before:
4 assembler - Number of assembler layout and relaxation steps
78563 assembler - Number of emitted assembler fragments
8693904 assembler - Number of emitted object file bytes
271223 assembler - Number of evaluated fixups
330771677 assembler - Number of fragment layouts
5958 assembler - Number of relaxed instructions
2508361 mcexpr - Number of MCExpr evaluations
real 0m26.123s
user 0m25.694s
sys 0m0.388s
after:
4 assembler - Number of assembler layout and relaxation steps
78563 assembler - Number of emitted assembler fragments
8693904 assembler - Number of emitted object file bytes
271223 assembler - Number of evaluated fixups
231507 assembler - Number of fragment layouts
5958 assembler - Number of relaxed instructions
2508361 mcexpr - Number of MCExpr evaluations
real 0m2.500s
user 0m2.113s
sys 0m0.273s
And yes, the outputs are identical :-)
llvm-svn: 121207
vpush instructions to save / restore VFP / NEON registers like this:
vpush {d8,d10,d11}
vpop {d8,d10,d11}
vpush and vpop do not allow gaps in the register list.
rdar://8728956
llvm-svn: 121197
uses of the function's blocks with undef. This code isn't needed,
because BasicBlock's destructor handles such uses. Also, undef isn't
correct, since blockaddresses may still be used for comparisons
with null.
llvm-svn: 121170
zextOrTrunc(), and APSInt methods extend(), extOrTrunc() and new method
trunc(), to be const and to return a new value instead of modifying the
object in place.
llvm-svn: 121120
(if available) as we go so that we get simple constantexprs not insane ones.
This fixes the failure of clang/test/CodeGenCXX/virtual-base-ctor.cpp
that the previous iteration of this patch had.
llvm-svn: 121111
actuall addresses in a .o file, so it is better to let the MachO writer compute
it.
This is good for two reasons. First, areas that shouldn't care about
addresses now don't have access to it. Second, the layout of each section
is independent. I should use this in a subsequent commit to speed it up.
Most of the patch is just removing the section address computation. The two
interesting parts are the change on how we handle padding in the end
of sections and how MachO can get the address of a-b when a and b are in
different sections.
Since now the expression evaluation normally doesn't know the section address,
it will think that a-b needs relocation and let the MachO writer know. Once
it has computed the section addresses, it calls back the expression evaluation
with the section addresses to resolve these expressions.
The remaining problem is the handling of padding. Currently it will create
a special alignment fragment at the end. Since that fragment doesn't update
the alignment of the section, it needs the real address to be computed.
Since now the layout will not compute a-b with a and b in different sections,
the only effect that the special alignment fragment has is update the
address size of the section. This can also be done by the MachO writer.
llvm-svn: 121076
optimization.
Consider:
static void foo() {
A = alloca
...
}
static void bar() {
B = alloca
...
call foo();
}
void main() {
bar()
}
The inliner proceeds bottom up, but lets pretend it decides not to inline foo
into bar. When it gets to main, it inlines bar into main(), and says "hey, I
just inlined an alloca "B" into main, lets remember that. Then it keeps going
and finds that it now contains a call to foo. It decides to inline foo into
main, and says "hey, foo has an alloca A, and I have an alloca B from another
inlined call site, lets reuse it". The problem with this of course, is that
the lifetime of A and B are nested, not disjoint.
Unfortunately I can't create a reasonable testcase for this: the one in the
PR is both huge and extremely sensitive, because you minor tweaks end up
causing foo to get inlined into bar too early. We already have tests for the
basic alloca merging optimization and this does not break them.
llvm-svn: 120995
memcpy's like:
memcpy(A, B)
memcpy(A, C)
we cannot delete the first memcpy as dead if A and C might be aliases.
If so, we actually get:
memcpy(A, B)
memcpy(A, A)
which is not correct to transform into:
memcpy(A, A)
This patch was heavily influenced by Jakub Staszak's patch in PR8728, thanks
Jakub!
llvm-svn: 120974
difficult on current ARM implementations for a few reasons.
1. Even though a single vmla has latency that is one cycle shorter than a pair
of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause
additional pipeline stall. So it's frequently better to single codegen
vmul + vadd.
2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to
stall for 4 cycles. We need to schedule them apart.
3. A vmla followed vmla is a special case. Obvious issuing back to back RAW
vmla + vmla is very bad. But this isn't ideal either:
vmul
vadd
vmla
Instead, we want to expand the second vmla:
vmla
vmul
vadd
Even with the 4 cycle vmul stall, the second sequence is still 2 cycles
faster.
Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough
but it isn't the optimial solution. This patch attempts to make it possible to
use vmla / vmls in cases where it is profitable.
A. Add missing isel predicates which cause vmla to be codegen'ed.
B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to
compute a fmul and a fmla.
C. Add additional isel checks for vmla, avoid cases where vmla is feeding into
fp instructions (except for the #3 exceptional case).
D. Add ARM hazard recognizer to model the vmla / vmls hazards.
E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the
vmla / vmls will trigger one of the special hazards.
Work in progress, only A+B are enabled.
llvm-svn: 120960
Also add asserts that the indices are valid in InsertValueInst::init(). ExtractValueInst already asserts when constructed with invalid indices.
llvm-svn: 120956
Should have no functional change other than the order of two transformations that are mutually-exclusive and the exact formatting of debug output.
