641 lines
20 KiB
LLVM
641 lines
20 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -force-streaming-compatible-sve < %s | FileCheck %s
|
|
|
|
target triple = "aarch64-unknown-linux-gnu"
|
|
|
|
;
|
|
; ASHR
|
|
;
|
|
|
|
define <4 x i8> @ashr_v4i8(<4 x i8> %op1, <4 x i8> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v4i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI0_0
|
|
; CHECK-NEXT: adrp x9, .LCPI0_1
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: ldr d2, [x8, :lo12:.LCPI0_0]
|
|
; CHECK-NEXT: ldr d3, [x9, :lo12:.LCPI0_1]
|
|
; CHECK-NEXT: lsl z0.h, p0/m, z0.h, z2.h
|
|
; CHECK-NEXT: asr z0.h, p0/m, z0.h, z2.h
|
|
; CHECK-NEXT: and z1.d, z1.d, z3.d
|
|
; CHECK-NEXT: asr z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <4 x i8> %op1, %op2
|
|
ret <4 x i8> %res
|
|
}
|
|
|
|
define <8 x i8> @ashr_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: asr z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <8 x i8> %op1, %op2
|
|
ret <8 x i8> %res
|
|
}
|
|
|
|
define <16 x i8> @ashr_v16i8(<16 x i8> %op1, <16 x i8> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: asr z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <16 x i8> %op1, %op2
|
|
ret <16 x i8> %res
|
|
}
|
|
|
|
define void @ashr_v32i8(<32 x i8>* %a, <32 x i8>* %b) #0 {
|
|
; CHECK-LABEL: ashr_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: asr z0.b, p0/m, z0.b, z2.b
|
|
; CHECK-NEXT: asr z1.b, p0/m, z1.b, z3.b
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <32 x i8>, <32 x i8>* %a
|
|
%op2 = load <32 x i8>, <32 x i8>* %b
|
|
%res = ashr <32 x i8> %op1, %op2
|
|
store <32 x i8> %res, <32 x i8>* %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x i16> @ashr_v2i16(<2 x i16> %op1, <2 x i16> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v2i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI4_0
|
|
; CHECK-NEXT: adrp x9, .LCPI4_1
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: ldr d2, [x8, :lo12:.LCPI4_0]
|
|
; CHECK-NEXT: ldr d3, [x9, :lo12:.LCPI4_1]
|
|
; CHECK-NEXT: lsl z0.s, p0/m, z0.s, z2.s
|
|
; CHECK-NEXT: asr z0.s, p0/m, z0.s, z2.s
|
|
; CHECK-NEXT: and z1.d, z1.d, z3.d
|
|
; CHECK-NEXT: asr z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <2 x i16> %op1, %op2
|
|
ret <2 x i16> %res
|
|
}
|
|
|
|
define <4 x i16> @ashr_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: asr z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <4 x i16> %op1, %op2
|
|
ret <4 x i16> %res
|
|
}
|
|
|
|
define <8 x i16> @ashr_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: asr z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <8 x i16> %op1, %op2
|
|
ret <8 x i16> %res
|
|
}
|
|
|
|
define void @ashr_v16i16(<16 x i16>* %a, <16 x i16>* %b) #0 {
|
|
; CHECK-LABEL: ashr_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: asr z0.h, p0/m, z0.h, z2.h
|
|
; CHECK-NEXT: asr z1.h, p0/m, z1.h, z3.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <16 x i16>, <16 x i16>* %a
|
|
%op2 = load <16 x i16>, <16 x i16>* %b
|
|
%res = ashr <16 x i16> %op1, %op2
|
|
store <16 x i16> %res, <16 x i16>* %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x i32> @ashr_v2i32(<2 x i32> %op1, <2 x i32> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: asr z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <2 x i32> %op1, %op2
|
|
ret <2 x i32> %res
|
|
}
|
|
|
|
define <4 x i32> @ashr_v4i32(<4 x i32> %op1, <4 x i32> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: asr z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <4 x i32> %op1, %op2
|
|
ret <4 x i32> %res
|
|
}
|
|
|
|
define void @ashr_v8i32(<8 x i32>* %a, <8 x i32>* %b) #0 {
