[X86] logical AND and OR in if-conditionals can turn to multiple branch instructions (llvm#162041)

manik-muk · akadutta · commit 6c12d87c9660 · 2025-10-14T16:21:23.000-05:00
Addresses llvm#160612
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -3624,6 +3624,16 @@ X86TargetLowering::getJumpConditionMergingParams(Instruction::BinaryOps Opc,
       match(Lhs, m_SpecificICmp(ICmpInst::ICMP_EQ, m_Value(), m_Value())) &&
       match(Rhs, m_SpecificICmp(ICmpInst::ICMP_EQ, m_Value(), m_Value())))
     BaseCost += 1;
+
+  // For OR conditions with EQ comparisons, prefer splitting into branches
+  // (unless CCMP is available). OR+EQ cannot be optimized via bitwise ops,
+  // unlike OR+NE which becomes (P|Q)!=0. Similarly, don't split signed
+  // comparisons (SLT, SGT) that can be optimized.
+  if (BaseCost >= 0 && !Subtarget.hasCCMP() && Opc == Instruction::Or &&
+      match(Lhs, m_SpecificICmp(ICmpInst::ICMP_EQ, m_Value(), m_Value())) &&
+      match(Rhs, m_SpecificICmp(ICmpInst::ICMP_EQ, m_Value(), m_Value())))
+    return {-1, -1, -1};
+
   return {BaseCost, BrMergingLikelyBias.getValue(),
           BrMergingUnlikelyBias.getValue()};
 }
diff --git a/llvm/test/CodeGen/X86/2007-08-09-IllegalX86-64Asm.ll b/llvm/test/CodeGen/X86/2007-08-09-IllegalX86-64Asm.ll
@@ -44,7 +44,7 @@ define ptr @ubyte_divmod(ptr %a, ptr %b) {
 ; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
 ; CHECK-NEXT:    callq __ubyte_convert_to_ctype
 ; CHECK-NEXT:    testl %eax, %eax
-; CHECK-NEXT:    js LBB0_6
+; CHECK-NEXT:    js LBB0_4
 ; CHECK-NEXT:  ## %bb.1: ## %cond_next.i
 ; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
 ; CHECK-NEXT:    movq %rbx, %rdi
@@ -53,84 +53,81 @@ define ptr @ubyte_divmod(ptr %a, ptr %b) {
 ; CHECK-NEXT:    sarl $31, %ecx
 ; CHECK-NEXT:    andl %eax, %ecx
 ; CHECK-NEXT:    cmpl $-2, %ecx
-; CHECK-NEXT:    je LBB0_10
+; CHECK-NEXT:    je LBB0_8
 ; CHECK-NEXT:  ## %bb.2: ## %cond_next.i
 ; CHECK-NEXT:    cmpl $-1, %ecx
-; CHECK-NEXT:    jne LBB0_3
-; CHECK-NEXT:  LBB0_8: ## %bb4
+; CHECK-NEXT:    jne LBB0_6
+; CHECK-NEXT:  LBB0_3: ## %bb4
 ; CHECK-NEXT:    movq _PyArray_API@GOTPCREL(%rip), %rax
 ; CHECK-NEXT:    movq (%rax), %rax
 ; CHECK-NEXT:    movq 16(%rax), %rax
-; CHECK-NEXT:    jmp LBB0_9
-; CHECK-NEXT:  LBB0_6: ## %_ubyte_convert2_to_ctypes.exit
+; CHECK-NEXT:    jmp LBB0_10
+; CHECK-NEXT:  LBB0_4: ## %_ubyte_convert2_to_ctypes.exit
 ; CHECK-NEXT:    cmpl $-2, %eax
-; CHECK-NEXT:    je LBB0_10
-; CHECK-NEXT:  ## %bb.7: ## %_ubyte_convert2_to_ctypes.exit
-; CHECK-NEXT:    cmpl $-1, %eax
 ; CHECK-NEXT:    je LBB0_8
-; CHECK-NEXT:  LBB0_3: ## %bb35
+; CHECK-NEXT:  ## %bb.5: ## %_ubyte_convert2_to_ctypes.exit
+; CHECK-NEXT:    cmpl $-1, %eax
+; CHECK-NEXT:    je LBB0_3
+; CHECK-NEXT:  LBB0_6: ## %bb35
 ; CHECK-NEXT:    movq _PyUFunc_API@GOTPCREL(%rip), %r14
