LDtocL, and other loads that roughly correspond to the TOC_ENTRY SDAG node,
represent loads from the TOC, which is invariant. As a result, these loads can
be hoisted out of loops, etc. In order to do this, we need to generate
GOT-style MMOs for TOC_ENTRY, which requires treating it as a legitimate memory
intrinsic node type. Once this is done, the MMO transfer is automatically
handled for TableGen-driven instruction selection, and for nodes generated
directly in PPCISelDAGToDAG, we need to transfer the MMOs manually.
Also, we were not transferring MMOs associated with pre-increment loads, so do
that too.
Lastly, this fixes an exposed bug where R30 was not added as a defined operand of
UpdateGBR.
This problem was highlighted by an example (used to generate the test case)
posted to llvmdev by Francois Pichet.
llvm-svn: 230553
We had somehow accumulated a few target-specific SDAG nodes dealing with PPC64
TOC access that were referenced only in TableGen patterns. The associated
(pseudo-)instructions are used, but are being generated directly. NFC.
llvm-svn: 230518
This adds support for the QPX vector instruction set, which is used by the
enhanced A2 cores on the IBM BG/Q supercomputers. QPX vectors are 256 bytes
wide, holding 4 double-precision floating-point values. Boolean values, modeled
here as <4 x i1> are actually also represented as floating-point values
(essentially { -1, 1 } for { false, true }). QPX shares many features with
Altivec and VSX, but is distinct from both of them. One major difference is
that, instead of adding completely-separate vector registers, QPX vector
registers are extensions of the scalar floating-point registers (lane 0 is the
corresponding scalar floating-point value). The operations supported on QPX
vectors mirrors that supported on the scalar floating-point values (with some
additional ones for permutations and logical/comparison operations).
I've been maintaining this support out-of-tree, as part of the bgclang project,
for several years. This is not the entire bgclang patch set, but is most of the
subset that can be cleanly integrated into LLVM proper at this time. Adding
this to the LLVM backend is part of my efforts to rebase bgclang to the current
LLVM trunk, but is independently useful (especially for codes that use LLVM as
a JIT in library form).
The assembler/disassembler test coverage is complete. The CodeGen test coverage
is not, but I've included some tests, and more will be added as follow-up work.
llvm-svn: 230413
See full discussion in http://reviews.llvm.org/D7491.
We now hide the add-immediate and call instructions together in a
separate pseudo-op, which is tagged to define GPR3 and clobber the
call-killed registers. The PPCTLSDynamicCall pass prior to RA now
expands this op into the two separate addi and call ops, with explicit
definitions of GPR3 on both instructions, and explicit clobbers on the
call instruction. The pass is now marked as requiring and preserving
the LiveIntervals and SlotIndexes analyses, and fixes these up after
the replacement sequences are introduced.
Self-hosting has been verified on LE P8 and BE P7 with various
optimization levels, etc. It has also been verified with the
--no-tls-optimize flag workaround removed.
llvm-svn: 228725
Unfortunately, even with the workaround of disabling the linker TLS
optimizations in Clang restored (which has already been done), this still
breaks self-hosting on my P7 machine (-O3 -DNDEBUG -mcpu=native).
Bill is currently working on an alternate implementation to address the TLS
issue in a way that also fully elides the linker bug (which, unfortunately,
this approach did not fully), so I'm reverting this now.
llvm-svn: 228460
This patch is a third attempt to properly handle the local-dynamic and
global-dynamic TLS models.
In my original implementation, calls to __tls_get_addr were hidden
from view until the asm-printer phase, at which point the underlying
branch-and-link instruction was created with proper relocations. This
mostly worked well, but I used some repellent techniques to ensure
that the TLS_GET_ADDR nodes at the SD and MI levels correctly received
input from GPR3 and produced output into GPR3. This proved to work
badly in the presence of multiple TLS variable accesses, with the
copies to and from GPR3 being scheduled incorrectly and generally
creating havoc.
In r221703, I addressed that problem by representing the calls to
__tls_get_addr as true calls during instruction lowering. This had
the advantage of removing all of the bad hacks and relying on the
existing call machinery to properly glue the copies in place. It
looked like this was going to be the right way to go.
However, as a side effect of the recent discovery of problems with
linker optimizations for TLS, we discovered cases of suboptimal code
generation with this strategy. The problem comes when tls_get_addr is
called for the same address, and there is a resulting CSE
opportunity. It turns out that in such cases MachineCSE will common
the addis/addi instructions that set up the input value to
tls_get_addr, but will not common the calls themselves. MachineCSE
does not have any machinery to common idempotent calls. This is
perfectly sensible, since presumably this would be done at the IR
level, and introducing calls in the back end isn't commonplace. In
any case, we end up with two calls to __tls_get_addr when one would
suffice, and that isn't good.
