[x86] Fix an amazing goof in the handling of sub, or, and xor lowering.

The comment for this code indicated that it should work similar to our
handling of add lowering above: if we see uses of an instruction other
than flag usage and store usage, it tries to avoid the specialized
X86ISD::* nodes that are designed for flag+op modeling and emits an
explicit test.

Problem is, only the add case actually did this. In all the other cases,
the logic was incomplete and inverted. Any time the value was used by
a store, we bailed on the specialized X86ISD node. All of this appears
to have been historical where we had different logic here. =/

Turns out, we have quite a few patterns designed around these nodes. We
should actually form them. I fixed the code to match what we do for add,
and it has quite a positive effect just within some of our test cases.
The only thing close to a regression I see is using:

  notl %r
  testl %r, %r

instead of:

  xorl -1, %r

But we can add a pattern or something to fold that back out. The
improvements seem more than worth this.

I've also worked with Craig to update the comments to no longer be
actively contradicted by the code. =[ Some of this still remains
a mystery to both Craig and myself, but this seems like a large step in
the direction of consistency and slightly more accurate comments.

Many thanks to Craig for help figuring out this nasty stuff.

Differential Revision: https://reviews.llvm.org/D37096

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@311737 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/X86/X86ISelLowering.cpp

@@ -16547,16 +16547,11 @@ SDValue X86TargetLowering::EmitTest(SDValue Op, unsigned X86CC, const SDLoc &dl,
   // non-casted variable when we check for possible users.
   switch (ArithOp.getOpcode()) {
   case ISD::ADD:
-    // Due to an isel shortcoming, be conservative if this add is likely to be
-    // selected as part of a load-modify-store instruction. When the root node
-    // in a match is a store, isel doesn't know how to remap non-chain non-flag
-    // uses of other nodes in the match, such as the ADD in this case. This
-    // leads to the ADD being left around and reselected, with the result being
-    // two adds in the output.  Alas, even if none our users are stores, that
-    // doesn't prove we're O.K.  Ergo, if we have any parents that aren't
-    // CopyToReg or SETCC, eschew INC/DEC.  A better fix seems to require
-    // climbing the DAG back to the root, and it doesn't seem to be worth the
-    // effort.
+    // We only want to rewrite this as a target-specific node with attached
+    // flags if there is a reasonable chance of either using that to do custom
+    // instructions selection that can fold some of the memory operands, or if
+    // only the flags are used. If there are other uses, leave the node alone
+    // and emit a test instruction.
     for (SDNode::use_iterator UI = Op.getNode()->use_begin(),
          UE = Op.getNode()->use_end(); UI != UE; ++UI)
       if (UI->getOpcode() != ISD::CopyToReg &&
@@ -16667,11 +16662,13 @@ SDValue X86TargetLowering::EmitTest(SDValue Op, unsigned X86CC, const SDLoc &dl,
   case ISD::SUB:
   case ISD::OR:
   case ISD::XOR:
-    // Due to the ISEL shortcoming noted above, be conservative if this op is
-    // likely to be selected as part of a load-modify-store instruction.
+    // Similar to ISD::ADD above, check if the uses will preclude useful
+    // lowering of the target-specific node.
     for (SDNode::use_iterator UI = Op.getNode()->use_begin(),
            UE = Op.getNode()->use_end(); UI != UE; ++UI)
-      if (UI->getOpcode() == ISD::STORE)
+      if (UI->getOpcode() != ISD::CopyToReg &&
+          UI->getOpcode() != ISD::SETCC &&
+          UI->getOpcode() != ISD::STORE)
         goto default_case;
 
     // Otherwise use a regular EFLAGS-setting instruction.