|
|
; CHECK-LABEL: ashr_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: asr z0.s, p0/m, z0.s, z2.s
|
|
; CHECK-NEXT: asr z1.s, p0/m, z1.s, z3.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <8 x i32>, <8 x i32>* %a
|
|
%op2 = load <8 x i32>, <8 x i32>* %b
|
|
%res = ashr <8 x i32> %op1, %op2
|
|
store <8 x i32> %res, <8 x i32>* %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x i64> @ashr_v1i64(<1 x i64> %op1, <1 x i64> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v1i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl1
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: asr z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <1 x i64> %op1, %op2
|
|
ret <1 x i64> %res
|
|
}
|
|
|
|
define <2 x i64> @ashr_v2i64(<2 x i64> %op1, <2 x i64> %op2) #0 {
|
|
; CHECK-LABEL: ashr_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: asr z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = ashr <2 x i64> %op1, %op2
|
|
ret <2 x i64> %res
|
|
}
|
|
|
|
define void @ashr_v4i64(<4 x i64>* %a, <4 x i64>* %b) #0 {
|
|
; CHECK-LABEL: ashr_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: asr z0.d, p0/m, z0.d, z2.d
|
|
; CHECK-NEXT: asr z1.d, p0/m, z1.d, z3.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <4 x i64>, <4 x i64>* %a
|
|
%op2 = load <4 x i64>, <4 x i64>* %b
|
|
%res = ashr <4 x i64> %op1, %op2
|
|
store <4 x i64> %res, <4 x i64>* %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; LSHR
|
|
;
|
|
|
|
define <4 x i8> @lshr_v4i8(<4 x i8> %op1, <4 x i8> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v4i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI14_0
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: ldr d2, [x8, :lo12:.LCPI14_0]
|
|
; CHECK-NEXT: and z1.d, z1.d, z2.d
|
|
; CHECK-NEXT: and z0.d, z0.d, z2.d
|
|
; CHECK-NEXT: lsr z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <4 x i8> %op1, %op2
|
|
ret <4 x i8> %res
|
|
}
|
|
|
|
define <8 x i8> @lshr_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsr z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <8 x i8> %op1, %op2
|
|
ret <8 x i8> %res
|
|
}
|
|
|
|
define <16 x i8> @lshr_v16i8(<16 x i8> %op1, <16 x i8> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsr z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <16 x i8> %op1, %op2
|
|
ret <16 x i8> %res
|
|
}
|
|
|
|
define void @lshr_v32i8(<32 x i8>* %a, <32 x i8>* %b) #0 {
|
|
; CHECK-LABEL: lshr_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsr z0.b, p0/m, z0.b, z2.b
|
|
; CHECK-NEXT: lsr z1.b, p0/m, z1.b, z3.b
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <32 x i8>, <32 x i8>* %a
|
|
%op2 = load <32 x i8>, <32 x i8>* %b
|
|
%res = lshr <32 x i8> %op1, %op2
|
|
store <32 x i8> %res, <32 x i8>* %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x i16> @lshr_v2i16(<2 x i16> %op1, <2 x i16> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v2i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI18_0
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: ldr d2, [x8, :lo12:.LCPI18_0]
|
|
; CHECK-NEXT: and z1.d, z1.d, z2.d
|
|
; CHECK-NEXT: and z0.d, z0.d, z2.d
|
|
; CHECK-NEXT: lsr z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <2 x i16> %op1, %op2
|
|
ret <2 x i16> %res
|
|
}
|
|
|
|
define <4 x i16> @lshr_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsr z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <4 x i16> %op1, %op2
|
|
ret <4 x i16> %res
|
|
}
|
|
|
|
define <8 x i16> @lshr_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsr z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <8 x i16> %op1, %op2
|
|
ret <8 x i16> %res
|
|
}
|
|
|
|
