llvm-mirror/test/CodeGen/AArch64/cond-sel.ll
Jessica Paquette 98b9ddd887 Recommit "Enable MachineOutliner by default under -Oz for AArch64"
Fixed the ASAN failure from before in r338148, so recommiting.

This patch enables the MachineOutliner by default in AArch64 under -Oz.

The MachineOutliner offers around a 4.5% improvement on the current -Oz code
size improvements.

We have done work into improving the debuggability of outlined code, so that
users of -Oz won't be surprised by the optimization. We have also been executing
the LLVM test suite and common external tests such as the SPEC suites
continuously with no issue. The outliner has a low compile-time overhead of
roughly 1%. At this point, the outliner would be a really good addition to the
-Oz pass pipeline!

llvm-svn: 338160
2018-07-27 20:18:27 +00:00

262 lines
8.2 KiB
LLVM

; RUN: llc -enable-machine-outliner=never -verify-machineinstrs < %s -mtriple=aarch64-none-linux-gnu -mcpu=cyclone | FileCheck %s
; RUN: llc -enable-machine-outliner=never -verify-machineinstrs < %s -mtriple=aarch64-none-linux-gnu -mattr=-fp-armv8 | FileCheck --check-prefix=CHECK-NOFP %s
@var32 = global i32 0
@var64 = global i64 0
define void @test_csel(i32 %lhs32, i32 %rhs32, i64 %lhs64) minsize {
; CHECK-LABEL: test_csel:
%tst1 = icmp ugt i32 %lhs32, %rhs32
%val1 = select i1 %tst1, i32 42, i32 52
store i32 %val1, i32* @var32
; CHECK-DAG: mov [[W52:w[0-9]+]], #{{52|0x34}}
; CHECK-DAG: mov [[W42:w[0-9]+]], #{{42|0x2a}}
; CHECK: csel {{w[0-9]+}}, [[W42]], [[W52]], hi
%rhs64 = sext i32 %rhs32 to i64
%tst2 = icmp sle i64 %lhs64, %rhs64
%val2 = select i1 %tst2, i64 %lhs64, i64 %rhs64
store i64 %val2, i64* @var64
; CHECK: sxtw [[EXT_RHS:x[0-9]+]], {{[wx]}}[[RHS:[0-9]+]]
; CHECK: cmp [[LHS:x[0-9]+]], w[[RHS]], sxtw
; CHECK: csel {{x[0-9]+}}, [[LHS]], [[EXT_RHS]], le
ret void
; CHECK: ret
}
define void @test_floatcsel(float %lhs32, float %rhs32, double %lhs64, double %rhs64) {
; CHECK-LABEL: test_floatcsel:
%tst1 = fcmp one float %lhs32, %rhs32
; CHECK: fcmp {{s[0-9]+}}, {{s[0-9]+}}
; CHECK-NOFP-NOT: fcmp
%val1 = select i1 %tst1, i32 42, i32 52
store i32 %val1, i32* @var32
; CHECK: mov [[W52:w[0-9]+]], #{{52|0x34}}
; CHECK: mov [[W42:w[0-9]+]], #{{42|0x2a}}
; CHECK: csel [[MAYBETRUE:w[0-9]+]], [[W42]], [[W52]], mi
; CHECK: csel {{w[0-9]+}}, [[W42]], [[MAYBETRUE]], gt
%tst2 = fcmp ueq double %lhs64, %rhs64
; CHECK: fcmp {{d[0-9]+}}, {{d[0-9]+}}
; CHECK-NOFP-NOT: fcmp
%val2 = select i1 %tst2, i64 9, i64 15
store i64 %val2, i64* @var64
; CHECK: orr w[[CONST15:[0-9]+]], wzr, #0xf
; CHECK: mov {{[wx]}}[[CONST9:[0-9]+]], #{{9|0x9}}
; CHECK: csel [[MAYBETRUE:x[0-9]+]], x[[CONST9]], x[[CONST15]], eq
; CHECK: csel {{x[0-9]+}}, x[[CONST9]], [[MAYBETRUE]], vs
ret void
; CHECK: ret
}
define void @test_csinc(i32 %lhs32, i32 %rhs32, i64 %lhs64) minsize {
; CHECK-LABEL: test_csinc:
; Note that commuting rhs and lhs in the select changes ugt to ule (i.e. hi to ls).
%tst1 = icmp ugt i32 %lhs32, %rhs32
%inc1 = add i32 %rhs32, 1
