Chris Lattner [Mon, 27 Apr 2009 07:35:58 +0000 (07:35 +0000)]
add an abbreviation for common PARM_VAR_DECL. All but 9 of the
parm var decls in leopard cocoa.h end up using this abbreviation,
which shrinks the bitcode file by about 50K: 7217736->7167120.
Before:
Block ID #12 (DECLS_BLOCK):
Num Instances: 1
Total Size: 2.23595e+07b/2.79494e+06B/698736W
% of file: 38.7233
Num SubBlocks: 0
Num Abbrevs: 0
Num Records: 139387
% Abbrev Recs: 0
After:
Block ID #12 (DECLS_BLOCK):
Num Instances: 1
Total Size: 2.02405e+07b/2.53006e+06B/632516W
% of file: 35.301
Num SubBlocks: 0
Num Abbrevs: 1
Num Records: 139387
% Abbrev Recs: 19.2902
Douglas Gregor [Mon, 27 Apr 2009 06:38:32 +0000 (06:38 +0000)]
Load most of the source manager's information lazily from the PCH
file. In particular, only eagerly load source location entries for
files and for the predefines buffer. Other buffers and
macro-instantiation source location entries are loaded lazily.
With the Cocoa-prefixed "Hello, World", we only load 815/26555 source
location entities. This halves the amount of user time we spend in
this "Hello, World" program with -fsyntax-only (down to .007s).
This optimization is part 1 of 2 for the source manager. This
eliminates most of the user time in loading a PCH file. We still spend
too much time initialize File structures (especially in the calls to
stat), so we need to either make the loading of source location
entries for files lazy or import the stat cache from the PTH
implementation.
Chris Lattner [Mon, 27 Apr 2009 01:46:12 +0000 (01:46 +0000)]
Change our silencing of C typedef redefinition handling to what we had
before r69391: typedef redefinition is an error by default, but if
*either* the old or new definition are from a system header, we silence
it.
Chris Lattner [Sun, 26 Apr 2009 22:26:21 +0000 (22:26 +0000)]
make llvm-bcanalyzer dump out PCH files symbolically. We should probably
eventually get an option to turn this off, but it is nice for looking at
statistics. For example, the types block now prints:
Block ID #11 (TYPES_BLOCK_ID):
Num Instances: 1
Total Size: 895100b/111888B/27971.9W
% of file: 1.55801
Num SubBlocks: 0
Num Abbrevs: 0
Num Records: 14899
% Abbrev Recs: 0
Douglas Gregor [Sun, 26 Apr 2009 22:20:50 +0000 (22:20 +0000)]
Some fixes for PCH (de-)serialization of Objective-C AST nodes:
- Deal with the Receiver/ClassInfo shared storage in ObjCMessageExpr
- Implement PCH support for ImplicitParamDecl
- Fix the handling of the body of an ObjCMethodDecl
- Several cast -> cast_or_null fixes
- Make Selector::getIdentifierInfoForSlot work for 1-argument, NULL
selectors.
- Make Selector::getAsString() work with NULL selectors.
- Fix the names of VisitObjCAtCatchStmt and VisitObjCAtFinallyStmt
in the PCH reader and writer; these were never getting called.
At this point, all of the pch-test tests pass for C and Objective-C.
Chris Lattner [Sun, 26 Apr 2009 18:22:24 +0000 (18:22 +0000)]
implement PR4077: [Linux kernel] inscrutable error on inline asm input/output constraint mismatch
Before we emitted:
$ clang t.c -S -m64
llvm: error: Unsupported asm: input constraint with a matching output constraint of incompatible type!
Now we produce:
$ clang t.c -S -m64
t.c:5:40: error: unsupported inline asm: input with type 'unsigned long' matching output with type 'int'
asm volatile("foo " : "=a" (a) :"0" (b));
~~~ ~^~
Douglas Gregor [Sun, 26 Apr 2009 03:52:11 +0000 (03:52 +0000)]
Simple little smoke-test script that tries to build PCH files and then
dump their contents for all of the compilable tests in Clang's
testsuite. All of the tests pass for C, but there are still many
failures for Objective-C.
Douglas Gregor [Sun, 26 Apr 2009 03:49:13 +0000 (03:49 +0000)]
When writing a PCH file, write multiple type and declaration blocks as
necessary and iterate until all types and declarations have been
written. This reduces the Cocoa.h PCH file size by about 4% (since we
don't write types we don't need), and fixes problems where writing a
declaration generates a new type.
This doesn't seem to have any impact on performance either way.
Douglas Gregor [Sun, 26 Apr 2009 02:02:08 +0000 (02:02 +0000)]
Add a new -ast-dump-full option that traverses the translation unit
declaration rather than printing through the HandleTopLevelDecl
action. Using this, one can deserialize an entire PCH file and dump
it.