define void @lshr_v16i16(<16 x i16>* %a, <16 x i16>* %b) #0 {
|
|
; CHECK-LABEL: lshr_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsr z0.h, p0/m, z0.h, z2.h
|
|
; CHECK-NEXT: lsr z1.h, p0/m, z1.h, z3.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <16 x i16>, <16 x i16>* %a
|
|
%op2 = load <16 x i16>, <16 x i16>* %b
|
|
%res = lshr <16 x i16> %op1, %op2
|
|
store <16 x i16> %res, <16 x i16>* %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x i32> @lshr_v2i32(<2 x i32> %op1, <2 x i32> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsr z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <2 x i32> %op1, %op2
|
|
ret <2 x i32> %res
|
|
}
|
|
|
|
define <4 x i32> @lshr_v4i32(<4 x i32> %op1, <4 x i32> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsr z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <4 x i32> %op1, %op2
|
|
ret <4 x i32> %res
|
|
}
|
|
|
|
define void @lshr_v8i32(<8 x i32>* %a, <8 x i32>* %b) #0 {
|
|
; CHECK-LABEL: lshr_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsr z0.s, p0/m, z0.s, z2.s
|
|
; CHECK-NEXT: lsr z1.s, p0/m, z1.s, z3.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <8 x i32>, <8 x i32>* %a
|
|
%op2 = load <8 x i32>, <8 x i32>* %b
|
|
%res = lshr <8 x i32> %op1, %op2
|
|
store <8 x i32> %res, <8 x i32>* %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x i64> @lshr_v1i64(<1 x i64> %op1, <1 x i64> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v1i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl1
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsr z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <1 x i64> %op1, %op2
|
|
ret <1 x i64> %res
|
|
}
|
|
|
|
define <2 x i64> @lshr_v2i64(<2 x i64> %op1, <2 x i64> %op2) #0 {
|
|
; CHECK-LABEL: lshr_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsr z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = lshr <2 x i64> %op1, %op2
|
|
ret <2 x i64> %res
|
|
}
|
|
|
|
define void @lshr_v4i64(<4 x i64>* %a, <4 x i64>* %b) #0 {
|
|
; CHECK-LABEL: lshr_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsr z0.d, p0/m, z0.d, z2.d
|
|
; CHECK-NEXT: lsr z1.d, p0/m, z1.d, z3.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <4 x i64>, <4 x i64>* %a
|
|
%op2 = load <4 x i64>, <4 x i64>* %b
|
|
%res = lshr <4 x i64> %op1, %op2
|
|
store <4 x i64> %res, <4 x i64>* %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; SHL
|
|
;
|
|
|
|
define <2 x i8> @shl_v2i8(<2 x i8> %op1, <2 x i8> %op2) #0 {
|
|
; CHECK-LABEL: shl_v2i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI28_0
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: ldr d2, [x8, :lo12:.LCPI28_0]
|
|
; CHECK-NEXT: and z1.d, z1.d, z2.d
|
|
; CHECK-NEXT: lsl z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <2 x i8> %op1, %op2
|
|
ret <2 x i8> %res
|
|
}
|
|
|
|
define <4 x i8> @shl_v4i8(<4 x i8> %op1, <4 x i8> %op2) #0 {
|
|
; CHECK-LABEL: shl_v4i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI29_0
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: ldr d2, [x8, :lo12:.LCPI29_0]
|
|
; CHECK-NEXT: and z1.d, z1.d, z2.d
|
|
; CHECK-NEXT: lsl z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <4 x i8> %op1, %op2
|
|
ret <4 x i8> %res
|
|
}
|
|
|
|
define <8 x i8> @shl_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
|
|
; CHECK-LABEL: shl_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsl z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <8 x i8> %op1, %op2
|
|
ret <8 x i8> %res
|
|
}
|
|
|
|
define <16 x i8> @shl_v16i8(<16 x i8> %op1, <16 x i8> %op2) #0 {
|
|
; CHECK-LABEL: shl_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsl z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <16 x i8> %op1, %op2
