Summary:
Straight-line strength reduction (SLSR) is implemented in GCC but not yet in
LLVM. It has proven to effectively simplify statements derived from an unrolled
loop, and can potentially benefit many other cases too. For example,
LLVM unrolls
#pragma unroll
foo (int i = 0; i < 3; ++i) {
sum += foo((b + i) * s);
}
into
sum += foo(b * s);
sum += foo((b + 1) * s);
sum += foo((b + 2) * s);
However, no optimizations yet reduce the internal redundancy of the three
expressions:
b * s
(b + 1) * s
(b + 2) * s
With SLSR, LLVM can optimize these three expressions into:
t1 = b * s
t2 = t1 + s
t3 = t2 + s
This commit is only an initial step towards implementing a series of such
optimizations. I will implement more (see TODO in the file commentary) in the
near future. This optimization is enabled for the NVPTX backend for now.
However, I am more than happy to push it to the standard optimization pipeline
after more thorough performance tests.
Test Plan: test/StraightLineStrengthReduce/slsr.ll
Reviewers: eliben, HaoLiu, meheff, hfinkel, jholewinski, atrick
Reviewed By: jholewinski, atrick
Subscribers: karthikthecool, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D7310
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228016 91177308-0d34-0410-b5e6-96231b3b80d8
This patch detects consecutive vector loads using the existing
EltsFromConsecutiveLoads() logic. This fixes:
http://llvm.org/bugs/show_bug.cgi?id=22329
This patch effectively reverts the tablegen additions of D6492 /
http://reviews.llvm.org/rL224344 ...which in hindsight were a horrible hack.
The test cases that were added with that patch are simply modified to load
from varying offsets of a base pointer. These loads did not match the existing
tablegen patterns.
A happy side effect of doing this optimization earlier is that we can now fold
the load into a math op where possible; this is shown in some of the updated
checks in the test file.
Differential Revision: http://reviews.llvm.org/D7303
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228006 91177308-0d34-0410-b5e6-96231b3b80d8
This can happen when a REV instruction is commuted.
The trick is not to define the _vi versions of instructions, which has these
consequences:
- code generation will always fail if a pseudo cannot be lowered
(very useful to catch bugs where an unsupported instruction somehow makes
it to the printer)
- ability to query if a pseudo can be lowered, which is done in commuteOpcode
to prevent REV from commuting to non-REV on VI
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227990 91177308-0d34-0410-b5e6-96231b3b80d8
This fixes a hang when using an empty geometry shader.
v2: - don't add s_nop when followed by s_waitcnt
- comestic changes
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227986 91177308-0d34-0410-b5e6-96231b3b80d8
r224330 introduced a bug by misinterpreting the "FeatureVectorUAMem" bit.
The commit log says that change did not affect anything, but that's not correct.
That change allowed SSE instructions to have unaligned mem operands folded into
math ops, and that's not allowed in the default specification for any SSE variant.
The bug is exposed when compiling for an AVX-capable CPU that had this feature
flag but without enabling AVX codegen. Another mistake in r224330 was not adding
the feature flag to all AVX CPUs; the AMD chips were excluded.
This is part of the fix for PR22371 ( http://llvm.org/bugs/show_bug.cgi?id=22371 ).
This feature bit is SSE-specific, so I've renamed it to "FeatureSSEUnalignedMem".
Changed the existing test case for the feature bit to reflect the new name and
renamed the test file itself to better reflect the feature.
Added runs to fold-vex.ll to check for the failing codegen.
Note that the feature bit is not set by default on any CPU because it may require a
configuration register setting to enable the enhanced unaligned behavior.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227983 91177308-0d34-0410-b5e6-96231b3b80d8
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.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227976 91177308-0d34-0410-b5e6-96231b3b80d8
This test was checking for lack of a "movaps" (an aligned load)
rather than a "movups" (an unaligned load). It also included
a store which complicated the checking.
Add specific CPU runs to prevent subtarget feature flag overrides
from inhibiting this optimization.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227972 91177308-0d34-0410-b5e6-96231b3b80d8
Improve EXTRACT_VECTOR_ELT DAG combine to catch conversion patterns
between x86mmx and i32 with more layers of indirection.
Before:
movq2dq %mm0, %xmm0
movd %xmm0, %eax
After:
movd %mm0, %eax
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227969 91177308-0d34-0410-b5e6-96231b3b80d8
LLVM ToT produces poor MMX code compared to 3.5. However, part of the previous
functionality can be achieved by using -x86-experimental-vector-widening-legalization.
Add tests to be sure we don't regress again.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227869 91177308-0d34-0410-b5e6-96231b3b80d8
ObjectLinkingLayer.
There are a two of overloads for addObject, one of which transfers ownership of
the underlying buffer to OrcMCJITReplacement. This commit makes the ownership
transfering version pass ownership down to the ObjectLinkingLayer in order to
prevent the issue described in r227778.
I think this commit will fix the sanitizer bot failures that necessitated the
removal of the load-object-a.ll regression test in r227785, so I'm reinstating
that test.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227845 91177308-0d34-0410-b5e6-96231b3b80d8
described by integer constants. This is a bit ugly, but if the source
language allows arbitrary type casting, the debug info must follow suit.