I presumed that the original design would have allowed commoning of
the machine-specific nodes that hid the __tls_get_addr calls, so as
suggested by Ulrich Weigand, I went back to that design and cleaned it
up so that the copies were properly held together by glue
nodes. However, it turned out that this didn't work either...the
presence of copies to physical registers kept the machine-specific
nodes from being commoned also.
All of which leads to the design presented here. This is a return to
the original design, except that no attempt is made to introduce
copies to and from GPR3 during instruction lowering. Virtual registers
are used until prior to register allocation. At that point, a special
pass is run that identifies the machine-specific nodes that hide the
tls_get_addr calls and introduces the copies to and from GPR3 around
them. The register allocator then coalesces these copies away. With
this design, MachineCSE succeeds in commoning tls_get_addr calls where
possible, and we get nice optimal code generation (better than GCC at
the moment, which does not common these calls).
One additional problem must be dealt with: After introducing the
mentions of the physical register GPR3, the aggressive anti-dependence
breaker sees opportunities to improve scheduling by selecting a
different register instead. Flags must be used on the instruction
descriptions to tell the anti-dependence breaker to keep its hands in
its pockets.
One thing missing from the original design was recording a definition
of the link register on the GET_TLS_ADDR nodes. Doing this was found
to be insufficient to force a stack frame to be created, which led to
looping behavior because two different LR values were stored at the
same address. This appears to have been an oversight in
PPCFrameLowering::determineFrameLayout(), which is repaired here.
Because MustSaveLR() returns true for calls to builtin_return_address,
this changed the expected behavior of
test/CodeGen/PowerPC/retaddr2.ll, which now stacks a frame but
formerly did not. I've fixed the test case to reflect this.
There are existing TLS tests to catch regressions; the checks in
test/CodeGen/PowerPC/tls-store2.ll proved to be too restrictive in the
face of instruction scheduling with these changes, so I fixed that
up.
I've added a new test case based on the PrettyStackTrace module that
demonstrated the original problem. This checks that we get correct
code generation and that CSE of the calls to __get_tls_addr has taken
place.
llvm-svn: 227976
isel is actually a cracked instruction on the P7/P8, and must start a dispatch
group. The scheduling model should reflect this so that we don't bunch too many
of them together when possible.
Thanks to Bill Schmidt and Pat Haugen for helping to sort this out.
llvm-svn: 227758
Function pointers under PPC64 ELFv1 (which is used on PPC64/Linux on the
POWER7, A2 and earlier cores) are really pointers to a function descriptor, a
structure with three pointers: the actual pointer to the code to which to jump,
the pointer to the TOC needed by the callee, and an environment pointer. We
used to chain these loads, and make them opaque to the rest of the optimizer,
so that they'd always occur directly before the call. This is not necessary,
and in fact, highly suboptimal on embedded cores. Once the function pointer is
known, the loads can be performed ahead of time; in fact, they can be hoisted
out of loops.
Now these function descriptors are almost always generated by the linker, and
thus the contents of the descriptors are invariant. As a result, by default,
we'll mark the associated loads as invariant (allowing them to be hoisted out
of loops). I've added a target feature to turn this off, however, just in case
someone needs that option (constructing an on-stack descriptor, casting it to a
function pointer, and then calling it cannot be well-defined C/C++ code, but I
can imagine some JIT-compilation system doing so).
Consider this simple test:
$ cat call.c
typedef void (*fp)();
void bar(fp x) {
for (int i = 0; i < 1600000000; ++i)
x();
}
$ cat main.c
typedef void (*fp)();
void bar(fp x);
void foo() {}
int main() {
bar(foo);
}
On the PPC A2 (the BG/Q supercomputer), marking the function-descriptor loads
as invariant brings the execution time down to ~8 seconds from ~32 seconds with
the loads in the loop.
The difference on the POWER7 is smaller. Compiling with:
gcc -std=c99 -O3 -mcpu=native call.c main.c : ~6 seconds [this is 4.8.2]
clang -O3 -mcpu=native call.c main.c : ~5.3 seconds
clang -O3 -mcpu=native call.c main.c -mno-invariant-function-descriptors : ~4 seconds
(looks like we'd benefit from additional loop unrolling here, as a first
guess, because this is faster with the extra loads)
The -mno-invariant-function-descriptors will be added to Clang shortly.
llvm-svn: 226207
Fill out our support for the floating-point status and control register
instructions (mcrfs and friends). As it turns out, these are necessary for
compiling src/test/harness_fp.h in TBB for PowerPC.