|
|
ret <16 x i8> %res
|
|
}
|
|
|
|
define void @shl_v32i8(<32 x i8>* %a, <32 x i8>* %b) #0 {
|
|
; CHECK-LABEL: shl_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsl z0.b, p0/m, z0.b, z2.b
|
|
; CHECK-NEXT: lsl z1.b, p0/m, z1.b, z3.b
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <32 x i8>, <32 x i8>* %a
|
|
%op2 = load <32 x i8>, <32 x i8>* %b
|
|
%res = shl <32 x i8> %op1, %op2
|
|
store <32 x i8> %res, <32 x i8>* %a
|
|
ret void
|
|
}
|
|
|
|
define <4 x i16> @shl_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
|
|
; CHECK-LABEL: shl_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsl z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <4 x i16> %op1, %op2
|
|
ret <4 x i16> %res
|
|
}
|
|
|
|
define <8 x i16> @shl_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
|
|
; CHECK-LABEL: shl_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsl z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <8 x i16> %op1, %op2
|
|
ret <8 x i16> %res
|
|
}
|
|
|
|
define void @shl_v16i16(<16 x i16>* %a, <16 x i16>* %b) #0 {
|
|
; CHECK-LABEL: shl_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsl z0.h, p0/m, z0.h, z2.h
|
|
; CHECK-NEXT: lsl z1.h, p0/m, z1.h, z3.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <16 x i16>, <16 x i16>* %a
|
|
%op2 = load <16 x i16>, <16 x i16>* %b
|
|
%res = shl <16 x i16> %op1, %op2
|
|
store <16 x i16> %res, <16 x i16>* %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x i32> @shl_v2i32(<2 x i32> %op1, <2 x i32> %op2) #0 {
|
|
; CHECK-LABEL: shl_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsl z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <2 x i32> %op1, %op2
|
|
ret <2 x i32> %res
|
|
}
|
|
|
|
define <4 x i32> @shl_v4i32(<4 x i32> %op1, <4 x i32> %op2) #0 {
|
|
; CHECK-LABEL: shl_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsl z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <4 x i32> %op1, %op2
|
|
ret <4 x i32> %res
|
|
}
|
|
|
|
define void @shl_v8i32(<8 x i32>* %a, <8 x i32>* %b) #0 {
|
|
; CHECK-LABEL: shl_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsl z0.s, p0/m, z0.s, z2.s
|
|
; CHECK-NEXT: lsl z1.s, p0/m, z1.s, z3.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <8 x i32>, <8 x i32>* %a
|
|
%op2 = load <8 x i32>, <8 x i32>* %b
|
|
%res = shl <8 x i32> %op1, %op2
|
|
store <8 x i32> %res, <8 x i32>* %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x i64> @shl_v1i64(<1 x i64> %op1, <1 x i64> %op2) #0 {
|
|
; CHECK-LABEL: shl_v1i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl1
|
|
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
|
|
; CHECK-NEXT: lsl z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <1 x i64> %op1, %op2
|
|
ret <1 x i64> %res
|
|
}
|
|
|
|
define <2 x i64> @shl_v2i64(<2 x i64> %op1, <2 x i64> %op2) #0 {
|
|
; CHECK-LABEL: shl_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; CHECK-NEXT: lsl z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; CHECK-NEXT: ret
|
|
%res = shl <2 x i64> %op1, %op2
|
|
ret <2 x i64> %res
|
|
}
|
|
|
|
define void @shl_v4i64(<4 x i64>* %a, <4 x i64>* %b) #0 {
|
|
; CHECK-LABEL: shl_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: ldp q2, q3, [x1]
|
|
; CHECK-NEXT: lsl z0.d, p0/m, z0.d, z2.d
|
|
; CHECK-NEXT: lsl z1.d, p0/m, z1.d, z3.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
%op1 = load <4 x i64>, <4 x i64>* %a
|
|
%op2 = load <4 x i64>, <4 x i64>* %b
|
|
%res = shl <4 x i64> %op1, %op2
|
|
store <4 x i64> %res, <4 x i64>* %a
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { "target-features"="+sve" }
|