (where this issue was discovered). No C regressions for x86_64-darwin. Ok for trunk ? There are more checks for MAIN_NAME_P, so this certainly isn't enough. And if it is a good idea then the whole check, whether a FUNCTION_DECL is considered 'main' should be put into a function in tree.[ch] and used everywhere. Note that what is 'main' is controlled by main_identifier_node, controlled by frontends. So - why is that not enough to control for Ada? Indeed, I think we could handle this issue in gigi for Ada. (I also think we don't want to handle crazy C code such as 'int my_main () asm (main)'. But, unless I missed something, doing this in gigi won't work with LTO. Well. To make this work in LTO the main function (thus, the program entry point) should be marked at cgraph level and all users of MAIN_NAME_P should instead check a flag on the cgraph node. Will write a predicate in tree.[ch]. Please instead transition main-ness to the graph. Ok, I will explore this way. Tristan.
Re: [PATCH] Change VECTOR_CST representation from TREE_LIST to TREE_VEC-like
On Thu, 15 Mar 2012, Richard Guenther wrote:
This removes the use of TREE_LISTs for VECTOR_CSTs and instead employs
a similar way of storing elements as TREE_VECs. I copied the
macro interface bits of the CONSTRUCTOR accesses and did a 1:1 transform
at most places to not let refactoring errors creep in (well, where
possible). I'm not sure if it's worth omitting the explicit length
field in favor of using that of the type - at least we are getting
consistency (no implicit zero elements) here for free.
Bootstrap and regtest on x86_64-unknown-linux-gnu running. I cannot
test sparc apart from compiling a cc1, which works.
Ok for the c-common and sparc bits? Any other comments?
Bootstrapped ok on x86_64-unknown-linux-gnu. Incremental patch to
fix issues revealed by testing below.
Richard.
Index: gcc/tree.c
===
--- gcc/tree.c.orig 2012-03-15 14:02:31.0 +0100
+++ gcc/tree.c 2012-03-15 13:51:27.0 +0100
@@ -1328,6 +1328,7 @@ build_vector_stat (tree type, tree *vals
v = ggc_alloc_zone_cleared_tree_node_stat (tree_zone, length PASS_MEM_STAT);
TREE_SET_CODE (v, VECTOR_CST);
+ TREE_CONSTANT (v) = 1;
TREE_TYPE (v) = type;
/* Iterate through elements and check for overflow. */
Index: gcc/varasm.c
===
--- gcc/varasm.c.orig 2012-03-15 14:02:31.0 +0100
+++ gcc/varasm.c2012-03-15 13:59:37.0 +0100
@@ -4596,8 +4596,8 @@ output_constant (tree exp, unsigned HOST
elt_size = GET_MODE_SIZE (inner);
- thissize = 0;
output_constant (VECTOR_CST_ELT (exp, 0), elt_size, align);
+ thissize = elt_size;
for (i = 1; i VECTOR_CST_NELTS (exp); ++i)
{
output_constant (VECTOR_CST_ELT (exp, i), elt_size, nalign);
[PATCH] Fix __builtin_ir{ound,int}{,f,l} expansion (PR middle-end/52592)
Hi!
If __builtin_ir{int,ound}{,f,l} expansion through optab fails,
we would end up generating a call to __builtin_ir{int,ound}{,f,l}
function (there is no ir{int,ound}{,f,l} in libm), which is wrong,
we should expand it as (int) lr{int,ound}{,f,l} in that case (similarly
to what we already do with __builtin_{i,l,ll}{ceil,floor}{,f,l}).
This has been reported as a failure to build mplayer on x86_64-linux
as well as a failure to build pulseaudio on s390x-linux.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux,
ok for trunk and 4.7.0?
2012-03-15 Jakub Jelinek ja...@redhat.com
Andrew Pinski apin...@cavium.com
PR middle-end/52592
* builtins.c (expand_builtin_int_roundingfn_2): If expanding
BUILT_IN_IR{INT,OUND}* using optab fails, emit lr{int,ound}*
calls instead of __builtin_ir{int,ound}*.
* gcc.dg/pr52592.c: New test.
--- gcc/builtins.c.jj 2012-02-12 15:46:32.0 +0100
+++ gcc/builtins.c 2012-03-15 10:20:59.335215596 +0100
@@ -2841,10 +2841,7 @@ expand_builtin_int_roundingfn_2 (tree ex
tree fndecl = get_callee_fndecl (exp);
tree arg;
enum machine_mode mode;
-
- /* There's no easy way to detect the case we need to set EDOM. */
- if (flag_errno_math)
-return NULL_RTX;
+ enum built_in_function fallback_fn = BUILT_IN_NONE;
if (!validate_arglist (exp, REAL_TYPE, VOID_TYPE))
gcc_unreachable ();
@@ -2854,47 +2851,79 @@ expand_builtin_int_roundingfn_2 (tree ex
switch (DECL_FUNCTION_CODE (fndecl))
{
CASE_FLT_FN (BUILT_IN_IRINT):
+ fallback_fn = BUILT_IN_LRINT;
+ /* FALLTHRU */
CASE_FLT_FN (BUILT_IN_LRINT):
CASE_FLT_FN (BUILT_IN_LLRINT):
- builtin_optab = lrint_optab; break;
+ builtin_optab = lrint_optab;
+ break;
CASE_FLT_FN (BUILT_IN_IROUND):
+ fallback_fn = BUILT_IN_LROUND;
+ /* FALLTHRU */
CASE_FLT_FN (BUILT_IN_LROUND):
CASE_FLT_FN (BUILT_IN_LLROUND):
- builtin_optab = lround_optab; break;
+ builtin_optab = lround_optab;
+ break;
default:
gcc_unreachable ();
}
+ /* There's no easy way to detect the case we need to set EDOM. */
+ if (flag_errno_math fallback_fn == BUILT_IN_NONE)
+return NULL_RTX;
+
/* Make a suitable register to place result in. */
mode = TYPE_MODE (TREE_TYPE (exp));
- target = gen_reg_rtx (mode);
+ /* There's no easy way to detect the case we need to set EDOM. */
+ if (!flag_errno_math)
+{
+ target = gen_reg_rtx (mode);
- /* Wrap the computation of the argument in a SAVE_EXPR, as we may
- need to expand the argument again. This way, we will not perform
- side-effects more the once. */
- CALL_EXPR_ARG (exp, 0) = arg = builtin_save_expr (arg);
+ /* Wrap the computation of the argument in a SAVE_EXPR, as we may
+need to expand the argument again. This way, we will not perform
+side-effects more the once. */
+ CALL_EXPR_ARG (exp, 0) = arg = builtin_save_expr (arg);
+
+ op0 = expand_expr (arg, NULL, VOIDmode, EXPAND_NORMAL);
+
+ start_sequence ();
+
+ if (expand_sfix_optab (target, op0, builtin_optab))
+ {
+ /* Output the entire sequence. */
+ insns = get_insns ();
+ end_sequence ();
+ emit_insn (insns);
+ return target;
+ }
+
+ /* If we were unable to expand via the builtin, stop the sequence
+(without outputting the insns) and call to the library function
+with the stabilized argument list. */
+ end_sequence ();
+}
- op0 = expand_expr (arg, NULL, VOIDmode, EXPAND_NORMAL);
+ if (fallback_fn != BUILT_IN_NONE)
+{
+ /* Fall back to rounding to long int. Use implicit_p 0 - for non-C99
+targets, (int) round (x) should never be transformed into
+BUILT_IN_IROUND and if __builtin_iround is called directly, emit
+a call to lround in the hope that the target provides at least some
+C99 functions. This should result in the best user experience for
+not full C99 targets. */
+ tree fallback_fndecl = mathfn_built_in_1 (TREE_TYPE (arg),
+ fallback_fn, 0);
- start_sequence ();
+ exp = build_call_nofold_loc (EXPR_LOCATION (exp),
+ fallback_fndecl, 1, arg);
- if (expand_sfix_optab (target, op0, builtin_optab))
-{
- /* Output the entire sequence. */
- insns = get_insns ();
- end_sequence ();
- emit_insn (insns);
- return target;
+ target = expand_call (exp, NULL_RTX, target == const0_rtx);
+ return convert_to_mode (mode, target, 0);
}
- /* If we were unable to expand via the builtin, stop the sequence
- (without outputting the insns) and call to the library function
- with the stabilized argument list. */
- end_sequence ();
-
target = expand_call (exp, target, target == const0_rtx);
return target;
---
[patch tree-optimization]: Fix for PR 45397 part 2 of 2
Hi,
this is the second part of the patch for this problem. It adds some
basic simplifications for ==/!=
comparisons for eliminating redudant operands.
It adds the following patterns:
-X ==/!= Z - X - Z ==/!= 0.
~X ==/!= Z ^ X - Z ==/!= ~0
X ==/!= X - Y - Y == 0
X ==/!= X + Y - Y == 0
X ==/!= X ^ Y - Y == 0
(X - Y) ==/!= (Z - Y) - X ==/!= Z
(Y - X) ==/!= (Y - Z) - X ==/!= Z
(X + Y) ==/!= (X + Z) - Y ==/!= Z
(X + Y) ==/!= (Z + X) - Y ==/!= Z
(X ^ Y) ==/!= (Z ^ X) - Y ==/!= Z
ChangeLog
2012-03-15 Kai Tietz kti...@redhat.com
PR tree-optimization/45397
* tree-ssa-forwprop.c (compare_equal_optimized_1): Add
simplification patterns for ==/!= comparison.
2012-03-15 Kai Tietz kti...@redhat.com
* gcc.dg/tree-ssa/pr45397-2.c: New test.
Regression tested for all languages (including Ada and Obj-C) on
x86_64-unknown-linux-gnu. Ok for apply?
Regards,
Kai
Index: gcc-trunk/gcc/tree-ssa-forwprop.c
===
--- gcc-trunk.orig/gcc/tree-ssa-forwprop.c
+++ gcc-trunk/gcc/tree-ssa-forwprop.c
@@ -381,6 +381,99 @@ compare_equal_optimize_1 (gimple stmt, e
|| !INTEGRAL_TYPE_P (type_outer))
return NULL_TREE;
+ /* Simplify -X ==/!= Z - X - Z ==/!= 0. */
+ if (TREE_CODE (op0) == NEGATE_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0))
+ TREE_CODE (op1) == MINUS_EXPR
+ TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op1, 0),
+ build_zero_cst (TREE_TYPE (op1)));
+
+ /* Simplify X - Z ==/!= -X - Z ==/!= 0. */
+ if (TREE_CODE (op1) == NEGATE_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op1, 0))
+ TREE_CODE (op0) == MINUS_EXPR
+ TREE_OPERAND (op1, 0) == TREE_OPERAND (op0, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op0, 0),
+ build_zero_cst (TREE_TYPE (op0)));
+
+ /* Simplify ~X ==/!= X ^ Y to Y ==/!= ~0. */
+ if (TREE_CODE (op0) == BIT_NOT_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0))
+ TREE_CODE (op1) == BIT_XOR_EXPR)
+{
+ if (TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op1, 0),
+ fold_build1 (BIT_NOT_EXPR,
+TREE_TYPE (op1),
+build_zero_cst (TREE_TYPE (op1;
+ if (TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 0))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op1, 1),
+ fold_build1 (BIT_NOT_EXPR,
+TREE_TYPE (op1),
+build_zero_cst (TREE_TYPE (op1;
+}
+
+ /* Simplify X ^ Y ==/!= ~X to Y ==/!= ~0. */
+ if (TREE_CODE (op1) == BIT_NOT_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op1, 0))
+ TREE_CODE (op0) == BIT_XOR_EXPR)
+{
+ if (TREE_OPERAND (op1, 0) == TREE_OPERAND (op0, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op0, 0),
+ fold_build1 (BIT_NOT_EXPR,
+TREE_TYPE (op0),
+build_zero_cst (TREE_TYPE (op0;
+ if (TREE_OPERAND (op1, 0) == TREE_OPERAND (op0, 0))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op0, 1),
+ fold_build1 (BIT_NOT_EXPR,
+TREE_TYPE (op0),
+build_zero_cst (TREE_TYPE (op0;
+}
+
+ /* For code being +, -, or ^-expression simplify (X code Y) ==/!= (Z code Y)
+ to (X ==/!= Z), and (X code Y) ==/!= (X code Z) to (Y ==/!= Z). */
+ if (TREE_CODE (op0) == TREE_CODE (op1)
+ (TREE_CODE (op0) == PLUS_EXPR
+ || TREE_CODE (op0) == MINUS_EXPR
+ || TREE_CODE (op0) == BIT_XOR_EXPR))
+{
+ /* Simplify (X code Y) ==/!= (X code Z) to Y ==/!= Z. */
+ if (TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 0)
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0)))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op0, 1),
+ TREE_OPERAND (op1, 1));
+ /* Simplify (X code Y) ==/!= (Z code X) to Y ==/!= Z, if code isn't
+minus operation. */
+ if (TREE_CODE (op0) != MINUS_EXPR
+ TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 1)
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0)))
+return fold_build2_loc
Re: [PATCH] Fix __builtin_ir{ound,int}{,f,l} expansion (PR middle-end/52592)
On Thu, Mar 15, 2012 at 2:06 PM, Jakub Jelinek ja...@redhat.com wrote:
Hi!
If __builtin_ir{int,ound}{,f,l} expansion through optab fails,
we would end up generating a call to __builtin_ir{int,ound}{,f,l}
function (there is no ir{int,ound}{,f,l} in libm), which is wrong,
we should expand it as (int) lr{int,ound}{,f,l} in that case (similarly
to what we already do with __builtin_{i,l,ll}{ceil,floor}{,f,l}).
This has been reported as a failure to build mplayer on x86_64-linux
as well as a failure to build pulseaudio on s390x-linux.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux,
ok for trunk and 4.7.0?
Yes.
Thanks,
Richard.
2012-03-15 Jakub Jelinek ja...@redhat.com
Andrew Pinski apin...@cavium.com
PR middle-end/52592
* builtins.c (expand_builtin_int_roundingfn_2): If expanding
BUILT_IN_IR{INT,OUND}* using optab fails, emit lr{int,ound}*
calls instead of __builtin_ir{int,ound}*.
* gcc.dg/pr52592.c: New test.
--- gcc/builtins.c.jj 2012-02-12 15:46:32.0 +0100
+++ gcc/builtins.c 2012-03-15 10:20:59.335215596 +0100
@@ -2841,10 +2841,7 @@ expand_builtin_int_roundingfn_2 (tree ex
tree fndecl = get_callee_fndecl (exp);
tree arg;
enum machine_mode mode;
-
- /* There's no easy way to detect the case we need to set EDOM. */
- if (flag_errno_math)
- return NULL_RTX;
+ enum built_in_function fallback_fn = BUILT_IN_NONE;
if (!validate_arglist (exp, REAL_TYPE, VOID_TYPE))
gcc_unreachable ();
@@ -2854,47 +2851,79 @@ expand_builtin_int_roundingfn_2 (tree ex
switch (DECL_FUNCTION_CODE (fndecl))
{
CASE_FLT_FN (BUILT_IN_IRINT):
+ fallback_fn = BUILT_IN_LRINT;
+ /* FALLTHRU */
CASE_FLT_FN (BUILT_IN_LRINT):
CASE_FLT_FN (BUILT_IN_LLRINT):
- builtin_optab = lrint_optab; break;
+ builtin_optab = lrint_optab;
+ break;
CASE_FLT_FN (BUILT_IN_IROUND):
+ fallback_fn = BUILT_IN_LROUND;
+ /* FALLTHRU */
CASE_FLT_FN (BUILT_IN_LROUND):
CASE_FLT_FN (BUILT_IN_LLROUND):
- builtin_optab = lround_optab; break;
+ builtin_optab = lround_optab;
+ break;
default:
gcc_unreachable ();
}
+ /* There's no easy way to detect the case we need to set EDOM. */
+ if (flag_errno_math fallback_fn == BUILT_IN_NONE)
+ return NULL_RTX;
+
/* Make a suitable register to place result in. */
mode = TYPE_MODE (TREE_TYPE (exp));
- target = gen_reg_rtx (mode);
+ /* There's no easy way to detect the case we need to set EDOM. */
+ if (!flag_errno_math)
+ {
+ target = gen_reg_rtx (mode);
- /* Wrap the computation of the argument in a SAVE_EXPR, as we may
- need to expand the argument again. This way, we will not perform
- side-effects more the once. */
- CALL_EXPR_ARG (exp, 0) = arg = builtin_save_expr (arg);
+ /* Wrap the computation of the argument in a SAVE_EXPR, as we may
+ need to expand the argument again. This way, we will not perform
+ side-effects more the once. */
+ CALL_EXPR_ARG (exp, 0) = arg = builtin_save_expr (arg);
+
+ op0 = expand_expr (arg, NULL, VOIDmode, EXPAND_NORMAL);
+
+ start_sequence ();
+
+ if (expand_sfix_optab (target, op0, builtin_optab))
+ {
+ /* Output the entire sequence. */
+ insns = get_insns ();
+ end_sequence ();
+ emit_insn (insns);
+ return target;
+ }
+
+ /* If we were unable to expand via the builtin, stop the sequence
+ (without outputting the insns) and call to the library function
+ with the stabilized argument list. */
+ end_sequence ();
+ }
- op0 = expand_expr (arg, NULL, VOIDmode, EXPAND_NORMAL);
+ if (fallback_fn != BUILT_IN_NONE)
+ {
+ /* Fall back to rounding to long int. Use implicit_p 0 - for non-C99
+ targets, (int) round (x) should never be transformed into
+ BUILT_IN_IROUND and if __builtin_iround is called directly, emit
+ a call to lround in the hope that the target provides at least some
+ C99 functions. This should result in the best user experience for
+ not full C99 targets. */
+ tree fallback_fndecl = mathfn_built_in_1 (TREE_TYPE (arg),
+ fallback_fn, 0);
- start_sequence ();
+ exp = build_call_nofold_loc (EXPR_LOCATION (exp),
+ fallback_fndecl, 1, arg);
- if (expand_sfix_optab (target, op0, builtin_optab))
- {
- /* Output the entire sequence. */
- insns = get_insns ();
- end_sequence ();
- emit_insn (insns);
- return target;
+ target = expand_call (exp, NULL_RTX, target == const0_rtx);
+ return convert_to_mode (mode, target, 0);
}
- /* If we were unable to expand via the builtin, stop the sequence
- (without outputting the insns) and call to the
Re: [patch tree-optimization]: Fix for PR 45397 part 1 of 2
On Thu, Mar 15, 2012 at 2:08 PM, Kai Tietz ktiet...@googlemail.com wrote
Hi,
The solution for this PR is a mix out of different issues. First is
of course the type-hoisting, but also
it shows some lacks in simplifications on integer-values, and on equal
and none-equal
comparisons.
