[CGP] Enable select in complex addr mode

[skatkov]

Summary

Enable select instruction handling in complex addr modes.

[skatkov]

To enable select I needed to update one test. Added test creator for the review whether he is ok with this change.

[skatkov]

To make a review easier, the generated code before this patch:

.globl  BinarySearch            # -- Begin function BinarySearch
.p2align        4, 0x90
.type   BinarySearch,@function
BinarySearch:                           # @BinarySearch
.cfi_startproc
# %bb.0:                                # %entry
movl    (%rsi), %eax
jmp     .LBB4_2
.p2align        4, 0x90
.LBB4_1:                                # %while.body
                                #   in Loop: Header=BB4_2 Depth=1
movl    %ecx, %eax
leaq    16(%rdx), %rcx
addq    $8, %rdx
btl     %eax, %edi
cmovaeq %rcx, %rdx
movq    (%rdx), %rdx
.LBB4_2:                                # %while.body
                                # =>This Inner Loop Header: Depth=1
movl    (%rdx), %ecx
cmpl    %ecx, %eax
ja      .LBB4_1
# %bb.3:                                # %while.end
retq
.Lfunc_end4:
.size   BinarySearch, .Lfunc_end4-BinarySearch
.cfi_endproc

The generated code after this patch:

.globl  BinarySearch            # -- Begin function BinarySearch
.p2align        4, 0x90
.type   BinarySearch,@function
BinarySearch:                           # @BinarySearch
.cfi_startproc
# %bb.0:                                # %entry
movl    (%rsi), %eax
jmp     .LBB4_2
.p2align        4, 0x90
.LBB4_1:                                # %while.body
                                #   in Loop: Header=BB4_2 Depth=1
movl    %ecx, %eax
xorl    %ecx, %ecx
btl     %eax, %edi
setae   %cl
movq    8(%rdx,%rcx,8), %rdx
.LBB4_2:                                # %while.body
                                # =>This Inner Loop Header: Depth=1
movl    (%rdx), %ecx
cmpl    %ecx, %eax
ja      .LBB4_1
# %bb.3:                                # %while.end
retq
.Lfunc_end4:
.size   BinarySearch, .Lfunc_end4-BinarySearch
.cfi_endproc

[reames]

The code change here is a clear improvement. Given no response from the test case author, LGTM.

llvm/trunk/lib/CodeGen/CodeGenPrepare.cpp

static cl::opt<bool> DisableComplexAddrModes(
     cl::desc("Disables combining addressing modes with different parts "
              "in optimizeMemoryInst."));
 
 static cl::opt<bool>
 AddrSinkNewPhis("addr-sink-new-phis", cl::Hidden, cl::init(false),
                 cl::desc("Allow creation of Phis in Address sinking."));
 
 static cl::opt<bool>
-AddrSinkNewSelects("addr-sink-new-select", cl::Hidden, cl::init(false),
+AddrSinkNewSelects("addr-sink-new-select", cl::Hidden, cl::init(true),
                    cl::desc("Allow creation of selects in Address sinking."));
 
 static cl::opt<bool> AddrSinkCombineBaseReg(
     "addr-sink-combine-base-reg", cl::Hidden, cl::init(true),
     cl::desc("Allow combining of BaseReg field in Address sinking."));
 
 static cl::opt<bool> AddrSinkCombineBaseGV(
     "addr-sink-combine-base-gv", cl::Hidden, cl::init(true),

llvm/trunk/test/CodeGen/X86/x86-cmov-converter.ll

for.body:                                         ; preds = %for.body.preheader, %for.body
   %cmp2 = icmp sgt i32 %1, %t.014
   %.t.0 = select i1 %cmp2, i32 %1, i32 %t.014
   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
   %exitcond = icmp eq i64 %indvars.iv.next, %wide.trip.count
   br i1 %exitcond, label %for.cond.cleanup, label %for.body
 }
 
 ; CHECK-LABEL: BinarySearch
-; CHECK: cmov
+; CHECK: set
 
 define i32 @BinarySearch(i32 %Mask, %struct.Node* nocapture readonly %Curr, %struct.Node* nocapture readonly %Next) #0 {
 entry:
   %Val8 = getelementptr inbounds %struct.Node, %struct.Node* %Curr, i64 0, i32 0
   %0 = load i32, i32* %Val8, align 8
   %Val19 = getelementptr inbounds %struct.Node, %struct.Node* %Next, i64 0, i32 0
   %1 = load i32, i32* %Val19, align 8
   %cmp10 = icmp ugt i32 %0, %1