[TargetLowering] Replace Log2_32_Ceil with Log2_32 in SimplifySetCC ctpop combine.

[craig.topper]

Summary

This combine can look through (trunc (ctpop X)). When doing this
it tries to make sure the trunc doesn't lose any information
from the ctpop. It does this by checking that the truncated type
has more bits that Log2_32_Ceil of the ctpop type. The Ceil is
unnecessary and pessimizes non-power of 2 types.

For example, ctpop of i256 requires 9 bits to represent the max
value of 256. But ctpop of i255 only requires 8 bits to represent
the max result of 255. Log2_32_Ceil of 256 and 255 both return 8
while Log2_32 returns 8 for 256 and 7 for 255

The code with popcnt enabled is a regression for this test case,
but it does match what already happens with i256 truncated to i9.
Since power of 2 is more likely, I don't think it should block
this change.

[spatel]

LGTM - thanks for explaining it.

llvm/test/CodeGen/X86/ctpop-combine.ll

164–165

Add 'nounwind' attribute to avoid cfi noise?

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

if (auto *N1C = dyn_cast<ConstantSDNode>(N1.getNode())) {
 
     SDValue CTPOP = N0;
     // Look through truncs that don't change the value of a ctpop.
     if (N0.hasOneUse() && N0.getOpcode() == ISD::TRUNCATE)
       CTPOP = N0.getOperand(0);
 
     if (CTPOP.hasOneUse() && CTPOP.getOpcode() == ISD::CTPOP &&
         (N0 == CTPOP ||
-         N0.getValueSizeInBits() > Log2_32_Ceil(CTPOP.getValueSizeInBits()))) {
+         N0.getValueSizeInBits() > Log2_32(CTPOP.getValueSizeInBits()))) {
       EVT CTVT = CTPOP.getValueType();
       SDValue CTOp = CTPOP.getOperand(0);
 
       // (ctpop x) u< 2 -> (x & x-1) == 0
       // (ctpop x) u> 1 -> (x & x-1) != 0
       if ((Cond == ISD::SETULT && C1 == 2) || (Cond == ISD::SETUGT && C1 == 1)){
         SDValue NegOne = DAG.getAllOnesConstant(dl, CTVT);
         SDValue Add = DAG.getNode(ISD::ADD, dl, CTVT, CTOp, NegOne);

llvm/test/CodeGen/X86/ctpop-combine.ll

 ; NO-POPCOUNT-NEXT:    movzbl %cl, %eax
 ; NO-POPCOUNT-NEXT:    retq
   %count = tail call i64 @llvm.ctpop.i64(i64 %x)
   %cmp = icmp ne i64 %count, 1
   %conv = zext i1 %cmp to i32
   ret i32 %conv
 }
 