 ; CHECK-NEXT:    movq (%r14), %rax
 ; CHECK-NEXT:    callq *216(%rax)
 ; CHECK-NEXT:    movzbl {{[0-9]+}}(%rsp), %edx
 ; CHECK-NEXT:    testb %dl, %dl
-; CHECK-NEXT:    je LBB0_4
-; CHECK-NEXT:  ## %bb.12: ## %cond_false.i
-; CHECK-NEXT:    setne %dil
+; CHECK-NEXT:    je LBB0_11
+; CHECK-NEXT:  ## %bb.7: ## %cond_false.i
 ; CHECK-NEXT:    movzbl {{[0-9]+}}(%rsp), %esi
 ; CHECK-NEXT:    movzbl %sil, %ecx
 ; CHECK-NEXT:    movl %ecx, %eax
 ; CHECK-NEXT:    divb %dl
 ; CHECK-NEXT:    movl %eax, %r15d
 ; CHECK-NEXT:    testb %cl, %cl
-; CHECK-NEXT:    setne %al
-; CHECK-NEXT:    testb %dil, %al
-; CHECK-NEXT:    jne LBB0_5
-; CHECK-NEXT:  LBB0_13: ## %cond_true.i200
-; CHECK-NEXT:    testb %dl, %dl
-; CHECK-NEXT:    jne LBB0_15
-; CHECK-NEXT:  ## %bb.14: ## %cond_true14.i
-; CHECK-NEXT:    movl $4, %edi
-; CHECK-NEXT:    callq _feraiseexcept
-; CHECK-NEXT:  LBB0_15: ## %ubyte_ctype_remainder.exit
-; CHECK-NEXT:    xorl %ebx, %ebx
-; CHECK-NEXT:    jmp LBB0_16
-; CHECK-NEXT:  LBB0_10: ## %bb17
+; CHECK-NEXT:    jne LBB0_12
+; CHECK-NEXT:    jmp LBB0_14
+; CHECK-NEXT:  LBB0_8: ## %bb17
 ; CHECK-NEXT:    callq _PyErr_Occurred
 ; CHECK-NEXT:    testq %rax, %rax
-; CHECK-NEXT:    jne LBB0_23
-; CHECK-NEXT:  ## %bb.11: ## %cond_next
+; CHECK-NEXT:    jne LBB0_27
+; CHECK-NEXT:  ## %bb.9: ## %cond_next
 ; CHECK-NEXT:    movq _PyArray_API@GOTPCREL(%rip), %rax
 ; CHECK-NEXT:    movq (%rax), %rax
 ; CHECK-NEXT:    movq 80(%rax), %rax
-; CHECK-NEXT:  LBB0_9: ## %bb4
+; CHECK-NEXT:  LBB0_10: ## %bb4
 ; CHECK-NEXT:    movq 96(%rax), %rax
 ; CHECK-NEXT:    movq %r14, %rdi
 ; CHECK-NEXT:    movq %rbx, %rsi
 ; CHECK-NEXT:    callq *40(%rax)
-; CHECK-NEXT:    jmp LBB0_24
-; CHECK-NEXT:  LBB0_4: ## %cond_true.i
+; CHECK-NEXT:    jmp LBB0_28
+; CHECK-NEXT:  LBB0_11: ## %cond_true.i
 ; CHECK-NEXT:    movl $4, %edi
 ; CHECK-NEXT:    callq _feraiseexcept
 ; CHECK-NEXT:    movzbl {{[0-9]+}}(%rsp), %edx
 ; CHECK-NEXT:    movzbl {{[0-9]+}}(%rsp), %esi
+; CHECK-NEXT:    xorl %r15d, %r15d
 ; CHECK-NEXT:    testb %sil, %sil
-; CHECK-NEXT:    sete %al
+; CHECK-NEXT:    je LBB0_14
+; CHECK-NEXT:  LBB0_12: ## %cond_false.i
 ; CHECK-NEXT:    testb %dl, %dl
-; CHECK-NEXT:    sete %cl
-; CHECK-NEXT:    xorl %r15d, %r15d
-; CHECK-NEXT:    orb %al, %cl
-; CHECK-NEXT:    jne LBB0_13
-; CHECK-NEXT:  LBB0_5: ## %cond_next17.i
+; CHECK-NEXT:    je LBB0_14
+; CHECK-NEXT:  ## %bb.13: ## %cond_next17.i
 ; CHECK-NEXT:    movzbl %sil, %eax
 ; CHECK-NEXT:    divb %dl
 ; CHECK-NEXT:    movzbl %ah, %ebx
-; CHECK-NEXT:  LBB0_16: ## %ubyte_ctype_remainder.exit
+; CHECK-NEXT:    jmp LBB0_18
+; CHECK-NEXT:  LBB0_14: ## %cond_true.i200
+; CHECK-NEXT:    testb %dl, %dl
+; CHECK-NEXT:    jne LBB0_17
+; CHECK-NEXT:  ## %bb.16: ## %cond_true14.i
+; CHECK-NEXT:    movl $4, %edi