test/CodeGen/X86/atomic-minmax-i6432.ll

@@ -15,29 +15,16 @@ define i64 @atomic_max_i64() nounwind {
 ; LINUX-NEXT:    .p2align 4, 0x90
 ; LINUX-NEXT:  .LBB0_1: # %atomicrmw.start
 ; LINUX-NEXT:    # =>This Inner Loop Header: Depth=1
-; LINUX-NEXT:    xorl %ecx, %ecx
 ; LINUX-NEXT:    cmpl %eax, %esi
+; LINUX-NEXT:    movl $0, %ecx
 ; LINUX-NEXT:    sbbl %edx, %ecx
-; LINUX-NEXT:    setl %cl
-; LINUX-NEXT:    andb $1, %cl
-; LINUX-NEXT:    movl %eax, %ebx
-; LINUX-NEXT:    jne .LBB0_3
-; LINUX-NEXT:  # BB#2: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB0_1 Depth=1
+; LINUX-NEXT:    movl $0, %ecx
+; LINUX-NEXT:    cmovll %edx, %ecx
 ; LINUX-NEXT:    movl $5, %ebx
-; LINUX-NEXT:  .LBB0_3: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB0_1 Depth=1
-; LINUX-NEXT:    testb %cl, %cl
-; LINUX-NEXT:    movl %edx, %ecx
-; LINUX-NEXT:    jne .LBB0_5
-; LINUX-NEXT:  # BB#4: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB0_1 Depth=1
-; LINUX-NEXT:    xorl %ecx, %ecx
-; LINUX-NEXT:  .LBB0_5: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB0_1 Depth=1
+; LINUX-NEXT:    cmovll %eax, %ebx
 ; LINUX-NEXT:    lock cmpxchg8b sc64
 ; LINUX-NEXT:    jne .LBB0_1
-; LINUX-NEXT:  # BB#6: # %atomicrmw.end
+; LINUX-NEXT:  # BB#2: # %atomicrmw.end
 ; LINUX-NEXT:    popl %esi
 ; LINUX-NEXT:    popl %ebx
 ; LINUX-NEXT:    retl
@@ -57,29 +44,16 @@ define i64 @atomic_max_i64() nounwind {
 ; PIC-NEXT:    .p2align 4, 0x90
 ; PIC-NEXT:  LBB0_1: ## %atomicrmw.start
 ; PIC-NEXT:    ## =>This Inner Loop Header: Depth=1
-; PIC-NEXT:    xorl %ecx, %ecx
 ; PIC-NEXT:    cmpl %eax, %edi
+; PIC-NEXT:    movl $0, %ecx
 ; PIC-NEXT:    sbbl %edx, %ecx
-; PIC-NEXT:    setl %cl
-; PIC-NEXT:    andb $1, %cl
-; PIC-NEXT:    movl %eax, %ebx
-; PIC-NEXT:    jne LBB0_3
-; PIC-NEXT:  ## BB#2: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB0_1 Depth=1
+; PIC-NEXT:    movl $0, %ecx
+; PIC-NEXT:    cmovll %edx, %ecx
 ; PIC-NEXT:    movl $5, %ebx
-; PIC-NEXT:  LBB0_3: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB0_1 Depth=1
-; PIC-NEXT:    testb %cl, %cl
-; PIC-NEXT:    movl %edx, %ecx
-; PIC-NEXT:    jne LBB0_5
-; PIC-NEXT:  ## BB#4: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB0_1 Depth=1
-; PIC-NEXT:    xorl %ecx, %ecx
-; PIC-NEXT:  LBB0_5: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB0_1 Depth=1
+; PIC-NEXT:    cmovll %eax, %ebx
 ; PIC-NEXT:    lock cmpxchg8b (%esi)
 ; PIC-NEXT:    jne LBB0_1
-; PIC-NEXT:  ## BB#6: ## %atomicrmw.end
+; PIC-NEXT:  ## BB#2: ## %atomicrmw.end
 ; PIC-NEXT:    popl %esi
 ; PIC-NEXT:    popl %edi
 ; PIC-NEXT:    popl %ebx
@@ -102,26 +76,13 @@ define i64 @atomic_min_i64() nounwind {
 ; LINUX-NEXT:    cmpl $7, %eax
 ; LINUX-NEXT:    movl %edx, %ecx
 ; LINUX-NEXT:    sbbl $0, %ecx
-; LINUX-NEXT:    setl %cl
-; LINUX-NEXT:    andb $1, %cl
-; LINUX-NEXT:    movl %eax, %ebx
-; LINUX-NEXT:    jne .LBB1_3
-; LINUX-NEXT:  # BB#2: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB1_1 Depth=1
+; LINUX-NEXT:    movl $0, %ecx
+; LINUX-NEXT:    cmovll %edx, %ecx
 ; LINUX-NEXT:    movl $6, %ebx
-; LINUX-NEXT:  .LBB1_3: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB1_1 Depth=1
-; LINUX-NEXT:    testb %cl, %cl
-; LINUX-NEXT:    movl %edx, %ecx
-; LINUX-NEXT:    jne .LBB1_5
-; LINUX-NEXT:  # BB#4: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB1_1 Depth=1
-; LINUX-NEXT:    xorl %ecx, %ecx
-; LINUX-NEXT:  .LBB1_5: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB1_1 Depth=1
+; LINUX-NEXT:    cmovll %eax, %ebx
 ; LINUX-NEXT:    lock cmpxchg8b sc64
 ; LINUX-NEXT:    jne .LBB1_1
-; LINUX-NEXT:  # BB#6: # %atomicrmw.end
+; LINUX-NEXT:  # BB#2: # %atomicrmw.end
 ; LINUX-NEXT:    popl %ebx
 ; LINUX-NEXT:    retl
 ;
@@ -141,26 +102,13 @@ define i64 @atomic_min_i64() nounwind {
 ; PIC-NEXT:    cmpl $7, %eax
 ; PIC-NEXT:    movl %edx, %ecx
 ; PIC-NEXT:    sbbl $0, %ecx
-; PIC-NEXT:    setl %cl
-; PIC-NEXT:    andb $1, %cl
-; PIC-NEXT:    movl %eax, %ebx
-; PIC-NEXT:    jne LBB1_3
-; PIC-NEXT:  ## BB#2: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB1_1 Depth=1
+; PIC-NEXT:    movl $0, %ecx
+; PIC-NEXT:    cmovll %edx, %ecx
 ; PIC-NEXT:    movl $6, %ebx
-; PIC-NEXT:  LBB1_3: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB1_1 Depth=1
-; PIC-NEXT:    testb %cl, %cl
-; PIC-NEXT:    movl %edx, %ecx
-; PIC-NEXT:    jne LBB1_5
-; PIC-NEXT:  ## BB#4: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB1_1 Depth=1
-; PIC-NEXT:    xorl %ecx, %ecx
-; PIC-NEXT:  LBB1_5: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB1_1 Depth=1
+; PIC-NEXT:    cmovll %eax, %ebx
 ; PIC-NEXT:    lock cmpxchg8b (%esi)
 ; PIC-NEXT:    jne LBB1_1
-; PIC-NEXT:  ## BB#6: ## %atomicrmw.end
+; PIC-NEXT:  ## BB#2: ## %atomicrmw.end
 ; PIC-NEXT:    popl %esi
 ; PIC-NEXT:    popl %ebx
