RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 19:23:23 +01:00
Code Issues Projects Releases Wiki Activity
87f0cf2715
llvm-mirror/test/Other/opt-pipeline-vector-passes.ll

92 lines
3.8 KiB
LLVM
Raw Normal View History

[NewPM] Pin various tests under Other/ to legacy PM These all are legacy PM-specific or have a corresponding NPM RUN line. Reviewed By: ychen Differential Revision: https://reviews.llvm.org/D86124
2020-08-18 02:48:04 +02:00
; RUN: opt -enable-new-pm=0 -O1 -debug-pass=Structure < %s -o /dev/null 2>&1 | FileCheck %s --check-prefixes=OLDPM_O1
; RUN: opt -enable-new-pm=0 -O2 -debug-pass=Structure < %s -o /dev/null 2>&1 | FileCheck %s --check-prefixes=OLDPM_O2
; RUN: opt -enable-new-pm=0 -O2 -extra-vectorizer-passes -debug-pass=Structure < %s -o /dev/null 2>&1 | FileCheck %s --check-prefixes=OLDPM_O2_EXTRA
; RUN: opt -enable-new-pm=0 -O1 -vectorize-loops=0 -debug-pass=Structure < %s -o /dev/null 2>&1 | FileCheck %s --check-prefixes=OLDPM_O1_FORCE_OFF
; RUN: opt -enable-new-pm=0 -O2 -vectorize-loops=0 -debug-pass=Structure < %s -o /dev/null 2>&1 | FileCheck %s --check-prefixes=OLDPM_O2_FORCE_OFF
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
; RUN: opt -disable-verify -debug-pass-manager -passes='default<O1>' -S %s 2>&1 | FileCheck %s --check-prefixes=NEWPM_O1
; RUN: opt -disable-verify -debug-pass-manager -passes='default<O2>' -S %s 2>&1 | FileCheck %s --check-prefixes=NEWPM_O2
[NewPM] Add ExtraVectorizerPasses support As it looks like NewPM generally is using SimpleLoopUnswitch instead of LoopUnswitch, this patch also use SimpleLoopUnswitch in the ExtraVectorizerPasses sequence (compared with LegacyPM which use the LoopUnswitch pass). Reviewed By: aeubanks Differential Revision: https://reviews.llvm.org/D95457
2021-01-23 00:54:04 +01:00
; RUN: opt -disable-verify -debug-pass-manager -passes='default<O2>' -extra-vectorizer-passes -S %s 2>&1 | FileCheck %s --check-prefixes=NEWPM_O2_EXTRA
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
; REQUIRES: asserts
; SLP does not run at -O1. Loop vectorization runs, but it only
; works on loops explicitly annotated with pragmas.
; OLDPM_O1-LABEL: Pass Arguments:
; OLDPM_O1: Loop Vectorization
; OLDPM_O1-NOT: SLP Vectorizer
[VectorCombine] position pass after SLP in the optimization pipeline rather than before There are 2 known problem patterns shown in the test diffs here: vector horizontal ops (an x86 specialization) and vector reductions. SLP has greater ability to match and fold those than vector-combine, so let SLP have first chance at that. This is a quick fix while we continue to improve vector-combine and possibly canonicalize to reduction intrinsics. In the longer term, we should improve matching of these patterns because if they were created in the "bad" forms shown here, then we would miss optimizing them. I'm not sure what is happening with alias analysis on the addsub test. The old pass manager now shows an extra line for that, and we see an improvement that comes from SLP vectorizing a store. I don't know what's missing with the new pass manager to make that happen. Strangely, I can't reproduce the behavior if I compile from C++ with clang and invoke the new PM with "-fexperimental-new-pass-manager". Differential Revision: https://reviews.llvm.org/D80236
2020-05-22 18:13:18 +02:00
; OLDPM_O1: Optimize scalar/vector ops
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
; Everything runs at -O2.