Eli Friedman [Sun, 26 Apr 2009 01:30:08 +0000 (01:30 +0000)]
Correct the order of the parameters to CheckAssignmentConstraints in
cleanup attribute checking. The difference isn't normally visible, but it
can make a difference...
Daniel Dunbar [Sun, 26 Apr 2009 01:10:38 +0000 (01:10 +0000)]
Driver: Add -std-default= option.
- This can be used to supply a default value for -std=; the idea is
that this can be used in conjunction with CCC_ADD_ARGS or
QA_OVERRIDE_GCC3_OPTIONS to change the default without having to
modify the build system.
Douglas Gregor [Sun, 26 Apr 2009 00:37:38 +0000 (00:37 +0000)]
Another shot at switching PCH on by default, now that we've cleaned up some bugs and improved performance. Will be reverted after Mr. Speedy gets done with it
Anders Carlsson [Sun, 26 Apr 2009 00:34:20 +0000 (00:34 +0000)]
When calling the cleanup function specified by __attribute__((cleanup)), make sure to bitcast the argument so it has the same type as the first argument of the cleanup function. Fixes <rdar://problem/6827047>.
Douglas Gregor [Sun, 26 Apr 2009 00:07:37 +0000 (00:07 +0000)]
Don't read all of the records in the PCH file's preprocessor block,
most of which are ignored. Instead, move the __COUNTER__ value out to
a PCH-level record (since it is handled eagerly) and move the header
file information into the SourceManager block (which is also,
currently, loaded eagerly).
This results in another 17% performance improvement in the
Cocoa-prefixed "Hello, World" with PCH.
Douglas Gregor [Sat, 25 Apr 2009 23:30:02 +0000 (23:30 +0000)]
Lazily load the controlling macros for all of the headers known in the
PCH file. In the Cocoa-prefixed "Hello, World" benchmark, this takes
us from reading 503 identifiers down to 37 and from 470 macros down to
4. It also results in an 8% performance improvement.
Chris Lattner [Sat, 25 Apr 2009 23:19:45 +0000 (23:19 +0000)]
"This fixes message sends to super in a way that both works with real code and passes the test in the test suite. It also fixes a crash when using recent versions of GNU libobjc and compiling modules that do not contain any constant strings but do contain a declaration of the constant string class and possible some other corner cases (thanks to Pete French for providing me with a test case for that one)."
Chris Lattner [Sat, 25 Apr 2009 21:59:05 +0000 (21:59 +0000)]
fix PR4073 by making designated initializer checking code use
VerifyIntegerConstantExpression instead of isIntegerConstantExpr.
This makes it ext-warn but tolerate things that fold to a constant
but that are not valid i-c-e's.
There must be a bug in the i-c-e computation though, because it
doesn't catch this case even with pedantic.
This also switches the later code to use EvaluateAsInt which is
simpler and handles everything that evaluate does.
Douglas Gregor [Sat, 25 Apr 2009 21:21:38 +0000 (21:21 +0000)]
Optimize the loading of an identifier from a PCH file when given the
identifier's ID. In this case, we know where the identifier's entry is
located in the hash table (it starts right before the identifier
string itself), so skip the hash table lookup and read the entry
directly. The performance improvement here is, gain, hard to quantify,
but it's the right thing to do.
Douglas Gregor [Sat, 25 Apr 2009 21:04:17 +0000 (21:04 +0000)]
PCH optimization for the identifier table, where we separate
"interesting" identifiers (e.g., those where the IdentifierInfo has
some useful information) from "uninteresting" identifiers (where the
IdentifierInfo is just a name). This makes the hash table smaller (so
searching in it should be faster) and, when loading "uninteresting"
identifiers, we skip the lookup in the hash table.
PCH file size is slightly smaller than before (since we don't emit the
contents of the uninteresting IdentifierInfo structures). The
Cocoa.h-prefixed "Hello, World" doesn't show any speedup, although
we're getting to the point where system noise is a bit issue.
Douglas Gregor [Sat, 25 Apr 2009 20:21:25 +0000 (20:21 +0000)]
Start implementing the PTH IdentifierInfo-saving trick in PCH,
allocating IdentifierInfos with a pointer into the string data stored
in the PCH file rather than having an entry in the identifier table's
string map. However, we don't actually get these savings at the
moment, because we go through the IdentifierTable when loading
identifiers from the on-disk hash table.
This commit is for record-keeping purposes only. I'll be reverting
this change (and the PCH layout tweak that preceded it) because it
appears that implementing this optimization will collide with another,
future optimization to reduce the size of the on-disk hash table for
identifiers. That optimization is likely to provide more benefit (with
less voodoo).