 ; PIC-NEXT:    retl
@@ -181,29 +129,16 @@ define i64 @atomic_umax_i64() nounwind {
 ; LINUX-NEXT:    .p2align 4, 0x90
 ; LINUX-NEXT:  .LBB2_1: # %atomicrmw.start
 ; LINUX-NEXT:    # =>This Inner Loop Header: Depth=1
-; LINUX-NEXT:    xorl %ecx, %ecx
 ; LINUX-NEXT:    cmpl %eax, %esi
+; LINUX-NEXT:    movl $0, %ecx
 ; LINUX-NEXT:    sbbl %edx, %ecx
-; LINUX-NEXT:    setb %cl
-; LINUX-NEXT:    andb $1, %cl
-; LINUX-NEXT:    movl %eax, %ebx
-; LINUX-NEXT:    jne .LBB2_3
-; LINUX-NEXT:  # BB#2: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB2_1 Depth=1
+; LINUX-NEXT:    movl $0, %ecx
+; LINUX-NEXT:    cmovbl %edx, %ecx
 ; LINUX-NEXT:    movl $7, %ebx
-; LINUX-NEXT:  .LBB2_3: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB2_1 Depth=1
-; LINUX-NEXT:    testb %cl, %cl
-; LINUX-NEXT:    movl %edx, %ecx
-; LINUX-NEXT:    jne .LBB2_5
-; LINUX-NEXT:  # BB#4: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB2_1 Depth=1
-; LINUX-NEXT:    xorl %ecx, %ecx
-; LINUX-NEXT:  .LBB2_5: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB2_1 Depth=1
+; LINUX-NEXT:    cmovbl %eax, %ebx
 ; LINUX-NEXT:    lock cmpxchg8b sc64
 ; LINUX-NEXT:    jne .LBB2_1
-; LINUX-NEXT:  # BB#6: # %atomicrmw.end
+; LINUX-NEXT:  # BB#2: # %atomicrmw.end
 ; LINUX-NEXT:    popl %esi
 ; LINUX-NEXT:    popl %ebx
 ; LINUX-NEXT:    retl
@@ -223,29 +158,16 @@ define i64 @atomic_umax_i64() nounwind {
 ; PIC-NEXT:    .p2align 4, 0x90
 ; PIC-NEXT:  LBB2_1: ## %atomicrmw.start
 ; PIC-NEXT:    ## =>This Inner Loop Header: Depth=1
-; PIC-NEXT:    xorl %ecx, %ecx
 ; PIC-NEXT:    cmpl %eax, %edi
+; PIC-NEXT:    movl $0, %ecx
 ; PIC-NEXT:    sbbl %edx, %ecx
-; PIC-NEXT:    setb %cl
-; PIC-NEXT:    andb $1, %cl
-; PIC-NEXT:    movl %eax, %ebx
-; PIC-NEXT:    jne LBB2_3
-; PIC-NEXT:  ## BB#2: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB2_1 Depth=1
+; PIC-NEXT:    movl $0, %ecx
+; PIC-NEXT:    cmovbl %edx, %ecx
 ; PIC-NEXT:    movl $7, %ebx
-; PIC-NEXT:  LBB2_3: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB2_1 Depth=1
-; PIC-NEXT:    testb %cl, %cl
-; PIC-NEXT:    movl %edx, %ecx
-; PIC-NEXT:    jne LBB2_5
-; PIC-NEXT:  ## BB#4: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB2_1 Depth=1
-; PIC-NEXT:    xorl %ecx, %ecx
-; PIC-NEXT:  LBB2_5: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB2_1 Depth=1
+; PIC-NEXT:    cmovbl %eax, %ebx
 ; PIC-NEXT:    lock cmpxchg8b (%esi)
 ; PIC-NEXT:    jne LBB2_1
-; PIC-NEXT:  ## BB#6: ## %atomicrmw.end
+; PIC-NEXT:  ## BB#2: ## %atomicrmw.end
 ; PIC-NEXT:    popl %esi
 ; PIC-NEXT:    popl %edi
 ; PIC-NEXT:    popl %ebx
@@ -268,26 +190,13 @@ define i64 @atomic_umin_i64() nounwind {
 ; LINUX-NEXT:    cmpl $9, %eax
 ; LINUX-NEXT:    movl %edx, %ecx
 ; LINUX-NEXT:    sbbl $0, %ecx
-; LINUX-NEXT:    setb %cl
-; LINUX-NEXT:    andb $1, %cl
-; LINUX-NEXT:    movl %eax, %ebx
-; LINUX-NEXT:    jne .LBB3_3
-; LINUX-NEXT:  # BB#2: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB3_1 Depth=1
+; LINUX-NEXT:    movl $0, %ecx
+; LINUX-NEXT:    cmovbl %edx, %ecx
 ; LINUX-NEXT:    movl $8, %ebx
-; LINUX-NEXT:  .LBB3_3: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB3_1 Depth=1
-; LINUX-NEXT:    testb %cl, %cl
-; LINUX-NEXT:    movl %edx, %ecx
-; LINUX-NEXT:    jne .LBB3_5
-; LINUX-NEXT:  # BB#4: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB3_1 Depth=1
-; LINUX-NEXT:    xorl %ecx, %ecx
-; LINUX-NEXT:  .LBB3_5: # %atomicrmw.start
-; LINUX-NEXT:    # in Loop: Header=BB3_1 Depth=1
+; LINUX-NEXT:    cmovbl %eax, %ebx
 ; LINUX-NEXT:    lock cmpxchg8b sc64
 ; LINUX-NEXT:    jne .LBB3_1
-; LINUX-NEXT:  # BB#6: # %atomicrmw.end
+; LINUX-NEXT:  # BB#2: # %atomicrmw.end
 ; LINUX-NEXT:    popl %ebx
 ; LINUX-NEXT:    retl
 ;
@@ -307,26 +216,13 @@ define i64 @atomic_umin_i64() nounwind {
 ; PIC-NEXT:    cmpl $9, %eax
 ; PIC-NEXT:    movl %edx, %ecx
 ; PIC-NEXT:    sbbl $0, %ecx
-; PIC-NEXT:    setb %cl
-; PIC-NEXT:    andb $1, %cl
-; PIC-NEXT:    movl %eax, %ebx
-; PIC-NEXT:    jne LBB3_3
-; PIC-NEXT:  ## BB#2: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB3_1 Depth=1
+; PIC-NEXT:    movl $0, %ecx
+; PIC-NEXT:    cmovbl %edx, %ecx
 ; PIC-NEXT:    movl $8, %ebx
-; PIC-NEXT:  LBB3_3: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB3_1 Depth=1
-; PIC-NEXT:    testb %cl, %cl
-; PIC-NEXT:    movl %edx, %ecx
-; PIC-NEXT:    jne LBB3_5
-; PIC-NEXT:  ## BB#4: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB3_1 Depth=1
-; PIC-NEXT:    xorl %ecx, %ecx
-; PIC-NEXT:  LBB3_5: ## %atomicrmw.start
-; PIC-NEXT:    ## in Loop: Header=BB3_1 Depth=1
+; PIC-NEXT:    cmovbl %eax, %ebx
 ; PIC-NEXT:    lock cmpxchg8b (%esi)
 ; PIC-NEXT:    jne LBB3_1
-; PIC-NEXT:  ## BB#6: ## %atomicrmw.end
+; PIC-NEXT:  ## BB#2: ## %atomicrmw.end
 ; PIC-NEXT:    popl %esi
 ; PIC-NEXT:    popl %ebx
 ; PIC-NEXT:    retl