; OLDPM_O2-LABEL: Pass Arguments:
; OLDPM_O2: Loop Vectorization
; OLDPM_O2: SLP Vectorizer
[VectorCombine] position pass after SLP in the optimization pipeline rather than before There are 2 known problem patterns shown in the test diffs here: vector horizontal ops (an x86 specialization) and vector reductions. SLP has greater ability to match and fold those than vector-combine, so let SLP have first chance at that. This is a quick fix while we continue to improve vector-combine and possibly canonicalize to reduction intrinsics. In the longer term, we should improve matching of these patterns because if they were created in the "bad" forms shown here, then we would miss optimizing them. I'm not sure what is happening with alias analysis on the addsub test. The old pass manager now shows an extra line for that, and we see an improvement that comes from SLP vectorizing a store. I don't know what's missing with the new pass manager to make that happen. Strangely, I can't reproduce the behavior if I compile from C++ with clang and invoke the new PM with "-fexperimental-new-pass-manager". Differential Revision: https://reviews.llvm.org/D80236
2020-05-22 18:13:18 +02:00
; OLDPM_O2: Optimize scalar/vector ops
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
[PassManager] restore early-cse to vector cleanup As noted in D80236 - the early-cse pass was included here before: D75145 / rG71a316883d50 But it got moved outside of the "extra" option there, then it got dropped while adjusting -vector-combine: rG6438ea45e053 rG57bb4787d72f So this is restoring the behavior and adding a test to prevent accidental changes again. I don't see an equivalent option for the new pass manager.
2020-06-14 15:50:10 +02:00
; Optionally run cleanup passes.
; OLDPM_O2_EXTRA-LABEL: Pass Arguments:
; OLDPM_O2_EXTRA: Loop Vectorization
; OLDPM_O2_EXTRA: Early CSE
; OLDPM_O2_EXTRA: Value Propagation
; OLDPM_O2_EXTRA: Combine redundant instructions
; OLDPM_O2_EXTRA: Loop Invariant Code Motion
; OLDPM_O2_EXTRA: Unswitch loops
; OLDPM_O2_EXTRA: Simplify the CFG
; OLDPM_O2_EXTRA: Combine redundant instructions
; OLDPM_O2_EXTRA: SLP Vectorizer
; OLDPM_O2_EXTRA: Early CSE
; OLDPM_O2_EXTRA: Optimize scalar/vector ops
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
; The loop vectorizer still runs at both -O1/-O2 even with the
; debug flag, but it only works on loops explicitly annotated
; with pragmas.
; OLDPM_O1_FORCE_OFF-LABEL: Pass Arguments:
; OLDPM_O1_FORCE_OFF: Loop Vectorization
; OLDPM_O1_FORCE_OFF-NOT: SLP Vectorizer
[VectorCombine] position pass after SLP in the optimization pipeline rather than before There are 2 known problem patterns shown in the test diffs here: vector horizontal ops (an x86 specialization) and vector reductions. SLP has greater ability to match and fold those than vector-combine, so let SLP have first chance at that. This is a quick fix while we continue to improve vector-combine and possibly canonicalize to reduction intrinsics. In the longer term, we should improve matching of these patterns because if they were created in the "bad" forms shown here, then we would miss optimizing them. I'm not sure what is happening with alias analysis on the addsub test. The old pass manager now shows an extra line for that, and we see an improvement that comes from SLP vectorizing a store. I don't know what's missing with the new pass manager to make that happen. Strangely, I can't reproduce the behavior if I compile from C++ with clang and invoke the new PM with "-fexperimental-new-pass-manager". Differential Revision: https://reviews.llvm.org/D80236
2020-05-22 18:13:18 +02:00
; OLDPM_O1_FORCE_OFF: Optimize scalar/vector ops
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
; OLDPM_O2_FORCE_OFF-LABEL: Pass Arguments:
; OLDPM_O2_FORCE_OFF: Loop Vectorization
; OLDPM_O2_FORCE_OFF: SLP Vectorizer
[VectorCombine] position pass after SLP in the optimization pipeline rather than before There are 2 known problem patterns shown in the test diffs here: vector horizontal ops (an x86 specialization) and vector reductions. SLP has greater ability to match and fold those than vector-combine, so let SLP have first chance at that. This is a quick fix while we continue to improve vector-combine and possibly canonicalize to reduction intrinsics. In the longer term, we should improve matching of these patterns because if they were created in the "bad" forms shown here, then we would miss optimizing them. I'm not sure what is happening with alias analysis on the addsub test. The old pass manager now shows an extra line for that, and we see an improvement that comes from SLP vectorizing a store. I don't know what's missing with the new pass manager to make that happen. Strangely, I can't reproduce the behavior if I compile from C++ with clang and invoke the new PM with "-fexperimental-new-pass-manager". Differential Revision: https://reviews.llvm.org/D80236
2020-05-22 18:13:18 +02:00
; OLDPM_O2_FORCE_OFF: Optimize scalar/vector ops
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
; There should be no difference with the new pass manager.
