separate tests for O2 and O3

Author: optimisan

llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll

@@ -1,7 +1,142 @@
-; RUN: llc -O3 -enable-new-pm -mtriple=amdgcn--amdhsa -disable-verify -print-pipeline-passes < %s 2>&1 \
+; RUN: llc -enable-new-pm -disable-verify -mtriple=amdgcn--amdhsa -print-pipeline-passes < %s 2>&1 \
+; RUN:   | FileCheck -check-prefix=GCN-O2 %s
+
+; RUN: llc -O3 -enable-new-pm -disable-verify -mtriple=amdgcn--amdhsa -print-pipeline-passes < %s 2>&1 \
 ; RUN:   | FileCheck -check-prefix=GCN-O3 %s
 
-; REQUIRES: asserts
+; GCN-O2: require<MachineModuleAnalysis>
+; GCN-O2: require<profile-summary>
+; GCN-O2: require<collector-metadata>
+; GCN-O2: pre-isel-intrinsic-lowering
+; GCN-O2: function(expand-large-div-rem
+; GCN-O2: expand-fp)
+; GCN-O2: amdgpu-remove-incompatible-functions
+; GCN-O2: amdgpu-printf-runtime-binding
+; GCN-O2: amdgpu-lower-ctor-dtor
+; GCN-O2: function(amdgpu-image-intrinsic-opt)
+; GCN-O2: expand-variadics
+; GCN-O2: amdgpu-always-inline
+; GCN-O2: always-inline
+; GCN-O2: amdgpu-export-kernel-runtime-handles
+; GCN-O2: amdgpu-sw-lower-lds
+; GCN-O2: amdgpu-lower-module-lds
+; GCN-O2: function(infer-address-spaces
+; GCN-O2: amdgpu-atomic-optimizer
+; GCN-O2: atomic-expand
+; GCN-O2: amdgpu-promote-alloca
+; GCN-O2: separate-const-offset-from-gep<>
+; GCN-O2: slsr
+; GCN-O2: early-cse<>
+; GCN-O2: nary-reassociate
+; GCN-O2: early-cse<>
+; GCN-O2: amdgpu-codegenprepare
+; GCN-O2: loop-mssa(loop-reduce)
+; GCN-O2: mergeicmps
+; GCN-O2: expand-memcmp
+; GCN-O2: gc-lowering
+; GCN-O2: lower-constant-intrinsics
+; GCN-O2: UnreachableBlockElimPass
+; GCN-O2: consthoist
+; GCN-O2: ReplaceWithVeclib
+; GCN-O2: partially-inline-libcalls
+; GCN-O2: ee-instrument<post-inline>
+; GCN-O2: scalarize-masked-mem-intrin
+; GCN-O2: ExpandReductionsPass
+; GCN-O2: early-cse<>
+; GCN-O2: amdgpu-lower-kernel-arguments)
+; GCN-O2: amdgpu-lower-buffer-fat-pointers
+; GCN-O2: cgscc(function(codegenprepare
+; GCN-O2: load-store-vectorizer
+; GCN-O2: lower-switch
+; GCN-O2: lower-invoke
+; GCN-O2: UnreachableBlockElimPass
+; GCN-O2: flatten-cfg
+; GCN-O2: sink
+; GCN-O2: amdgpu-late-codegenprepare
+; GCN-O2: amdgpu-unify-divergent-exit-nodes
+; GCN-O2: fix-irreducible
+; GCN-O2: unify-loop-exits
+; GCN-O2: StructurizeCFGPass
+; GCN-O2: amdgpu-annotate-uniform
+; GCN-O2: si-annotate-control-flow
+; GCN-O2: amdgpu-rewrite-undef-for-phi
+; GCN-O2: lcssa))
+; GCN-O2: amdgpu-perf-hint
+; GCN-O2: cgscc(function(require<uniformity>
+; GCN-O2: callbr-prepare
+; GCN-O2: safe-stack
+; GCN-O2: stack-protector))
+; GCN-O2: cgscc(function(machine-function(amdgpu-isel
+; GCN-O2: si-fix-sgpr-copies
+; GCN-O2: si-i1-copies
+; GCN-O2: finalize-isel
+; GCN-O2: early-tailduplication
+; GCN-O2: opt-phis
+; GCN-O2: stack-coloring
+; GCN-O2: localstackalloc
+; GCN-O2: dead-mi-elimination
+; GCN-O2: early-machinelicm
+; GCN-O2: machine-cse
+; GCN-O2: machine-sink
+; GCN-O2: peephole-opt
+; GCN-O2: dead-mi-elimination
+; GCN-O2: si-fold-operands
+; GCN-O2: gcn-dpp-combine
+; GCN-O2: si-load-store-opt
+; GCN-O2: si-peephole-sdwa
+; GCN-O2: early-machinelicm
+; GCN-O2: machine-cse
+; GCN-O2: si-fold-operands
+; GCN-O2: dead-mi-elimination
+; GCN-O2: si-shrink-instructions
+; GCN-O2: detect-dead-lanes
+; GCN-O2: InitUndefPass
+; GCN-O2: ProcessImplicitDefsPass
+; GCN-O2: unreachable-mbb-elimination
+; GCN-O2: require<live-vars>
+; GCN-O2: require<machine-loops>
+; GCN-O2: phi-node-elimination
+; GCN-O2: two-address-instruction
+; GCN-O2: register-coalescer
+; GCN-O2: rename-independent-subregs
+; GCN-O2: machine-scheduler
+; GCN-O2: greedy<all>
+; GCN-O2: amdgpu-nsa-reassign
+; GCN-O2: VirtRegRewriterPass
+; GCN-O2: stack-slot-coloring
+; GCN-O2: machine-cp
+; GCN-O2: machinelicm
+; GCN-O2: si-fix-vgpr-copies
+; GCN-O2: si-optimize-exec-masking
+; GCN-O2: remove-redundant-debug-values
+; GCN-O2: fixup-statepoint-caller-saved
+; GCN-O2: PostRAMachineSinkingPass
+; GCN-O2: ShrinkWrapPass
+; GCN-O2: PrologEpilogInserterPass
+; GCN-O2: branch-folder
+; GCN-O2: tailduplication
+; GCN-O2: machine-latecleanup
+; GCN-O2: machine-cp
+; GCN-O2: post-ra-pseudos
+; GCN-O2: postmisched
+; GCN-O2: block-placement
+; GCN-O2: fentry-insert
+; GCN-O2: xray-instrumentation
+; GCN-O2: patchable-function
+; GCN-O2: gcn-create-vopd
+; GCN-O2: si-memory-legalizer
+; GCN-O2: si-insert-waitcnts
+; GCN-O2: si-late-branch-lowering
+; GCN-O2: si-pre-emit-peephole
+; GCN-O2: post-RA-hazard-rec
+; GCN-O2: AMDGPUWaitSGPRHazardsPass
+; GCN-O2: amdgpu-insert-delay-alu
+; GCN-O2: branch-relaxation
+; GCN-O2: remove-loads-into-fake-uses
+; GCN-O2: live-debug-values
+; GCN-O2: machine-sanmd
+; GCN-O2: stack-frame-layout)
+; GCN-O2: invalidate<machine-function-info>))
 
 ; GCN-O3: require<MachineModuleAnalysis>
 ; GCN-O3: require<profile-summary>