Thanks to Raf Schietekat for reporting the issue!
llvm-svn: 226070
Patch by Kit Barton.
Support for the ICBT instruction is currently present, but limited to
embedded processors. This change adds a new FeatureICBT that can be used
to identify whether the ICBT instruction is available on a specific processor.
Two new tests are added:
* Positive test to ensure the icbt instruction is present when using
-mcpu=pwr8
* Negative test to ensure the icbt instruction is not generated when
using -mcpu=pwr7
Both test cases use the Prefetch opcode in LLVM. They are based on the
ppc64-prefetch.ll test case.
llvm-svn: 226033
We really need a separate 64-bit version of this instruction so that it can be
marked as clobbering LR8 (instead of just LR). No change in functionality
(although the verifier might be slightly happier), however, it is required for
stackmap/patchpoint support. Thus, this will be covered by stackmap test cases
once those are added.
llvm-svn: 225804
Remove the README.txt entry regarding register allocation of CR logical ops,
and replace it with a FIXME in PPCInstrInfo.td. The text in the README.txt was
not really accurate, and thanks goes to Pat Haugen (and Bill Schmidt) from IBM
for clarifying what was intended and highlighting the relevant text in the ISA
specification.
llvm-svn: 225325
Newer POWER cores, and the A2, support the cmpb instruction. This instruction
compares its operands, treating each of the 8 bytes in the GPRs separately,
returning a 'mask' result of 0 (for false) or -1 (for true) in each byte.
Code generation support is added, in the form of a PPCISelDAGToDAG
DAG-preprocessing routine, that recognizes patterns close to what the
instruction computes (either exactly, or related by a constant masking
operation), and generates the cmpb instruction (along with any necessary
constant masking operation). This can be expanded if use cases arise.
llvm-svn: 225106
On non-Darwin PPC64, the TOC reload needs to come directly after the bctrl
instruction (for indirect calls) because the 'bctrl/ld 2, 40(1)' instruction
sequence is interpreted by the unwinding code in libgcc. To make sure these
occur as a pair, as with other pairings interpreted by the linker, fuse the two
instructions into one instruction (for code generation only).
In the future, we might wish to do this by emitting CFI directives instead,
but this solution is simpler, and mirrors what GCC does. Additional discussion
on this point is contained in the PR.
Fixes PR22015.
llvm-svn: 224788
We've long supported readcyclecounter on PPC64, but it is easier there (the
read of the 64-bit time-base register can be accomplished via a single
instruction). This now provides an implementation for PPC32 as well. On PPC32,
the time-base register is still 64 bits, but can only be read 32 bits at a time
via two separate SPRs. The ISA manual explains how to do this properly (it
involves re-reading the upper bits and looping if the counter has wrapped while
being read).
This requires PPC to implement a custom integer splitting legalization for the
READCYCLECOUNTER node, turning it into a target-specific SDAG node, which then
gets turned into a pseudo-instruction, which is then expanded to the necessary
sequence (which has three SPR reads, the comparison and the branch).
Thanks to Paul Hargrove for pointing out to me that this was still unimplemented.
llvm-svn: 223161
Add assembler support for the fixed-point cache-inhibited load/store
instructions. These are hypervisor-level only, so don't get too excited ;)
Fixes PR21650.
llvm-svn: 222976
The attn instruction is not part of the Power ISA, but is documented in the A2
user manual, and is accepted by the GNU assembler for the A2 and the POWER4+.
Reported as part of PR21650.
llvm-svn: 222712
Summary:
Large-model was added first. With the addition of support for multiple PIC
models in LLVM, now add small-model PIC for 32-bit PowerPC, SysV4 ABI. This
generates more optimal code, for shared libraries with less than about 16380
data objects.
Test Plan: Test cases added or updated
Reviewers: joerg, hfinkel
Reviewed By: hfinkel
Subscribers: jholewinski, mcrosier, emaste, llvm-commits
Differential Revision: http://reviews.llvm.org/D5399
llvm-svn: 221791
My original support for the general dynamic and local dynamic TLS
models contained some fairly obtuse hacks to generate calls to
__tls_get_addr when lowering a TargetGlobalAddress. Rather than
generating real calls, special GET_TLS_ADDR nodes were used to wrap
the calls and only reveal them at assembly time. I attempted to
provide correct parameter and return values by chaining CopyToReg and
CopyFromReg nodes onto the GET_TLS_ADDR nodes, but this was also not
fully correct. Problems were seen with two back-to-back stores to TLS
variables, where the call sequences ended up overlapping with unhappy
results. Additionally, since these weren't real calls, the proper
register side effects of a call were not recorded, so clobbered values
were kept live across the calls.