 define i1 @ctpop_trunc_non_power2(i255 %x) {
-; POPCOUNT-LABEL: ctpop_trunc_non_power2:
-; POPCOUNT:       # %bb.0:
-; POPCOUNT-NEXT:    movabsq $9223372036854775807, %rax # imm = 0x7FFFFFFFFFFFFFFF
-; POPCOUNT-NEXT:    andq %rcx, %rax
-; POPCOUNT-NEXT:    popcntq %rsi, %rcx
-; POPCOUNT-NEXT:    popcntq %rdi, %rsi
-; POPCOUNT-NEXT:    addl %ecx, %esi
-; POPCOUNT-NEXT:    popcntq %rax, %rax
-; POPCOUNT-NEXT:    popcntq %rdx, %rcx
-; POPCOUNT-NEXT:    addl %eax, %ecx
-; POPCOUNT-NEXT:    addl %esi, %ecx
-; POPCOUNT-NEXT:    cmpb $1, %cl
-; POPCOUNT-NEXT:    sete %al
-; POPCOUNT-NEXT:    retq
-;
-; NO-POPCOUNT-LABEL: ctpop_trunc_non_power2:
-; NO-POPCOUNT:       # %bb.0:
-; NO-POPCOUNT-NEXT:    movabsq $9223372036854775807, %rax # imm = 0x7FFFFFFFFFFFFFFF
-; NO-POPCOUNT-NEXT:    andq %rcx, %rax
-; NO-POPCOUNT-NEXT:    movq %rsi, %rcx
-; NO-POPCOUNT-NEXT:    shrq %rcx
-; NO-POPCOUNT-NEXT:    movabsq $6148914691236517205, %r8 # imm = 0x5555555555555555
-; NO-POPCOUNT-NEXT:    andq %r8, %rcx
-; NO-POPCOUNT-NEXT:    subq %rcx, %rsi
-; NO-POPCOUNT-NEXT:    movabsq $3689348814741910323, %r11 # imm = 0x3333333333333333
-; NO-POPCOUNT-NEXT:    movq %rsi, %rcx
-; NO-POPCOUNT-NEXT:    andq %r11, %rcx
-; NO-POPCOUNT-NEXT:    shrq $2, %rsi
-; NO-POPCOUNT-NEXT:    andq %r11, %rsi
-; NO-POPCOUNT-NEXT:    addq %rcx, %rsi
-; NO-POPCOUNT-NEXT:    movq %rsi, %rcx
-; NO-POPCOUNT-NEXT:    shrq $4, %rcx
-; NO-POPCOUNT-NEXT:    addq %rsi, %rcx
-; NO-POPCOUNT-NEXT:    movabsq $1085102592571150095, %r9 # imm = 0xF0F0F0F0F0F0F0F
-; NO-POPCOUNT-NEXT:    andq %r9, %rcx
-; NO-POPCOUNT-NEXT:    movabsq $72340172838076673, %r10 # imm = 0x101010101010101
-; NO-POPCOUNT-NEXT:    imulq %r10, %rcx
-; NO-POPCOUNT-NEXT:    shrq $56, %rcx
-; NO-POPCOUNT-NEXT:    movq %rdi, %rsi
-; NO-POPCOUNT-NEXT:    shrq %rsi
-; NO-POPCOUNT-NEXT:    andq %r8, %rsi
-; NO-POPCOUNT-NEXT:    subq %rsi, %rdi
-; NO-POPCOUNT-NEXT:    movq %rdi, %rsi
-; NO-POPCOUNT-NEXT:    andq %r11, %rsi
-; NO-POPCOUNT-NEXT:    shrq $2, %rdi
-; NO-POPCOUNT-NEXT:    andq %r11, %rdi
-; NO-POPCOUNT-NEXT:    addq %rsi, %rdi
-; NO-POPCOUNT-NEXT:    movq %rdi, %rsi
-; NO-POPCOUNT-NEXT:    shrq $4, %rsi
-; NO-POPCOUNT-NEXT:    addq %rdi, %rsi
-; NO-POPCOUNT-NEXT:    andq %r9, %rsi
-; NO-POPCOUNT-NEXT:    imulq %r10, %rsi
-; NO-POPCOUNT-NEXT:    shrq $56, %rsi
-; NO-POPCOUNT-NEXT:    addl %ecx, %esi
-; NO-POPCOUNT-NEXT:    movq %rax, %rcx
-; NO-POPCOUNT-NEXT:    shrq %rcx
-; NO-POPCOUNT-NEXT:    movabsq $1537228672809129301, %rdi # imm = 0x1555555555555555
-; NO-POPCOUNT-NEXT:    andq %rcx, %rdi
