near the GOT, which new doesn't do. So break out the allocate into a new function.
Also move GOT index handling into JITResolver. This lets it update the mapping when a Lazy
function is JITed. It doesn't managed the table, just the mapping. Note that this is
still non-ideal, as any function that takes a function address should also take a GOT
index, but that is a lot of changes. The relocation resolve process updates any GOT entry
it sees is out of date.
llvm-svn: 22537
* Made bullet points start with a verb and lowercase, since they are not
complete sentences
* Cleaned up grammar, removed extraneous verbosity
llvm-svn: 22485
This is the first incremental patch to implement this feature. It adds no
functionality to LLVM but setup up the information needed from targets in
order to implement the optimization correctly. Each target needs to specify
the maximum number of store operations for conversion of the llvm.memset,
llvm.memcpy, and llvm.memmove intrinsics into a sequence of store operations.
The limit needs to be chosen at the threshold of performance for such an
optimization (generally smallish). The target also needs to specify whether
the target can support unaligned stores for multi-byte store operations.
This helps ensure the optimization doesn't generate code that will trap on
an alignment errors.
More patches to follow.
llvm-svn: 22468
for the command line options. This helps with situations where the executable
name sought is too generic and a more meaningful name needs to be used for
the command line options. It also helps satisfy picky project leaders.
llvm-svn: 22461
* FIND_STD_PROGRAM will find a program in the path or using --with options
and verify that the path/bin/program is executable. Also allows checking
for include files and libraries. If found, USE_PROGRAM is set, otherwise
its not set. Also sets PROGRAM_BIN (bin directory), and PROGRAM_DIR (top
level directory). If headers are found, sets PROGRAM_INC. If libraries
are found, sets PROGRAM_LIB.
* CHECK_PROGRAM_SANITY can be used to run a program with some option that
only produces information output and requires no input. If the output
matches a regular expression, the program passes the sanity check.
Otherwise, an error occurs.
llvm-svn: 22458
vector that represents the .o file at once, build up a vector for each
section of the .o file. This is needed because the .o file writer needs
to be able to switch between sections as it emits them (e.g. switch
between the .text section and the .rel section when emitting code).
This patch has no functionality change.
llvm-svn: 22453
legalizer to eliminate them. With this comes the expected code quality
improvements, such as, for this:
double foo(unsigned short X) { return X; }
we now generate this:
_foo:
subl $4, %esp
movzwl 8(%esp), %eax
movl %eax, (%esp)
fildl (%esp)
addl $4, %esp
ret
instead of this:
_foo:
subl $4, %esp
movw 8(%esp), %ax
movzwl %ax, %eax ;; Load not folded into this.
movl %eax, (%esp)
fildl (%esp)
addl $4, %esp
ret
-Chris
llvm-svn: 22449
SelectionDAGLegalize::ExpandLegalUINT_TO_FP method.
Add a new method, PromoteLegalUINT_TO_FP, which allows targets to request
that UINT_TO_FP operations be promoted to a larger input type. This is
useful for targets that have some UINT_TO_FP or SINT_TO_FP operations but
not all of them (like X86).
The same should be done with SINT_TO_FP, but this patch does not do that
yet.
llvm-svn: 22447
printed as part of the opcode. This allows something like
cmp${cc}ss in the x86 backed to be printed as cmpltss, cmpless, etc.
depending on what the value of $cc is.
llvm-svn: 22439
Add parenthesis around the value being negated; that way, if the value
begins with a minus sign (e.g. negative integer), we won't generate a
C predecrement operator by mistake.
llvm-svn: 22437
GRAPHVIZ will contain the path to the program if its found (or "echo Graphviz"
if not) and the #define HAVE_GRAPHVIZ will be defined if its found.
llvm-svn: 22424
It is not safe to call LegalizeOp on something that has already been legalized.
Instead, just force another iteration of legalization.
This could affect all platforms but X86, as this codepath is dynamically
dead on X86 (ISD::MEMSET and friends are legal).
llvm-svn: 22419
JIT can run against a multi-threaded program without getting its data
structures messed up. Also had to add the examples directory to the path
for the tests so that ParallelJIT can be found.
llvm-svn: 22415
Add an example program that utilizes multiple threads in the JIT to process
work. This was used by Evan Jones as the original test case for ensuring
that the ExecutionEngine was thread safe.
Original source by Evan Jones (adapted from other LLVM JIT examples) and
made LLVM style compliant by Reid Spencer.
llvm-svn: 22411
This patch completes the changes for making lli thread-safe. Here's the list
of changes:
* The Support/ThreadSupport* files were removed and replaced with the
MutexGuard.h file since all ThreadSupport* declared was a Mutex Guard.
