The motivation is that the update script has at least two deviations (`<...>@GOT`/`<...>@PLT`/ and not hiding pointer arithmetics) from what pretty much all the checklines were generated with, and most of the tests are still not updated, so each time one of the non-up-to-date tests is updated to see the effect of the code change, there is a lot of noise. Instead of having to deal with that each time, let's just deal with everything at once. This has been done via: ``` cd llvm-project/llvm/test/CodeGen/X86 grep -rl "; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py" | xargs -L1 <...>/llvm-project/llvm/utils/update_llc_test_checks.py --llc-binary <...>/llvm-project/build/bin/llc ``` Not all tests were regenerated, however.
69 lines
2.2 KiB
LLVM
69 lines
2.2 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -o - %s | FileCheck %s
|
|
target triple = "x86_64-unknown-unknown"
|
|
|
|
; select with and i1/or i1 condition should be implemented as a series of 2
|
|
; cmovs, not by producing two conditions and using and on them.
|
|
|
|
define dso_local i32 @select_and(i32 %a0, i32 %a1, float %a2, float %a3, i32 %a4, i32 %a5) {
|
|
; CHECK-LABEL: select_and:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: movl %edx, %eax
|
|
; CHECK-NEXT: ucomiss %xmm0, %xmm1
|
|
; CHECK-NEXT: cmovbel %ecx, %eax
|
|
; CHECK-NEXT: cmpl %esi, %edi
|
|
; CHECK-NEXT: cmovael %ecx, %eax
|
|
; CHECK-NEXT: retq
|
|
%cmp0 = icmp ult i32 %a0, %a1
|
|
%cmp1 = fcmp olt float %a2, %a3
|
|
%and = and i1 %cmp0, %cmp1
|
|
%res = select i1 %and, i32 %a4, i32 %a5
|
|
ret i32 %res
|
|
}
|
|
|
|
; select with and i1 condition should be implemented as a series of 2 cmovs, not
|
|
; by producing two conditions and using and on them.
|
|
|
|
define dso_local i32 @select_or(i32 %a0, i32 %a1, float %a2, float %a3, i32 %a4, i32 %a5) {
|
|
; CHECK-LABEL: select_or:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: movl %ecx, %eax
|
|
; CHECK-NEXT: ucomiss %xmm0, %xmm1
|
|
; CHECK-NEXT: cmoval %edx, %eax
|
|
; CHECK-NEXT: cmpl %esi, %edi
|
|
; CHECK-NEXT: cmovbl %edx, %eax
|
|
; CHECK-NEXT: retq
|
|
%cmp0 = icmp ult i32 %a0, %a1
|
|
%cmp1 = fcmp olt float %a2, %a3
|
|
%and = or i1 %cmp0, %cmp1
|
|
%res = select i1 %and, i32 %a4, i32 %a5
|
|
ret i32 %res
|
|
}
|
|
|
|
; If one of the conditions is materialized as a 0/1 value anyway, then the
|
|
; sequence of 2 cmovs should not be used.
|
|
|
|
@var32 = dso_local global i32 0
|
|
define dso_local i32 @select_noopt(i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %a4) {
|
|
; CHECK-LABEL: select_noopt:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: movl %ecx, %eax
|
|
; CHECK-NEXT: cmpl %esi, %edi
|
|
; CHECK-NEXT: setb %cl
|
|
; CHECK-NEXT: cmpl %edx, %esi
|
|
; CHECK-NEXT: setb %dl
|
|
; CHECK-NEXT: orb %cl, %dl
|
|
; CHECK-NEXT: movzbl %dl, %ecx
|
|
; CHECK-NEXT: movl %ecx, var32(%rip)
|
|
; CHECK-NEXT: testb %cl, %cl
|
|
; CHECK-NEXT: cmovel %r8d, %eax
|
|
; CHECK-NEXT: retq
|
|
%cmp0 = icmp ult i32 %a0, %a1
|
|
%cmp1 = icmp ult i32 %a1, %a2
|
|
%or = or i1 %cmp0, %cmp1
|
|
%zero_one = zext i1 %or to i32
|
|
store volatile i32 %zero_one, i32* @var32
|
|
%res = select i1 %or, i32 %a3, i32 %a4
|
|
ret i32 %res
|
|
}
|