The first patch adds to forward-propagation the ability to do type-hoisting
for some conversion operations and do simplification for inner binary
operations on it.
Most important part is here the adjustment of constant integer-values
in statement-lists
for a truncation.
I limited that patch to handle in compare_equal_optimize_1 only
bitwise-and operations
inner a truncation-cast. Of course for bitwise-xor/or operations some
more simplifications
are possible.
This patch just does the type-hoisting part. In a second patch I add
to compare_equal_optimize_1
the ability for further required simplifications for fixing this problem.
This looks like to match unbound pattern sizes and thus does not fit
into the forwprop machinery. Instead it was suggested elsewhere
that promoting / demoting registers should be done in a separate pass
where you can compute a lattice of used bits and apply a transform
based on that lattice and target information (according to PROMOTE_MODE
for example).
Richard.
ChangeLog
2012-03-15 Kai Tietz kti...@redhat.com
PR tree-optimization/45397
* tree-ssa-forwprop.c (compare_equal_optimize_1): New
function.
(combine_cond_expr_cond): Use compare_equal_optimize_1
function.
(truncate_integers): New function.
(combine_inner_conversion): Likewise.
(ssa_forward_propagate_and_combine): Use for conversions
combine_inner_conversion function.
2012-03-15 Kai Tietz kti...@redhat.com
* gcc.dg/tree-ssa/pr45397-1.c: New testcase.
Regression tested for all languages (including Ada and Obj-C) on
x86_64-unknown-linux-gnu. Ok for apply?
Regards,
Kai
Index: gcc-trunk/gcc/tree-ssa-forwprop.c
===
--- gcc-trunk.orig/gcc/tree-ssa-forwprop.c
+++ gcc-trunk/gcc/tree-ssa-forwprop.c
@@ -362,6 +362,150 @@ rhs_to_tree (tree type, gimple stmt)
gcc_unreachable ();
}
+/* This function does simplifications of comparison OP0 ==/!= OP1
while integral
+ typed operands
+ On success new statement's TREE is returned, otherwise NULL_TREE. */
+
+static tree
+compare_equal_optimize_1 (gimple stmt, enum tree_code code, tree type,
+ tree op0, tree op1)
+{
+ gimple_stmt_iterator gsi;
+ tree type_outer;
+ tree type_inner, op_inner;
+ tree op1_l, op1_r, tem;
+ gimple newop;
+
+ type_outer = TREE_TYPE (op0);
+ if ((code != EQ_EXPR code != NE_EXPR)
+ || !INTEGRAL_TYPE_P (type_outer))
+ return NULL_TREE;
+
+ /* If OP0 isn't a conversion, then check if OP1 might be one. If so
+ swap arguments, otherwise return NULL_TREE. */
+ if (!CONVERT_EXPR_P (op0))
+ {
+ if (!CONVERT_EXPR_P (op1))
+ return NULL_TREE;
+ tem = op0;
+ op0 = op1;
+ op1 = tem;
+ }
+
+ op_inner = TREE_OPERAND (op0, 0);
+ type_inner = TREE_TYPE (op_inner);
+
+ /* Operations only apply to integral typed operands of cast. */
+ if (!INTEGRAL_TYPE_P (type_inner))
+ return NULL_TREE;
+
+ /* If second operand is also a type-conversion, check that underlying
operand
+ is integral typed. */
+ if (CONVERT_EXPR_P (op1)
+ !INTEGRAL_TYPE_P (TREE_TYPE (TREE_OPERAND (op1, 0
+ return NULL_TREE;
+
+ /* Simplify ((type) X ==/!= (type) X) to true/false, if X has no
side-effects
+ and is integral typed. */
+ if (CONVERT_EXPR_P (op1)
+ TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 0)
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0)))
+ return fold_convert (type, (code == EQ_EXPR ? boolean_true_node
+ : boolean_false_node));
+
+ /* Simplify (type) X ==/!= (type) Y to X ==/!= Y, if types of X and Y are
+ compatible and type-precision of X is smaller or equal to TYPE's. */
+ if (CONVERT_EXPR_P (op1)
+ TYPE_PRECISION (type_inner) = TYPE_PRECISION (type_outer))
+ {
+ tem = TREE_OPERAND (op1, 0);
+ if (!useless_type_conversion_p (type_inner, TREE_TYPE (tem)))
+ return NULL_TREE;
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ op_inner, tem);
+ }
+
+ /* Verify that for pattern 'OP0 = (type) X' the type of X is of
integral kind,
+ is unsigned, and has smaller or equal precision to type TYPE. */
+ if (TYPE_PRECISION (type_inner) TYPE_PRECISION (type_outer)
+ || !TYPE_UNSIGNED (type_inner))
+ return NULL_TREE;
+
+ /* Simplify (type) X ==/!= CST to X ==/!= CST' with CST'=(type-of-X) CST.
*/
+ if (TREE_CODE (op1) == INTEGER_CST)
+ {
+ tree new_cst = fold_convert (type_inner,
Re: [patch tree-optimization]: Fix for PR 45397 part 2 of 2
On Thu, Mar 15, 2012 at 2:09 PM, Kai Tietz ktiet...@googlemail.com wrote:
Hi,
this is the second part of the patch for this problem. It adds some
basic simplifications for ==/!=
comparisons for eliminating redudant operands.
It adds the following patterns:
-X ==/!= Z - X - Z ==/!= 0.
~X ==/!= Z ^ X - Z ==/!= ~0
X ==/!= X - Y - Y == 0
X ==/!= X + Y - Y == 0
X ==/!= X ^ Y - Y == 0
(X - Y) ==/!= (Z - Y) - X ==/!= Z
(Y - X) ==/!= (Y - Z) - X ==/!= Z
(X + Y) ==/!= (X + Z) - Y ==/!= Z
(X + Y) ==/!= (Z + X) - Y ==/!= Z
(X ^ Y) ==/!= (Z ^ X) - Y ==/!= Z
Can you re-base this patch to work without the previous one? Also
please coordinate with Andrew. Note that all of these(?) simplifications
are already done by fold_comparison which we could share if you'd split
out the EXPR_P op0/op1 cases with separated operands/code.
Richard.
ChangeLog
2012-03-15 Kai Tietz kti...@redhat.com
PR tree-optimization/45397
* tree-ssa-forwprop.c (compare_equal_optimized_1): Add
simplification patterns for ==/!= comparison.
2012-03-15 Kai Tietz kti...@redhat.com
* gcc.dg/tree-ssa/pr45397-2.c: New test.
Regression tested for all languages (including Ada and Obj-C) on
x86_64-unknown-linux-gnu. Ok for apply?
Regards,
Kai
Index: gcc-trunk/gcc/tree-ssa-forwprop.c
===
--- gcc-trunk.orig/gcc/tree-ssa-forwprop.c
+++ gcc-trunk/gcc/tree-ssa-forwprop.c
@@ -381,6 +381,99 @@ compare_equal_optimize_1 (gimple stmt, e
|| !INTEGRAL_TYPE_P (type_outer))
return NULL_TREE;
+ /* Simplify -X ==/!= Z - X - Z ==/!= 0. */
+ if (TREE_CODE (op0) == NEGATE_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0))
+ TREE_CODE (op1) == MINUS_EXPR
+ TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op1, 0),
+ build_zero_cst (TREE_TYPE (op1)));
+
+ /* Simplify X - Z ==/!= -X - Z ==/!= 0. */
+ if (TREE_CODE (op1) == NEGATE_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op1, 0))
+ TREE_CODE (op0) == MINUS_EXPR
+ TREE_OPERAND (op1, 0) == TREE_OPERAND (op0, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op0, 0),
+ build_zero_cst (TREE_TYPE (op0)));
+
+ /* Simplify ~X ==/!= X ^ Y to Y ==/!= ~0. */
+ if (TREE_CODE (op0) == BIT_NOT_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0))
+ TREE_CODE (op1) == BIT_XOR_EXPR)
+ {
+ if (TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op1, 0),
+ fold_build1 (BIT_NOT_EXPR,
+ TREE_TYPE (op1),
+ build_zero_cst (TREE_TYPE
(op1;
+ if (TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 0))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op1, 1),
+ fold_build1 (BIT_NOT_EXPR,
+ TREE_TYPE (op1),
+ build_zero_cst (TREE_TYPE
(op1;
+ }
+
+ /* Simplify X ^ Y ==/!= ~X to Y ==/!= ~0. */
+ if (TREE_CODE (op1) == BIT_NOT_EXPR
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op1, 0))
+ TREE_CODE (op0) == BIT_XOR_EXPR)
+ {
+ if (TREE_OPERAND (op1, 0) == TREE_OPERAND (op0, 1))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op0, 0),
+ fold_build1 (BIT_NOT_EXPR,
+ TREE_TYPE (op0),
+ build_zero_cst (TREE_TYPE
(op0;
+ if (TREE_OPERAND (op1, 0) == TREE_OPERAND (op0, 0))
+ return fold_build2_loc (gimple_location (stmt), code, type,
+ TREE_OPERAND (op0, 1),
+ fold_build1 (BIT_NOT_EXPR,
+ TREE_TYPE (op0),
+ build_zero_cst (TREE_TYPE
(op0;
+ }
+
+ /* For code being +, -, or ^-expression simplify (X code Y) ==/!= (Z code
Y)
+ to (X ==/!= Z), and (X code Y) ==/!= (X code Z) to (Y ==/!= Z). */
+ if (TREE_CODE (op0) == TREE_CODE (op1)
+ (TREE_CODE (op0) == PLUS_EXPR
+ || TREE_CODE (op0) == MINUS_EXPR
+ || TREE_CODE (op0) == BIT_XOR_EXPR))
+ {
+ /* Simplify (X code Y) ==/!= (X code Z) to Y ==/!= Z. */
+ if (TREE_OPERAND (op0, 0) == TREE_OPERAND (op1, 0)
+ !TREE_SIDE_EFFECTS (TREE_OPERAND (op0, 0)))
+ return
Re: [patch tree-optimization]: Fix for PR 45397 part 2 of 2
2012/3/15 Richard Guenther richard.guent...@gmail.com: On Thu, Mar 15, 2012 at 2:09 PM, Kai Tietz ktiet...@googlemail.com wrote: Hi, this is the second part of the patch for this problem. It adds some basic simplifications for ==/!= comparisons for eliminating redudant operands. It adds the following patterns: -X ==/!= Z - X - Z ==/!= 0. ~X ==/!= Z ^ X - Z ==/!= ~0 X ==/!= X - Y - Y == 0 X ==/!= X + Y - Y == 0 X ==/!= X ^ Y - Y == 0 (X - Y) ==/!= (Z - Y) - X ==/!= Z (Y - X) ==/!= (Y - Z) - X ==/!= Z (X + Y) ==/!= (X + Z) - Y ==/!= Z (X + Y) ==/!= (Z + X) - Y ==/!= Z (X ^ Y) ==/!= (Z ^ X) - Y ==/!= Z Can you re-base this patch to work without the previous one? Also please coordinate with Andrew. Note that all of these(?) simplifications are already done by fold_comparison which we could share if you'd split out the EXPR_P op0/op1 cases with separated operands/code. Richard. Hmm, fold_comparison doesn't do the same thing as it checks for possible overflow. This is true for comparisons not being ==/!= or having operands of none-integral-type. But for ==/!= with integral typed arguments the overflow doesn't matter at all. And exactly this is what patch implements here. This optimization of course is just desired in non-AST form, as we otherwise loose information in FE. Therefore I didn't added it to fold_const. I can rework the patch so that it works without the other one. Regards, Kai
Re: [patch tree-optimization]: Fix for PR 45397 part 1 of 2
2012/3/15 Richard Guenther richard.guent...@gmail.com: On Thu, Mar 15, 2012 at 2:08 PM, Kai Tietz ktiet...@googlemail.com wrote Hi, The solution for this PR is a mix out of different issues. First is of course the type-hoisting, but also it shows some lacks in simplifications on integer-values, and on equal and none-equal comparisons. The first patch adds to forward-propagation the ability to do type-hoisting for some conversion operations and do simplification for inner binary operations on it. Most important part is here the adjustment of constant integer-values in statement-lists for a truncation. I limited that patch to handle in compare_equal_optimize_1 only bitwise-and operations inner a truncation-cast. Of course for bitwise-xor/or operations some more simplifications are possible. This patch just does the type-hoisting part. In a second patch I add to compare_equal_optimize_1 the ability for further required simplifications for fixing this problem. This looks like to match unbound pattern sizes and thus does not fit into the forwprop machinery. Instead it was suggested elsewhere that promoting / demoting registers should be done in a separate pass where you can compute a lattice of used bits and apply a transform based on that lattice and target information (according to PROMOTE_MODE for example). Richard. Well, the integer truncation part might be something for a separate pass. It could then also take care that within single-use gimple-statements the integral-constant is always on right-hand-side of first statement of an +, -, |, ^, , and mul. But the cast-hoisting code itself is not unbound AFAICS and has fixed pattern size. Regards, Kai
Re: [patch tree-optimization]: Fix for PR 45397 part 1 of 2
On Thu, Mar 15, 2012 at 02:53:10PM +0100, Kai Tietz wrote: This looks like to match unbound pattern sizes and thus does not fit into the forwprop machinery. Instead it was suggested elsewhere that promoting / demoting registers should be done in a separate pass where you can compute a lattice of used bits and apply a transform based on that lattice and target information (according to PROMOTE_MODE for example). Well, the integer truncation part might be something for a separate pass. It could then also take care that within single-use gimple-statements the integral-constant is always on right-hand-side of first statement of an +, -, |, ^, , and mul. But the cast-hoisting code itself is not unbound AFAICS and has fixed pattern size. The type demotion is PR45397/PR47477 among other PRs. I'd just walk from the narrowing integer conversion stmts recursively through the def stmts, see if they can be narrowed, note it, and finally if everything or significant portion of the stmts can be demoted (if not all, with some narrowing integer conversion stmt inserted), do it all together. Jakub
Re: [patch tree-optimization]: Fix for PR 45397 part 1 of 2
On Thu, Mar 15, 2012 at 3:00 PM, Jakub Jelinek ja...@redhat.com wrote: On Thu, Mar 15, 2012 at 02:53:10PM +0100, Kai Tietz wrote: This looks like to match unbound pattern sizes and thus does not fit into the forwprop machinery. Instead it was suggested elsewhere that promoting / demoting registers should be done in a separate pass where you can compute a lattice of used bits and apply a transform based on that lattice and target information (according to PROMOTE_MODE for example). Well, the integer truncation part might be something for a separate pass. It could then also take care that within single-use gimple-statements the integral-constant is always on right-hand-side of first statement of an +, -, |, ^, , and mul. But the cast-hoisting code itself is not unbound AFAICS and has fixed pattern size. Can you split that part out then please? The type demotion is PR45397/PR47477 among other PRs. I'd just walk from the narrowing integer conversion stmts recursively through the def stmts, see if they can be narrowed, note it, and finally if everything or significant portion of the stmts can be demoted (if not all, with some narrowing integer conversion stmt inserted), do it all together. For PROMOTE_MODE targets you'd promote but properly mask out constants (to make them cheaper to generate, for example). You'd also take advantate of targets that can do zero/sign-extending loads without extra cost (ISTR that's quite important for some SPEC 2k6 benchmark on x86_64). Richard. Jakub
Re: [patch tree-optimization]: Fix for PR 45397 part 2 of 2
On Thu, Mar 15, 2012 at 2:46 PM, Kai Tietz ktiet...@googlemail.com wrote: 2012/3/15 Richard Guenther richard.guent...@gmail.com: On Thu, Mar 15, 2012 at 2:09 PM, Kai Tietz ktiet...@googlemail.com wrote: Hi, this is the second part of the patch for this problem. It adds some basic simplifications for ==/!= comparisons for eliminating redudant operands. It adds the following patterns: -X ==/!= Z - X - Z ==/!= 0. ~X ==/!= Z ^ X - Z ==/!= ~0 X ==/!= X - Y - Y == 0 X ==/!= X + Y - Y == 0 X ==/!= X ^ Y - Y == 0 (X - Y) ==/!= (Z - Y) - X ==/!= Z (Y - X) ==/!= (Y - Z) - X ==/!= Z (X + Y) ==/!= (X + Z) - Y ==/!= Z (X + Y) ==/!= (Z + X) - Y ==/!= Z (X ^ Y) ==/!= (Z ^ X) - Y ==/!= Z Can you re-base this patch to work without the previous one? Also please coordinate with Andrew. Note that all of these(?) simplifications are already done by fold_comparison which we could share if you'd split out the EXPR_P op0/op1 cases with separated operands/code. Richard. Hmm, fold_comparison doesn't do the same thing as it checks for possible overflow. This is true for comparisons not being ==/!= or having operands of none-integral-type. But for ==/!= with integral typed arguments the overflow doesn't matter at all. And exactly this is what patch implements here. fold_comparison does not check for overflow for ==/!=. This optimization of course is just desired in non-AST form, as we otherwise loose information in FE. Therefore I didn't added it to fold_const. Which pieces are not already in fold-const btw? forwprop already re-constructs trees for the defs of the lhs/rhs of a comparison. Richard. I can rework the patch so that it works without the other one. Regards, Kai
Re: [PATCH] Change VECTOR_CST representation from TREE_LIST to TREE_VEC-like
* config/sparc/sparc.c (sparc_handle_vis_mul8x16): Adjust interface
and implementation.