The proper thing to do is to lower these into calls in the first
place. This is relatively straightforward; see the changes to
PPCTargetLowering::LowerGlobalTLSAddress() in PPCISelLowering.cpp.
The changes here are standard call lowering, except that we need to
track the fact that these calls will require a relocation. This is
done by adding a machine operand flag of MO_TLSLD or MO_TLSGD to the
TargetGlobalAddress operand that appears earlier in the sequence.
The calls to LowerCallTo() eventually find their way to
LowerCall_64SVR4() or LowerCall_32SVR4(), which call FinishCall(),
which calls PrepareCall(). In PrepareCall(), we detect the calls to
__tls_get_addr and immediately snag the TargetGlobalTLSAddress with
the annotated relocation information. This becomes an extra operand
on the call following the callee, which is expected for nodes of type
tlscall. We change the call opcode to CALL_TLS for this case. Back
in FinishCall(), we change it again to CALL_NOP_TLS for 64-bit only,
since we require a TOC-restore nop following the call for the 64-bit
ABIs.
During selection, patterns in PPCInstrInfo.td and PPCInstr64Bit.td
convert the CALL_TLS nodes into BL_TLS nodes, and convert the
CALL_NOP_TLS nodes into BL8_NOP_TLS nodes. This replaces the code
removed from PPCAsmPrinter.cpp, as the BL_TLS or BL8_NOP_TLS
nodes can now be emitted normally using their patterns and the
associated printTLSCall print method.
Finally, as a result of these changes, all references to get-tls-addr
in its various guises are no longer used, so they have been removed.
There are existing TLS tests to verify the changes haven't messed
anything up). I've added one new test that verifies that the problem
with the original code has been fixed.
llvm-svn: 221703
Summary:
hwsync is only required for seq_cst fences, acquire and release one can use
the cheaper lwsync.
Test Plan: Added some cases to atomics.ll + make check-all
Reviewers: jfb, wschmidt
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5317
llvm-svn: 218995
Older Book-E cores, such as the PPC 440, support only msync (which has the same
encoding as sync 0), but not any of the other sync forms. Newer Book-E cores,
however, do support sync, and for performance reasons we should allow the use
of the more-general form.
This refactors msync use into its own feature group so that it applies by
default only to older Book-E cores (of the relevant cores, we only have
definitions for the PPC440/450 currently).
llvm-svn: 218923
Summary:
Atomic loads and store of up to the native size (32 bits, or 64 for PPC64)
can be lowered to a simple load or store instruction (as the synchronization
is already handled by AtomicExpand, and the atomicity is guaranteed thanks to
the alignment requirements of atomic accesses). This is exactly what this patch
does. Previously, these were implemented by complex
load-linked/store-conditional loops.. an obvious performance problem.
For example, this patch turns
```
define void @store_i8_unordered(i8* %mem) {
store atomic i8 42, i8* %mem unordered, align 1
ret void
}
```
from
```
_store_i8_unordered: ; @store_i8_unordered
; BB#0:
rlwinm r2, r3, 3, 27, 28
li r4, 42
xori r5, r2, 24
rlwinm r2, r3, 0, 0, 29
li r3, 255
slw r4, r4, r5
slw r3, r3, r5
and r4, r4, r3
LBB4_1: ; =>This Inner Loop Header: Depth=1
lwarx r5, 0, r2
andc r5, r5, r3
or r5, r4, r5
stwcx. r5, 0, r2
bne cr0, LBB4_1
; BB#2:
blr
```
into
```
_store_i8_unordered: ; @store_i8_unordered
; BB#0:
li r2, 42
stb r2, 0(r3)
blr
```
which looks like a pretty clear win to me.
Test Plan:
fixed the tests + new test for indexed accesses + make check-all
Reviewers: jfb, wschmidt, hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5587
llvm-svn: 218922
Summary:
This patch makes use of AtomicExpandPass in Power for inserting fences around
atomic as part of an effort to remove fence insertion from SelectionDAGBuilder.
As a big bonus, it lets us use sync 1 (lightweight sync, often used by the mnemonic
lwsync) instead of sync 0 (heavyweight sync) in many cases.
I also added a test, as there was no test for the barriers emitted by the Power
backend for atomic loads and stores.
Test Plan: new test + make check-all
Reviewers: jfb
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D5180
llvm-svn: 218331
Adds code generation support for dcbtst (data cache prefetch for write) and
icbt (instruction cache prefetch for read - Book E cores only).
We still end up with a 'cannot select' error for the non-supported prefetch
intrinsic forms. This will be fixed in a later commit.
Fixes PR20692.
llvm-svn: 216339
Nemanja Ivanovic