Internally, it now stores the ConstantInt*s as Constant*s, and actual undef values instead of nulls.
llvm-svn: 120946
result. This allows us to compile:
void *test12(long count) {
return new int[count];
}
into:
test12:
movl $4, %ecx
movq %rdi, %rax
mulq %rcx
movq $-1, %rdi
cmovnoq %rax, %rdi
jmp __Znam ## TAILCALL
instead of:
test12:
movl $4, %ecx
movq %rdi, %rax
mulq %rcx
seto %cl
testb %cl, %cl
movq $-1, %rdi
cmoveq %rax, %rdi
jmp __Znam
Of course it would be even better if the regalloc inverted the cmov to 'cmovoq',
which would eliminate the need for the 'movq %rdi, %rax'.
llvm-svn: 120936
backend that they were all implemented except umul. This one fell back
to the default implementation that did a hi/lo multiply and compared the
top. Fix this to check the overflow flag that the 'mul' instruction
sets, so we can avoid an explicit test. Now we compile:
void *func(long count) {
return new int[count];
}
into:
__Z4funcl: ## @_Z4funcl
movl $4, %ecx ## encoding: [0xb9,0x04,0x00,0x00,0x00]
movq %rdi, %rax ## encoding: [0x48,0x89,0xf8]
mulq %rcx ## encoding: [0x48,0xf7,0xe1]
seto %cl ## encoding: [0x0f,0x90,0xc1]
testb %cl, %cl ## encoding: [0x84,0xc9]
movq $-1, %rdi ## encoding: [0x48,0xc7,0xc7,0xff,0xff,0xff,0xff]
cmoveq %rax, %rdi ## encoding: [0x48,0x0f,0x44,0xf8]
jmp __Znam ## TAILCALL
instead of:
__Z4funcl: ## @_Z4funcl
movl $4, %ecx ## encoding: [0xb9,0x04,0x00,0x00,0x00]
movq %rdi, %rax ## encoding: [0x48,0x89,0xf8]
mulq %rcx ## encoding: [0x48,0xf7,0xe1]
testq %rdx, %rdx ## encoding: [0x48,0x85,0xd2]
movq $-1, %rdi ## encoding: [0x48,0xc7,0xc7,0xff,0xff,0xff,0xff]
cmoveq %rax, %rdi ## encoding: [0x48,0x0f,0x44,0xf8]
jmp __Znam ## TAILCALL
Other than the silly seto+test, this is using the o bit directly, so it's going in the right
direction.
llvm-svn: 120935
foo = a - b
.long foo
instead of just
.long a - b
First, on darwin9 64 bits the assembler produces the wrong result. Second,
if "a" is the end of the section all darwin assemblers (9, 10 and mc) will not
consider a - b to be a constant but will if the dummy foo is created.
Split how we handle these cases. The first one is something MC should take care
of. The second one has to be handled by the caller.
llvm-svn: 120889
doing that if the target is darwin10 or newer.
This fixes
*) Direct object emission was producing objects without the workaround on
darwin9.
*) Assembly printing was producing objects with the workaround on linux.
llvm-svn: 120866
darwin9 linker, what is needed to avoid it and where to get more information.
Also make the workaround simpler. Just the regular end_sequence we normally
create is more than 4 bytes.
Tested by building cctools and ld64 from darwin9 on a darwin10 system and using
those. I checked that I was able to reproduce the bootstrap failure when
the the workaround was disabled.
llvm-svn: 120854
The StrongPHIElimination pass did not work, and nobody has worked on it for two
years.
A rewrite is underway, so I am leaving this shell pass instead of deleting it
completely.
llvm-svn: 120830
editing of the current interval.
These methods may cause coalescing, there are corresponding set*Unchecked
methods for editing without coalescing. The non-coalescing methods are useful
for applying monotonic transforms to all keys or values in a map without
accidentally coalescing transformed and untransformed intervals.
llvm-svn: 120829
<MCInst 2251 <MCOperand Reg:70> <MCOperand Reg:66> <MCOperand Imm:0> <MCOperand Reg:0> <MCOperand Imm:14> <MCOperand Reg:0>>
Notice that the "reg" here is 0, which is an invalid register. Put a check in
the code for this to prevent crashing.
llvm-svn: 120766
contain only data. Handle them specially instead of using AddSectionToTheEnd.
This moves a hack from the generic assembler to the elf writer. It is also
a bit faster and should make other improvements easier.
llvm-svn: 120683
Scan the MachineFunction for DBG_VALUE instructions, and replace them with a
data structure similar to LiveIntervals. The live range of a DBG_VALUE is
determined by propagating it down the dominator tree until a new DBG_VALUE is
found. When a DBG_VALUE lives in a register, its live range is confined to the
live range of the register's value.
LiveDebugVariables runs before coalescing, so DBG_VALUEs are not artificially
extended when registers are joined.
The missing half will recreate DBG_VALUE instructions from the intervals when
register allocation is complete.
The pass is disabled by default. It can be enabled with the temporary command
line option -live-debug-variables.
llvm-svn: 120636
The user (i.e. whoever generated a call to the intrinsic in the first place) is
essentially asking for a particular instruction to be placed in the assembler.
If that instruction won't execute on the target machine, that's their problem
not ours. Two buildbots with processors that don't support SSE3 were barfing
on the apm.ll test in CodeGen/X86 because of this assertion.
llvm-svn: 120574