%val1 = select i1 %tst1, i32 %inc1, i32 %lhs32
store volatile i32 %val1, i32* @var32
; CHECK: cmp [[LHS:w[0-9]+]], [[RHS:w[0-9]+]]
; CHECK: csinc {{w[0-9]+}}, [[LHS]], [[RHS]], ls
%rhs2 = add i32 %rhs32, 42
%tst2 = icmp sle i32 %lhs32, %rhs2
%inc2 = add i32 %rhs32, 1
%val2 = select i1 %tst2, i32 %lhs32, i32 %inc2
store volatile i32 %val2, i32* @var32
; CHECK: cmp [[LHS:w[0-9]+]], {{w[0-9]+}}
; CHECK: csinc {{w[0-9]+}}, [[LHS]], {{w[0-9]+}}, le
; Note that commuting rhs and lhs in the select changes ugt to ule (i.e. hi to ls).
%rhs3 = sext i32 %rhs32 to i64
%tst3 = icmp ugt i64 %lhs64, %rhs3
%inc3 = add i64 %rhs3, 1
%val3 = select i1 %tst3, i64 %inc3, i64 %lhs64
store volatile i64 %val3, i64* @var64
; CHECK: cmp [[LHS:x[0-9]+]], {{w[0-9]+}}
; CHECK: csinc {{x[0-9]+}}, [[LHS]], {{x[0-9]+}}, ls
%rhs4 = zext i32 %rhs32 to i64
%tst4 = icmp sle i64 %lhs64, %rhs4
%inc4 = add i64 %rhs4, 1
%val4 = select i1 %tst4, i64 %lhs64, i64 %inc4
store volatile i64 %val4, i64* @var64
; CHECK: cmp [[LHS:x[0-9]+]], {{w[0-9]+}}
; CHECK: csinc {{x[0-9]+}}, [[LHS]], {{x[0-9]+}}, le
ret void
; CHECK: ret
}
define void @test_csinv(i32 %lhs32, i32 %rhs32, i64 %lhs64) minsize {
; CHECK-LABEL: test_csinv:
; Note that commuting rhs and lhs in the select changes ugt to ule (i.e. hi to ls).
%tst1 = icmp ugt i32 %lhs32, %rhs32
%inc1 = xor i32 -1, %rhs32
%val1 = select i1 %tst1, i32 %inc1, i32 %lhs32
store volatile i32 %val1, i32* @var32
; CHECK: cmp [[LHS:w[0-9]+]], [[RHS:w[0-9]+]]
; CHECK: csinv {{w[0-9]+}}, [[LHS]], [[RHS]], ls
%rhs2 = add i32 %rhs32, 42
%tst2 = icmp sle i32 %lhs32, %rhs2
%inc2 = xor i32 -1, %rhs32
%val2 = select i1 %tst2, i32 %lhs32, i32 %inc2
store volatile i32 %val2, i32* @var32
; CHECK: cmp [[LHS:w[0-9]+]], {{w[0-9]+}}
; CHECK: csinv {{w[0-9]+}}, [[LHS]], {{w[0-9]+}}, le
; Note that commuting rhs and lhs in the select changes ugt to ule (i.e. hi to ls).
%rhs3 = sext i32 %rhs32 to i64
%tst3 = icmp ugt i64 %lhs64, %rhs3
%inc3 = xor i64 -1, %rhs3
%val3 = select i1 %tst3, i64 %inc3, i64 %lhs64
store volatile i64 %val3, i64* @var64
; CHECK: cmp [[LHS:x[0-9]+]], {{w[0-9]+}}
; CHECK: csinv {{x[0-9]+}}, [[LHS]], {{x[0-9]+}}, ls
%rhs4 = zext i32 %rhs32 to i64
%tst4 = icmp sle i64 %lhs64, %rhs4
%inc4 = xor i64 -1, %rhs4
%val4 = select i1 %tst4, i64 %lhs64, i64 %inc4
store volatile i64 %val4, i64* @var64
; CHECK: cmp [[LHS:x[0-9]+]], {{w[0-9]+}}
; CHECK: csinv {{x[0-9]+}}, [[LHS]], {{x[0-9]+}}, le
ret void
; CHECK: ret
}
define void @test_csinv0(i32 %lhs32, i32 %rhs32, i64 %lhs64, i64 %rhs64) minsize {
; CHECK-LABEL: test_csinv0:
%tst1 = icmp ugt i32 %lhs32, %rhs32
%val1 = select i1 %tst1, i32 0, i32 -1
store volatile i32 %val1, i32* @var32
; CHECK: cmp [[LHS:w[0-9]+]], [[RHS:w[0-9]+]]
; CHECK: csetm {{w[0-9]+}}, ls
%rhs2 = add i32 %rhs32, 42
%tst2 = icmp sle i32 %lhs32, %rhs2
%val2 = select i1 %tst2, i32 -1, i32 %rhs2
store volatile i32 %val2, i32* @var32
; CHECK: cmp [[LHS2:w[0-9]+]], [[RHS2:w[0-9]+]]
; CHECK: csinv {{w[0-9]+}}, [[RHS2]], wzr, gt