test/CodeGen/X86/atomic128.ll

@@ -165,26 +165,13 @@ define void @fetch_and_min(i128* %p, i128 %bits) {
 ; CHECK-NEXT:    cmpq %rax, %rsi
 ; CHECK-NEXT:    movq %r8, %rcx
 ; CHECK-NEXT:    sbbq %rdx, %rcx
-; CHECK-NEXT:    setge %cl
-; CHECK-NEXT:    andb $1, %cl
-; CHECK-NEXT:    movq %rax, %rbx
-; CHECK-NEXT:    jne LBB5_3
-; CHECK-NEXT:  ## BB#2: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB5_1 Depth=1
-; CHECK-NEXT:    movq %rsi, %rbx
-; CHECK-NEXT:  LBB5_3: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB5_1 Depth=1
-; CHECK-NEXT:    testb %cl, %cl
-; CHECK-NEXT:    movq %rdx, %rcx
-; CHECK-NEXT:    jne LBB5_5
-; CHECK-NEXT:  ## BB#4: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB5_1 Depth=1
 ; CHECK-NEXT:    movq %r8, %rcx
-; CHECK-NEXT:  LBB5_5: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB5_1 Depth=1
+; CHECK-NEXT:    cmovgeq %rdx, %rcx
+; CHECK-NEXT:    movq %rsi, %rbx
+; CHECK-NEXT:    cmovgeq %rax, %rbx
 ; CHECK-NEXT:    lock cmpxchg16b (%rdi)
 ; CHECK-NEXT:    jne LBB5_1
-; CHECK-NEXT:  ## BB#6: ## %atomicrmw.end
+; CHECK-NEXT:  ## BB#2: ## %atomicrmw.end
 ; CHECK-NEXT:    movq %rax, {{.*}}(%rip)
 ; CHECK-NEXT:    movq %rdx, _var+{{.*}}(%rip)
 ; CHECK-NEXT:    popq %rbx
@@ -211,26 +198,13 @@ define void @fetch_and_max(i128* %p, i128 %bits) {
 ; CHECK-NEXT:    cmpq %rsi, %rax
 ; CHECK-NEXT:    movq %rdx, %rcx
 ; CHECK-NEXT:    sbbq %r8, %rcx
-; CHECK-NEXT:    setge %cl
-; CHECK-NEXT:    andb $1, %cl
-; CHECK-NEXT:    movq %rax, %rbx
-; CHECK-NEXT:    jne LBB6_3
-; CHECK-NEXT:  ## BB#2: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB6_1 Depth=1
-; CHECK-NEXT:    movq %rsi, %rbx
-; CHECK-NEXT:  LBB6_3: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB6_1 Depth=1
-; CHECK-NEXT:    testb %cl, %cl
-; CHECK-NEXT:    movq %rdx, %rcx
-; CHECK-NEXT:    jne LBB6_5
-; CHECK-NEXT:  ## BB#4: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB6_1 Depth=1
 ; CHECK-NEXT:    movq %r8, %rcx
-; CHECK-NEXT:  LBB6_5: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB6_1 Depth=1
+; CHECK-NEXT:    cmovgeq %rdx, %rcx
+; CHECK-NEXT:    movq %rsi, %rbx
+; CHECK-NEXT:    cmovgeq %rax, %rbx
 ; CHECK-NEXT:    lock cmpxchg16b (%rdi)
 ; CHECK-NEXT:    jne LBB6_1
-; CHECK-NEXT:  ## BB#6: ## %atomicrmw.end
+; CHECK-NEXT:  ## BB#2: ## %atomicrmw.end
 ; CHECK-NEXT:    movq %rax, {{.*}}(%rip)
 ; CHECK-NEXT:    movq %rdx, _var+{{.*}}(%rip)
 ; CHECK-NEXT:    popq %rbx
@@ -257,26 +231,13 @@ define void @fetch_and_umin(i128* %p, i128 %bits) {
 ; CHECK-NEXT:    cmpq %rax, %rsi
 ; CHECK-NEXT:    movq %r8, %rcx
 ; CHECK-NEXT:    sbbq %rdx, %rcx
-; CHECK-NEXT:    setae %cl
-; CHECK-NEXT:    andb $1, %cl
-; CHECK-NEXT:    movq %rax, %rbx
-; CHECK-NEXT:    jne LBB7_3
-; CHECK-NEXT:  ## BB#2: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB7_1 Depth=1
-; CHECK-NEXT:    movq %rsi, %rbx
-; CHECK-NEXT:  LBB7_3: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB7_1 Depth=1
-; CHECK-NEXT:    testb %cl, %cl
-; CHECK-NEXT:    movq %rdx, %rcx
-; CHECK-NEXT:    jne LBB7_5
-; CHECK-NEXT:  ## BB#4: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB7_1 Depth=1
 ; CHECK-NEXT:    movq %r8, %rcx
-; CHECK-NEXT:  LBB7_5: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB7_1 Depth=1
+; CHECK-NEXT:    cmovaeq %rdx, %rcx
+; CHECK-NEXT:    movq %rsi, %rbx
+; CHECK-NEXT:    cmovaeq %rax, %rbx
 ; CHECK-NEXT:    lock cmpxchg16b (%rdi)
 ; CHECK-NEXT:    jne LBB7_1
-; CHECK-NEXT:  ## BB#6: ## %atomicrmw.end
+; CHECK-NEXT:  ## BB#2: ## %atomicrmw.end
 ; CHECK-NEXT:    movq %rax, {{.*}}(%rip)
 ; CHECK-NEXT:    movq %rdx, _var+{{.*}}(%rip)
 ; CHECK-NEXT:    popq %rbx
@@ -303,26 +264,13 @@ define void @fetch_and_umax(i128* %p, i128 %bits) {
 ; CHECK-NEXT:    cmpq %rax, %rsi
 ; CHECK-NEXT:    movq %r8, %rcx
 ; CHECK-NEXT:    sbbq %rdx, %rcx
-; CHECK-NEXT:    setb %cl
-; CHECK-NEXT:    andb $1, %cl
-; CHECK-NEXT:    movq %rax, %rbx
-; CHECK-NEXT:    jne LBB8_3
-; CHECK-NEXT:  ## BB#2: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB8_1 Depth=1
-; CHECK-NEXT:    movq %rsi, %rbx
-; CHECK-NEXT:  LBB8_3: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB8_1 Depth=1
-; CHECK-NEXT:    testb %cl, %cl
-; CHECK-NEXT:    movq %rdx, %rcx
-; CHECK-NEXT:    jne LBB8_5
-; CHECK-NEXT:  ## BB#4: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB8_1 Depth=1
 ; CHECK-NEXT:    movq %r8, %rcx
-; CHECK-NEXT:  LBB8_5: ## %atomicrmw.start
-; CHECK-NEXT:    ## in Loop: Header=BB8_1 Depth=1
+; CHECK-NEXT:    cmovbq %rdx, %rcx
+; CHECK-NEXT:    movq %rsi, %rbx
+; CHECK-NEXT:    cmovbq %rax, %rbx
 ; CHECK-NEXT:    lock cmpxchg16b (%rdi)
 ; CHECK-NEXT:    jne LBB8_1
-; CHECK-NEXT:  ## BB#6: ## %atomicrmw.end
+; CHECK-NEXT:  ## BB#2: ## %atomicrmw.end
 ; CHECK-NEXT:    movq %rax, {{.*}}(%rip)
 ; CHECK-NEXT:    movq %rdx, _var+{{.*}}(%rip)
 ; CHECK-NEXT:    popq %rbx