+; CHECK-NEXT:    callq _feraiseexcept
+; CHECK-NEXT:  LBB0_17: ## %ubyte_ctype_remainder.exit
+; CHECK-NEXT:    xorl %ebx, %ebx
+; CHECK-NEXT:  LBB0_18: ## %ubyte_ctype_remainder.exit
 ; CHECK-NEXT:    movq (%r14), %rax
 ; CHECK-NEXT:    callq *224(%rax)
 ; CHECK-NEXT:    testl %eax, %eax
-; CHECK-NEXT:    je LBB0_19
-; CHECK-NEXT:  ## %bb.17: ## %cond_true61
+; CHECK-NEXT:    je LBB0_21
+; CHECK-NEXT:  ## %bb.19: ## %cond_true61
 ; CHECK-NEXT:    movl %eax, %ebp
 ; CHECK-NEXT:    movq (%r14), %rax
 ; CHECK-NEXT:    movq _.str5@GOTPCREL(%rip), %rdi
@@ -139,8 +136,8 @@ define ptr @ubyte_divmod(ptr %a, ptr %b) {
 ; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rcx
 ; CHECK-NEXT:    callq *200(%rax)
 ; CHECK-NEXT:    testl %eax, %eax
-; CHECK-NEXT:    js LBB0_23
-; CHECK-NEXT:  ## %bb.18: ## %cond_next73
+; CHECK-NEXT:    js LBB0_27
+; CHECK-NEXT:  ## %bb.20: ## %cond_next73
 ; CHECK-NEXT:    movl $1, {{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    movq (%r14), %rax
 ; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rsi
@@ -149,45 +146,45 @@ define ptr @ubyte_divmod(ptr %a, ptr %b) {
 ; CHECK-NEXT:    movl %ebp, %edx
 ; CHECK-NEXT:    callq *232(%rax)
 ; CHECK-NEXT:    testl %eax, %eax
-; CHECK-NEXT:    jne LBB0_23
-; CHECK-NEXT:  LBB0_19: ## %cond_next89
+; CHECK-NEXT:    jne LBB0_27
+; CHECK-NEXT:  LBB0_21: ## %cond_next89
 ; CHECK-NEXT:    movl $2, %edi
 ; CHECK-NEXT:    callq _PyTuple_New
 ; CHECK-NEXT:    testq %rax, %rax
-; CHECK-NEXT:    je LBB0_23
-; CHECK-NEXT:  ## %bb.20: ## %cond_next97
+; CHECK-NEXT:    je LBB0_27
+; CHECK-NEXT:  ## %bb.22: ## %cond_next97
 ; CHECK-NEXT:    movq %rax, %r14
 ; CHECK-NEXT:    movq _PyArray_API@GOTPCREL(%rip), %r12
 ; CHECK-NEXT:    movq (%r12), %rax
 ; CHECK-NEXT:    movq 200(%rax), %rdi
 ; CHECK-NEXT:    xorl %esi, %esi
 ; CHECK-NEXT:    callq *304(%rdi)
 ; CHECK-NEXT:    testq %rax, %rax
-; CHECK-NEXT:    je LBB0_21
-; CHECK-NEXT:  ## %bb.25: ## %cond_next135
+; CHECK-NEXT:    je LBB0_25
+; CHECK-NEXT:  ## %bb.23: ## %cond_next135
 ; CHECK-NEXT:    movb %r15b, 16(%rax)
 ; CHECK-NEXT:    movq %rax, 24(%r14)
 ; CHECK-NEXT:    movq (%r12), %rax
 ; CHECK-NEXT:    movq 200(%rax), %rdi
 ; CHECK-NEXT:    xorl %esi, %esi
 ; CHECK-NEXT:    callq *304(%rdi)
 ; CHECK-NEXT:    testq %rax, %rax
-; CHECK-NEXT:    je LBB0_21
-; CHECK-NEXT:  ## %bb.26: ## %cond_next182
+; CHECK-NEXT:    je LBB0_25
+; CHECK-NEXT:  ## %bb.24: ## %cond_next182
 ; CHECK-NEXT:    movb %bl, 16(%rax)
 ; CHECK-NEXT:    movq %rax, 32(%r14)
 ; CHECK-NEXT:    movq %r14, %rax
-; CHECK-NEXT:    jmp LBB0_24
-; CHECK-NEXT:  LBB0_21: ## %cond_true113
+; CHECK-NEXT:    jmp LBB0_28
+; CHECK-NEXT:  LBB0_25: ## %cond_true113
 ; CHECK-NEXT:    decq (%r14)
-; CHECK-NEXT:    jne LBB0_23
-; CHECK-NEXT:  ## %bb.22: ## %cond_true126
+; CHECK-NEXT:    jne LBB0_27