For example:
void foo() {
float a;
*(int *)&a = 0;
}
For the curious: SROA replaces the float alloca with an i32 alloca, which
is then optimized away and described via dbg.value(i32 0, ...).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227827 91177308-0d34-0410-b5e6-96231b3b80d8
This is true for SI only. CI+ supports unaligned memory accesses,
but this requires driver support, so for now we disallow unaligned
accesses for all GCN targets.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227822 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
Previously it only avoided optimizing signed comparisons to 0.
Sometimes the DAGCombiner will optimize the unsigned comparisons
to 0 before it gets to the peephole pass, but sometimes it doesn't.
Fix for PR22373.
Test Plan: test/CodeGen/ARM/sub-cmp-peephole.ll
Reviewers: jfb, manmanren
Subscribers: aemerson, llvm-commits
Differential Revision: http://reviews.llvm.org/D7274
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227809 91177308-0d34-0410-b5e6-96231b3b80d8
The commit r225977 uncovered this bug. The problem was that the vectorizer tried to
read the second operand of an already deleted instruction.
The bug didn't show up before r225977 because the freed memory still contained a non-null pointer.
With r225977 deletion of instructions is delayed and the read operand pointer is always null.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227800 91177308-0d34-0410-b5e6-96231b3b80d8
The VSX store instructions were also picking up an implicit "may read" from the
default pattern, which was an intrinsic (and we don't currently have a way of
specifying write-only intrinsics).
This was causing MI verification to fail for VSX spill restores.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227759 91177308-0d34-0410-b5e6-96231b3b80d8
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.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227758 91177308-0d34-0410-b5e6-96231b3b80d8
This moves the transformation introduced in r223757 into a separate MI pass.
This allows it to cover many more cases (not only cases where there must be a
reserved call frame), and perform rudimentary call folding. It still doesn't
have a heuristic, so it is enabled only for optsize/minsize, with stack
alignment <= 8, where it ought to be a fairly clear win.
(Re-commit of r227728)
Differential Revision: http://reviews.llvm.org/D6789
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227752 91177308-0d34-0410-b5e6-96231b3b80d8
The TOC base pointer is passed in r2, and we normally reserve this register so
that we can depend on it being there. However, for leaf functions, and
specifically those leaf functions that don't do any TOC access of their own
(which is generally due to accessing the constant pool, using TLS, etc.),
we can treat r2 as an ordinary callee-saved register (it must be callee-saved
because, for local direct calls, the linker will not insert any save/restore
code).
The allocation order has been changed slightly for PPC64/ELF systems to put r2
at the end of the list (while leaving it near the beginning for Darwin systems
to prevent unnecessary output changes). While r2 is allocatable, using it still
requires spill/restore traffic, and thus comes at the end of the list.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227745 91177308-0d34-0410-b5e6-96231b3b80d8
This moves the transformation introduced in r223757 into a separate MI pass.
This allows it to cover many more cases (not only cases where there must be a
reserved call frame), and perform rudimentary call folding. It still doesn't
have a heuristic, so it is enabled only for optsize/minsize, with stack
alignment <= 8, where it ought to be a fairly clear win.
Differential Revision: http://reviews.llvm.org/D6789
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227728 91177308-0d34-0410-b5e6-96231b3b80d8
This should be sufficient to replace the initial (minor) function pass
pipeline in Clang with the new pass manager. I'll probably add an (off
by default) flag to do that just to ensure we can get extra testing.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227726 91177308-0d34-0410-b5e6-96231b3b80d8
I've added RUN lines both to the basic test for EarlyCSE and the
target-specific test, as this serves as a nice test that the TTI layer
in the new pass manager is in fact working well.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227725 91177308-0d34-0410-b5e6-96231b3b80d8
over declarations.
This is both quite unproductive and causes things to crash, for example
domtree would just assert.
I've added a declaration and a domtree run to the basic high-level tests
for the new pass manager.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227724 91177308-0d34-0410-b5e6-96231b3b80d8
produce it.
This adds a function to the TargetMachine that produces this analysis
via a callback for each function. This in turn faves the way to produce
a *different* TTI per-function with the correct subtarget cached.
I've also done the necessary wiring in the opt tool to thread the target
machine down and make it available to the pass registry so that we can
construct this analysis from a target machine when available.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227721 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
CUDA driver can unroll loops when jit-compiling PTX. To prevent CUDA
driver from unrolling a loop marked with llvm.loop.unroll.disable is not
unrolled by CUDA driver, we need to emit .pragma "nounroll" at the
header of that loop.
This patch also extracts getting unroll metadata from loop ID metadata
into a shared helper function.
Test Plan: test/CodeGen/NVPTX/nounroll.ll
Reviewers: eliben, meheff, jholewinski
Reviewed By: jholewinski
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D7041
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227703 91177308-0d34-0410-b5e6-96231b3b80d8
aggregate or scalar, the debug info needs to refer to the absolute offset
(relative to the entire variable) instead of storing the offset inside
the smaller aggregate.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227702 91177308-0d34-0410-b5e6-96231b3b80d8
This patch adds shuffle mask decodes for integer zero extends (pmovzx** and movq xmm,xmm) and scalar float/double loads/moves (movss/movsd).
Also adds shuffle mask decodes for integer loads (movd/movq).
Differential Revision: http://reviews.llvm.org/D7228
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227688 91177308-0d34-0410-b5e6-96231b3b80d8
