In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled.

  Simplify Consecutive Merge Store Candidate Search

  Now that address aliasing is much less conservative, push through
  simplified store merging search which only checks for parallel stores
  through the chain subgraph. This is cleaner as the separation of
  non-interfering loads/stores from the store-merging logic.

  Whem merging stores, search up the chain through a single load, and
  finds all possible stores by looking down from through a load and a
  TokenFactor to all stores visited. This improves the quality of the
  output SelectionDAG and generally the output CodeGen (with some
  exceptions).

  Additional Minor Changes:

    1. Finishes removing unused AliasLoad code
    2. Unifies the the chain aggregation in the merged stores across
       code paths
    3. Re-add the Store node to the worklist after calling
       SimplifyDemandedBits.
    4. Increase GatherAllAliasesMaxDepth from 6 to 18. That number is
       arbitrary, but seemed sufficient to not cause regressions in
       tests.

  This finishes the change Matt Arsenault started in r246307 and
  jyknight's original patch.

  Many tests required some changes as memory operations are now
  reorderable. Some tests relying on the order were changed to use
  volatile memory operations

  Noteworthy tests:

    CodeGen/AArch64/argument-blocks.ll -
      It's not entirely clear what the test_varargs_stackalign test is
      supposed to be asserting, but the new code looks right.

    CodeGen/AArch64/arm64-memset-inline.lli -
    CodeGen/AArch64/arm64-stur.ll -
    CodeGen/ARM/memset-inline.ll -
      The backend now generates *worse* code due to store merging
      succeeding, as we do do a 16-byte constant-zero store efficiently.

    CodeGen/AArch64/merge-store.ll -
      Improved, but there still seems to be an extraneous vector insert
      from an element to itself?

    CodeGen/PowerPC/ppc64-align-long-double.ll -
      Worse code emitted in this case, due to the improved store->load
      forwarding.

    CodeGen/X86/dag-merge-fast-accesses.ll -
    CodeGen/X86/MergeConsecutiveStores.ll -
    CodeGen/X86/stores-merging.ll -
    CodeGen/Mips/load-store-left-right.ll -
      Restored correct merging of non-aligned stores

    CodeGen/AMDGPU/promote-alloca-stored-pointer-value.ll -
      Improved. Correctly merges buffer_store_dword calls

    CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll -
      Improved. Sidesteps loading a stored value and merges two stores

    CodeGen/X86/pr18023.ll -
      This test has been removed, as it was asserting incorrect
      behavior. Non-volatile stores *CAN* be moved past volatile loads,
      and now are.

    CodeGen/X86/vector-idiv.ll -
    CodeGen/X86/vector-lzcnt-128.ll -
      It's basically impossible to tell what these tests are actually
      testing. But, looks like the code got better due to the memory
      operations being recognized as non-aliasing.

    CodeGen/X86/win32-eh.ll -
      Both loads of the securitycookie are now merged.

    CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot-compute.ll -
      This test appears to work but no longer exhibits the spill
      behavior.

Reviewers: arsenm, hfinkel, tstellarAMD, nhaehnle, jyknight

Subscribers: wdng, nhaehnle, nemanjai, arsenm, weimingz, niravd, RKSimon, aemerson, qcolombet, resistor, tstellarAMD, t.p.northover, spatel

Differential Revision: https://reviews.llvm.org/D14834

llvm-svn: 282600

Author: niravhdave

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -824,7 +824,7 @@ TargetLoweringBase::TargetLoweringBase(const TargetMachine &tm) : TM(tm) {
   MinFunctionAlignment = 0;
   PrefFunctionAlignment = 0;
   PrefLoopAlignment = 0;
-  GatherAllAliasesMaxDepth = 6;
+  GatherAllAliasesMaxDepth = 18;
   MinStackArgumentAlignment = 1;
   MinimumJumpTableEntries = 4;
   // TODO: the default will be switched to 0 in the next commit, along

llvm/lib/CodeGen/TargetLoweringBase.cpp

@@ -446,16 +446,6 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(const TargetMachine &TM,
   setSelectIsExpensive(false);
   PredictableSelectIsExpensive = false;
 