; This is tested more thoroughly in other test files.
; NEWPM_O1-LABEL: Running pass: LoopVectorizePass
; NEWPM_O1-NOT: Running pass: SLPVectorizerPass
[VectorCombine] position pass after SLP in the optimization pipeline rather than before There are 2 known problem patterns shown in the test diffs here: vector horizontal ops (an x86 specialization) and vector reductions. SLP has greater ability to match and fold those than vector-combine, so let SLP have first chance at that. This is a quick fix while we continue to improve vector-combine and possibly canonicalize to reduction intrinsics. In the longer term, we should improve matching of these patterns because if they were created in the "bad" forms shown here, then we would miss optimizing them. I'm not sure what is happening with alias analysis on the addsub test. The old pass manager now shows an extra line for that, and we see an improvement that comes from SLP vectorizing a store. I don't know what's missing with the new pass manager to make that happen. Strangely, I can't reproduce the behavior if I compile from C++ with clang and invoke the new PM with "-fexperimental-new-pass-manager". Differential Revision: https://reviews.llvm.org/D80236
2020-05-22 18:13:18 +02:00
; NEWPM_O1: Running pass: VectorCombinePass
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
; NEWPM_O2-LABEL: Running pass: LoopVectorizePass
; NEWPM_O2: Running pass: SLPVectorizerPass
[VectorCombine] position pass after SLP in the optimization pipeline rather than before There are 2 known problem patterns shown in the test diffs here: vector horizontal ops (an x86 specialization) and vector reductions. SLP has greater ability to match and fold those than vector-combine, so let SLP have first chance at that. This is a quick fix while we continue to improve vector-combine and possibly canonicalize to reduction intrinsics. In the longer term, we should improve matching of these patterns because if they were created in the "bad" forms shown here, then we would miss optimizing them. I'm not sure what is happening with alias analysis on the addsub test. The old pass manager now shows an extra line for that, and we see an improvement that comes from SLP vectorizing a store. I don't know what's missing with the new pass manager to make that happen. Strangely, I can't reproduce the behavior if I compile from C++ with clang and invoke the new PM with "-fexperimental-new-pass-manager". Differential Revision: https://reviews.llvm.org/D80236
2020-05-22 18:13:18 +02:00
; NEWPM_O2: Running pass: VectorCombinePass
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
[NewPM] Add ExtraVectorizerPasses support As it looks like NewPM generally is using SimpleLoopUnswitch instead of LoopUnswitch, this patch also use SimpleLoopUnswitch in the ExtraVectorizerPasses sequence (compared with LegacyPM which use the LoopUnswitch pass). Reviewed By: aeubanks Differential Revision: https://reviews.llvm.org/D95457
2021-01-23 00:54:04 +01:00
; NEWPM_O2_EXTRA-LABEL: Running pass: LoopVectorizePass
; NEWPM_O2_EXTRA: Running pass: EarlyCSEPass
; NEWPM_O2_EXTRA: Running pass: CorrelatedValuePropagationPass
; NEWPM_O2_EXTRA: Running pass: InstCombinePass
; NEWPM_O2_EXTRA: Running pass: LICMPass
; NEWPM_O2_EXTRA: Running pass: SimpleLoopUnswitchPass
; NEWPM_O2_EXTRA: Running pass: SimplifyCFGPass
; NEWPM_O2_EXTRA: Running pass: InstCombinePass
; NEWPM_O2_EXTRA: Running pass: SLPVectorizerPass
; NEWPM_O2_EXTRA: Running pass: EarlyCSEPass
; NEWPM_O2_EXTRA: Running pass: VectorCombinePass
define i64 @f(i1 %cond) {
entry:
br label %loop
loop:
%i = phi i64 [ 0, %entry ], [ %inc, %loop ]
%inc = add i64 %i, 1
br i1 %cond, label %loop, label %exit
exit:
ret i64 %i
[PassManager] add tests for vector pass enabling; NFC
2020-02-28 23:49:22 +01:00
}
Reference in New Issue Copy Permalink