(sparc_fold_builtin): Adjust.
OK modulo:
/* Multiply the vector elements in ELTS0 to the elements in ELTS1 as
specified by FNCODE. All of the elements in ELTS0 and ELTS1 lists must be
integer !constants. A tree list with the results of the
multiplications is stored !to the array *N_ELTS, and each element in
the list is of INNER_TYPE. */
! static void
! sparc_handle_vis_mul8x16 (tree *n_elts, int fncode, tree inner_type,
! tree elts0, tree elts1)
{
The prototype and comment must be adjusted:
/* Multiply the VECTOR_CSTs CST0 and CST1 as specified by FNCODE and put
the result into the array N_ELTS, whose elements are of INNER_TYPE. */
static void
sparc_handle_vis_mul8x16 (tree *n_elts, int fncode, tree inner_type,
tree cst0, tree cst1)
--
Eric Botcazou
Re: Remove obsolete IRIX 6.5 support
Ian Lance Taylor i...@google.com writes:
Rainer Orth r...@cebitec.uni-bielefeld.de writes:
* I'm removing IRIX-specific parts of libgo. Given that libgo is
imported from upstream (and supposed to work or made work on the 4.7
branch), I don't know if this a good idea.
Yeah, it's not. Just send me the mainline patches to libgo rather than
applying them yourself, I'll apply them to the upstream repository and
commit to mainline.
Here's the libgo part of the patch. Given that IRIX 6.5 Go support
remains on the 4.7 branch (and I hope to fix it up sufficiently to be
useful after 4.7.0 is released), I wonder if it's a good idea to apply
it upstream or better handle libgo like the other upstream libraries
(boehm-gc, libffi) which I didn't touch for the removal.
Thanks.
Rainer
2012-03-10 Rainer Orth r...@cebitec.uni-bielefeld.de
libgo:
* configure.ac (is_irix): Remove.
(mips-sgi-irix6.5*): Don't set OSCFLAGS.
* configure: Regenerate.
* Makefile.am [LIBGO_IS_IRIX]: Remove.
* Makefile.in: Regenerate.
* mksysinfo.sh [__sgi__]: Remove.
(__timespec): Remove.
* go/net/http/cgi/host.go (osDefaultInheritEnv): Remove irix
handling.
* go/syscall/socket_irix.go: Remove.
* go/time/zoneinfo_unix.go (zoneDirs): Remove IRIX 6 support.
diff --git a/libgo/Makefile.am b/libgo/Makefile.am
--- a/libgo/Makefile.am
+++ b/libgo/Makefile.am
@@ -646,12 +646,6 @@ go_net_sock_file = go/net/sock_linux.go
go_net_sockopt_file = go/net/sockopt_linux.go
go_net_sockoptip_file = go/net/sockoptip_linux.go
else
-if LIBGO_IS_IRIX
-go_net_cgo_file = go/net/cgo_linux.go
-go_net_sock_file = go/net/sock_linux.go
-go_net_sockopt_file = go/net/sockopt_linux.go
-go_net_sockoptip_file = go/net/sockoptip_linux.go
-else
if LIBGO_IS_SOLARIS
go_net_cgo_file = go/net/cgo_linux.go
go_net_sock_file = go/net/sock_linux.go
@@ -671,7 +665,6 @@ go_net_sockoptip_file = go/net/sockoptip
endif
endif
endif
-endif
if LIBGO_IS_LINUX
go_net_sendfile_file = go/net/sendfile_linux.go
@@ -754,9 +747,6 @@ else
if LIBGO_IS_SOLARIS
go_os_sys_file = go/os/sys_uname.go
else
-if LIBGO_IS_IRIX
-go_os_sys_file = go/os/sys_uname.go
-else
if LIBGO_IS_RTEMS
go_os_sys_file = go/os/sys_uname.go
else
@@ -764,7 +754,6 @@ go_os_sys_file = go/os/sys_bsd.go
endif
endif
endif
-endif
if LIBGO_IS_SOLARIS
go_os_stat_file = go/os/stat_solaris.go
@@ -862,12 +851,8 @@ go_sync_files = \
if LIBGO_IS_SOLARIS
go_syslog_file = go/log/syslog/syslog_libc.go
else
-if LIBGO_IS_IRIX
-go_syslog_file = go/log/syslog/syslog_libc.go
-else
go_syslog_file = go/log/syslog/syslog_unix.go
endif
-endif
go_log_syslog_files = \
go/log/syslog/syslog.go \
@@ -1464,13 +1449,9 @@ else
if LIBGO_IS_SOLARIS
syscall_socket_file = go/syscall/socket_solaris.go
else
-if LIBGO_IS_IRIX
-syscall_socket_file = go/syscall/socket_irix.go
-else
syscall_socket_file = go/syscall/socket_bsd.go
endif
endif
-endif
# Support for uname.
if LIBGO_IS_SOLARIS
diff --git a/libgo/configure.ac b/libgo/configure.ac
--- a/libgo/configure.ac
+++ b/libgo/configure.ac
@@ -126,7 +126,6 @@ AC_SUBST(go_include)
is_darwin=no
is_freebsd=no
-is_irix=no
is_linux=no
is_netbsd=no
is_rtems=no
@@ -135,7 +134,6 @@ GOOS=unknown
case ${host} in
*-*-darwin*) is_darwin=yes; GOOS=darwin ;;
*-*-freebsd*) is_freebsd=yes; GOOS=freebsd ;;
- *-*-irix6*)is_irix=yes;GOOS=irix ;;
*-*-linux*)is_linux=yes; GOOS=linux ;;
*-*-netbsd*) is_netbsd=yes; GOOS=netbsd ;;
*-*-rtems*)is_rtems=yes; GOOS=rtems ;;
@@ -143,7 +141,6 @@ case ${host} in
esac
AM_CONDITIONAL(LIBGO_IS_DARWIN, test $is_darwin = yes)
AM_CONDITIONAL(LIBGO_IS_FREEBSD, test $is_freebsd = yes)
-AM_CONDITIONAL(LIBGO_IS_IRIX, test $is_irix = yes)
AM_CONDITIONAL(LIBGO_IS_LINUX, test $is_linux = yes)
AM_CONDITIONAL(LIBGO_IS_NETBSD, test $is_netbsd = yes)
AM_CONDITIONAL(LIBGO_IS_RTEMS, test $is_rtems = yes)
@@ -285,11 +282,6 @@ AC_SUBST(GO_SYSCALL_OS_ARCH_FILE)
dnl Special flags used to generate sysinfo.go.
OSCFLAGS=-D_GNU_SOURCE -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64
case $target in
-mips-sgi-irix6.5*)
- # IRIX 6 needs _XOPEN_SOURCE=500 for the XPG5 version of struct
- # msghdr in sys/socket.h.
- OSCFLAGS=$OSCFLAGS -D_XOPEN_SOURCE=500
- ;;
*-*-solaris2.[[89]])
# Solaris 8/9 need this so struct msghdr gets the msg_control
# etc. fields in sys/socket.h (_XPG4_2).
diff --git a/libgo/go/net/http/cgi/host.go b/libgo/go/net/http/cgi/host.go
--- a/libgo/go/net/http/cgi/host.go
+++ b/libgo/go/net/http/cgi/host.go
@@ -35,7 +35,6 @@ var osDefaultInheritEnv = map[string][]s
darwin: {DYLD_LIBRARY_PATH},
freebsd: {LD_LIBRARY_PATH},
hpux:{LD_LIBRARY_PATH, SHLIB_PATH},
- irix:{LD_LIBRARY_PATH, LD_LIBRARYN32_PATH, LD_LIBRARY64_PATH},
linux: {LD_LIBRARY_PATH},
openbsd: {LD_LIBRARY_PATH},
solaris: {LD_LIBRARY_PATH, LD_LIBRARY_PATH_32, LD_LIBRARY_PATH_64},
diff --git
Re: Remove obsolete Solaris 8 support
Ian Lance Taylor i...@google.com writes: Rainer Orth r...@cebitec.uni-bielefeld.de writes: libgo: * configure.ac (OSCFLAGS): Remove *-*-solaris2.8 handling. (libgo_cv_lib_makecontext_stack_top): Remove sparc*-*-solaris2.8* handling. * configure: Regenerate. As with the Irix 5 changes, just send the libgo patches to me rather than committing them directly to the repository. Thanks. Here it is, with the same caveat as for the IRIX 6.5 removal. Rainer diff --git a/libgo/configure.ac b/libgo/configure.ac --- a/libgo/configure.ac +++ b/libgo/configure.ac @@ -282,8 +282,8 @@ AC_SUBST(GO_SYSCALL_OS_ARCH_FILE) dnl Special flags used to generate sysinfo.go. OSCFLAGS=-D_GNU_SOURCE -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 case $target in -*-*-solaris2.[[89]]) - # Solaris 8/9 need this so struct msghdr gets the msg_control +*-*-solaris2.9) + # Solaris 9 needs this so struct msghdr gets the msg_control # etc. fields in sys/socket.h (_XPG4_2). OSCFLAGS=$OSCFLAGS -D_XOPEN_SOURCE=500 -D_XOPEN_SOURCE_EXTENDED -D__EXTENSIONS__ ;; @@ -623,7 +623,7 @@ fi dnl Check if makecontext expects the uc_stack member of ucontext to point dnl to the top of the stack. case $target in - sparc*-*-solaris2.[[89]]*) + sparc*-*-solaris2.9*) libgo_cv_lib_makecontext_stack_top=yes ;; *) -- - Rainer Orth, Center for Biotechnology, Bielefeld University
Re: [patch tree-optimization]: Fix for PR 45397 part 2 of 2
2012/3/15 Richard Guenther richard.guent...@gmail.com: On Thu, Mar 15, 2012 at 2:46 PM, Kai Tietz ktiet...@googlemail.com wrote: 2012/3/15 Richard Guenther richard.guent...@gmail.com: On Thu, Mar 15, 2012 at 2:09 PM, Kai Tietz ktiet...@googlemail.com wrote: Hi, this is the second part of the patch for this problem. It adds some basic simplifications for ==/!= comparisons for eliminating redudant operands. It adds the following patterns: -X ==/!= Z - X - Z ==/!= 0. ~X ==/!= Z ^ X - Z ==/!= ~0 X ==/!= X - Y - Y == 0 X ==/!= X + Y - Y == 0 X ==/!= X ^ Y - Y == 0 (X - Y) ==/!= (Z - Y) - X ==/!= Z (Y - X) ==/!= (Y - Z) - X ==/!= Z (X + Y) ==/!= (X + Z) - Y ==/!= Z (X + Y) ==/!= (Z + X) - Y ==/!= Z (X ^ Y) ==/!= (Z ^ X) - Y ==/!= Z Can you re-base this patch to work without the previous one? Also please coordinate with Andrew. Note that all of these(?) simplifications are already done by fold_comparison which we could share if you'd split out the EXPR_P op0/op1 cases with separated operands/code. Richard. Hmm, fold_comparison doesn't do the same thing as it checks for possible overflow. This is true for comparisons not being ==/!= or having operands of none-integral-type. But for ==/!= with integral typed arguments the overflow doesn't matter at all. And exactly this is what patch implements here. fold_comparison does not check for overflow for ==/!=. This optimization of course is just desired in non-AST form, as we otherwise loose information in FE. Therefore I didn't added it to fold_const. Which pieces are not already in fold-const btw? forwprop already re-constructs trees for the defs of the lhs/rhs of a comparison. Richard. I have tried to use here instead a call to fold_build2 instead, and I had to notice that it didn't optimized a single case (beside the - and ~ case on both sides). I see in fold const for example in the pattern 'X +- C1 CMP Y +- C2' to 'X CMP Y +- C2 +- C1' explicit the check for it. ... /* Transform comparisons of the form X +- C1 CMP Y +- C2 to X CMP Y +- C2 +- C1 for signed X, Y. This is valid if the resulting offset is smaller in absolute value than the original one. */ if (TYPE_OVERFLOW_UNDEFINED (TREE_TYPE (arg0)) (TREE_CODE (arg0) == PLUS_EXPR || TREE_CODE (arg0) == MINUS_EXPR) ... The same for pattern X +- C1 CMP C2 to X CMP C2 +- C1. The cases for '(X + Y) ==/!= (Z + X)' and co have the same issue or are simply not present. Sorry fold_const doesn't cover this at all. Kai
Re: The state of glibc libm
Hello! SSE ABI entries for i?86 in glibc were rejected. ?I proposed them like 4-5 years ago to make -mfpmath=sse not suck. With the new libm hopefully this can be revisited. But there's the ABI and there's the internal implementation. My point was just that relying on x87 fully again does not really make sense anymore in 2012. That's true, the return value shuffle to/from the x87 stack might not be too bad for performance, nor argument passing on the stack. Can this issue be solved with alternative function entry points, in the same way ICC does? This way, we can pass values in SSE, not on the stack. Uros.
Re: Remove obsolete Solaris 8 support
If we are, we'll never learn if this code is needed on anything beyond Solaris 8 and keep this cruft around basically forever. So what? This file is a big kludge and there is no value whatsoever in it being elegant or particularly readable or even efficient. I'll at least try a bootstrap with the nframes = 2 code ripped out with Solaris 9 FCS libc and libthread (and of course the whole rest: latest Solaris 9, 10 and 11) and see if anything breaks. To make things clear upfront: I'll oppose any change that removes something. -- Eric Botcazou
struct siginfo vs. siginfo_t (was: GNU C Library master sources branch, master, updated. glibc-2.15-229-g4efeffc)
Hi!
On 26 Feb 2012 18:17:52 -, drep...@sourceware.org wrote:
http://sources.redhat.com/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=4efeffc1d583597e4f52985b9747269e47b754e2
commit 4efeffc1d583597e4f52985b9747269e47b754e2
Author: Ulrich Drepper drep...@gmail.com
Date: Sun Feb 26 13:17:27 2012 -0500
Fix up POSIX testing in conformtest
[...]
+ * sysdeps/unix/sysv/linux/bits/siginfo.h: Don't name siginfo_t
+ struct. [...]
[...]
diff --git a/sysdeps/unix/sysv/linux/bits/siginfo.h
b/sysdeps/unix/sysv/linux/bits/siginfo.h
index ecef39d..0635e2f 100644
--- a/sysdeps/unix/sysv/linux/bits/siginfo.h
+++ b/sysdeps/unix/sysv/linux/bits/siginfo.h
[...]
@@ -47,7 +47,7 @@ typedef union sigval
# define __SI_PAD_SIZE ((__SI_MAX_SIZE / sizeof (int)) - 3)
# endif
-typedef struct siginfo
+typedef struct
{
int si_signo;/* Signal number. */
int si_errno;/* If non-zero, an errno value associated with
[...]
