Craig Topper 2f472871bc [X86] Add support for passing 'prefer-vector-width' function attribute into X86Subtarget and exposing via X86's getRegisterWidth TTI interface. ...

This will cause the vectorizers to do some limiting of the vector widths they create. This is not a strict limit. There are reasons I know of that the loop vectorizer will generate larger vectors for.

I've written this in such a way that the interface will only return a properly supported width(0/128/256/512) even if the attribute says something funny like 384 or 10.

This has been split from D41895 with the remainder in a follow up commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323015 91177308-0d34-0410-b5e6-96231b3b80d8

2018-01-20 00:26:08 +00:00

..

already-vectorized.ll

[Loop Vectorize] Added a separate metadata

2017-08-20 10:32:41 +00:00

assume.ll

…

avx1.ll

…

avx512.ll

[X86] Add support for passing 'prefer-vector-width' function attribute into X86Subtarget and exposing via X86's getRegisterWidth TTI interface.

2018-01-20 00:26:08 +00:00

consecutive-ptr-cg-bug.ll

Move Transforms/LoopVectorize/consecutive-ptr-cg-bug.ll into the X86 subdirectory

2017-12-16 05:10:20 +00:00

consecutive-ptr-uniforms.ll

Support arbitrary address space pointers in masked gather/scatter intrinsics.

2017-05-03 12:28:54 +00:00

constant-vector-operand.ll

…

conversion-cost.ll

…

cost-model.ll

[X86] Enable interleaved memory access by default

2016-10-20 21:04:31 +00:00

float-induction-x86.ll

[SimplifyCFG] use pass options and remove the latesimplifycfg pass

2017-10-28 18:43:07 +00:00

force-ifcvt.ll

…

fp32_to_uint32-cost-model.ll

…

fp64_to_uint32-cost-model.ll

Revert "r306473 - re-commit r306336: Enable vectorizer-maximize-bandwidth by default."

2017-07-01 03:24:09 +00:00

fp_to_sint8-cost-model.ll

…

funclet.ll

…

gather_scatter.ll

Support arbitrary address space pointers in masked gather/scatter intrinsics.

2017-05-03 12:28:54 +00:00

gather-cost.ll

Second attempt at r285517.

2016-10-31 13:17:31 +00:00

gather-vs-interleave.ll

[Loop Vectorizer] Cost-based decision for vectorization form of memory instruction.

2017-02-08 19:25:23 +00:00

gcc-examples.ll

Revert "r306473 - re-commit r306336: Enable vectorizer-maximize-bandwidth by default."

2017-07-01 03:24:09 +00:00

illegal-parallel-loop-uniform-write.ll

…

imprecise-through-phis.ll

…

int128_no_gather.ll

[X86] Remove PCOMMIT instruction support since Intel has deprecated this instruction with no plans to release products with it.

2017-02-08 05:45:39 +00:00

interleaving.ll

[X86] enable memory interleaving for X86\SLM arch.

2017-01-25 09:14:48 +00:00

lit.local.cfg

…

masked_load_store.ll

Revert "r306473 - re-commit r306336: Enable vectorizer-maximize-bandwidth by default."

2017-07-01 03:24:09 +00:00

max-mstore.ll

…

metadata-enable.ll

Changed basic cost of store operation on X86

2017-08-20 12:34:29 +00:00

min-trip-count-switch.ll

…

mul_slm_16bit.ll

[X86] fixing failed test in commit: r291657

2017-01-11 09:03:11 +00:00

no_fpmath_with_hotness.ll

Revert "r306473 - re-commit r306336: Enable vectorizer-maximize-bandwidth by default."

2017-07-01 03:24:09 +00:00

no_fpmath.ll

Revert "r306473 - re-commit r306336: Enable vectorizer-maximize-bandwidth by default."

2017-07-01 03:24:09 +00:00

no-vector.ll

…

parallel-loops-after-reg2mem.ll

…

parallel-loops.ll

…

powof2div.ll

…

pr34438.ll

LoopVectorize: MaxVF should not be larger than the loop trip count

2017-09-04 08:35:13 +00:00

propagate-metadata.ll

…

ptr-indvar-crash.ll

…

rauw-bug.ll

…

reduction-crash.ll

Revert "r306473 - re-commit r306336: Enable vectorizer-maximize-bandwidth by default."

2017-07-01 03:24:09 +00:00

reduction-small-size.ll

[LV] Ignore the cost of values that will not appear in the vectorized loop

2017-12-12 08:57:43 +00:00

reg-usage-debug.ll

[LoopVectorizer] When estimating reg usage, unused insts may "end" another use

2016-11-15 14:27:33 +00:00

reg-usage.ll

[LV] Avoid computing the register usage for default VF. NFC

2017-09-13 19:35:45 +00:00

register-assumption.ll

…

scatter_crash.ll

Support arbitrary address space pointers in masked gather/scatter intrinsics.

2017-05-03 12:28:54 +00:00

slm-no-vectorize.ll

llvm/test/Transforms/LoopVectorize/X86/slm-no-vectorize.ll: -debug is available in +Asserts.

2017-07-02 14:25:27 +00:00

small-size.ll

Improve profile-guided heuristics to use estimated trip count.

2017-06-19 18:48:58 +00:00

strided_load_cost.ll

Currently isLikelyComplexAddressComputation tries to figure out if the given stride seems to be 'complex' and need some extra cost for address computation handling.

2017-01-05 14:03:41 +00:00

struct-store.ll

…

svml-calls-finite.ll

[TLI] Add mapping for various '__<func>_finite' forms of the math routines to SVML routines

2017-05-12 22:11:26 +00:00

svml-calls.ll

…

tripcount.ll

…

uint64_to_fp64-cost-model.ll

…

uniform_load.ll

…

uniform-phi.ll

…

uniformshift.ll

…

unroll_selection.ll

…

unroll-pm.ll

…

unroll-small-loops.ll

…

veclib-calls.ll

…

vect.omp.force.ll

…

vect.omp.force.small-tc.ll

[LV] Optimize for size when vectorizing loops with tiny trip count

2017-06-30 08:02:35 +00:00

vector_max_bandwidth.ll

[LV] Clamp the VF to the trip count

2017-09-12 16:32:45 +00:00

vector_ptr_load_store.ll

…

vector-scalar-select-cost.ll

…

vectorization-remarks-loopid-dbg.ll

[Loop Vectorize] Added a separate metadata

2017-08-20 10:32:41 +00:00

vectorization-remarks-missed.ll

[Loop Vectorize] Added a separate metadata

2017-08-20 10:32:41 +00:00

vectorization-remarks-profitable.ll

[LV] These remark should have been missed remarks

2017-03-01 04:31:15 +00:00

vectorization-remarks.ll

[Loop Vectorize] Added a separate metadata

2017-08-20 10:32:41 +00:00

vectorize-only-for-real.ll

…

x86_fp80-vector-store.ll

…

x86-predication.ll

[LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe

2017-11-14 12:09:30 +00:00