-; NO-POPCOUNT-NEXT:    subq %rdi, %rax
-; NO-POPCOUNT-NEXT:    movq %rax, %rcx
-; NO-POPCOUNT-NEXT:    andq %r11, %rcx
-; NO-POPCOUNT-NEXT:    shrq $2, %rax
-; NO-POPCOUNT-NEXT:    andq %r11, %rax
-; NO-POPCOUNT-NEXT:    addq %rcx, %rax
-; NO-POPCOUNT-NEXT:    movq %rax, %rcx
-; NO-POPCOUNT-NEXT:    shrq $4, %rcx
-; NO-POPCOUNT-NEXT:    addq %rax, %rcx
-; NO-POPCOUNT-NEXT:    andq %r9, %rcx
-; NO-POPCOUNT-NEXT:    imulq %r10, %rcx
-; NO-POPCOUNT-NEXT:    shrq $56, %rcx
-; NO-POPCOUNT-NEXT:    movq %rdx, %rax
-; NO-POPCOUNT-NEXT:    shrq %rax
-; NO-POPCOUNT-NEXT:    andq %r8, %rax
-; NO-POPCOUNT-NEXT:    subq %rax, %rdx
-; NO-POPCOUNT-NEXT:    movq %rdx, %rax
-; NO-POPCOUNT-NEXT:    andq %r11, %rax
-; NO-POPCOUNT-NEXT:    shrq $2, %rdx
-; NO-POPCOUNT-NEXT:    andq %r11, %rdx
-; NO-POPCOUNT-NEXT:    addq %rax, %rdx
-; NO-POPCOUNT-NEXT:    movq %rdx, %rax
-; NO-POPCOUNT-NEXT:    shrq $4, %rax
-; NO-POPCOUNT-NEXT:    addq %rdx, %rax
-; NO-POPCOUNT-NEXT:    andq %r9, %rax
-; NO-POPCOUNT-NEXT:    imulq %r10, %rax
-; NO-POPCOUNT-NEXT:    shrq $56, %rax
-; NO-POPCOUNT-NEXT:    addl %ecx, %eax
-; NO-POPCOUNT-NEXT:    addl %esi, %eax
-; NO-POPCOUNT-NEXT:    cmpb $1, %al
-; NO-POPCOUNT-NEXT:    sete %al
-; NO-POPCOUNT-NEXT:    retq
+; CHECK-LABEL: ctpop_trunc_non_power2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    pushq %rbx
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset %rbx, -16
+; CHECK-NEXT:    movabsq $9223372036854775807, %r8 # imm = 0x7FFFFFFFFFFFFFFF
+; CHECK-NEXT:    movq %rcx, %r9
+; CHECK-NEXT:    andq %r8, %r9
+; CHECK-NEXT:    movq %rdi, %r11
+; CHECK-NEXT:    addq $-1, %r11
+; CHECK-NEXT:    movq %rsi, %r10
+; CHECK-NEXT:    adcq $-1, %r10
+; CHECK-NEXT:    movq %rdx, %rax
+; CHECK-NEXT:    adcq $-1, %rax
+; CHECK-NEXT:    movq %rcx, %rbx
+; CHECK-NEXT:    adcq %r8, %rbx
+; CHECK-NEXT:    andq %rdi, %r11
+; CHECK-NEXT:    andq %rdx, %rax
+; CHECK-NEXT:    orq %r11, %rax
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    andq %r8, %rbx
+; CHECK-NEXT:    andq %rcx, %rbx
+; CHECK-NEXT:    orq %r10, %rbx
+; CHECK-NEXT:    orq %rax, %rbx
+; CHECK-NEXT:    sete %cl
+; CHECK-NEXT:    orq %rdx, %rdi
+; CHECK-NEXT:    orq %rsi, %r9
+; CHECK-NEXT:    orq %rdi, %r9
+; CHECK-NEXT:    setne %al
+; CHECK-NEXT:    andb %cl, %al
+; CHECK-NEXT:    popq %rbx
+; CHECK-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-NEXT:    retq
   %a = call i255 @llvm.ctpop.i255(i255 %x)
   %b = trunc i255 %a to i8 ; largest value from ctpop is 255, fits in 8 bits.
   %c = icmp eq i8 %b, 1
   ret i1 %c
 }
 declare i255 @llvm.ctpop.i255(i255)