-  // We want to find all load dependencies for long chains of stores to enable
-  // merging into very wide vectors. The problem is with vectors with > 4
-  // elements. MergeConsecutiveStores will attempt to merge these because x8/x16
-  // vectors are a legal type, even though we have to split the loads
-  // usually. When we can more precisely specify load legality per address
-  // space, we should be able to make FindBetterChain/MergeConsecutiveStores
-  // smarter so that they can figure out what to do in 2 iterations without all
-  // N > 4 stores on the same chain.
-  GatherAllAliasesMaxDepth = 16;
-
   // FIXME: Need to really handle these.
   MaxStoresPerMemcpy  = 4096;
   MaxStoresPerMemmove = 4096;

llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp

@@ -62,7 +62,7 @@ define i64 @test_hfa_ignores_gprs([7 x float], [2 x float] %in, i64, i64 %res) {
 ; but should go in an 8-byte aligned slot.
 define void @test_varargs_stackalign() {
 ; CHECK-LABEL: test_varargs_stackalign:
-; CHECK-DARWINPCS: stp {{w[0-9]+}}, {{w[0-9]+}}, [sp, #16]
+; CHECK-DARWINPCS: str {{x[0-9]+}}, [sp, #16]
 
   call void(...) @callee([3 x float] undef, [2 x float] [float 1.0, float 2.0])
   ret void

llvm/test/CodeGen/AArch64/argument-blocks.ll

@@ -6,17 +6,13 @@
 define void @fn9(i32 %a1, i32 %a2, i32 %a3, i32 %a4, i32 %a5, i32 %a6, i32 %a7, i32 %a8, i32 %a9, ...) nounwind noinline ssp {
 ; CHECK-LABEL: fn9:
 ; 9th fixed argument
-; CHECK: ldr {{w[0-9]+}}, [sp, #64]
-; CHECK: add [[ARGS:x[0-9]+]], sp, #72
-; CHECK: add {{x[0-9]+}}, [[ARGS]], #8
+; CHECK: add x[[ADDR:[0-9]+]], sp, #72
 ; First vararg
-; CHECK: ldr {{w[0-9]+}}, [sp, #72]
-; CHECK: add {{x[0-9]+}}, {{x[0-9]+}}, #8
+; CHECK-DAG: ldr {{w[0-9]+}}, [sp, #72]
 ; Second vararg
-; CHECK: ldr {{w[0-9]+}}, [{{x[0-9]+}}]
-; CHECK: add {{x[0-9]+}}, {{x[0-9]+}}, #8
+; CHECK-DAG: ldr {{w[0-9]+}}, [x[[ADDR]]]
 ; Third vararg
-; CHECK: ldr {{w[0-9]+}}, [{{x[0-9]+}}]
+; CHECK-DAG: ldr {{w[0-9]+}}, [x[[ADDR]]], #8
   %1 = alloca i32, align 4
   %2 = alloca i32, align 4
   %3 = alloca i32, align 4

llvm/test/CodeGen/AArch64/arm64-abi-varargs.ll

@@ -205,10 +205,7 @@ declare i32 @args_i32(i32, i32, i32, i32, i32, i32, i32, i32, i16 signext, i32,
 define i32 @test8(i32 %argc, i8** nocapture %argv) nounwind {
 entry:
 ; CHECK-LABEL: test8
-; CHECK: strb {{w[0-9]+}}, [sp, #3]
-; CHECK: strb wzr, [sp, #2]
-; CHECK: strb {{w[0-9]+}}, [sp, #1]
-; CHECK: strb wzr, [sp]
+; CHECK: str w8, [sp]
 ; CHECK: bl
 ; FAST-LABEL: test8
 ; FAST: strb {{w[0-9]+}}, [sp]

llvm/test/CodeGen/AArch64/arm64-abi.ll

@@ -9,11 +9,15 @@ entry:
   ret void
 }
 
+; FIXME: This shouldn't need to load in a zero value to store
+;        (e.g. stp xzr,xzr [sp, #16])
+
 define void @t2() nounwind ssp {
 entry:
 ; CHECK-LABEL: t2:
+; CHECK: movi v0.2d, #0000000000000000
+; CHECK: stur q0, [sp, #16]
 ; CHECK: strh wzr, [sp, #32]
-; CHECK: stp xzr, xzr, [sp, #16]
 ; CHECK: str xzr, [sp, #8]
   %buf = alloca [26 x i8], align 1
   %0 = getelementptr inbounds [26 x i8], [26 x i8]* %buf, i32 0, i32 0

llvm/test/CodeGen/AArch64/arm64-memset-inline.ll

@@ -47,11 +47,14 @@ define void @foo5(i8* %p, i32 %val) nounwind {
   ret void
 }
 