+; CHECK-NEXT:  ## %bb.26: ## %cond_true126
 ; CHECK-NEXT:    movq 8(%r14), %rax
 ; CHECK-NEXT:    movq %r14, %rdi
 ; CHECK-NEXT:    callq *48(%rax)
-; CHECK-NEXT:  LBB0_23: ## %UnifiedReturnBlock
+; CHECK-NEXT:  LBB0_27: ## %UnifiedReturnBlock
 ; CHECK-NEXT:    xorl %eax, %eax
-; CHECK-NEXT:  LBB0_24: ## %UnifiedReturnBlock
+; CHECK-NEXT:  LBB0_28: ## %UnifiedReturnBlock
 ; CHECK-NEXT:    addq $32, %rsp
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    popq %r12
diff --git a/llvm/test/CodeGen/X86/pr160612.ll b/llvm/test/CodeGen/X86/pr160612.ll
@@ -0,0 +1,74 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -O2 | FileCheck %s
+
+; Test for issue #160612: OR conditions in branches should use multiple branches
+; instead of materializing booleans with SETCC when no special optimizations apply.
+
+declare void @subroutine_foo()
+declare void @subroutine_bar()
+
+; Original issue: (x == 0 || y == 0) was generating SETCC + TEST + BRANCH
+; instead of using two conditional branches directly.
+define void @func_a(i32 noundef %x, i32 noundef %y) {
+; CHECK-LABEL: func_a:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    testl %edi, %edi
+; CHECK-NEXT:    je subroutine_foo@PLT # TAILCALL
+; CHECK-NEXT:  # %bb.1: # %entry
+; CHECK-NEXT:    testl %esi, %esi
+; CHECK-NEXT:    jne subroutine_bar@PLT # TAILCALL
+; CHECK-NEXT:  # %bb.2: # %if.then
+; CHECK-NEXT:    jmp subroutine_foo@PLT # TAILCALL
+entry:
+  %cmp = icmp eq i32 %x, 0
+  %cmp1 = icmp eq i32 %y, 0
+  %or.cond = or i1 %cmp, %cmp1
+  br i1 %or.cond, label %if.then, label %if.else
+
+if.then:
+  tail call void @subroutine_foo()
+  br label %if.end
+
+if.else:
+  tail call void @subroutine_bar()
+  br label %if.end
+
+if.end:
+  ret void
+}
+
+; Reference implementation that already generated optimal code.
+; This should continue to generate the same optimal code.
+define void @func_b(i32 noundef %x, i32 noundef %y) {
+; CHECK-LABEL: func_b:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    testl %edi, %edi
+; CHECK-NEXT:    je subroutine_foo@PLT # TAILCALL
+; CHECK-NEXT:  # %bb.1: # %if.else
+; CHECK-NEXT:    testl %esi, %esi
+; CHECK-NEXT:    je subroutine_foo@PLT # TAILCALL
+; CHECK-NEXT:  # %bb.2: # %if.else3
+; CHECK-NEXT:    jmp subroutine_bar@PLT # TAILCALL
+entry:
+  %cmp = icmp eq i32 %x, 0
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  tail call void @subroutine_foo()
+  br label %if.end4
+
+if.else:
+  %cmp1 = icmp eq i32 %y, 0
+  br i1 %cmp1, label %if.then2, label %if.else3
+
+if.then2:
+  tail call void @subroutine_foo()
+  br label %if.end4
+
+if.else3:
+  tail call void @subroutine_bar()
+  br label %if.end4
+
+if.end4:
+  ret void
+}
diff --git a/llvm/test/CodeGen/X86/x86-shrink-wrap-unwind.ll b/llvm/test/CodeGen/X86/x86-shrink-wrap-unwind.ll