This change breaks GCC:
In file included from
/scratch/tschwing/FM_sh-linux-gnu-mk2/src/gcc-mainline/libgcc/unwind-dw2.c:377:0:
./md-unwind-support.h: In function 'sh_fallback_frame_state':
./md-unwind-support.h:182:17: error: field 'info' has incomplete type
In my case, this is really libgcc/config/sh/linux-unwind.h:
[...]
181struct rt_sigframe {
182 struct siginfo info;
183 struct ucontext uc;
184} *rt_ = context-cfa;
[...]
There are more such cases:
boehm-gc/os_dep.c:# define SIGINFO_T struct siginfo
libgcc/config/alpha/linux-unwind.h: struct siginfo info;
libgcc/config/bfin/linux-unwind.h:struct siginfo *pinfo;
libgcc/config/bfin/linux-unwind.h:struct siginfo info;
libgcc/config/i386/linux-unwind.h:struct siginfo *pinfo;
libgcc/config/i386/linux-unwind.h:struct siginfo info;
libgcc/config/ia64/linux-unwind.h:struct siginfo *info;
libgcc/config/ia64/linux-unwind.h:struct siginfo *info;
libgcc/config/mips/linux-unwind.h:struct siginfo info;
libgcc/config/pa/linux-unwind.h:struct siginfo info;
libgcc/config/sh/linux-unwind.h: struct siginfo *pinfo;
libgcc/config/sh/linux-unwind.h: struct siginfo info;
libgcc/config/sh/linux-unwind.h: struct siginfo info;
libgcc/config/tilepro/linux-unwind.h:struct siginfo info;
libgcc/config/xtensa/linux-unwind.h:struct siginfo info;
(Is there really nobody doing nightly testing of GCC against glibc master
branch on x86, which would have shown this earlier?)
Replacing struct siginfo with siginfo_t in sh/linux-unwind.h makes the
build pass. Is this the desired approach; shall we apply this for all
cases listed above (and submit to Boehm GC upstream)? I wonder -- if GCC
breaks, how much software out in the wild is going to break once this
glibc change ripples through?
Grüße,
Thomas
pgpkHs9szr65i.pgp
Description: PGP signature
Re: [google] Add -gfission support to GCC (issue5754090)
I wasn't trying to be pompous! It's just our project name, but I thought fission to be quite appropriate for what it does. How does -gsplit or -gsplit-dwarf work for you? Or -gsplit-debug? -g is already supposed to convey the debug, so I think that -gsplit-dwarf is the best proposal (and we already have -gstrict-dwarf for example). Thanks. -- Eric Botcazou
Re: [Patch, libfortran] RFC: Shared vtables, constification
PING! (At this point, obviously for trunk only) On Mon, Feb 13, 2012 at 20:20, Janne Blomqvist blomqvist.ja...@gmail.com wrote: Hi, the attached patch changes the low-level libgfortran IO dispatching mechanism to use shared vtables for each stream type, instead of all the function pointers being replicated for each unit. This is similar to e.g. how the C++ frontend implements vtables. The benefits are: - Slightly smaller heap memory overhead for each unit as only the vtable pointer needs to be stored, and slightly faster unit initialization as only the vtable pointer needs to be setup instead of all the function pointers in the stream struct. - Looking at unix.o with readelf, one sees Relocation section '.rela.data.rel.ro.local.mem_vtable' at offset 0x15550 contains 8 entries: and similarly for the other vtables; according to http://www.airs.com/blog/archives/189 this means that after relocation the page where this data resides may be marked read-only. The downside is that the sizes of the .text and .data sections are increased. Before: text data bss dec hex filename 1116991 6664 592 1124247 112797 ./x86_64-unknown-linux-gnu/libgfortran/.libs/libgfortran.so After: text data bss dec hex filename 1117487 6936 592 1125015 112a97 ./x86_64-unknown-linux-gnu/libgfortran/.libs/libgfortran.so The data section increase is due to the vtables, the text increase is, I guess, due to the extra pointer dereference when calling the IO functions. Regtested on x86_64-unknown-linux-gnu, Ok for trunk, or 4.8? 2012-02-13 Janne Blomqvist j...@gcc.gnu.org * io/unix.h (struct stream): Rename to stream_vtable. (struct stream): New struct definition. (sread): Dereference vtable pointer. (swrite): Likewise. (sseek): Likewise. (struncate): Likewise. (sflush): Likewise. (sclose): Likewise. * io/unix.c (raw_vtable): New variable. (buf_vtable): Likewise. (mem_vtable): Likewise. (mem4_vtable): Likewise. (raw_init): Assign vtable pointer. (buf_init): Likewise. (open_internal): Likewise. (open_internal4): Likewise. -- Janne Blomqvist -- Janne Blomqvist
Re: struct siginfo vs. siginfo_t (was: GNU C Library master sources branch, master, updated. glibc-2.15-229-g4efeffc)
On Thu, Mar 15, 2012 at 11:05 AM, Thomas Schwinge
tho...@codesourcery.com wrote:
Hi!
On 26 Feb 2012 18:17:52 -, drep...@sourceware.org wrote:
http://sources.redhat.com/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=4efeffc1d583597e4f52985b9747269e47b754e2
commit 4efeffc1d583597e4f52985b9747269e47b754e2
Author: Ulrich Drepper drep...@gmail.com
Date: Sun Feb 26 13:17:27 2012 -0500
Fix up POSIX testing in conformtest
[...]
+ * sysdeps/unix/sysv/linux/bits/siginfo.h: Don't name siginfo_t
+ struct. [...]
[...]
diff --git a/sysdeps/unix/sysv/linux/bits/siginfo.h
b/sysdeps/unix/sysv/linux/bits/siginfo.h
index ecef39d..0635e2f 100644
--- a/sysdeps/unix/sysv/linux/bits/siginfo.h
+++ b/sysdeps/unix/sysv/linux/bits/siginfo.h
[...]
@@ -47,7 +47,7 @@ typedef union sigval
# define __SI_PAD_SIZE ((__SI_MAX_SIZE / sizeof (int)) - 3)
# endif
-typedef struct siginfo
+typedef struct
{
int si_signo; /* Signal number. */
int si_errno; /* If non-zero, an errno value associated with
[...]
This change breaks GCC:
In file included from
/scratch/tschwing/FM_sh-linux-gnu-mk2/src/gcc-mainline/libgcc/unwind-dw2.c:377:0:
./md-unwind-support.h: In function 'sh_fallback_frame_state':
./md-unwind-support.h:182:17: error: field 'info' has incomplete type
In my case, this is really libgcc/config/sh/linux-unwind.h:
[...]
181 struct rt_sigframe {
182 struct siginfo info;
183 struct ucontext uc;
184 } *rt_ = context-cfa;
[...]
POSIX says you get siginto_t *not* struct siginfo, please fix the code.
(Is there really nobody doing nightly testing of GCC against glibc master
branch on x86, which would have shown this earlier?)
I don't test building GCC against glibc master unless I'm making a
potentially ABI breaking change.
We should be rebuilding *all* of userspace when glibc changes. It
would be nice if we setup an OpenEmbedded system to rebuild as much of
x86-64 userspace as possible against a new glibc and check for
regressions.
Replacing struct siginfo with siginfo_t in sh/linux-unwind.h makes the
build pass. Is this the desired approach; shall we apply this for all
cases listed above (and submit to Boehm GC upstream)? I wonder -- if GCC
breaks, how much software out in the wild is going to break once this
glibc change ripples through?
You can't know until you try building a full distribution.
Cheers,
Carlos.
Re: [PATCH] gfortran testsuite: implicitly cleanup-modules
On Thu, Mar 15, 2012 at 1:39 PM, Bernhard Reutner-Fischer
rep.dot@gmail.com wrote:
On Tue, Mar 13, 2012 at 01:30:29PM -0700, Mike Stump wrote:
On Mar 13, 2012, at 9:38 AM, Bernhard Reutner-Fischer wrote:
Could some of the testsuite maintainers please eyeball?
I've eyed it, the only thing that stood out was:
-foreach testcase [lsort [glob -nocomplain $srcdir/$subdir/*.F]] {
- if ![runtest_file_p $runtests $testcase] then {
- continue
- }
- fortran-torture $testcase
-}
-
-foreach testcase [lsort [glob -nocomplain $srcdir/$subdir/*.f90]] {
- if ![runtest_file_p $runtests $testcase] then {
- continue
- }
- fortran-torture $testcase
-}
which, I hope does what you want.
It replaces a manually unrolled loop with a loop over the
testcase-extensions and works as expected. Or did you mean something
else?
Assuming you like those parts, Ok.
committed as r185430.
You forgot to add fortran-modules.exp :(
That breaks final testing result it seems (at least).
Richard.
Since it touches non-fortran, please watch for and respond to any problems it
might cause.
Of course, as always.
PS: I saw that the cleanup-tree-dump is also a bit redundant.
When looking at e.g. gcc/testsuite/gcc.dg/pr52578.c one would think that
scan-tree-dump-times might be able to automagically collect which
dumpfiles to cleanup as last step in dg-final.
One could go one step further and even implicitly add the corresponding
dg-options according to the expected scan-tree-dump files although this
might not be worth it iff there are passes that may produce dump-files
different to their pass name (did not look if this is possible). One
would have to strip the pass-number off the dump-file for passes that
are run more than once, too. Just a thought..
Thanks and cheers,
Re: The state of glibc libm
Uros Bizjak ubiz...@gmail.com writes: Hello! SSE ABI entries for i?86 in glibc were rejected. ?I proposed them like 4-5 years ago to make -mfpmath=sse not suck. With the new libm hopefully this can be revisited. But there's the ABI and there's the internal implementation. My point was just that relying on x87 fully again does not really make sense anymore in 2012. That's true, the return value shuffle to/from the x87 stack might not be too bad for performance, nor argument passing on the stack. Can this issue be solved with alternative function entry points, in the same way ICC does? This way, we can pass values in SSE, not on the stack. Passing values in SSE is a big win, especially on some CPUs like Atom. You can just annotate math.h (possibly with different entry names, so that someone not including it doesn't get totally bogus results) -Andi -- a...@linux.intel.com -- Speaking for myself only
Re: [doc] GCC 4.7 Solaris updates to install.texi
Gerald Pfeifer ger...@pfeifer.com writes:
On Mon, 12 Mar 2012, Rainer Orth wrote:
Tested with make doc/gccinstall.info doc/gccinstall.pdf, ok for mainline
and 4.7 branch?
+Sun does not ship a C compiler with Solaris 2 before Solaris 10, though
+you can download the Sun Studio compilers for free. In Solaris 10 and
+11, GCC 3.4.3 is available in @command{/usr/sfw/bin/gcc}. Solaris 11
+also provides GCC 4.5.2 in @command{/usr/gcc/4.5/bin/gcc}. Alternatively,
I see Richi has approved, just wonder whether as @command would
be more appropriate than in @command, twice?
I think so. Ok for mainline and 4.7 branch?
Rainer
2012-03-15 Rainer Orth r...@cebitec.uni-bielefeld.de
* doc/install.texi (Specific, *-*-solaris2*): Improve wording.
# HG changeset patch
# Parent a1e4c800ecf4ef89c2393b7471ebdf9bbcad40e1
Improve wording in GCC 4.7 Solaris update to install.texi
diff --git a/gcc/doc/install.texi b/gcc/doc/install.texi
--- a/gcc/doc/install.texi
+++ b/gcc/doc/install.texi
@@ -4071,8 +4071,8 @@ been removed in GCC 4.6.
Sun does not ship a C compiler with Solaris 2 before Solaris 10, though
you can download the Sun Studio compilers for free. In Solaris 10 and
-11, GCC 3.4.3 is available in @command{/usr/sfw/bin/gcc}. Solaris 11
-also provides GCC 4.5.2 in @command{/usr/gcc/4.5/bin/gcc}. Alternatively,
+11, GCC 3.4.3 is available as @command{/usr/sfw/bin/gcc}. Solaris 11
+also provides GCC 4.5.2 as @command{/usr/gcc/4.5/bin/gcc}. Alternatively,
you can install a pre-built GCC to bootstrap and install GCC. See the
@uref{binaries.html,,binaries page} for details.
--
-
Rainer Orth, Center for Biotechnology, Bielefeld University
Re: [PATCH] Fix PRs 52080, 52097 and 48124, rewrite bitfield expansion, enable the C++ memory model wrt bitfields everywhere
1. is easy, see patch below. 2. is much harder - we need to compute the bit-offset relative to the bitfield group start, thus in get_bit_range we do /* Compute the adjustment to bitpos from the offset of the field relative to the representative. */ offset = size_diffop (DECL_FIELD_OFFSET (field), DECL_FIELD_OFFSET (repr)); bitoffset = (tree_low_cst (offset, 1) * BITS_PER_UNIT + tree_low_cst (DECL_FIELD_BIT_OFFSET (field), 1) - tree_low_cst (DECL_FIELD_BIT_OFFSET (repr), 1)); and we cannot generally assume offset is zero (well, maybe we could arrange that though, at least we could assert that for all fields in a bitfield group DECL_FIELD_OFFSET is the same?). I'm skeptical. Any suggestion? Apart from trying to make sure that offset will be zero by construction? Or by simply not handling bitfields properly that start at a variable offset? Computing the offset in stor-layout.c and storing it in DECL_INITIAL? The finish_bitfield_representative hunk implements the fix for 1., the rest the proposed zero-by-construction solution for 2. Thanks! -- Eric Botcazou
Re: struct siginfo vs. siginfo_t (was: GNU C Library master sources branch, master, updated. glibc-2.15-229-g4efeffc)
On Thu, Mar 15, 2012 at 8:57 AM, Carlos O'Donell
car...@systemhalted.org wrote:
On Thu, Mar 15, 2012 at 11:05 AM, Thomas Schwinge
tho...@codesourcery.com wrote:
Hi!
On 26 Feb 2012 18:17:52 -, drep...@sourceware.org wrote:
http://sources.redhat.com/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=4efeffc1d583597e4f52985b9747269e47b754e2
commit 4efeffc1d583597e4f52985b9747269e47b754e2
Author: Ulrich Drepper drep...@gmail.com
Date: Sun Feb 26 13:17:27 2012 -0500
Fix up POSIX testing in conformtest
[...]
+ * sysdeps/unix/sysv/linux/bits/siginfo.h: Don't name siginfo_t
+ struct. [...]
[...]
diff --git a/sysdeps/unix/sysv/linux/bits/siginfo.h
b/sysdeps/unix/sysv/linux/bits/siginfo.h
index ecef39d..0635e2f 100644
--- a/sysdeps/unix/sysv/linux/bits/siginfo.h
+++ b/sysdeps/unix/sysv/linux/bits/siginfo.h
[...]
@@ -47,7 +47,7 @@ typedef union sigval
# define __SI_PAD_SIZE ((__SI_MAX_SIZE / sizeof (int)) - 3)
# endif
-typedef struct siginfo
+typedef struct
{
int si_signo; /* Signal number. */
int si_errno; /* If non-zero, an errno value associated with
[...]
This change breaks GCC:
In file included from
/scratch/tschwing/FM_sh-linux-gnu-mk2/src/gcc-mainline/libgcc/unwind-dw2.c:377:0:
./md-unwind-support.h: In function 'sh_fallback_frame_state':
./md-unwind-support.h:182:17: error: field 'info' has incomplete type
In my case, this is really libgcc/config/sh/linux-unwind.h:
[...]
181 struct rt_sigframe {
182 struct siginfo info;
183 struct ucontext uc;
184 } *rt_ = context-cfa;
[...]
POSIX says you get siginto_t *not* struct siginfo, please fix the code.
struct siginfo may not work correctly for Linux/x32
and siginfo_t does since siginfo_t has an alignment
attribute, which isn't applied to struct siginfo.
--
H.J.
Re: [patch tree-optimization]: Fix for PR 45397 part 1 of 2
Hi,
On Thu, 15 Mar 2012, Richard Guenther wrote:
The type demotion is PR45397/PR47477 among other PRs. I'd just walk
from the narrowing integer conversion stmts recursively through the
def stmts, see if they can be narrowed, note it, and finally if
everything or significant portion of the stmts can be demoted (if not
all, with some narrowing integer conversion stmt inserted), do it all
together.
For PROMOTE_MODE targets you'd promote but properly mask out
constants (to make them cheaper to generate, for example). You'd
also take advantate of targets that can do zero/sign-extending loads
without extra cost (ISTR that's quite important for some SPEC 2k6
benchmark on x86_64).
gamess. I still have an old proof of concept patch doing type promotion.
Probably doesn't apply anymore, and it's using too broad predicates (it
simple-mindedly extends to the largest type see in an expression tree).
But I think that basic idea of it is sound.
Ciao,
Michael.
Index: passes.c
===
--- passes.c(revision 159226)
+++ passes.c(working copy)
@@ -831,6 +831,7 @@ init_optimization_passes (void)
NEXT_PASS (pass_all_optimizations);
{
struct opt_pass **p = pass_all_optimizations.pass.sub;
+ extern struct gimple_opt_pass pass_bprop_extends;
NEXT_PASS (pass_remove_cgraph_callee_edges);
/* Initial scalar cleanups before alias computation.
