one-byte NOPs. If the processor actually executes those NOPs, as it sometimes
does with aligned bundling, this can have a performance impact. From my
micro-benchmarks run on my one machine, a 15-byte NOP followed by twelve
one-byte NOPs is about 20% worse than a 15 followed by a 12. This patch
changes NOP emission to emit as many 15-byte (the maximum) as possible followed
by at most one shorter NOP.
llvm-svn: 176464
* Only apply divide bypass optimization when not optimizing for size.
* Fixed bug caused by constant for 0 value of type Int32,
used dividend type to generate the constant instead.
* For atom x86-64 apply the divide bypass to use 16-bit divides instead of
64-bit divides when operand values are small enough.
* Added lit tests for 64-bit divide bypass.
Patch by Tyler Nowicki!
llvm-svn: 176442
The VDUP instruction source register doesn't allow a non-constant lane
index, so make sure we don't construct a ARM::VDUPLANE node asking it to
do so.
rdar://13328063
http://llvm.org/bugs/show_bug.cgi?id=13963
llvm-svn: 176413
This adds minimalistic support for PHI nodes to llvm.objectsize() evaluation
fingers crossed so that it does break clang boostrap again..
llvm-svn: 176408
This matters for example in following matrix multiply:
int **mmult(int rows, int cols, int **m1, int **m2, int **m3) {
int i, j, k, val;
for (i=0; i<rows; i++) {
for (j=0; j<cols; j++) {
val = 0;
for (k=0; k<cols; k++) {
val += m1[i][k] * m2[k][j];
}
m3[i][j] = val;
}
}
return(m3);
}
Taken from the test-suite benchmark Shootout.
We estimate the cost of the multiply to be 2 while we generate 9 instructions
for it and end up being quite a bit slower than the scalar version (48% on my
machine).
Also, properly differentiate between avx1 and avx2. On avx-1 we still split the
vector into 2 128bits and handle the subvector muls like above with 9
instructions.
Only on avx-2 will we have a cost of 9 for v4i64.
I changed the test case in test/Transforms/LoopVectorize/X86/avx1.ll to use an
add instead of a mul because with a mul we now no longer vectorize. I did
verify that the mul would be indeed more expensive when vectorized with 3
kernels:
for (i ...)
r += a[i] * 3;
for (i ...)
m1[i] = m1[i] * 3; // This matches the test case in avx1.ll
and a matrix multiply.
In each case the vectorized version was considerably slower.
radar://13304919
llvm-svn: 176403
The LoopVectorizer often runs multiple times on the same function due to inlining.
When this happens the loop vectorizer often vectorizes the same loops multiple times, increasing code size and adding unneeded branches.
With this patch, the vectorizer during vectorization puts metadata on scalar loops and marks them as 'already vectorized' so that it knows to ignore them when it sees them a second time.
PR14448.
llvm-svn: 176399
Fix the way resources are counted. I'm taking some time to cleanup the
way MachineScheduler handles in-order machine resources. Eventually
we'll need more PPC/Atom test cases in tree.
llvm-svn: 176390
The sys::fs::is_directory() check is unnecessary because, if the filename is
a directory, the function will fail anyway with the same error code returned.
Remove the check to avoid an unnecessary stat call.
Someone needs to review on windows and see if the check is necessary there or not.
llvm-svn: 176386
This patch eliminates the need to emit a constant move instruction when this
pattern is matched:
(select (setgt a, Constant), T, F)
The pattern above effectively turns into this:
(conditional-move (setlt a, Constant + 1), F, T)
llvm-svn: 176384
detail.
The was this test was written, it was relying on an implementation detail
(fixups) and hence was very brittle (relying, among other things, on the
exact ordering of statistics printed by MC).
The test was rewritten to check a more observable output difference. While it
doesn't cover 100% of the things the original test covered, it's a good
practice to write regression tests this way. If we want to check that
internal details and invariants hold, such tests should be expressed as unit
tests.
llvm-svn: 176377
The make (all) target takes care of creating lit configs and auto-generating
tests. The problem with the original 'lit.site.cfg' target is it's not
recursive and doesn't fully create everything necessary for testing
clang-tools-extra.
llvm-svn: 176374
This reduces the time actually spent doing string to ID conversion and shows a 10% improvement in compile time for a particularly bad case that involves ARM Neon intrinsics (these have many overloads).
Patch by Jean-Luc Duprat!
llvm-svn: 176365
- ISD::SHL/SRL/SRA must have either both scalar or both vector operands
but TLI.getShiftAmountTy() so far only return scalar type. As a
result, backend logic assuming that breaks.
- Rename the original TLI.getShiftAmountTy() to
TLI.getScalarShiftAmountTy() and re-define TLI.getShiftAmountTy() to
return target-specificed scalar type or the same vector type as the
1st operand.
- Fix most TICG logic assuming TLI.getShiftAmountTy() a simple scalar
type.
llvm-svn: 176364
dispatch code. As far as I can tell the thumb2 code is behaving as expected.
I was able to compile and run the associated test case for both arm and thumb1.
rdar://13066352
llvm-svn: 176363
v2: based on Michels patch, but now allows copying of all registers sizes.
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
llvm-svn: 176346