+;; FIXME: Again, with the writing of a quadword zero...
+
 define void @foo(%struct.X* nocapture %p) nounwind optsize ssp {
 ; CHECK-LABEL: foo:
 ; CHECK-NOT: str
-; CHECK: stur    xzr, [x0, #12]
-; CHECK-NEXT: stur    xzr, [x0, #4]
+; CHECK: stur    q0, [x0, #4]
+; CHECK-FIXME: stur    xzr, [x0, #12]
+; CHECK-FIXME-NEXT: stur    xzr, [x0, #4]
 ; CHECK-NEXT: ret
   %B = getelementptr inbounds %struct.X, %struct.X* %p, i64 0, i32 1
   %val = bitcast i64* %B to i8*

llvm/test/CodeGen/AArch64/arm64-stur.ll

@@ -4,8 +4,9 @@
 @g0 = external global <3 x float>, align 16
 @g1 = external global <3 x float>, align 4
 
-; CHECK: ldr s[[R0:[0-9]+]], {{\[}}[[R1:x[0-9]+]]{{\]}}, #4
-; CHECK: ld1{{\.?s?}} { v[[R0]]{{\.?s?}} }[1], {{\[}}[[R1]]{{\]}}
+; CHECK: ldr q[[R0:[0-9]+]], {{\[}}[[R1:x[0-9]+]], :lo12:g0
+;; TODO: this next line seems like a redundant no-op move?
+; CHECK: ins     v0.s[1], v0.s[1]
 ; CHECK: str d[[R0]]
 
 define void @blam() {

llvm/test/CodeGen/AArch64/merge-store.ll

@@ -1,5 +1,4 @@
-; RUN: llc --combiner-alias-analysis=false < %s | FileCheck %s
-; RUN: llc --combiner-alias-analysis=true  < %s | FileCheck %s
+; RUN: llc < %s | FileCheck %s
 
 ; This test checks that we do not merge stores together which have
 ; dependencies through their non-chain operands (e.g. one store is the

llvm/test/CodeGen/AArch64/vector_merge_dep_check.ll

@@ -88,12 +88,9 @@ define void @load_v4i8_to_v4f32_unaligned(<4 x float> addrspace(1)* noalias %out
 ; SI-DAG: v_cvt_f32_ubyte2_e32
 ; SI-DAG: v_cvt_f32_ubyte3_e32
 
-; SI-DAG: v_lshrrev_b32_e32 v{{[0-9]+}}, 24
-; SI-DAG: v_lshrrev_b32_e32 v{{[0-9]+}}, 16
 ; SI-DAG: v_lshlrev_b32_e32 v{{[0-9]+}}, 16
 ; SI-DAG: v_lshlrev_b32_e32 v{{[0-9]+}}, 8
 ; SI-DAG: v_and_b32_e32 v{{[0-9]+}}, 0xffff,
-; SI-DAG: v_and_b32_e32 v{{[0-9]+}}, 0xff00,
 ; SI-DAG: v_add_i32
 
 ; SI: buffer_store_dwordx4

llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll

@@ -1,13 +1,21 @@
-; RUN: llc -O0 -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=+amdgpu-debugger-insert-nops -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -O0 -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=+amdgpu-debugger-insert-nops -verify-machineinstrs < %s | FileCheck %s --check-prefix=CHECK
+; RUN: llc -O0 -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=+amdgpu-debugger-insert-nops -verify-machineinstrs < %s | FileCheck %s --check-prefix=CHECKNOP
 
-; CHECK: test01.cl:2:{{[0-9]+}}
-; CHECK-NEXT: s_nop 0
+; This test expects that we have one instance for each line in some order with "s_nop 0" instances after each.
 