They ensure memory accesses are not indirect wherever possible. */
@@ -838,6 +839,7 @@ init_optimization_passes (void)
NEXT_PASS (pass_update_address_taken);
NEXT_PASS (pass_rename_ssa_copies);
NEXT_PASS (pass_complete_unrolli);
+ NEXT_PASS (pass_bprop_extends);
NEXT_PASS (pass_ccp);
NEXT_PASS (pass_forwprop);
NEXT_PASS (pass_call_cdce);
Index: tree-ssa-ccp.c
===
--- tree-ssa-ccp.c (revision 159226)
+++ tree-ssa-ccp.c (working copy)
@@ -1999,3 +1999,263 @@ struct gimple_opt_pass pass_fold_builtin
| TODO_update_ssa /* todo_flags_finish */
}
};
+
+#if 1
+static bool
+promote_through_insn_p (gimple def, tree *prhs1, tree *prhs2)
+{
+ tree lhs, rhs1, rhs2;
+ if (!is_gimple_assign (def))
+return false;
+ lhs = gimple_assign_lhs (def);
+ rhs1 = rhs2 = NULL;
+ switch (gimple_assign_rhs_class (def))
+{
+ case GIMPLE_SINGLE_RHS:
+ rhs1 = gimple_assign_rhs1 (def);
+ if (TREE_CODE (rhs1) != SSA_NAME)
+ return false;
+ break;
+ case GIMPLE_UNARY_RHS:
+ rhs1 = gimple_assign_rhs1 (def);
+ if (TREE_TYPE (gimple_expr_type (def)) != TREE_TYPE (rhs1))
+ return false;
+ break;
+ case GIMPLE_BINARY_RHS:
+ rhs1 = gimple_assign_rhs1 (def);
+ rhs2 = gimple_assign_rhs2 (def);
+
+ switch (gimple_assign_rhs_code (def))
+ {
+ case LSHIFT_EXPR: case RSHIFT_EXPR:
+ case LROTATE_EXPR: case RROTATE_EXPR:
+ rhs2 = NULL;
+ if (TREE_TYPE (lhs) != TREE_TYPE (rhs1))
+ return false;
+ break;
+ case PLUS_EXPR: case MINUS_EXPR:
+ case MULT_EXPR: case EXACT_DIV_EXPR:
+ case TRUNC_DIV_EXPR: case CEIL_DIV_EXPR:
+ case FLOOR_DIV_EXPR: case ROUND_DIV_EXPR:
+ case TRUNC_MOD_EXPR: case CEIL_MOD_EXPR:
+ case FLOOR_MOD_EXPR: case ROUND_MOD_EXPR:
+ case RDIV_EXPR:
+ case MIN_EXPR: case MAX_EXPR:
+ case BIT_IOR_EXPR: case BIT_XOR_EXPR: case BIT_AND_EXPR:
+ if (TREE_TYPE (lhs) != TREE_TYPE (rhs1)
+ || TREE_TYPE (lhs) != TREE_TYPE (rhs2))
+ return false;
+ break;
+ default:
+ return false;
+ }
+ break;
+ default:
+ return false;
+}
+ if (rhs1 TREE_CODE (rhs1) != SSA_NAME)
+rhs1 = NULL;
+ if (prhs1)
+*prhs1 = rhs1;
+ if (rhs2 TREE_CODE (rhs2) != SSA_NAME)
+rhs2 = NULL;
+ if (prhs2)
+*prhs2 = rhs2;
+ return true;
+}
+
+static tree
+get_extended_version (tree newtype, tree name, bool force)
+{
+ tree ret = TREE_CHAIN (name);
+ tree rhs1, rhs2;
+ gimple def;
+ /* If we already have a version of NAME, try to use it. If it
+ doesn't match in type, fail. */
+ if (ret)
+{
+ if (TREE_TYPE (ret) == newtype)
+ return ret;
+ else
+ return NULL_TREE;
+}
+ def = SSA_NAME_DEF_STMT (name);
+ /* If we can propagate through our defining insn, try to do that. */
+ if (promote_through_insn_p (def, rhs1, rhs2))
+{
+ gimple stmt;
+ tree extrhs1, extrhs2;
+ gimple_stmt_iterator gsi;
+ enum tree_code code;
+ if (rhs1)
+ {
+ extrhs1 = get_extended_version (newtype, rhs1, true);
+ if (!extrhs1)
+ /* ??? We could force here. */
+ return NULL_TREE;
+ }
+ else
+ extrhs1 =
Re: [ping] Vectorizer patches for 4.8
Richard Guenther wrote: On Thu, Mar 8, 2012 at 3:56 PM, Ulrich Weigand uweig...@de.ibm.com wrote: Ira Rosen posted a couple of vectorizer patches intended for 4.8: =A0 http://gcc.gnu.org/ml/gcc-patches/2012-02/msg00191.html =A0 http://gcc.gnu.org/ml/gcc-patches/2012-02/msg00223.html OK to commit the two patches to mainline? Ok. I've checked them in now, thanks! Who will be able to help with problems with those patches? I'll do my best to address any issues that may come up ... Bye, Ulrich -- Dr. Ulrich Weigand GNU Toolchain for Linux on System z and Cell BE ulrich.weig...@de.ibm.com
Re: [PATCH] Fix PRs 52080, 52097 and 48124, rewrite bitfield expansion, enable the C++ memory model wrt bitfields everywhere
Computing the offset in stor-layout.c and storing it in DECL_INITIAL? Ugh. I just realized that the DECL_BIT_FIELD_REPRESENTATIVE is built during layout... but is overloaded with DECL_QUALIFIER. That's probably the source of the miscompilation I talked about earlier. Can we delay it until after gimplification ? Then we could use DECL_INITIAL. -- Eric Botcazou
Re: [PATCH] Change VECTOR_CST representation from TREE_LIST to TREE_VEC-like
On Thu, 15 Mar 2012, Richard Guenther wrote: c-family/ * c-pretty-print.c (pp_c_initializer_list): Adjust. The c-family changes are OK. -- Joseph S. Myers jos...@codesourcery.com
Re: [patch, gcc RFA] dg-extract-results.sh: Handle KFAILs.
On 03/14/2012 08:15 PM, Mike Stump wrote: On Mar 14, 2012, at 10:21 AM, Doug Evans wrote: The results of running the testsuite in parallel should match the results when run serially. This patch adds KFAIL counts so that happens. [There's still a nit that the order of the results don't precisely match, but that's a separate issue.] I will check this into the gdb tree if there are no objections. Any reason not to apply it to the gcc tree as well? I don't know that the gcc tree has any of the known stuff, though, that's not on purpose or by design, just no one has done it, I think. Still, kfail is standard DejaGnu, not a GDB invention. It'd be nice not to need to fork the script for this. -- Pedro Alves
Re: RFA: consolidate DWARF strings into libiberty
Sigh, libiberty is supposed to be a portability library, not a kitchen-sink for common stuff. Should I give up that premise? Or should we consider a common dwarf2 helper library, and move even more of the dwarf2 code into it? First, you'll notice that the first constant for a given enum is defined using a separate macro name. I couldn't think of a better way to avoid the trailing ',' in an enum warning -- one element must be distinguished, so I chose the first. I typically terminate the list with a FOO_MAX enum, which is going to need to be handled differently anyway. Also, you'll see that the names passed to the macros start with an _. Beware you're not violating (or causing to violate) some ISO namespace requirement for identifiers starting with underscores.
Re: RFA: consolidate DWARF strings into libiberty
DJ == DJ Delorie d...@redhat.com writes: DJ Sigh, libiberty is supposed to be a portability library, not a DJ kitchen-sink for common stuff. Should I give up that premise? Or DJ should we consider a common dwarf2 helper library, and move even more DJ of the dwarf2 code into it? My reasoning was: I would have put it somewhere else, but there is nowhere else to put it. Making a new library for this one file seemed like overkill. Finally, there is already stuff in libiberty not related to portability. E.g., hashtab or the demangler. Tom First, you'll notice that the first constant for a given enum is Tom defined using a separate macro name. I couldn't think of a better Tom way to avoid the trailing ',' in an enum warning -- one element Tom must be distinguished, so I chose the first. DJ I typically terminate the list with a FOO_MAX enum, which is going DJ to need to be handled differently anyway. I usually do that to, but I wanted this change not to affect the contents of the defined enums. Tom Also, you'll see that the names passed to the macros start with an _. DJ Beware you're not violating (or causing to violate) some ISO namespace DJ requirement for identifiers starting with underscores. Yeah, I didn't consider that. I guess I can just put the whole DW_TAG_ prefix in there. That isn't a big deal. Or if you have some other suggestion, I can implement it. Tom
Re: [patch, libffi] Sync merge libffi
Iain, I have posted at http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg01799.html the regtests on powerpc-apple-darwin9 with the patch. I still get the following failures FAIL: libffi.call/err_bad_abi.c -O0 -W -Wall execution test FAIL: libffi.call/err_bad_abi.c -O2 execution test FAIL: libffi.call/err_bad_abi.c -O3 execution test FAIL: libffi.call/err_bad_abi.c -Os execution test FAIL: libffi.call/err_bad_abi.c -O2 -fomit-frame-pointer execution test TIA Dominique
Re: RFA: consolidate DWARF strings into libiberty
Finally, there is already stuff in libiberty not related to portability. E.g., hashtab or the demangler. Yeah, I know, hence my Should I give up that premise? I guess I can just put the whole DW_TAG_ prefix in there. That isn't a big deal. Or if you have some other suggestion, I can implement it. If the macros always prepend something to it, it should be safe enough, despite technically violating the spirit of the rule. I have no other suggestions. But given you've already had the issue with and, you're already seeing problems anyway, so maybe that's a sign that it's not actually safe enough ?
Re: RFA: consolidate DWARF strings into libiberty
On 03/15/2012 06:48 PM, DJ Delorie wrote: Finally, there is already stuff in libiberty not related to portability. E.g., hashtab or the demangler. Yeah, I know, hence my Should I give up that premise? Wouldn't it make sense to eventually switch everything to gnulib for portability instead? With that in mind, I predict we'll end up with libiberty composed of only the odd stuff. :-) -- Pedro Alves
Re: [patch, gcc RFA] dg-extract-results.sh: Handle KFAILs.
On Mar 15, 2012, at 11:09 AM, Pedro Alves wrote: Still, kfail is standard DejaGnu, not a GDB invention. It'd be nice not to need to fork the script for this. The change is fine for the gcc tree.
Re: RFA: consolidate DWARF strings into libiberty
On Thu, Mar 15, 2012 at 12:41:54PM -0600, Tom Tromey wrote: I guess I can just put the whole DW_TAG_ prefix in there. That isn't a big deal. Or if you have some other suggestion, I can implement it. Yeah, I think the either the whole OP_TAG (DW_TAG_foobar, ...), or OP_TAG (TAG_foobar, ...) would be safer and nicer. Jakub
[Patch, fortran] Use BUILT_IN_IROUND
Hi,
since some time GCC has BUILT_IN_IROUND{F,,L}, similar to lround() and
llround() but the result is returned as an integer. As there is no
corresponding libc function, this builtin is expanded to lround{f,l}
except when -ffast-math is used, in which case it enables slightly
shorter and faster code to be generated inline. Attached patch enables
this builtin in the gfortran frontend. For the testcase
function my_nint(x)
implicit none
real :: x
integer :: my_nint
my_nint = nint(x)
end function my_nint
compiled with -O2 -ffast-math on x86-64, the difference between the
assembler output of trunk and trunk+patch:
--- iround.trunk.O2fastmath.s 2012-03-15 20:12:40.045069324 +0200
+++ iround.s2012-03-15 20:24:19.501320278 +0200
@@ -12,7 +12,7 @@ my_nint_:
movss .LC0(%rip), %xmm0
orps%xmm1, %xmm0
addss %xmm2, %xmm0
- cvttss2siq %xmm0, %rax
+ cvttss2si %xmm0, %eax
ret
.cfi_endproc
.LFE0:
Comparing the size of the object files, in the patched version the
text section is one byte shorter.
Regtested on x86_64-unknown-linux-gnu, Ok for trunk?
(As an aside, there were some recent problems with __builtin_iround(),
see PR 52592, but they seem to be fixed now so I think this should be
safe)
2012-03-15 Janne Blomqvist j...@gcc.gnu.org
* f95-lang.c (gfc_init_builtin_functions): Initialize
BUILT_IN_IROUND.
* mathbuiltins.def: Add IROUND.
* trans-intrinsic.c (build_round_expr): Use BUILT_IN_IROUND if
type size matches.
(gfc_build_intrinsic_lib_fndecls): Build iround functions.
--
Janne Blomqvist
diff --git a/gcc/fortran/f95-lang.c b/gcc/fortran/f95-lang.c
index 05b598f..3f28e67 100644
--- a/gcc/fortran/f95-lang.c
+++ b/gcc/fortran/f95-lang.c
@@ -773,7 +773,11 @@ gfc_init_builtin_functions (void)
gfc_define_builtin (__builtin_fmodf, mfunc_float[1],
BUILT_IN_FMODF, fmodf, ATTR_CONST_NOTHROW_LEAF_LIST);
- /* lround{f,,l} and llround{f,,l} */
+ /* iround{f,,l}, lround{f,,l} and llround{f,,l} */
+ ftype = build_function_type_list (integer_type_node,
+float_type_node, NULL_TREE);
+ gfc_define_builtin(__builtin_iroundf, ftype, BUILT_IN_IROUNDF,
+ iroundf, ATTR_CONST_NOTHROW_LEAF_LIST);
ftype = build_function_type_list (long_integer_type_node,
float_type_node, NULL_TREE);
gfc_define_builtin (__builtin_lroundf, ftype, BUILT_IN_LROUNDF,
@@ -783,6 +787,10 @@ gfc_init_builtin_functions (void)
gfc_define_builtin (__builtin_llroundf, ftype, BUILT_IN_LLROUNDF,
llroundf, ATTR_CONST_NOTHROW_LEAF_LIST);
+ ftype = build_function_type_list (integer_type_node,
+double_type_node, NULL_TREE);
+ gfc_define_builtin(__builtin_iround, ftype, BUILT_IN_IROUND,
+ iround, ATTR_CONST_NOTHROW_LEAF_LIST);
ftype = build_function_type_list (long_integer_type_node,
double_type_node, NULL_TREE);
gfc_define_builtin (__builtin_lround, ftype, BUILT_IN_LROUND,
@@ -792,6 +800,10 @@ gfc_init_builtin_functions (void)
gfc_define_builtin (__builtin_llround, ftype, BUILT_IN_LLROUND,
llround, ATTR_CONST_NOTHROW_LEAF_LIST);
+ ftype = build_function_type_list (integer_type_node,
+long_double_type_node, NULL_TREE);
+ gfc_define_builtin(__builtin_iroundl, ftype, BUILT_IN_IROUNDL,
+ iroundl, ATTR_CONST_NOTHROW_LEAF_LIST);
ftype = build_function_type_list (long_integer_type_node,
long_double_type_node, NULL_TREE);
gfc_define_builtin (__builtin_lroundl, ftype, BUILT_IN_LROUNDL,
diff --git a/gcc/fortran/mathbuiltins.def b/gcc/fortran/mathbuiltins.def
index b0bcc1f..f6d9586 100644
--- a/gcc/fortran/mathbuiltins.def
+++ b/gcc/fortran/mathbuiltins.def
@@ -64,6 +64,7 @@ OTHER_BUILTIN (FMOD, fmod, 2, true)
OTHER_BUILTIN (FREXP, frexp, frexp, false)
OTHER_BUILTIN (LLROUND, llround, llround, true)
OTHER_BUILTIN (LROUND,lround,lround, true)
+OTHER_BUILTIN (IROUND, iround,iround, true)
OTHER_BUILTIN (NEXTAFTER, nextafter, 2, true)
OTHER_BUILTIN (POW, pow, 1, true)
OTHER_BUILTIN (ROUND, round, 1, true)
diff --git a/gcc/fortran/trans-intrinsic.c b/gcc/fortran/trans-intrinsic.c
index ac9f507..5e54d8e 100644
--- a/gcc/fortran/trans-intrinsic.c
+++ b/gcc/fortran/trans-intrinsic.c
@@ -376,28 +376,24 @@ build_round_expr (tree arg, tree restype)
{
tree argtype;
tree fn;
- bool longlong;
int argprec, resprec;
argtype = TREE_TYPE (arg);
argprec = TYPE_PRECISION (argtype);
resprec = TYPE_PRECISION (restype);
- /* Depending on the type of the result, choose the long int intrinsic
- (lround family) or long long intrinsic (llround). We might also
- need to convert the result afterwards. */
- if (resprec = LONG_TYPE_SIZE)
-
Re: RFA: consolidate DWARF strings into libiberty
DJ == DJ Delorie d...@redhat.com writes: Tom Finally, there is already stuff in libiberty not related to Tom portability. E.g., hashtab or the demangler. DJ Yeah, I know, hence my Should I give up that premise? Yeah. I am not sure there will ever be enough shared code to warrant a new library, particularly because adding a new library is so expensive -- not just the configury stuff but also adding it to the link lines in the Makefiles of all the tools that might need it. I suppose if I had my wish list implemented here, it would be to remove the portability stuff from libiberty in favor of gnulib, and keep libiberty as a higher-level library. Tom I guess I can just put the whole DW_TAG_ prefix in there. That Tom isn't a big deal. Or if you have some other suggestion, I can Tom implement it. DJ If the macros always prepend something to it, it should be safe DJ enough, despite technically violating the spirit of the rule. I have DJ no other suggestions. DJ But given you've already had the issue with and, you're already DJ seeing problems anyway, so maybe that's a sign that it's not actually DJ safe enough ? I think this arises from a bug in cpp, since it issues the -Wc++-compat warning even if the token in question is only used as an argument to stringizing. This seems mistaken -- but it seemed advisable to work around it. Tom
Re: RFA: consolidate DWARF strings into libiberty
Jakub == Jakub Jelinek ja...@redhat.com writes: Jakub On Thu, Mar 15, 2012 at 12:41:54PM -0600, Tom Tromey wrote: I guess I can just put the whole DW_TAG_ prefix in there. That isn't a big deal. Or if you have some other suggestion, I can implement it. Jakub Yeah, I think the either the whole OP_TAG (DW_TAG_foobar, ...), or Jakub OP_TAG (TAG_foobar, ...) would be safer and nicer. Ok, no problem. I will make that change. Tom
Re: [patch, libffi] Sync merge libffi
Hi Dominique, On 15 Mar 2012, at 18:46, Dominique Dhumieres wrote: I have posted at http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg01799.html the regtests on powerpc-apple-darwin9 with the patch. I still get the following failures thanks - I think the priority is to get this (unwind) patch sorted out ... (Note that I've confirmed with David that the unwind patch is local to Darwin, so I will ping that in due course). --- ... I was aware that these (new) tests failed: FAIL: libffi.call/err_bad_abi.c -O0 -W -Wall execution test FAIL: libffi.call/err_bad_abi.c -O2 execution test FAIL: libffi.call/err_bad_abi.c -O3 execution test FAIL: libffi.call/err_bad_abi.c -Os execution test FAIL: libffi.call/err_bad_abi.c -O2 -fomit-frame-pointer execution test David Edelsohn was kind enough to point out that a patch for this has been posted to libffi-discuss, http://sourceware.org/ml/libffi-discuss/2012/msg00070.html unfortunately my schedule is a little hectic right now, and I've not had time to try it. cheers Iain
Preserve pointer types in ivopts
Currently, tree-ssa-loop-ivopts assumes that pointers and integers can
be used interchangeably. It prefers to compute everything in unsigned
integer types rather than pointer types.