; Note that commuting rhs and lhs in the select changes ugt to ule (i.e. hi to ls).
%rhs3 = mul i64 %rhs64, 19
%tst3 = icmp ugt i64 %lhs64, %rhs3
%val3 = select i1 %tst3, i64 %rhs3, i64 -1
store volatile i64 %val3, i64* @var64
; CHECK: cmp [[LHS3:x[0-9]+]], [[RHS3:x[0-9]+]]
; CHECK: csinv {{x[0-9]+}}, [[RHS3]], xzr, hi
ret void
; CHECK: ret
}
define void @test_csneg(i32 %lhs32, i32 %rhs32, i64 %lhs64) minsize {
; CHECK-LABEL: test_csneg:
; Note that commuting rhs and lhs in the select changes ugt to ule (i.e. hi to ls).
%tst1 = icmp ugt i32 %lhs32, %rhs32
%inc1 = sub i32 0, %rhs32
%val1 = select i1 %tst1, i32 %inc1, i32 %lhs32
store volatile i32 %val1, i32* @var32
; CHECK: cmp [[LHS:w[0-9]+]], [[RHS:w[0-9]+]]
; CHECK: csneg {{w[0-9]+}}, [[LHS]], [[RHS]], ls
%rhs2 = add i32 %rhs32, 42
%tst2 = icmp sle i32 %lhs32, %rhs2
%inc2 = sub i32 0, %rhs32
%val2 = select i1 %tst2, i32 %lhs32, i32 %inc2
store volatile i32 %val2, i32* @var32
; CHECK: cmp [[LHS:w[0-9]+]], {{w[0-9]+}}
; CHECK: csneg {{w[0-9]+}}, [[LHS]], {{w[0-9]+}}, le
; Note that commuting rhs and lhs in the select changes ugt to ule (i.e. hi to ls).
%rhs3 = sext i32 %rhs32 to i64
%tst3 = icmp ugt i64 %lhs64, %rhs3
%inc3 = sub i64 0, %rhs3
%val3 = select i1 %tst3, i64 %inc3, i64 %lhs64
store volatile i64 %val3, i64* @var64
; CHECK: cmp [[LHS:x[0-9]+]], {{w[0-9]+}}
; CHECK: csneg {{x[0-9]+}}, [[LHS]], {{x[0-9]+}}, ls
%rhs4 = zext i32 %rhs32 to i64
%tst4 = icmp sle i64 %lhs64, %rhs4
%inc4 = sub i64 0, %rhs4
%val4 = select i1 %tst4, i64 %lhs64, i64 %inc4
store volatile i64 %val4, i64* @var64
; CHECK: cmp [[LHS:x[0-9]+]], {{w[0-9]+}}
; CHECK: csneg {{x[0-9]+}}, [[LHS]], {{x[0-9]+}}, le
ret void
; CHECK: ret
}
define void @test_cset(i32 %lhs, i32 %rhs, i64 %lhs64) {
; CHECK-LABEL: test_cset:
; N.b. code is not optimal here (32-bit csinc would be better) but
; incoming DAG is too complex
%tst1 = icmp eq i32 %lhs, %rhs
%val1 = zext i1 %tst1 to i32
store i32 %val1, i32* @var32
; CHECK: cmp {{w[0-9]+}}, {{w[0-9]+}}
; CHECK: cset {{w[0-9]+}}, eq
%rhs64 = sext i32 %rhs to i64
%tst2 = icmp ule i64 %lhs64, %rhs64
%val2 = zext i1 %tst2 to i64
store i64 %val2, i64* @var64
; CHECK: cset {{w[0-9]+}}, ls
ret void
; CHECK: ret
}
define void @test_csetm(i32 %lhs, i32 %rhs, i64 %lhs64) {
; CHECK-LABEL: test_csetm:
%tst1 = icmp eq i32 %lhs, %rhs
%val1 = sext i1 %tst1 to i32
store i32 %val1, i32* @var32
; CHECK: cmp {{w[0-9]+}}, {{w[0-9]+}}
; CHECK: csetm {{w[0-9]+}}, eq
%rhs64 = sext i32 %rhs to i64
%tst2 = icmp ule i64 %lhs64, %rhs64
%val2 = sext i1 %tst2 to i64
store i64 %val2, i64* @var64
; CHECK: csetm {{x[0-9]+}}, ls
ret void
; CHECK: ret
}
define <1 x i1> @test_wide_comparison(i32 %in) {
; CHECK-LABEL: test_wide_comparison:
; CHECK: cmp w0, #1234
; CHECK: cset
%tmp = icmp sgt i32 %in, 1234
%res = select i1 %tmp, <1 x i1> <i1 1>, <1 x i1> zeroinitializer
ret <1 x i1> %res
}
define i32 @test_select_undef() {
; CHECK-LABEL: test_select_undef:
; CHECK: ret
%res = select i1 undef, i32 0, i32 42
ret i32 %res
}