-; CHECK: test01.cl:3:{{[0-9]+}}
-; CHECK-NEXT: s_nop 0
+; Check that each line appears at least once
+; CHECK-DAG: test01.cl:2:3
+; CHECK-DAG: test01.cl:3:3
+; CHECK-DAG: test01.cl:4:3
 
-; CHECK: test01.cl:4:{{[0-9]+}}
-; CHECK-NEXT: s_nop 0
+
+; Check that each of each of the lines consists of the line output, followed by "s_nop 0"
+; CHECKNOP: test01.cl:{{[234]}}:3
+; CHECKNOP-NEXT: s_nop 0
+; CHECKNOP: test01.cl:{{[234]}}:3
+; CHECKNOP-NEXT: s_nop 0
+; CHECKNOP: test01.cl:{{[234]}}:3
+; CHECKNOP-NEXT: s_nop 0
 
 ; CHECK: test01.cl:5:{{[0-9]+}}
 ; CHECK-NEXT: s_nop 0
@@ -21,7 +29,7 @@ entry:
   call void @llvm.dbg.declare(metadata i32 addrspace(1)** %A.addr, metadata !17, metadata !18), !dbg !19
   %0 = load i32 addrspace(1)*, i32 addrspace(1)** %A.addr, align 4, !dbg !20
   %arrayidx = getelementptr inbounds i32, i32 addrspace(1)* %0, i32 0, !dbg !20
-  store i32 1, i32 addrspace(1)* %arrayidx, align 4, !dbg !21
+  store i32 1, i32 addrspace(1)* %arrayidx, align 4, !dbg !20
   %1 = load i32 addrspace(1)*, i32 addrspace(1)** %A.addr, align 4, !dbg !22
   %arrayidx1 = getelementptr inbounds i32, i32 addrspace(1)* %1, i32 1, !dbg !22
   store i32 2, i32 addrspace(1)* %arrayidx1, align 4, !dbg !23

llvm/test/CodeGen/AMDGPU/debugger-insert-nops.ll

@@ -1,8 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-NOAA %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-NOAA %s
-
-; RUN: llc -march=amdgcn -verify-machineinstrs -combiner-alias-analysis -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -combiner-alias-analysis -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
+; RUN: llc -march=amdgcn -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
+; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
 
 ; This test is mostly to test DAG store merging, so disable the vectorizer.
 ; Run with devices with different unaligned load restrictions.
@@ -149,17 +146,10 @@ define void @merge_global_store_4_constants_f32(float addrspace(1)* %out) #0 {
   ret void
 }
 
-; FIXME: Should be able to merge this
 ; GCN-LABEL: {{^}}merge_global_store_4_constants_mixed_i32_f32:
-; GCN-NOAA: buffer_store_dword v
-; GCN-NOAA: buffer_store_dword v
-; GCN-NOAA: buffer_store_dword v
-; GCN-NOAA: buffer_store_dword v
-
-; GCN-AA: buffer_store_dwordx2
-; GCN-AA: buffer_store_dword v
-; GCN-AA: buffer_store_dword v
-
+; GCN: buffer_store_dwordx2
+; GCN: buffer_store_dword v
+; GCN: buffer_store_dword v
 ; GCN: s_endpgm
 define void @merge_global_store_4_constants_mixed_i32_f32(float addrspace(1)* %out) #0 {
   %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1
@@ -478,17 +468,9 @@ define void @merge_global_store_4_adjacent_loads_i8_natural_align(i8 addrspace(1
   ret void
 }
 
-; This works once AA is enabled on the subtarget
 ; GCN-LABEL: {{^}}merge_global_store_4_vector_elts_loads_v4i32:
 ; GCN: buffer_load_dwordx4 [[LOAD:v\[[0-9]+:[0-9]+\]]]
-
-; GCN-NOAA: buffer_store_dword v
-; GCN-NOAA: buffer_store_dword v
-; GCN-NOAA: buffer_store_dword v
-; GCN-NOAA: buffer_store_dword v
-
-; GCN-AA: buffer_store_dwordx4 [[LOAD]]
-
+; GCN: buffer_store_dwordx4 [[LOAD]]
 ; GCN: s_endpgm
 define void @merge_global_store_4_vector_elts_loads_v4i32(i32 addrspace(1)* %out, <4 x i32> addrspace(1)* %in) #0 {
   %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1

llvm/test/CodeGen/AMDGPU/merge-stores.ll

@@ -32,10 +32,10 @@
 ; HSA-ELT4-DAG: buffer_store_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:24{{$}}
 ; HSA-ELT4-DAG: buffer_store_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:28{{$}}
 