On a new target that I'm working on, this assumption is problematic;
casting a pointer to an integer and doing arithmetic on the integer
would be much too expensive to be useful. I came up with the patch
below, which makes ivopts try harder to preserve pointer types.
tree-affine is changed to keep track of base pointers, and do arithmetic
(in particular subtraction) properly if base pointers are involved.
Bootstrapped and regression tested with all languages on i686-linux. I
get FAILs in:
* gcc.dg/guality/pr43051-1.c
A somewhat contrived testcase; gdb now produces optimized out
messages which I think is acceptable? In any case I remain to be
convinced that this testcase demonstrates any real problem.
* gcc.dg/tree-ssa/reassoc-19.c scan-tree-dump-times reassoc2 \+ 0
This seems to fail because we do POINTER_PLUS (ptr, NEG offset))
rather than (char *)((int)ptr - offset). As far as I can tell this
is also not a real problem, but I don't know how to adjust the
testcase.
Comments? Ok?
Bernd
* tree-affine.c (aff_combination_zero): Initialize baseptr.
(aff_combination_add): Handle baseptrs. Abort if both are set.
(aff_combination_diff): New function.
(tree_to_aff_combination_1): Renamed from tree_to_aff_combination.
Remove code to handle pointers; changed to call itself recursively.
(tree_to_aff_combination): New function. Handle the pointer cases
formerly found in the function of the same name, and use
tree_to_aff_combination_1 to compute the offsets.
(aff_combination_to_tree): Build a POINTER_PLUS_EXPR around the
offset if the baseptr is nonnull.
* tree-affine.h (struct affine_tree_combination): New member baseptr.
(aff_combination_diff): Declare.
* tree-predcom.c (determine_offset, valid_initialier_p): Use
aff_combination_diff and return false if it fails.
* tree-ssa-loop-ivopts.c (determine_base_object): If an non-pointer
is cast to a pointer, return the cast.
(add_candidate_1): Use sizetype for steps of a pointer-type iv.
(add_old_iv_candidates): Only add a zero-base pointer candidate if
the precision of pointers and sizetype is equal.
(get_computation_aff): Don't convert steps to pointer types.
Ensure pointers are not scaled. Use aff_combination_diff for
subtraction.
(ptr_difference_cost, difference_cost): Use aff_combination_diff and
return infinite_cost if it fails.
(get_loop_invariant_expr_id): Likewise, returning -1 on failure.
(get_computation_cost_at): Fail if bad pointer expressions would be
generated.
(rewrite_use_nonlinear_expr): Use POINTER_PLUS_EXPR if necessary.
* tree-ssa-address.c (addr_to_parts): Look for a baseptr in the
aff_tree.
* tree-ssa-loop-im.c (mem_refs_may_alias_p): Use aff_combination_diff.
testsuite/
* gcc.dg/tree-ssa/loop-4.c: Scan only for real MEMs, not addresses of
them.
Index: gcc/tree-ssa-loop-im.c
===
--- gcc/tree-ssa-loop-im.c (revision 184938)
+++ gcc/tree-ssa-loop-im.c (working copy)
@@ -1772,8 +1772,8 @@ mem_refs_may_alias_p (tree mem1, tree me
get_inner_reference_aff (mem2, off2, size2);
aff_combination_expand (off1, ttae_cache);
aff_combination_expand (off2, ttae_cache);
- aff_combination_scale (off1, double_int_minus_one);
- aff_combination_add (off2, off1);
+ if (!aff_combination_diff (off2, off1))
+return true;
if (aff_comb_cannot_overlap_p (off2, size1, size2))
return false;
Index: gcc/testsuite/gcc.dg/tree-ssa/loop-4.c
===
--- gcc/testsuite/gcc.dg/tree-ssa/loop-4.c (revision 184938)
+++ gcc/testsuite/gcc.dg/tree-ssa/loop-4.c (working copy)
@@ -37,7 +37,7 @@ void xxx(void)
/* { dg-final { scan-tree-dump-times \\* \[^\\n\\r\]*= 0 optimized } } */
/* { dg-final { scan-tree-dump-times \[^\\n\\r\]*= \\* 0 optimized } } */
-/* { dg-final { scan-tree-dump-times MEM 1 optimized } } */
+/* { dg-final { scan-tree-dump-times \[^\]MEM 1 optimized } } */
/* And the original induction variable should be eliminated. */
Index: gcc/tree-ssa-loop-ivopts.c
===
--- gcc/tree-ssa-loop-ivopts.c (revision 184938)
+++ gcc/tree-ssa-loop-ivopts.c (working copy)
@@ -879,15 +879,21 @@ static tree
determine_base_object (tree expr)
{
enum tree_code code = TREE_CODE (expr);
+ tree type = TREE_TYPE (expr);
tree base, obj;
/* If this is a pointer casted to any type, we need to determine
the base object for the pointer; so handle conversions before
throwing away
Re: [PATCH] gfortran testsuite: implicitly cleanup-modules
On Thu, Mar 15, 2012 at 05:56:32PM +0100, Bernhard Reutner-Fischer wrote: On Thu, Mar 15, 2012 at 04:57:12PM +0100, Richard Guenther wrote: On Thu, Mar 15, 2012 at 1:39 PM, Bernhard Reutner-Fischer rep.dot@gmail.com wrote: committed as r185430. You forgot to add fortran-modules.exp :( committed as r185439. I am very sorry for that.. Even with that file in, libgomp and libitm make check still fail, can't find fortran-modules.exp. Jakub
Re: libgo patch RFA: Export {enter,exit}syscall
On Tue, Mar 13, 2012 at 4:01 PM, Ian Lance Taylor i...@google.com wrote: The cooperative threading model used by Go works by calling entersyscall whenever we are about to make a call to a C function that may block. That was not being done for a call to getaddrinfo used when doing a DNS lookup. This patch fixes that problem by exporting the entersyscall and exitsyscall functions from the syscall package. Exporting a function in Go is done by naming it with a capital letter, so this patch simply consistently renames the entersyscall and exitsyscall functions and all their uses. This will also be useful for SWIG, which faces a similar issue. Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu. Committed to mainline. OK for 4.7 branch? Approved for 4.7 by Jakub offline. Committed to 4.7 branch. Ian
[PATCH] Use vpermpd instead of vpermq for V4DF AVX2 permutations (PR target/52568)
Hi!
This patch let's -mavx2 use vpermpd instead of vpermq for
V4DFmode __builtin_shuffle (e.g. {1, 2, 3, 0}).
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2012-03-15 Jakub Jelinek ja...@redhat.com
PR target/52568
* config/i386/sse.md (UNSPEC_VPERMDF): Remove.
(avx2_permv4df): Remove.
(avx2_permv4di): Macroize into...
(avx2_permmode): ... this using VI8F_256 iterator.
(avx2_permv4di_1): Macroize into...
(avx2_permmode_1): ... this using VI8F_256 iterator.
--- gcc/config/i386/sse.md.jj 2012-03-14 09:39:41.0 +0100
+++ gcc/config/i386/sse.md 2012-03-15 12:47:56.649703448 +0100
@@ -80,7 +80,6 @@ (define_c_enum unspec [
;; For AVX2 support
UNSPEC_VPERMSI
- UNSPEC_VPERMDF
UNSPEC_VPERMSF
UNSPEC_VPERMTI
UNSPEC_GATHER
@@ -11889,19 +11888,6 @@ (define_insn avx2_permvarv8si
(set_attr prefix vex)
(set_attr mode OI)])
-(define_insn avx2_permv4df
- [(set (match_operand:V4DF 0 register_operand =x)
- (unspec:V4DF
- [(match_operand:V4DF 1 register_operand xm)
- (match_operand:SI 2 const_0_to_255_operand n)]
- UNSPEC_VPERMDF))]
- TARGET_AVX2
- vpermpd\t{%2, %1, %0|%0, %1, %2}
- [(set_attr type sselog)
- (set_attr prefix_extra 1)
- (set_attr prefix vex)
- (set_attr mode OI)])
-
(define_insn avx2_permvarv8sf
[(set (match_operand:V8SF 0 register_operand =x)
(unspec:V8SF
@@ -11914,25 +11900,25 @@ (define_insn avx2_permvarv8sf
(set_attr prefix vex)
(set_attr mode OI)])
-(define_expand avx2_permv4di
- [(match_operand:V4DI 0 register_operand )
- (match_operand:V4DI 1 nonimmediate_operand )
+(define_expand avx2_permmode
+ [(match_operand:VI8F_256 0 register_operand )
+ (match_operand:VI8F_256 1 nonimmediate_operand )
(match_operand:SI 2 const_0_to_255_operand )]
TARGET_AVX2
{
int mask = INTVAL (operands[2]);
- emit_insn (gen_avx2_permv4di_1 (operands[0], operands[1],
- GEN_INT ((mask 0) 3),
- GEN_INT ((mask 2) 3),
- GEN_INT ((mask 4) 3),
- GEN_INT ((mask 6) 3)));
+ emit_insn (gen_avx2_permmode_1 (operands[0], operands[1],
+ GEN_INT ((mask 0) 3),
+ GEN_INT ((mask 2) 3),
+ GEN_INT ((mask 4) 3),
+ GEN_INT ((mask 6) 3)));
DONE;
})
-(define_insn avx2_permv4di_1
- [(set (match_operand:V4DI 0 register_operand =x)
- (vec_select:V4DI
- (match_operand:V4DI 1 nonimmediate_operand xm)
+(define_insn avx2_permmode_1
+ [(set (match_operand:VI8F_256 0 register_operand =x)
+ (vec_select:VI8F_256
+ (match_operand:VI8F_256 1 nonimmediate_operand xm)
(parallel [(match_operand 2 const_0_to_3_operand )
(match_operand 3 const_0_to_3_operand )
(match_operand 4 const_0_to_3_operand )
@@ -11945,11 +11931,11 @@ (define_insn avx2_permv4di_1
mask |= INTVAL (operands[4]) 4;
mask |= INTVAL (operands[5]) 6;
operands[2] = GEN_INT (mask);
- return vpermq\t{%2, %1, %0|%0, %1, %2};
+ return vpermssemodesuffix\t{%2, %1, %0|%0, %1, %2};
}
[(set_attr type sselog)
(set_attr prefix vex)
- (set_attr mode OI)])
+ (set_attr mode sseinsnmode)])
(define_insn avx2_permv2ti
[(set (match_operand:V4DI 0 register_operand =x)
Jakub
[PATCH] Improve AVX V4DF and V8SF permutations (PR target/52568)
Hi!
As noted in the PR, we can vectorize e.g. V4DFmode
__builtin_shuffle (, {1, 2, 3, 0}) in 3 insns, some intra-lane
permutation, followed by swapping of the lanes (vperm2f128) and
finally vblend{pd,ps} that merges in the registers with non-swapped
and swapped lanes.
Bootstrapped/regtested on x86_64-linux and i686-linux, tested
additionally with
GCC_TEST_RUN_EXPENSIVE=1 make check-gcc
RUNTESTFLAGS='--target_board=unix\{-m32/-mavx,-m64/-mavx\}
dg-torture.exp=vshuf*'
Ok for trunk?
2012-03-15 Jakub Jelinek ja...@redhat.com
PR target/52568
* config/i386/i386.c (expand_vec_perm_vperm2f128_vblend): New
function.
(ix86_expand_vec_perm_const_1): Use it.
* gcc.dg/torture/vshuf-4.inc: Add two new tests.
* gcc.dg/torture/vshuf-8.inc: Likewise.
* gcc.dg/torture/vshuf-16.inc: Likewise.
* gcc.dg/torture/vshuf-32.inc: Likewise.