test/CodeGen/X86/cmov.ll

@@ -208,7 +208,8 @@ define i8 @test7(i1 inreg %c, i8 inreg %a, i8 inreg %b) nounwind {
 define i32 @smin(i32 %x) {
 ; CHECK-LABEL: smin:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    xorl $-1, %edi
+; CHECK-NEXT:    notl %edi
+; CHECK-NEXT:    testl %edi, %edi
 ; CHECK-NEXT:    movl $-1, %eax
 ; CHECK-NEXT:    cmovsl %edi, %eax
 ; CHECK-NEXT:    retq

test/CodeGen/X86/cmp.ll

@@ -388,13 +388,14 @@ define void @test20(i32 %bf.load, i8 %x1, i8* %b_addr) {
 ; CHECK-LABEL: test20:
 ; CHECK:       ## BB#0: ## %entry
 ; CHECK-NEXT:    xorl %eax, %eax ## encoding: [0x31,0xc0]
-; CHECK-NEXT:    andl $16777215, %edi ## encoding: [0x81,0xe7,0xff,0xff,0xff,0x00]
+; CHECK-NEXT:    testl $16777215, %edi ## encoding: [0xf7,0xc7,0xff,0xff,0xff,0x00]
 ; CHECK-NEXT:    ## imm = 0xFFFFFF
 ; CHECK-NEXT:    setne %al ## encoding: [0x0f,0x95,0xc0]
 ; CHECK-NEXT:    movzbl %sil, %ecx ## encoding: [0x40,0x0f,0xb6,0xce]
 ; CHECK-NEXT:    addl %eax, %ecx ## encoding: [0x01,0xc1]
 ; CHECK-NEXT:    setne (%rdx) ## encoding: [0x0f,0x95,0x02]
-; CHECK-NEXT:    testl %edi, %edi ## encoding: [0x85,0xff]
+; CHECK-NEXT:    testl $16777215, %edi ## encoding: [0xf7,0xc7,0xff,0xff,0xff,0x00]
+; CHECK-NEXT:    ## imm = 0xFFFFFF
 ; CHECK-NEXT:    setne {{.*}}(%rip) ## encoding: [0x0f,0x95,0x05,A,A,A,A]
 ; CHECK-NEXT:    ## fixup A - offset: 3, value: _d-4, kind: reloc_riprel_4byte
 ; CHECK-NEXT:    retq ## encoding: [0xc3]