The implementation of MutexGuard.h is now based on sys::Mutex which hides
its implementation and makes it unnecessary to have the -NoSupport.h and
-PThreads.h versions of ThreadSupport.
* All places in ExecutionEngine that previously referred to "Mutex" now
refer to sys::Mutex
* All places in ExecutionEngine that previously referred to "MutexLocker"
now refer to MutexGuard (this is frivolous but I believe the technically
correct name for such a class is "Guard" not a "Locker").
These changes passed all of llvm-test. All we need now are some test cases
that actually use multiple threads.
llvm-svn: 22404
Add a Mutex class for thread synchronization in a platform-independent way.
The current implementation only supports pthreads. Win32 use of Critical
Sections will be added later. The design permits other threading models to
be used if (and only if) pthreads is not available.
llvm-svn: 22403
Make sure that -lpthread gets added to LIBS variable which puts it at the
end of the tools' link commands, if libpthread.a is found.
Add a test for pthread.h so we can use #ifdef HAVE_PTHREAD_H
llvm-svn: 22401
Implement the X86 Subtarget.
This consolidates the checks for target triple, and setting options based
on target triple into one place. This allows us to convert the asm printer
and isel over from being littered with "forDarwin", "forCygwin", etc. into
just having the appropriate flags for each subtarget feature controlling
the code for that feature.
This patch also implements indirect external and weak references in the
X86 pattern isel, for darwin. Next up is to convert over the asm printers
to use this new interface.
llvm-svn: 22389
Instead of emitting a JIT stub that looks like this:
internal void %l1_main_entry_2E_ce_wrapper(int) {
header:
%resolver = call sbyte* %getPointerToNamedFunction( sbyte* getelementptr ([20 x sbyte]* %l1_main_entry_2E_ce_name, int 0, int 0) ) ; <sbyte*> [#uses=1]
%resolverCast = cast sbyte* %resolver to void (int)* ; <void (int)*> [#uses=1]
call void %resolverCast( int %0 )
ret void
}
Emit one that looks like this:
internal void %l1_main_entry_2E_ce_wrapper(int) {
Entry:
%fpcache = load void (int)** %l1_main_entry_2E_ce.fpcache ; <void (int)*> [#uses=2]
%isNull = seteq void (int)* %fpcache, null ; <bool> [#uses=1]
br bool %isNull, label %lookupfp, label %usecache
usecache: ; preds = %lookupfp, %Entry
%fp = phi void (int)* [ %resolverCast, %lookupfp ], [ %fpcache, %Entry ] ; <void (int)*> [#uses=1]
call void %fp( int %0 )
ret void
lookupfp: ; preds = %Entry
%resolver = call sbyte* %getPointerToNamedFunction( sbyte* getelementptr ([20 x sbyte]* %l1_main_entry_2E_ce_name, int 0, int 0) ) ; <sbyte*> [#uses=1]
%resolverCast = cast sbyte* %resolver to void (int)* ; <void (int)*> [#uses=2]
store void (int)* %resolverCast, void (int)** %l1_main_entry_2E_ce.fpcache
br label %usecache
}
This makes the JIT debugger *MUCH* faster on large programs, as
getPointerToNamedFunction takes time linear with the size of the program, and
before we would call it every time a function in the text module was called from
the safe module (ouch!).
llvm-svn: 22387
allows objdump to know which function we are emitting to:
00000000 <foo>: <----
0: b8 01 00 00 00 mov $0x1,%eax
5: 03 44 24 04 add 0x4(%esp,1),%eax
9: c3 ret
... and allows .o files to be useful for linking :)
llvm-svn: 22378
Add a *VERY INITIAL* machine code emitter class. This is enough to take
this C function:
int foo(int X) { return X +1; }
and make objdump produce the following:
$ objdump -d t-llvm.o
t-llvm.o: file format elf32-i386
Disassembly of section .text:
00000000 <.text>:
0: b8 01 00 00 00 mov $0x1,%eax
5: 03 44 24 04 add 0x4(%esp,1),%eax
9: c3 ret
Anything using branches or refering to the constant pool or requiring
relocations will not work yet.
llvm-svn: 22375
This is the last MVTSDNode.
This allows us to eliminate a bunch of special case code for handling
MVTSDNodes.
Also, remove some uses of dyn_cast that should really be cast (which is
cheaper in a release build).
llvm-svn: 22368
MVTSDNode class. This class is used to provide an operand to operators
that require an extra type. We start by converting FP_ROUND_INREG and
SIGN_EXTEND_INREG over to using it.
llvm-svn: 22364
Chris Lattner