--- gcc/config/i386/i386.c.jj 2012-03-14 09:39:41.0 +0100
+++ gcc/config/i386/i386.c 2012-03-15 16:29:05.091015505 +0100
@@ -36627,6 +36627,73 @@ expand_vec_perm_interleave3 (struct expa
return true;
}
+/* A subroutine of ix86_expand_vec_perm_builtin_1. Try to implement
+ a single vector permutation using a single intra-lane vector
+ permutation, vperm2f128 swapping the lanes and vblend* insn blending
+ the non-swapped and swapped vectors together. */
+
+static bool
+expand_vec_perm_vperm2f128_vblend (struct expand_vec_perm_d *d)
+{
+ struct expand_vec_perm_d dfirst, dsecond;
+ unsigned i, j, msk, nelt = d-nelt, nelt2 = nelt / 2;
+ rtx seq;
+ bool ok;
+ rtx (*blend) (rtx, rtx, rtx, rtx) = NULL;
+
+ if (!TARGET_AVX
+ || TARGET_AVX2
+ || (d-vmode != V8SFmode d-vmode != V4DFmode)
+ || d-op0 != d-op1)
+return false;
+
+ dfirst = *d;
+ for (i = 0; i nelt; i++)
+dfirst.perm[i] = 0xff;
+ for (i = 0, msk = 0; i nelt; i++)
+{
+ j = (d-perm[i] nelt2) ? i | nelt2 : i ~nelt2;
+ if (dfirst.perm[j] != 0xff dfirst.perm[j] != d-perm[i])
+ return false;
+ dfirst.perm[j] = d-perm[i];
+ if (j != i)
+ msk |= (1 i);
+}
+ for (i = 0; i nelt; i++)
+if (dfirst.perm[i] == 0xff)
+ dfirst.perm[i] = i;
+
+ if (!d-testing_p)
+dfirst.target = gen_reg_rtx (dfirst.vmode);
+
+ start_sequence ();
+ ok = expand_vec_perm_1 (dfirst);
+ seq = get_insns ();
+ end_sequence ();
+
+ if (!ok)
+return false;
+
+ if (d-testing_p)
+return true;
+
+ emit_insn (seq);
+
+ dsecond = *d;
+ dsecond.op0 = dfirst.target;
+ dsecond.op1 = dfirst.target;
+ dsecond.target = gen_reg_rtx (dsecond.vmode);
+ for (i = 0; i nelt; i++)
+dsecond.perm[i] = i ^ nelt2;
+
+ ok = expand_vec_perm_1 (dsecond);
+ gcc_assert (ok);
+
+ blend = d-vmode == V8SFmode ? gen_avx_blendps256 : gen_avx_blendpd256;
+ emit_insn (blend (d-target, dfirst.target, dsecond.target, GEN_INT (msk)));
+ return true;
+}
+
/* A subroutine of expand_vec_perm_even_odd_1. Implement the double-word
permutation with two pshufb insns and an ior. We should have already
failed all two instruction sequences. */
@@ -37278,6 +37345,9 @@ ix86_expand_vec_perm_const_1 (struct exp
if (expand_vec_perm_interleave3 (d))
return true;
+ if (expand_vec_perm_vperm2f128_vblend (d))
+return true;
+
/* Try sequences of four instructions. */
if (expand_vec_perm_vpshufb2_vpermq (d))
--- gcc/testsuite/gcc.dg/torture/vshuf-4.inc.jj 2011-11-10 18:08:58.0
+0100
+++ gcc/testsuite/gcc.dg/torture/vshuf-4.inc2012-03-15 16:49:17.796455812
+0100
@@ -21,7 +21,9 @@ T (17,1, 3, 5, 7) \
T (18, 3, 3, 3, 3) \
T (19, 3, 2, 1, 0) \
T (20, 0, 4, 1, 5) \
-T (21, 2, 6, 3, 7)
+T (21, 2, 6, 3, 7) \
+T (22, 1, 2, 3, 0) \
+T (23, 2, 1, 0, 3)
#define EXPTESTS \
T (116,1, 2, 4, 3) \
T (117,7, 3, 3, 0) \
--- gcc/testsuite/gcc.dg/torture/vshuf-8.inc.jj 2011-11-10 18:08:58.0
+0100
+++ gcc/testsuite/gcc.dg/torture/vshuf-8.inc2012-03-15 16:50:25.313099258
+0100
@@ -21,7 +21,9 @@ T (17,1, 3, 5, 7, 9, 11, 13, 15) \
T (18, 3, 3, 3, 3, 3, 3, 3, 3) \
T (19, 7, 6, 5, 4, 3, 2, 1, 0) \
T (20, 0, 8, 1, 9, 2, 10, 3, 11) \
-T (21, 4, 12, 5, 13, 6, 14, 7, 15)
+T (21, 4, 12, 5, 13, 6, 14, 7, 15) \
+T (22, 1, 2, 3, 4, 5, 6, 7, 0) \
+T (23, 6, 5, 4, 3, 2, 1, 0, 7)
#define EXPTESTS \
T (116,9, 3, 9, 4, 7, 0, 0, 6) \
T (117,4, 14, 12, 8, 9, 6, 0, 10) \
--- gcc/testsuite/gcc.dg/torture/vshuf-16.inc.jj2011-11-10
18:08:58.0 +0100
+++ gcc/testsuite/gcc.dg/torture/vshuf-16.inc 2012-03-15 16:51:15.383835439
+0100
@@ -21,7 +21,9 @@ T (17,1, 3, 5, 7, 9, 11, 13, 15, 17, 19
T (18, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3) \
T (19, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0) \
T (20, 0, 16, 1, 17, 2, 18, 3, 19, 4, 20, 5, 21, 6, 22, 7, 23) \
-T (21, 8, 24, 9, 25, 10, 26, 11, 27, 12, 28, 13, 29, 14, 30, 15, 31)
+T (21, 8, 24, 9, 25, 10, 26, 11, 27, 12, 28, 13, 29, 14, 30, 15, 31) \
+T
Re: [Patch, fortran] Use BUILT_IN_IROUND
Janne Blomqvist wrote:
since some time GCC has BUILT_IN_IROUND{F,,L}, similar to lround() and
llround() but the result is returned as an integer.
Regtested on x86_64-unknown-linux-gnu, Ok for trunk?
OK. Thanks for the patch! Nit: Could you check mathbuiltins.def - at
least in the diff, iround seems to be misaligned (one missing).
Tobias
2012-03-15 Janne Blomqvistj...@gcc.gnu.org
* f95-lang.c (gfc_init_builtin_functions): Initialize
BUILT_IN_IROUND.
* mathbuiltins.def: Add IROUND.
* trans-intrinsic.c (build_round_expr): Use BUILT_IN_IROUND if
type size matches.
(gfc_build_intrinsic_lib_fndecls): Build iround functions.
Re: [Patch, fortran] Use BUILT_IN_IROUND
On Thu, Mar 15, 2012 at 22:14, Tobias Burnus bur...@net-b.de wrote:
Janne Blomqvist wrote:
since some time GCC has BUILT_IN_IROUND{F,,L}, similar to lround() and
llround() but the result is returned as an integer.
Regtested on x86_64-unknown-linux-gnu, Ok for trunk?
OK. Thanks for the patch! Nit: Could you check mathbuiltins.def - at least
in the diff, iround seems to be misaligned (one missing).
Ah, a tab had sneaked in, fixed. Committed the fixed patch as r185442.
Thanks for the quick review!
--
Janne Blomqvist
[wwwdocs] Buildstat update for 4.6
Latest results for 4.6.x -tgc Testresults for 4.6.2: hppa2.0w-hp-hpux11.11 i386-pc-solaris2.8 i686-pc-linux-gnu powerpc-apple-darwin8.11.0 sparc-sun-solaris2.8 (2) x86_64-apple-darwin10.8.0 x86_64-apple-darwin11.3.0 x86_64-unknown-linux-gnu Testresults for 4.6.2: powerpc-ibm-aix5.3.0.0 (2) x86_64-unknown-linux-gnu Index: buildstat.html === RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-4.6/buildstat.html,v retrieving revision 1.9 diff -u -r1.9 buildstat.html --- buildstat.html 14 Dec 2011 10:57:49 - 1.9 +++ buildstat.html 15 Mar 2012 21:11:11 - @@ -78,6 +78,7 @@ tdhppa2.0w-hp-hpux11.11/td tdnbsp;/td tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg00172.html;4.6.3/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg03199.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-06/msg03306.html;4.6.1/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-03/msg02630.html;4.6.0/a @@ -107,6 +108,7 @@ tdi386-pc-solaris2.8/td tdnbsp;/td tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg01757.html;4.6.3/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg03286.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-07/msg00139.html;4.6.1/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-04/msg00175.html;4.6.0/a, @@ -165,6 +167,7 @@ tdi686-pc-linux-gnu/td tdnbsp;/td tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg00163.html;4.6.3/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg03395.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg03062.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg00899.html;4.6.1/a, @@ -203,6 +206,7 @@ tdpowerpc-apple-darwin8.11.0/td tdnbsp;/td tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg00497.html;4.6.3/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg03272.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-07/msg01092.html;4.6.1/a /td @@ -218,6 +222,15 @@ /tr tr +tdpowerpc-ibm-aix5.3.0.0/td +tdnbsp;/td +tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-02/msg00142.html;4.6.2/a, +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-02/msg00436.html;4.6.2/a +/td +/tr + +tr tdpowerpc-ibm-aix6.1.0.0/td tdnbsp;/td tdTest results: @@ -260,6 +273,8 @@ tdsparc-sun-solaris2.8/td tdnbsp;/td tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg01811.html;4.6.3/a, +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg00337.html;4.6.3/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-11/msg01044.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-11/msg00683.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg03536.html;4.6.2/a, @@ -325,6 +340,7 @@ tdx86_64-apple-darwin10.8.0/td tdnbsp;/td tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg01724.html;4.6.3/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg02994.html;4.6.2/a /td /tr @@ -338,9 +354,19 @@ /tr tr +tdx86_64-apple-darwin11.3.0/td +tdnbsp;/td +tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg00118.html;4.6.3/a +/td +/tr + +tr tdx86_64-unknown-linux-gnu/td tdnbsp;/td tdTest results: +a href=http://gcc.gnu.org/ml/gcc-testresults/2012-03/msg01688.html;4.6.3/a, +a href=http://gcc.gnu.org/ml/gcc-testresults/2011-12/msg00968.html;4.6.2/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-06/msg03135.html;4.6.1/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-06/msg01380.html;4.6.0/a, a href=http://gcc.gnu.org/ml/gcc-testresults/2011-05/msg03091.html;4.6.0/a,
C++ PATCH for c++/52582 (ICE with devirtualization on PPC)
jakub and richi analyzed this bug; normally we set DECL_EXTERNAL on all
functions with vague linkage, but we were failing to do so for the
synthesized destructor, which caused problems with devirtualization.
Fixed thus.
Jakub tested the patch. Applying to trunk; will apply to 4.7 branch
after 4.7.0.
commit a5e3b73aaee13ed672641faee53989785cffed83
Author: Jason Merrill ja...@redhat.com
Date: Wed Mar 14 14:24:22 2012 -0400
PR c++/52582
* method.c (implicitly_declare_fn): Set DECL_EXTERNAL.
diff --git a/gcc/cp/method.c b/gcc/cp/method.c
index 0718916..0d4793e 100644
--- a/gcc/cp/method.c
+++ b/gcc/cp/method.c
@@ -1593,6 +1593,7 @@ implicitly_declare_fn (special_function_kind kind, tree type, bool const_p)
DECL_DELETED_FN (fn) = deleted_p;
DECL_DECLARED_CONSTEXPR_P (fn) = constexpr_p;
}
+ DECL_EXTERNAL (fn) = true;
DECL_NOT_REALLY_EXTERN (fn) = 1;
DECL_DECLARED_INLINE_P (fn) = 1;
gcc_assert (!TREE_USED (fn));
diff --git a/gcc/testsuite/g++.dg/torture/pr52582.C b/gcc/testsuite/g++.dg/torture/pr52582.C
new file mode 100644
index 000..1b93fc4
--- /dev/null
+++ b/gcc/testsuite/g++.dg/torture/pr52582.C
@@ -0,0 +1,23 @@
+// PR c++/52582
+
+inline void *operator new (__SIZE_TYPE__, void *p) throw ()
+{
+ return p;
+}
+struct B
+{
+ virtual ~B ();
+ B ();
+};
+struct A : B
+{
+ A () : B () {}
+ virtual void bar ();
+};
+void
+foo ()
+{
+ char a[64];
+ B *b = new (a) A ();
+ b-~B ();
+}
[pph] add tests and -fpph-check (issue5820069)
This patch mostly adds several test cases reduced from full-scale attempts to
use PPH.
c?anonymous* -- problems handling anonymous/tagless types
c?features* -- problems with benign macro redefinitions
x?tmpldfltparm* -- inappropriately merging default template arguments
It also add an option -fpph-check to refine our check for headers compatible
with PPH. It implies -fprimary-system-header-okay and checks for the main
source file missing a guard. Generating a pph file implies the same check.
Tested on x64.
Index: gcc/c-family/ChangeLog.pph
2012-03-15 Lawrence Crowl cr...@google.com
* c.opt (-fpph-check): New.
* c-opts.c (c_common_handle_option): Add OPT_fpph_check.
(case OPT__output_pph_): Also imply -fpph-check.
* c-common.h (bool pph_check_main_missing_guard): New.
* c-common.c (bool pph_check_main_missing_guard): New.
(const char *pph_out_file): Explicitly initialize to NULL.
Index: gcc/testsuite/ChangeLog.pph
2012-03-15 Lawrence Crowl cr...@google.com
* g++.dg/pph/c0anonymous.h: New.
* g++.dg/pph/c1anonymous1.h: New.
* g++.dg/pph/c1anonymous2.h: New.
* g++.dg/pph/c5features1.h: New.
* g++.dg/pph/c5features2.h: New.
* g++.dg/pph/c7features.cc: New.
* g++.dg/pph/d8dupguard.cc: Add xfail-if comment.
* g++.dg/pph/x0tmpldfltparm.h: New.
* g++.dg/pph/x1tmpldfltparm.cc: New.
* g++.dg/pph/y9overload.cc: Add xfail-if comment.
Index: gcc/cp/ChangeLog.pph
2012-03-15 Lawrence Crowl cr...@google.com
* pph.h (pph_check_main_guarded): New.
* pph-core.c (pph_check_main_guarded): New.
* pph-out.c (pph_writer_finish): Factor out guard check into
pph_check_main_guarded.
* parser.c (c_parse_file): Add guard check for non-pph compiles.
Index: gcc/c-family/c.opt
===
--- gcc/c-family/c.opt (revision 185442)
+++ gcc/c-family/c.opt (working copy)
@@ -981,6 +981,10 @@ fplan9-extensions
C ObjC Var(flag_plan9_extensions)
Enable Plan 9 language extensions
+fpph-check
+C++
+-fpph-checkCheck a header for minimal compatibility with PPH.
+
fpph-debug=
C++ Joined RejectNegative UInteger Var(flag_pph_debug)
-fpph-debug=N Enable debugging output at level N from PPH support
Index: gcc/c-family/c-opts.c
===
--- gcc/c-family/c-opts.c (revision 185442)
+++ gcc/c-family/c-opts.c (working copy)
@@ -404,6 +404,7 @@ c_common_handle_option (size_t scode, co
case OPT__output_pph_:
pph_out_file = arg;
cpp_opts-primary_system_header_okay = true;
+ pph_check_main_missing_guard = true;
break;
case OPT_A:
@@ -814,6 +815,11 @@ c_common_handle_option (size_t scode, co
set_struct_debug_option (global_options, loc, arg);
break;
+case OPT_fpph_check:
+ cpp_opts-primary_system_header_okay = true;
+ pph_check_main_missing_guard = true;
+ break;
+
case OPT_fpph_hdr_:
add_pph_header_map (arg);
break;
Index: gcc/c-family/c-common.c
===
--- gcc/c-family/c-common.c (revision 185442)
+++ gcc/c-family/c-common.c (working copy)
@@ -194,7 +194,11 @@ const char *pch_file;
/* The file name to which we should write a preparsed header, or
NULL if no header will be written in this compile. */
-const char *pph_out_file;
+const char *pph_out_file = NULL;
+
+/* Whether or not we should check for a guard on the main input file. */
+
+bool pph_check_main_missing_guard = false;
/* Nonzero if an ISO standard was selected. It rejects macros in the
user's namespace. */
Index: gcc/c-family/c-common.h
===
--- gcc/c-family/c-common.h (revision 185442)
+++ gcc/c-family/c-common.h (working copy)
@@ -585,6 +585,10 @@ extern const char *pch_file;
extern const char *pph_out_file;
+/* Whether or not we should check for a guard on the main input file. */
+
+extern bool pph_check_main_missing_guard;
+
/* Return true if we have any map from INCLUDE to PPH file. */
extern bool
Index: gcc/testsuite/g++.dg/pph/c1anonymous1.h
===
--- gcc/testsuite/g++.dg/pph/c1anonymous1.h (revision 0)
+++ gcc/testsuite/g++.dg/pph/c1anonymous1.h (revision 0)
@@ -0,0 +1,11 @@
+// {xfail-if ANONYMOUS MERGING { *-*-* } { -fpph-map=pph.map } }
+// { dg-bogus c0anonymous.h:4:16: error: 'anon_t' has a previous declaration
here { xfail *-*-* } 0 }
+
+#ifndefC1ANONYMOUS
+#defineC1ANONYMOUS
+
+#include c0anonymous.h
+
+enum { first, second }; // { dg-bogus 'anon_t' referred to as enum {
xfail *-*-* } }
+
+#endif
Index: gcc/testsuite/g++.dg/pph/c1anonymous2.h
Re: Preserve pointer types in ivopts
Hi, Currently, tree-ssa-loop-ivopts assumes that pointers and integers can be used interchangeably. It prefers to compute everything in unsigned integer types rather than pointer types. On a new target that I'm working on, this assumption is problematic; casting a pointer to an integer and doing arithmetic on the integer would be much too expensive to be useful. I came up with the patch below, which makes ivopts try harder to preserve pointer types. tree-affine is changed to keep track of base pointers, and do arithmetic (in particular subtraction) properly if base pointers are involved. Bootstrapped and regression tested with all languages on i686-linux. I get FAILs in: * gcc.dg/guality/pr43051-1.c A somewhat contrived testcase; gdb now produces optimized out messages which I think is acceptable? In any case I remain to be convinced that this testcase demonstrates any real problem. * gcc.dg/tree-ssa/reassoc-19.c scan-tree-dump-times reassoc2 \+ 0 This seems to fail because we do POINTER_PLUS (ptr, NEG offset)) rather than (char *)((int)ptr - offset). As far as I can tell this is also not a real problem, but I don't know how to adjust the testcase. Comments? Ok? the reason unsigned integer types are prefered is that possible overflows during the computation have defined semantics. With pointer types, the intermediate steps of the computations could have undefined behavior, possibly confusing further optimizations. Is the patch with this regard? Zdenek
Re: [Patch, libfortran] RFC: Shared vtables, constification
On 03/15/2012 11:42 AM, Janne Blomqvist wrote: PING! (At this point, obviously for trunk only) Yes, OK for trunk. On Mon, Feb 13, 2012 at 20:20, Janne Blomqvist blomqvist.ja...@gmail.com wrote: Hi, the attached patch changes the low-level libgfortran IO dispatching mechanism to use shared vtables for each stream type, instead of all the function pointers being replicated for each unit. This is similar to e.g. how the C++ frontend implements vtables. The benefits are: - Slightly smaller heap memory overhead for each unit as only the vtable pointer needs to be stored, and slightly faster unit initialization as only the vtable pointer needs to be setup instead of all the function pointers in the stream struct. - Looking at unix.o with readelf, one sees Relocation section '.rela.data.rel.ro.local.mem_vtable' at offset 0x15550 contains 8 entries: and similarly for the other vtables; according to http://www.airs.com/blog/archives/189 this means that after relocation the page where this data resides may be marked read-only. The downside is that the sizes of the .text and .data sections are increased. Before: textdata bss dec hex filename 11169916664 592 1124247 112797 ./x86_64-unknown-linux-gnu/libgfortran/.libs/libgfortran.so After: textdata bss dec hex filename 11174876936 592 1125015 112a97 ./x86_64-unknown-linux-gnu/libgfortran/.libs/libgfortran.so The data section increase is due to the vtables, the text increase is, I guess, due to the extra pointer dereference when calling the IO functions. Regtested on x86_64-unknown-linux-gnu, Ok for trunk, or 4.8? 2012-02-13 Janne Blomqvistj...@gcc.gnu.org * io/unix.h (struct stream): Rename to stream_vtable. (struct stream): New struct definition. (sread): Dereference vtable pointer. (swrite): Likewise. (sseek): Likewise. (struncate): Likewise. (sflush): Likewise. (sclose): Likewise. * io/unix.c (raw_vtable): New variable. (buf_vtable): Likewise. (mem_vtable): Likewise. (mem4_vtable): Likewise. (raw_init): Assign vtable pointer. (buf_init): Likewise. (open_internal): Likewise. (open_internal4): Likewise. -- Janne Blomqvist
Re: struct siginfo vs. siginfo_t (was: GNU C Library master sources branch, master, updated. glibc-2.15-229-g4efeffc)
On Thursday 15 March 2012 11:57:00 Carlos O'Donell wrote: We should be rebuilding *all* of userspace when glibc changes. It would be nice if we setup an OpenEmbedded system to rebuild as much of x86-64 userspace as possible against a new glibc and check for regressions. emerge -e world -mike signature.asc Description: This is a digitally signed message part.