test/CodeGen/X86/select.ll

@@ -166,19 +166,15 @@ define void @test5(i1 %c, <2 x i16> %a, <2 x i16> %b, <2 x i16>* %p) nounwind {
 ; MCU-LABEL: test5:
 ; MCU:       # BB#0:
 ; MCU-NEXT:    pushl %esi
-; MCU-NEXT:    andb $1, %al
+; MCU-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; MCU-NEXT:    testb $1, %al
 ; MCU-NEXT:    jne .LBB4_2
 ; MCU-NEXT:  # BB#1:
+; MCU-NEXT:    movw {{[0-9]+}}(%esp), %cx
 ; MCU-NEXT:    movw {{[0-9]+}}(%esp), %dx
 ; MCU-NEXT:  .LBB4_2:
-; MCU-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; MCU-NEXT:    testb %al, %al
-; MCU-NEXT:    jne .LBB4_4
-; MCU-NEXT:  # BB#3:
-; MCU-NEXT:    movw {{[0-9]+}}(%esp), %cx
-; MCU-NEXT:  .LBB4_4:
-; MCU-NEXT:    movw %dx, (%esi)
 ; MCU-NEXT:    movw %cx, 2(%esi)
+; MCU-NEXT:    movw %dx, (%esi)
 ; MCU-NEXT:    popl %esi
 ; MCU-NEXT:    retl
   %x = select i1 %c, <2 x i16> %a, <2 x i16> %b
@@ -292,81 +288,69 @@ define x86_fp80 @test7(i32 %tmp8) nounwind {
 define void @test8(i1 %c, <6 x i32>* %dst.addr, <6 x i32> %src1,<6 x i32> %src2) nounwind {
 ; GENERIC-LABEL: test8:
 ; GENERIC:       ## BB#0:
-; GENERIC-NEXT:    andb $1, %dil
+; GENERIC-NEXT:    testb $1, %dil
 ; GENERIC-NEXT:    jne LBB7_1
 ; GENERIC-NEXT:  ## BB#2:
-; GENERIC-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; GENERIC-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; GENERIC-NEXT:    jmp LBB7_3
-; GENERIC-NEXT:  LBB7_1:
 ; GENERIC-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; GENERIC-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; GENERIC-NEXT:    movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; GENERIC-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; GENERIC-NEXT:  LBB7_3:
-; GENERIC-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; GENERIC-NEXT:    testb %dil, %dil
-; GENERIC-NEXT:    jne LBB7_4
-; GENERIC-NEXT:  ## BB#5:
+; GENERIC-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; GENERIC-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; GENERIC-NEXT:    movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; GENERIC-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; GENERIC-NEXT:    jmp LBB7_3
+; GENERIC-NEXT:  LBB7_1:
+; GENERIC-NEXT:    movd %r9d, %xmm0
+; GENERIC-NEXT:    movd %r8d, %xmm1
+; GENERIC-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; GENERIC-NEXT:    movd %ecx, %xmm2
+; GENERIC-NEXT:    movd %edx, %xmm0
+; GENERIC-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; GENERIC-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; GENERIC-NEXT:    movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; GENERIC-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; GENERIC-NEXT:    movd {{.*#+}} xmm3 = mem[0],zero,zero,zero
 ; GENERIC-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; GENERIC-NEXT:    jmp LBB7_6
-; GENERIC-NEXT:  LBB7_4:
-; GENERIC-NEXT:    movd %r9d, %xmm1
-; GENERIC-NEXT:    movd %r8d, %xmm2
-; GENERIC-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; GENERIC-NEXT:    movd %ecx, %xmm3
-; GENERIC-NEXT:    movd %edx, %xmm1
-; GENERIC-NEXT:  LBB7_6:
-; GENERIC-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; GENERIC-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; GENERIC-NEXT:  LBB7_3:
+; GENERIC-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; GENERIC-NEXT:    pcmpeqd %xmm2, %xmm2
-; GENERIC-NEXT:    paddd %xmm2, %xmm1
 ; GENERIC-NEXT:    paddd %xmm2, %xmm0
-; GENERIC-NEXT:    movq %xmm0, 16(%rsi)
-; GENERIC-NEXT:    movdqa %xmm1, (%rsi)
+; GENERIC-NEXT:    paddd %xmm2, %xmm1
+; GENERIC-NEXT:    movq %xmm1, 16(%rsi)
+; GENERIC-NEXT:    movdqa %xmm0, (%rsi)
 ; GENERIC-NEXT:    retq
 ; GENERIC-NEXT:    ## -- End function
 ;
 ; ATOM-LABEL: test8:
 ; ATOM:       ## BB#0:
-; ATOM-NEXT:    andb $1, %dil
+; ATOM-NEXT:    testb $1, %dil
 ; ATOM-NEXT:    jne LBB7_1
 ; ATOM-NEXT:  ## BB#2:
-; ATOM-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; ATOM-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; ATOM-NEXT:    jmp LBB7_3
-; ATOM-NEXT:  LBB7_1:
-; ATOM-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; ATOM-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; ATOM-NEXT:  LBB7_3:
-; ATOM-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; ATOM-NEXT:    testb %dil, %dil
-; ATOM-NEXT:    jne LBB7_4
-; ATOM-NEXT:  ## BB#5:
 ; ATOM-NEXT:    movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; ATOM-NEXT:    movd {{.*#+}} xmm3 = mem[0],zero,zero,zero
 ; ATOM-NEXT:    movd {{.*#+}} xmm4 = mem[0],zero,zero,zero
+; ATOM-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; ATOM-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; ATOM-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; ATOM-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; ATOM-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; ATOM-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
-; ATOM-NEXT:    jmp LBB7_6
-; ATOM-NEXT:  LBB7_4:
-; ATOM-NEXT:    movd %r9d, %xmm1
+; ATOM-NEXT:    jmp LBB7_3
+; ATOM-NEXT:  LBB7_1:
+; ATOM-NEXT:    movd %r9d, %xmm0
 ; ATOM-NEXT:    movd %r8d, %xmm2
-; ATOM-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; ATOM-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; ATOM-NEXT:    movd %ecx, %xmm3
-; ATOM-NEXT:    movd %edx, %xmm1
-; ATOM-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; ATOM-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
-; ATOM-NEXT:  LBB7_6:
+; ATOM-NEXT:    movd %edx, %xmm0
+; ATOM-NEXT:    movd {{.*#+}} xmm4 = mem[0],zero,zero,zero
+; ATOM-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; ATOM-NEXT:  LBB7_3:
+; ATOM-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; ATOM-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; ATOM-NEXT:    pcmpeqd %xmm2, %xmm2
+; ATOM-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
 ; ATOM-NEXT:    paddd %xmm2, %xmm0
 ; ATOM-NEXT:    paddd %xmm2, %xmm1
-; ATOM-NEXT:    movq %xmm0, 16(%rsi)
-; ATOM-NEXT:    movdqa %xmm1, (%rsi)
+; ATOM-NEXT:    movq %xmm1, 16(%rsi)
+; ATOM-NEXT:    movdqa %xmm0, (%rsi)
 ; ATOM-NEXT:    retq
 ; ATOM-NEXT:    ## -- End function
 ;
@@ -376,68 +360,67 @@ define void @test8(i1 %c, <6 x i32>* %dst.addr, <6 x i32> %src1,<6 x i32> %src2)
 ; MCU-NEXT:    pushl %ebx
 ; MCU-NEXT:    pushl %edi
 ; MCU-NEXT:    pushl %esi
-; MCU-NEXT:    andb $1, %al
+; MCU-NEXT:    testb $1, %al
 ; MCU-NEXT:    jne .LBB7_1
 ; MCU-NEXT:  # BB#2:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ecx
-; MCU-NEXT:    movl (%ecx), %ecx
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; MCU-NEXT:    movl (%eax), %eax
 ; MCU-NEXT:    je .LBB7_5
 ; MCU-NEXT:  .LBB7_4:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %esi
-; MCU-NEXT:    movl (%esi), %esi
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ecx
+; MCU-NEXT:    movl (%ecx), %ecx
 ; MCU-NEXT:    je .LBB7_8
 ; MCU-NEXT:  .LBB7_7:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %edi
-; MCU-NEXT:    movl (%edi), %edi
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %esi
+; MCU-NEXT:    movl (%esi), %esi
 ; MCU-NEXT:    je .LBB7_11
 ; MCU-NEXT:  .LBB7_10:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ebx
-; MCU-NEXT:    movl (%ebx), %ebx
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %edi
+; MCU-NEXT:    movl (%edi), %edi
 ; MCU-NEXT:    je .LBB7_14
 ; MCU-NEXT:  .LBB7_13:
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ebx
+; MCU-NEXT:    movl (%ebx), %ebx
+; MCU-NEXT:    je .LBB7_17
+; MCU-NEXT:  .LBB7_16:
 ; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ebp
-; MCU-NEXT:    jmp .LBB7_15
+; MCU-NEXT:    jmp .LBB7_18
 ; MCU-NEXT:  .LBB7_1:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ecx
-; MCU-NEXT:    movl (%ecx), %ecx
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; MCU-NEXT:    movl (%eax), %eax
 ; MCU-NEXT:    jne .LBB7_4
 ; MCU-NEXT:  .LBB7_5:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %esi
-; MCU-NEXT:    movl (%esi), %esi
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ecx
+; MCU-NEXT:    movl (%ecx), %ecx
 ; MCU-NEXT:    jne .LBB7_7
 ; MCU-NEXT:  .LBB7_8:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %edi
-; MCU-NEXT:    movl (%edi), %edi
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %esi
+; MCU-NEXT:    movl (%esi), %esi
 ; MCU-NEXT:    jne .LBB7_10
 ; MCU-NEXT:  .LBB7_11:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ebx
-; MCU-NEXT:    movl (%ebx), %ebx
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %edi
+; MCU-NEXT:    movl (%edi), %edi
 ; MCU-NEXT:    jne .LBB7_13
 ; MCU-NEXT:  .LBB7_14:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ebp
-; MCU-NEXT:  .LBB7_15:
-; MCU-NEXT:    movl (%ebp), %ebp
-; MCU-NEXT:    testb %al, %al
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ebx
+; MCU-NEXT:    movl (%ebx), %ebx
 ; MCU-NEXT:    jne .LBB7_16
-; MCU-NEXT:  # BB#17:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; MCU-NEXT:    jmp .LBB7_18
-; MCU-NEXT:  .LBB7_16:
-; MCU-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; MCU-NEXT:  .LBB7_17:
+; MCU-NEXT:    leal {{[0-9]+}}(%esp), %ebp
 ; MCU-NEXT:  .LBB7_18:
-; MCU-NEXT:    movl (%eax), %eax
-; MCU-NEXT:    decl %eax
+; MCU-NEXT:    movl (%ebp), %ebp
 ; MCU-NEXT:    decl %ebp
 ; MCU-NEXT:    decl %ebx
 ; MCU-NEXT:    decl %edi
 ; MCU-NEXT:    decl %esi
 ; MCU-NEXT:    decl %ecx
-; MCU-NEXT:    movl %ecx, 20(%edx)
-; MCU-NEXT:    movl %esi, 16(%edx)
-; MCU-NEXT:    movl %edi, 12(%edx)
-; MCU-NEXT:    movl %ebx, 8(%edx)
-; MCU-NEXT:    movl %ebp, 4(%edx)
-; MCU-NEXT:    movl %eax, (%edx)
+; MCU-NEXT:    decl %eax
+; MCU-NEXT:    movl %eax, 20(%edx)
+; MCU-NEXT:    movl %ecx, 16(%edx)
+; MCU-NEXT:    movl %esi, 12(%edx)
+; MCU-NEXT:    movl %edi, 8(%edx)
+; MCU-NEXT:    movl %ebx, 4(%edx)
+; MCU-NEXT:    movl %ebp, (%edx)
 ; MCU-NEXT:    popl %esi
 ; MCU-NEXT:    popl %edi
 ; MCU-NEXT:    popl %ebx