Chris Lattner [Sat, 25 Apr 2009 18:38:18 +0000 (18:38 +0000)]
with the fixes for better invalid decl/type propagation, this code
is no longer needed: a function type and a function declarator are
always known to line up.
Douglas Gregor [Sat, 25 Apr 2009 18:35:21 +0000 (18:35 +0000)]
Write the declaration and type offset arrays into the bitstream as
blobs, so that we don't need to do any work to get these arrays into
memory at PCH load time.
This gives another 19% performance improvement to the Cocoa-prefixed
"Hello, World!".
Douglas Gregor [Sat, 25 Apr 2009 17:48:32 +0000 (17:48 +0000)]
Load the selector table lazily from the PCH file.
This results in a 10% speedup on the Cocoa-prefixed "Hello, World!",
all of which is (not surprisingly) user time. There was a tiny
reduction in the size of the PCH file for Cocoa.h, because certain
selectors aren't being written twice.
I'm using two new tricks here that I'd like to replicate elsewhere:
(1) The selectors not used in the global method pool are packed into
the blob after the global method pool's on-disk hash table and
stored as keys, so that all selectors are in the same blob.
(2) We record the offsets of each selector key when we write it into
the global method pool (or after it, in the same blob). The offset
table is written as a blob, so that we don't need to pack/unpack a
SmallVector with its contents.
Chris Lattner [Sat, 25 Apr 2009 08:47:54 +0000 (08:47 +0000)]
Change SemaType's "GetTypeForDeclarator" and "ConvertDeclSpecToType" to
always return a non-null QualType + error bit. This fixes a bunch of
cases that didn't check for null result (and could thus crash) and eliminates
some crappy code scattered throughout sema.
This also improves the diagnostics in the recursive struct case to eliminate
a bogus second error. It also cleans up the case added to function.c by forming
a proper function type even though the declarator is erroneous, allowing the
parameter to be added to the function. Before:
t.c:2:1: error: unknown type name 'unknown_type'
unknown_type f(void*P)
^
t.c:4:3: error: use of undeclared identifier 'P'
P+1;
^
After:
t.c:2:1: error: unknown type name 'unknown_type'
unknown_type f(void*P)
^
Chris Lattner [Sat, 25 Apr 2009 08:06:05 +0000 (08:06 +0000)]
This is a pretty big cleanup for how invalid decl/type are handle.
This gets rid of a bunch of random InvalidDecl bools in sema, changing
us to use the following approach:
1. When analyzing a declspec or declarator, if an error is found, we
set a bit in Declarator saying that it is invalid.
2. Once the Decl is created by sema, we immediately set the isInvalid
bit on it from what is in the declarator. From this point on, sema
consistently looks at and sets the bit on the decl.
This gives a very clear separation of concerns and simplifies a bunch
of code. In addition to this, this patch makes these changes:
1. it renames DeclSpec::getInvalidType() -> isInvalidType().
2. various "merge" functions no longer return bools: they just set the
invalid bit on the dest decl if invalid.
3. The ActOnTypedefDeclarator/ActOnFunctionDeclarator/ActOnVariableDeclarator
methods now set invalid on the decl returned instead of returning an
invalid bit byref.
4. In SemaType, refering to a typedef that was invalid now propagates the
bit into the resultant type. Stuff declared with the invalid typedef
will now be marked invalid.
5. Various methods like CheckVariableDeclaration now return void and set the
invalid bit on the decl they check.
There are a few minor changes to tests with this, but the only major bad
result is test/SemaCXX/constructor-recovery.cpp. I'll take a look at this
next.
Chris Lattner [Sat, 25 Apr 2009 06:12:16 +0000 (06:12 +0000)]
fix PR4049, a crash on invalid, by making sema install the right number of
parameters in a functiondecl, even if the decl is invalid and has a confusing
Declarator. On the testcase, we now emit one beautiful diagnostic:
t.c:2:1: error: unknown type name 'unknown_type'
unknown_type f(void*)
^
GCC 4.0 produces:
t.c:2: error: syntax error before ‘f’
t.c: In function ‘f’:
t.c:2: error: parameter name omitted
and GCC 4.2:
t.c:2: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘f’
Chris Lattner [Sat, 25 Apr 2009 06:03:53 +0000 (06:03 +0000)]
rename getNumParmVarDeclsFromType back to getNumParams(),
remove a special case that was apparently for typeof() and
generalize the code in SemaDecl that handles typedefs to
handle any sugar type (including typedef, typeof, etc).
Improve comment to make it more clear what is going on.
Chris Lattner [Sat, 25 Apr 2009 05:51:56 +0000 (05:51 +0000)]
in:
typedef void foo(void);
We get a typedef for a functiontypeproto with no arguments, not
one with one argument and type void. This means the code being
removed in SemaDecl is dead.
projects / clang / log