Re: [PATCH] Use vpermpd instead of vpermq for V4DF AVX2 permutations (PR target/52568)
On 03/15/12 13:05, Jakub Jelinek wrote:
Hi!
This patch let's -mavx2 use vpermpd instead of vpermq for
V4DFmode __builtin_shuffle (e.g. {1, 2, 3, 0}).
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2012-03-15 Jakub Jelinek ja...@redhat.com
PR target/52568
* config/i386/sse.md (UNSPEC_VPERMDF): Remove.
(avx2_permv4df): Remove.
(avx2_permv4di): Macroize into...
(avx2_permmode): ... this using VI8F_256 iterator.
(avx2_permv4di_1): Macroize into...
(avx2_permmode_1): ... this using VI8F_256 iterator.
Ok.
r~
Re: [PATCH] Improve AVX V4DF and V8SF permutations (PR target/52568)
On 03/15/12 13:09, Jakub Jelinek wrote:
Hi!
As noted in the PR, we can vectorize e.g. V4DFmode
__builtin_shuffle (, {1, 2, 3, 0}) in 3 insns, some intra-lane
permutation, followed by swapping of the lanes (vperm2f128) and
finally vblend{pd,ps} that merges in the registers with non-swapped
and swapped lanes.
Bootstrapped/regtested on x86_64-linux and i686-linux, tested
additionally with
GCC_TEST_RUN_EXPENSIVE=1 make check-gcc
RUNTESTFLAGS='--target_board=unix\{-m32/-mavx,-m64/-mavx\}
dg-torture.exp=vshuf*'
Ok for trunk?
2012-03-15 Jakub Jelinek ja...@redhat.com
PR target/52568
* config/i386/i386.c (expand_vec_perm_vperm2f128_vblend): New
function.
(ix86_expand_vec_perm_const_1): Use it.
* gcc.dg/torture/vshuf-4.inc: Add two new tests.
* gcc.dg/torture/vshuf-8.inc: Likewise.
* gcc.dg/torture/vshuf-16.inc: Likewise.
* gcc.dg/torture/vshuf-32.inc: Likewise.
Ok.
r~
Re: Preserve pointer types in ivopts
On Fri, Mar 16, 2012 at 12:20:44AM +0100, Bernd Schmidt wrote: On 03/16/2012 12:16 AM, Jakub Jelinek wrote: On Fri, Mar 16, 2012 at 12:03:08AM +0100, Bernd Schmidt wrote: On 03/15/2012 11:12 PM, Zdenek Dvorak wrote: the reason unsigned integer types are prefered is that possible overflows during the computation have defined semantics. With pointer types, the intermediate steps of the computations could have undefined behavior, possibly confusing further optimizations. Is the patch with this regard? It's trying to use sizetype for pointer offset computations. As far as I can tell that's supposed to be an unsigned type, so it should be OK. I think the final POINTER_PLUS_EXPRs we make can't overflow in valid programs. In the IL before ivopts it shouldn't for valid programs, but what ivopts makes out of it often would, that is why it uses unsigned integers instead. Well, what are our rules for whether overflow on POINTER_PLUS_EXPR is defined or not? A quick search through the headers and docs doesn't turn up anything. Would there be a downside to defining it as wrapping? Can you show an example where a POINTER_PLUS_EXPR produced by ivopts would overflow? Don't have a testcase right now, I've just seen IVOPTS many times in the past initialize an IV with start of array minus some constant, end of array plus some constant or similar (which is fine if the IV is unsigned integer of the size of a pointer, but certainly wouldn't be fine if the IV had pointer type). For pointer arithmetics in the IL we assume the C requirements, pointer arithmetics can be performed only within the same object, so for int a[10]; both of the following are invalid, even in the IL: int *p = a - 1; int *q = a + 11; Jakub
Re: Preserve pointer types in ivopts
Hi,
Well, what are our rules for whether overflow on POINTER_PLUS_EXPR is
defined or not? A quick search through the headers and docs doesn't turn
up anything. Would there be a downside to defining it as wrapping?
Can you show an example where a POINTER_PLUS_EXPR produced by ivopts
would overflow?
Don't have a testcase right now, I've just seen IVOPTS many times in the
past initialize an IV with start of array minus some constant, end of array
plus some constant or similar (which is fine if the IV is unsigned integer
of the size of a pointer, but certainly wouldn't be fine if the IV had
pointer type). For pointer arithmetics in the IL we assume the C
requirements, pointer arithmetics can be performed only within the same
object, so for
int a[10];
both of the following are invalid, even in the IL:
int *p = a - 1;
int *q = a + 11;
for example, something like this:
int a[1000];
void foo(int n)
{
int i, *p = a;
for (i = 8; i n; i++)
*p++ = 10;
}
ivopts may decide to change this to:
int a[1000];
void foo(int n)
{
int i, *p = a - 8;
for (i = 8; i n; i++)
p[i] = 10;
}
which may require one less adition, depending on the available addressing modes.
Of course, as written above this has undefined behavior; hence, the casts to
unsigned integer,
Zdenek
Re: Preserve pointer types in ivopts
On 03/16/2012 12:44 AM, Jakub Jelinek wrote: For pointer arithmetics in the IL we assume the C requirements, pointer arithmetics can be performed only within the same object, so for int a[10]; both of the following are invalid, even in the IL: int *p = a - 1; int *q = a + 11; Ok, so what's the solution? Add a second POINTER_PLUS_EXPR code with defined overflow behaviour, or add a flag to it? Bernd
Re: Preserve pointer types in ivopts
On Thu, Mar 15, 2012 at 5:09 PM, Bernd Schmidt ber...@codesourcery.com wrote: On 03/16/2012 12:44 AM, Jakub Jelinek wrote: For pointer arithmetics in the IL we assume the C requirements, pointer arithmetics can be performed only within the same object, so for int a[10]; both of the following are invalid, even in the IL: int *p = a - 1; int *q = a + 11; Ok, so what's the solution? Add a second POINTER_PLUS_EXPR code with defined overflow behaviour, or add a flag to it? We should have one for PLUS_EXPR also. There was some movement on that on a branch that Richard Guenther did but I don't know if he is going to work on it further. I have been noticing more and more the need for this feature while working on my tree combiner branch, that I might try to see if Richard's branch can be revisited. Thanks, Andrew Pinski
Re: Ping: Re: [patch middle-end]: Fix PR/48814 - [4.4/4.5/4.6/4.7 Regression] Incorrect scalar increment result
On 15 March 2012 15:40, Richard Guenther wrote: On Thu, Mar 15, 2012 at 4:22 PM, Kai Tietz ktiet...@googlemail.com wrote: Richard, ping. I think now could be a good time for applying the patch you have for this issue as we are in stage 1. It will still regress the two libstdc++ testcases (well, I guess so at least). Jonathan - you didn't answer my reply to your question? Would it be ok to apply this patch with leaving the regressions in-place, to be investigated by libstdc++ folks? Sorry, I've either forgotten or missed the reply - but if you think the problem is in libstdc++ then certainly go ahead and apply it, I'll investigate the libstdc++ problems (and ask for help if they defeat me!)
[PATCH, ARM, 4.6] backport PR pch/45979
Hi there. This patch backports my PCH on ARM EABI fix[1] for pch/PR45979 to the 4.6 branch. This fixes PCH support on ARM and tidies up the random pch testsuite failures that are seen between runs. OK for 4.6? -- Michael [1] http://gcc.gnu.org/ml/gcc-patches/2011-05/msg00017.html gcc/ 2012-03-16 Michael Hope michael.h...@linaro.org Backport from mainline 2011-05-05 Michael Hope michael.h...@linaro.org PR pch/45979 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for __ARM_EABI__ hosts. diff --git a/gcc/config/host-linux.c b/gcc/config/host-linux.c index 47ce3ea..ec61055 100644 --- a/gcc/config/host-linux.c +++ b/gcc/config/host-linux.c @@ -84,6 +84,8 @@ # define TRY_EMPTY_VM_SPACE0x6000 #elif defined(__mc68000__) # define TRY_EMPTY_VM_SPACE0x4000 +#elif defined(__ARM_EABI__) +# define TRY_EMPTY_VM_SPACE 0x6000 #else # define TRY_EMPTY_VM_SPACE0 #endif
[google/4.6] Fix problems with -gfission (issue5844043)
For google/gcc-4_6 branch.
This patch fixes several problems with -gfission:
- Bad index for range list in the compile unit DIE.
- DW_AT_ranges attribute for compile unit in the wrong file.
- Incorrect size for skeleton type unit DIEs.
- Wrote location expression using DW_OP_addr to DWO file.
- Emitted skeleton debug section even when there is no debug info.
Tested: bootstrap, gcc regression tests, hand testing on -gfission
test cases.
include/
2012-03-15 Sterling Augustine saugust...@google.com
Cary Coutant ccout...@google.com
* dwarf2.h (enum dwarf_location_atom): Add
DW_OP_GNU_addr_index.
gcc/
2012-03-15 Sterling Augustine saugust...@google.com
Cary Coutant ccout...@google.com
* dwarf2out.c (dwarf_stack_op_name): Add DW_OP_GNU_addr_index.
(size_of_loc_descr): Likewise.
(output_loc_operands): Likewise.
(output_loc_operands_raw): Likewise.
(new_addr_loc_descr): New function.
(add_AT_range_list): Store index in AT_index.
(size_of_die): For range_list, use AT_index for index value.
(add_top_level_skeleton_die_attrs): Don't add DW_AT_stmt_list here.
(get_skeleton_type_unit): New function.
(output_skeleton_debug_sections): Add comp_unit parameter; adjust
caller. Don't allocate new comp unit DIE here. Move allocation
of debug_skeleton_info_section_label and
debug_skeleton_abbrev_section_label to dwarf2out_init. Call
get_skeleton_type_unit.
(output_comdat_type_unit): Remove assert; call get_skeleton_type_unit.
(mem_loc_descriptor): Call new_addr_loc_desc.
(loc_descriptor): Likewise.
(loc_list_from_tree): Likewise.
(add_const_value_attribute): Likewise.
(dwarf2out_init): Allocate debug_skeleton_info_section_label and
debug_skeleton_abbrev_section_label.
(output_indirect_string): Check for DW_FORM_strp instead of label
and refcount.
(output_addr_table): Add case for dw_val_class_loc.
(resolve_addr_in_expr): Handle DW_OP_addr_index.
(hash_loc_operands): Likewise.
(compare_loc_operands): Likewise.
(dwarf2out_finish): Allocate skeleton compile unit DIE here; add
range lists, DW_AT_stmt_list, and DW_AT_macro_info to it instead
of dwo compile unit DIE. Output skeleton debug info section only
if there is debug info.
Index: include/dwarf2.h
===
--- include/dwarf2.h(revision 185451)
+++ include/dwarf2.h(working copy)
@@ -547,6 +547,8 @@ enum dwarf_location_atom
DW_OP_GNU_uninit = 0xf0,
DW_OP_GNU_encoded_addr = 0xf1,
DW_OP_GNU_implicit_pointer = 0xf2,
+/* Extension for Fission. See http://gcc.gnu.org/wiki/DebugFission. */
+DW_OP_GNU_addr_index = 0xfb,
/* HP extensions. */
DW_OP_HP_unknown = 0xe0, /* Ouch, the same as GNU_push_tls_address. */
DW_OP_HP_is_value= 0xe1,
Index: gcc/dwarf2out.c
===
--- gcc/dwarf2out.c (revision 185451)
+++ gcc/dwarf2out.c (working copy)
@@ -4779,6 +4779,8 @@ dwarf_stack_op_name (unsigned int op)
return DW_OP_GNU_encoded_addr;
case DW_OP_GNU_implicit_pointer:
return DW_OP_GNU_implicit_pointer;
+case DW_OP_GNU_addr_index:
+ return DW_OP_GNU_addr_index;
default:
return OP_unknown;
@@ -4897,6 +4899,9 @@ size_of_loc_descr (dw_loc_descr_ref loc)
case DW_OP_addr:
size += DWARF2_ADDR_SIZE;
break;
+case DW_OP_GNU_addr_index:
+ size += size_of_uleb128 (loc-dw_loc_oprnd1.v.val_unsigned);
+ break;
case DW_OP_const1u:
case DW_OP_const1s:
size += 1;
@@ -5275,6 +5280,11 @@ output_loc_operands (dw_loc_descr_ref lo
}
break;
+case DW_OP_GNU_addr_index:
+ dw2_asm_output_data_uleb128 (loc-dw_loc_oprnd1.v.val_unsigned,
+ (address index));
+ break;
+
case DW_OP_GNU_implicit_pointer:
{
char label[MAX_ARTIFICIAL_LABEL_BYTES
@@ -5343,6 +5353,7 @@ output_loc_operands_raw (dw_loc_descr_re
switch (loc-dw_loc_opc)
{
case DW_OP_addr:
+case DW_OP_GNU_addr_index:
case DW_OP_implicit_value:
/* We cannot output addresses in .cfi_escape, only bytes. */
gcc_unreachable ();
@@ -6314,6 +6325,7 @@ static inline dw_loc_descr_ref AT_loc (d
static void add_AT_loc_list (dw_die_ref, enum dwarf_attribute,
dw_loc_list_ref);
static inline dw_loc_list_ref AT_loc_list (dw_attr_ref);
+static unsigned int add_addr_table_entry (dw_attr_node *);
static void add_AT_addr (dw_die_ref, enum dwarf_attribute, rtx);
static inline rtx AT_addr (dw_attr_ref);
static void add_AT_lbl_id (dw_die_ref, enum dwarf_attribute, const char *);
@@ -6571,6 +6583,31 @@ static bool generic_type_p (tree);
static void
Re: [patch] Fix non-standard Ada bootstrap failure on IA-64
On Mar 9, 2012, Eric Botcazou ebotca...@adacore.com wrote: It does that only in case the -g0 build would add the same locs to the table. Only the DEBUG_INSN_P setting_insn locs are there just in -g builds and not in -g0 ones. If that's really supposed to work like so, then this is the bug, because the non-legitimate expression is present only with -g and its location promoted. It could be the case that an alternate, legitimate representation of the same expression is used elsewhere, and at the point that goes into cselib, the previously debug-only entry should become a regular entry in the cselib table. Is this what you observe? If that's it, maybe we need to somehow mark debug-only locs in loc lists, and get some code other than VTA to skip the debug-only locs, or somethink like that. -- Alexandre Oliva, freedom fighterhttp://FSFLA.org/~lxoliva/ You must be the change you wish to see in the world. -- Gandhi Be Free! -- http://FSFLA.org/ FSF Latin America board member Free Software Evangelist Red Hat Brazil Compiler Engineer
