On processors supporting vector registers and SIMD instructions, enable i128 as legal type in VRs. This allows many operations to be implemented via native instructions directly in VRs (including add, subtract, logical operations and shifts). For a few other operations (e.g. multiply and divide, as well as atomic operations), we need to move the i128 value back to a GPR pair to use the corresponding instruction there. Overall, this is still beneficial. The patch includes the following LLVM changes: - Enable i128 as legal type - Set up legal operations (in SystemZInstrVector.td) - Custom expansion for i128 add/subtract with carry - Custom expansion for i128 comparisons and selects - Support for moving i128 to/from GPR pairs when required - Handle 128-bit integer constant values everywhere - Use i128 as intrinsic operand type where appropriate - Updated and new test cases In addition, clang builtins are updated to reflect the intrinsic operand type changes (which also improves compatibility with GCC).
238 lines
6.7 KiB
LLVM
238 lines
6.7 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
|
|
; Test 128-bit comparisons in vector registers on z13
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 -verify-machineinstrs | FileCheck %s
|
|
|
|
; Equality comparison.
|
|
define i64 @f1(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: vceqgs %v0, %v1, %v0
|
|
; CHECK-NEXT: locgrnhe %r4, %r5
|
|
; CHECK-NEXT: lgr %r2, %r4
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp eq i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Inequality comparison.
|
|
define i64 @f2(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: vceqgs %v0, %v1, %v0
|
|
; CHECK-NEXT: locgre %r4, %r5
|
|
; CHECK-NEXT: lgr %r2, %r4
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp ne i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Signed greater-than comparison.
|
|
define i64 @f3(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: vecg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB2_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB2_2:
|
|
; CHECK-NEXT: locgrl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp sgt i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Signed less-than comparison.
|
|
define i64 @f4(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r3), 3
|
|
; CHECK-NEXT: vecg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB3_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB3_2:
|
|
; CHECK-NEXT: locgrl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp slt i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Signed greater-or-equal comparison.
|
|
define i64 @f5(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r3), 3
|
|
; CHECK-NEXT: vecg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB4_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB4_2:
|
|
; CHECK-NEXT: locgrnl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp sge i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Signed less-or-equal comparison.
|
|
define i64 @f6(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: vecg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB5_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB5_2:
|
|
; CHECK-NEXT: locgrnl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp sle i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Unsigned greater-than comparison.
|
|
define i64 @f7(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: veclg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB6_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB6_2:
|
|
; CHECK-NEXT: locgrl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp ugt i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Unsigned less-than comparison.
|
|
define i64 @f8(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f8:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r3), 3
|
|
; CHECK-NEXT: veclg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB7_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB7_2:
|
|
; CHECK-NEXT: locgrl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp ult i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Unsigned greater-or-equal comparison.
|
|
define i64 @f9(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f9:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r3), 3
|
|
; CHECK-NEXT: veclg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB8_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB8_2:
|
|
; CHECK-NEXT: locgrnl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp uge i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Unsigned less-or-equal comparison.
|
|
define i64 @f10(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f10:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: veclg %v0, %v1
|
|
; CHECK-NEXT: jlh .LBB9_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vchlgs %v0, %v1, %v0
|
|
; CHECK-NEXT: .LBB9_2:
|
|
; CHECK-NEXT: locgrnl %r5, %r4
|
|
; CHECK-NEXT: lgr %r2, %r5
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp ule i128 %value1, %value2
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Use VTM for "x & y == 0" comparison.
|
|
define i64 @f11(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f11:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: vtm %v1, %v0
|
|
; CHECK-NEXT: locgrnhe %r4, %r5
|
|
; CHECK-NEXT: lgr %r2, %r4
|
|
; CHECK-NEXT: br %r14
|
|
%and = and i128 %value1, %value2
|
|
%cond = icmp eq i128 %and, 0
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Use VTM for "x & y != 0" comparison.
|
|
define i64 @f12(i128 %value1, i128 %value2, i64 %a, i64 %b) {
|
|
; CHECK-LABEL: f12:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r3), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r2), 3
|
|
; CHECK-NEXT: vtm %v1, %v0
|
|
; CHECK-NEXT: locgre %r4, %r5
|
|
; CHECK-NEXT: lgr %r2, %r4
|
|
; CHECK-NEXT: br %r14
|
|
%and = and i128 %value1, %value2
|
|
%cond = icmp ne i128 %and, 0
|
|
%res = select i1 %cond, i64 %a, i64 %b
|
|
ret i64 %res
|
|
}
|
|
|
|
; Select between i128 values.
|
|
define i128 @f13(i64 %value1, i64 %value2, i128 %a, i128 %b) {
|
|
; CHECK-LABEL: f13:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r5), 3
|
|
; CHECK-NEXT: cgrje %r3, %r4, .LBB12_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vl %v1, 0(%r6), 3
|
|
; CHECK-NEXT: vaq %v0, %v0, %v1
|
|
; CHECK-NEXT: .LBB12_2:
|
|
; CHECK-NEXT: vst %v0, 0(%r2), 3
|
|
; CHECK-NEXT: br %r14
|
|
%cond = icmp eq i64 %value1, %value2
|
|
%sum = add i128 %a, %b
|
|
%res = select i1 %cond, i128 %a, i128 %sum
|
|
ret i128 %res
|
|
}
|
|
|