-; HSA-ELT4: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen{{$}}
-; HSA-ELT4: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:4{{$}}
-; HSA-ELT4: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:8{{$}}
-; HSA-ELT4: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:12{{$}}
+; HSA-ELT4-DAG: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen{{$}}
+; HSA-ELT4-DAG: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:4{{$}}
+; HSA-ELT4-DAG: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:8{{$}}
+; HSA-ELT4-DAG: buffer_load_dword {{v[0-9]+}}, v{{[0-9]+}}, s[0:3], s9 offen offset:12{{$}}
 define void @private_elt_size_v4i32(<4 x i32> addrspace(1)* %out, i32 addrspace(1)* %index.array) #0 {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()

llvm/test/CodeGen/AMDGPU/private-element-size.ll

@@ -156,9 +156,8 @@ define void @reorder_global_load_local_store_global_load(i32 addrspace(1)* %out,
 
 ; FUNC-LABEL: @reorder_local_offsets
 ; CI: ds_read2_b32 {{v\[[0-9]+:[0-9]+\]}}, {{v[0-9]+}} offset0:100 offset1:102
-; CI: ds_write2_b32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}} offset0:3 offset1:100
-; CI: ds_read_b32 {{v[0-9]+}}, {{v[0-9]+}} offset:12
-; CI: ds_write_b32 {{v[0-9]+}}, {{v[0-9]+}} offset:408
+; CI-DAG: ds_write2_b32 {{v[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset0:3 offset1:100
+; CI-DAG: ds_write_b32 {{v[0-9]+}}, {{v[0-9]+}} offset:408
 ; CI: buffer_store_dword
 ; CI: s_endpgm
 define void @reorder_local_offsets(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* noalias nocapture readnone %gptr, i32 addrspace(3)* noalias nocapture %ptr0) #0 {
@@ -180,12 +179,12 @@ define void @reorder_local_offsets(i32 addrspace(1)* nocapture %out, i32 addrspa
 }
 
 ; FUNC-LABEL: @reorder_global_offsets
-; CI: buffer_load_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:400
-; CI: buffer_load_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:408
-; CI: buffer_store_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:12
-; CI: buffer_store_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:400
-; CI: buffer_store_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:408
-; CI: buffer_load_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:12
+; CI-DAG: buffer_load_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:400
+; CI-DAG: buffer_load_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:408
+; CI-DAG: buffer_store_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:12
+; CI-DAG: buffer_store_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:400
+; CI-DAG: buffer_store_dword {{v[0-9]+}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:408
+; CI: buffer_store_dword
 ; CI: s_endpgm
 define void @reorder_global_offsets(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* noalias nocapture readnone %gptr, i32 addrspace(1)* noalias nocapture %ptr0) #0 {
   %ptr1 = getelementptr inbounds i32, i32 addrspace(1)* %ptr0, i32 3

llvm/test/CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll

@@ -3,6 +3,13 @@
 ; RUN: llc -march=amdgcn -mcpu=hawaii -mtriple=amdgcn-unknown-amdhsa -mattr=+vgpr-spilling -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CIHSA -check-prefix=HSA %s
 ; RUN: llc -march=amdgcn -mcpu=fiji -mtriple=amdgcn-unknown-amdhsa -mattr=+vgpr-spilling -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VIHSA -check-prefix=HSA %s
 
+; FIXME: this fails because the load generated from extractelement is
+;; now properly recognized as forwardable to the value stored in
+;; insertelement, and thus the loads/stores drop away entirely. This
+;; makes the intended test, of running out of registers, not occur.
+
+;; XFAIL: *
+
 ; This ends up using all 256 registers and requires register
 ; scavenging which will fail to find an unsued register.
 

llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot-compute.ll

@@ -1,6 +1,12 @@
 ; RUN: llc -march=amdgcn -mcpu=tahiti -mattr=+vgpr-spilling -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
 ; RUN: llc -march=amdgcn -mcpu=fiji -mattr=+vgpr-spilling -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
 