test/CodeGen/X86/tbm_patterns.ll

@@ -27,10 +27,9 @@ define i32 @test_x86_tbm_bextri_u32_m(i32* nocapture %a) nounwind {
 define i32 @test_x86_tbm_bextri_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_bextri_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    shrl $4, %edi
-; CHECK-NEXT:    andl $4095, %edi # imm = 0xFFF
-; CHECK-NEXT:    cmovel %esi, %edi
-; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    bextr $3076, %edi, %eax # imm = 0xC04
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = lshr i32 %a, 4
   %t1 = and i32 %t0, 4095
@@ -63,10 +62,10 @@ define i64 @test_x86_tbm_bextri_u64_m(i64* nocapture %a) nounwind {
 define i64 @test_x86_tbm_bextri_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_bextri_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    shrl $4, %edi
-; CHECK-NEXT:    andl $4095, %edi # imm = 0xFFF
-; CHECK-NEXT:    cmoveq %rsi, %rdi
-; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    bextr $3076, %edi, %eax # imm = 0xC04
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    cmovneq %rax, %rsi
+; CHECK-NEXT:    movq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = lshr i64 %a, 4
   %t1 = and i64 %t0, 4095
@@ -88,9 +87,8 @@ define i32 @test_x86_tbm_blcfill_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_blcfill_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcfill_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; CHECK-NEXT:    leal 1(%rdi), %eax
-; CHECK-NEXT:    andl %edi, %eax
+; CHECK-NEXT:    blcfill %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
 ; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = add i32 %a, 1
@@ -113,8 +111,8 @@ define i64 @test_x86_tbm_blcfill_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_blcfill_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcfill_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leaq 1(%rdi), %rax
-; CHECK-NEXT:    andq %rdi, %rax
+; CHECK-NEXT:    blcfill %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
 ; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = add i64 %a, 1
@@ -138,10 +136,8 @@ define i32 @test_x86_tbm_blci_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_blci_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blci_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; CHECK-NEXT:    leal 1(%rdi), %eax
-; CHECK-NEXT:    notl %eax
-; CHECK-NEXT:    orl %edi, %eax
+; CHECK-NEXT:    blci %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
 ; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = add i32 1, %a
@@ -166,9 +162,8 @@ define i64 @test_x86_tbm_blci_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_blci_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blci_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leaq 1(%rdi), %rax
-; CHECK-NEXT:    notq %rax
-; CHECK-NEXT:    orq %rdi, %rax
+; CHECK-NEXT:    blci %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
 ; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = add i64 1, %a
@@ -213,11 +208,8 @@ define i32 @test_x86_tbm_blcic_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_blcic_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcic_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; CHECK-NEXT:    leal 1(%rdi), %eax
-; CHECK-NEXT:    movl %edi, %ecx
-; CHECK-NEXT:    notl %ecx
-; CHECK-NEXT:    andl %ecx, %eax
+; CHECK-NEXT:    blcic %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
 ; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = xor i32 %a, -1
@@ -242,9 +234,8 @@ define i64 @test_x86_tbm_blcic_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_blcic_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcic_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leaq 1(%rdi), %rax
-; CHECK-NEXT:    notq %rdi
-; CHECK-NEXT:    andq %rdi, %rax
+; CHECK-NEXT:    blcic %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
 ; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = xor i64 %a, -1
@@ -268,9 +259,8 @@ define i32 @test_x86_tbm_blcmsk_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_blcmsk_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcmsk_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; CHECK-NEXT:    leal 1(%rdi), %eax
-; CHECK-NEXT:    xorl %edi, %eax
+; CHECK-NEXT:    blcmsk %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
 ; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = add i32 %a, 1
@@ -293,8 +283,8 @@ define i64 @test_x86_tbm_blcmsk_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_blcmsk_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcmsk_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leaq 1(%rdi), %rax
-; CHECK-NEXT:    xorq %rdi, %rax
+; CHECK-NEXT:    blcmsk %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
 ; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = add i64 %a, 1
@@ -317,9 +307,8 @@ define i32 @test_x86_tbm_blcs_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_blcs_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcs_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; CHECK-NEXT:    leal 1(%rdi), %eax
-; CHECK-NEXT:    orl %edi, %eax
+; CHECK-NEXT:    blcs %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
 ; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = add i32 %a, 1
@@ -342,8 +331,8 @@ define i64 @test_x86_tbm_blcs_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_blcs_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blcs_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leaq 1(%rdi), %rax
-; CHECK-NEXT:    orq %rdi, %rax
+; CHECK-NEXT:    blcs %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
 ; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = add i64 %a, 1
@@ -366,9 +355,8 @@ define i32 @test_x86_tbm_blsfill_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_blsfill_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blsfill_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; CHECK-NEXT:    leal -1(%rdi), %eax
-; CHECK-NEXT:    orl %edi, %eax
+; CHECK-NEXT:    blsfill %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
 ; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = add i32 %a, -1
@@ -391,8 +379,8 @@ define i64 @test_x86_tbm_blsfill_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_blsfill_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blsfill_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leaq -1(%rdi), %rax
-; CHECK-NEXT:    orq %rdi, %rax
+; CHECK-NEXT:    blsfill %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
 ; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = add i64 %a, -1
@@ -416,12 +404,9 @@ define i32 @test_x86_tbm_blsic_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_blsic_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blsic_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    movl %edi, %eax
-; CHECK-NEXT:    notl %eax
-; CHECK-NEXT:    decl %edi
-; CHECK-NEXT:    orl %eax, %edi
-; CHECK-NEXT:    cmovel %esi, %edi
-; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    blsic %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = xor i32 %a, -1
   %t1 = add i32 %a, -1
@@ -445,12 +430,9 @@ define i64 @test_x86_tbm_blsic_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_blsic_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_blsic_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    movq %rdi, %rax
-; CHECK-NEXT:    notq %rax
-; CHECK-NEXT:    decq %rdi
-; CHECK-NEXT:    orq %rax, %rdi
-; CHECK-NEXT:    cmoveq %rsi, %rdi
-; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    blsic %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = xor i64 %a, -1
   %t1 = add i64 %a, -1
@@ -474,12 +456,9 @@ define i32 @test_x86_tbm_t1mskc_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_t1mskc_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_t1mskc_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    movl %edi, %eax
-; CHECK-NEXT:    notl %eax
-; CHECK-NEXT:    incl %edi
-; CHECK-NEXT:    orl %eax, %edi
-; CHECK-NEXT:    cmovel %esi, %edi
-; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    t1mskc %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = xor i32 %a, -1
   %t1 = add i32 %a, 1
@@ -503,12 +482,9 @@ define i64 @test_x86_tbm_t1mskc_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_t1mskc_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_t1mskc_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    movq %rdi, %rax
-; CHECK-NEXT:    notq %rax
-; CHECK-NEXT:    incq %rdi
-; CHECK-NEXT:    orq %rax, %rdi
-; CHECK-NEXT:    cmoveq %rsi, %rdi
-; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    t1mskc %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = xor i64 %a, -1
   %t1 = add i64 %a, 1
@@ -532,12 +508,9 @@ define i32 @test_x86_tbm_tzmsk_u32(i32 %a) nounwind {
 define i32 @test_x86_tbm_tzmsk_u32_z(i32 %a, i32 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_tzmsk_u32_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    movl %edi, %eax
-; CHECK-NEXT:    notl %eax
-; CHECK-NEXT:    decl %edi
-; CHECK-NEXT:    andl %eax, %edi
-; CHECK-NEXT:    cmovel %esi, %edi
-; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    tzmsk %edi, %eax
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    cmovel %esi, %eax
 ; CHECK-NEXT:    retq
   %t0 = xor i32 %a, -1
   %t1 = add i32 %a, -1
@@ -561,12 +534,9 @@ define i64 @test_x86_tbm_tzmsk_u64(i64 %a) nounwind {
 define i64 @test_x86_tbm_tzmsk_u64_z(i64 %a, i64 %b) nounwind {
 ; CHECK-LABEL: test_x86_tbm_tzmsk_u64_z:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    movq %rdi, %rax
-; CHECK-NEXT:    notq %rax
-; CHECK-NEXT:    decq %rdi
-; CHECK-NEXT:    andq %rax, %rdi
-; CHECK-NEXT:    cmoveq %rsi, %rdi
-; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    tzmsk %rdi, %rax
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmoveq %rsi, %rax
 ; CHECK-NEXT:    retq
   %t0 = xor i64 %a, -1
   %t1 = add i64 %a, -1