+;; FIXME: this fails because the load generated from extractelement is
+;; now properly recognized as forwardable to the value stored in
+;; insertelement, and thus the loads/stores drop away entirely. This
+;; makes the intended test, of running out of registers, not occur.
+;; XFAIL: *
+
 ; This ends up using all 255 registers and requires register
 ; scavenging which will fail to find an unsued register.
 

llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll

@@ -12,7 +12,8 @@ define void @test_byval_8_bytes_alignment(i32 %i, ...) {
 entry:
 ; CHECK: sub       sp, sp, #12
 ; CHECK: sub       sp, sp, #4
-; CHECK: stmib     sp, {r1, r2, r3}
+; CHECK: add       r0, sp, #4
+; CHECK: stm       sp, {r0, r1, r2, r3}
   %g = alloca i8*
   %g1 = bitcast i8** %g to i8*
   call void @llvm.va_start(i8* %g1)

llvm/test/CodeGen/ARM/2012-10-04-AAPCS-byval-align8.ll

@@ -51,12 +51,12 @@ entry:
 ; REALIGN: vld1.64 {{{d[0-9]+, d[0-9]+}}}, [r[[R2]]:128]
 
 
+; REALIGN: vst1.64 {{{d[0-9]+, d[0-9]+}}}, [r[[R1]]:128]
 ; REALIGN: orr r[[R2:[0-9]+]], r[[R1:[0-9]+]], #48
 ; REALIGN: vst1.64 {{{d[0-9]+, d[0-9]+}}}, [r[[R2]]:128]
 ; REALIGN: orr r[[R2:[0-9]+]], r[[R1]], #32
 ; REALIGN: vst1.64 {{{d[0-9]+, d[0-9]+}}}, [r[[R2]]:128]
-; REALIGN: orr r[[R2:[0-9]+]], r[[R1]], #16
-; REALIGN: vst1.64 {{{d[0-9]+, d[0-9]+}}}, [r[[R2]]:128]
+; REALIGN: orr r[[R1:[0-9]+]], r[[R1]], #16
 ; REALIGN: vst1.64 {{{d[0-9]+, d[0-9]+}}}, [r[[R1]]:128]
 
 ; REALIGN: add r[[R1:[0-9]+]], r[[R0:0]], #48

llvm/test/CodeGen/ARM/alloc-no-stack-realign.ll

@@ -9,8 +9,6 @@ entry:
 ; CHECK-LABEL: t:
 ; CHECK: vpop {d8}
 ; CHECK-NOT: vpopne
-; CHECK: pop {r7, pc}
-; CHECK: vpop {d8}
 ; CHECK: pop {r7, pc}
   br i1 undef, label %if.else, label %if.then
 

llvm/test/CodeGen/ARM/ifcvt10.ll

@@ -3,9 +3,15 @@
 define void @t1(i8* nocapture %c) nounwind optsize {
 entry:
 ; CHECK-LABEL: t1:
+
+;; FIXME: like with arm64-memset-inline.ll, learning how to merge
+;; stores made this code worse, since it now uses a vector move,
+;; instead of just using an strd instruction taking two registers.
+
+; CHECK: vmov.i32 d16, #0x0
+; CHECK: vst1.32 {d16}, [r0:64]!
 ; CHECK: movs r1, #0
-; CHECK: strd r1, r1, [r0]
-; CHECK: str r1, [r0, #8]
+; CHECK: str r1, [r0]
   call void @llvm.memset.p0i8.i64(i8* %c, i8 0, i64 12, i32 8, i1 false)
   ret void
 }

llvm/test/CodeGen/ARM/memset-inline.ll

@@ -6,9 +6,9 @@ define void @multiple_store() {
 ; CHECK: movs [[VAL:r[0-9]+]], #42
 ; CHECK: movt r[[BASE1]], #15
 
-; CHECK: str [[VAL]], [r[[BASE1]]]
-; CHECK: str [[VAL]], [r[[BASE1]], #24]
-; CHECK: str.w [[VAL]], [r[[BASE1]], #42]
+; CHECK-DAG: str [[VAL]], [r[[BASE1]]]
+; CHECK-DAG: str [[VAL]], [r[[BASE1]], #24]
+; CHECK-DAG: str.w [[VAL]], [r[[BASE1]], #42]
 
 ; CHECK: movw r[[BASE2:[0-9]+]], #20394
 ; CHECK: movt r[[BASE2]], #18