907 lines
29 KiB
LLVM
907 lines
29 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"
|
|
|
|
;
|
|
; sext i1 -> i32
|
|
;
|
|
|
|
; NOTE: Covers the scenario where a SIGN_EXTEND_INREG is required, whose inreg
|
|
; type's element type is not byte based and thus cannot be lowered directly to
|
|
; an SVE instruction.
|
|
define void @sext_v8i1_v8i32(<8 x i1> %a, <8 x i32>* %out) #0 {
|
|
; CHECK-LABEL: sext_v8i1_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI0_0
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: uunpklo z2.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: ldr q1, [x8, :lo12:.LCPI0_0]
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: lsl z2.s, p0/m, z2.s, z1.s
|
|
; CHECK-NEXT: lsl z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: asr z2.s, p0/m, z2.s, z1.s
|
|
; CHECK-NEXT: asr z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: stp q2, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <8 x i1> %a to <8 x i32>
|
|
store <8 x i32> %b, <8 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; sext i3 -> i64
|
|
;
|
|
|
|
; NOTE: Covers the scenario where a SIGN_EXTEND_INREG is required, whose inreg
|
|
; type's element type is not power-of-2 based and thus cannot be lowered
|
|
; directly to an SVE instruction.
|
|
define void @sext_v4i3_v4i64(<4 x i3> %a, <4 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v4i3_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI1_0
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z2.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: ldr q1, [x8, :lo12:.LCPI1_0]
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: lsl z2.d, p0/m, z2.d, z1.d
|
|
; CHECK-NEXT: lsl z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: asr z2.d, p0/m, z2.d, z1.d
|
|
; CHECK-NEXT: asr z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: stp q2, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <4 x i3> %a to <4 x i64>
|
|
store <4 x i64> %b, <4 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; sext i8 -> i16
|
|
;
|
|
|
|
define void @sext_v16i8_v16i16(<16 x i8> %a, <16 x i16>* %out) #0 {
|
|
; CHECK-LABEL: sext_v16i8_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sunpklo z1.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <16 x i8> %a to <16 x i16>
|
|
store <16 x i16>%b, <16 x i16>* %out
|
|
ret void
|
|
}
|
|
|
|
; NOTE: Extra 'add' is to prevent the extend being combined with the load.
|
|
define void @sext_v32i8_v32i16(<32 x i8>* %in, <32 x i16>* %out) #0 {
|
|
; CHECK-LABEL: sext_v32i8_v32i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.b, z0.b, z0.b
|
|
; CHECK-NEXT: sunpklo z2.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.b, z1.b, z1.b
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: sunpklo z3.h, z1.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z1.h, z1.b
|
|
; CHECK-NEXT: stp q2, q0, [x1]
|
|
; CHECK-NEXT: stp q3, q1, [x1, #32]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i8>, <32 x i8>* %in
|
|
%b = add <32 x i8> %a, %a
|
|
%c = sext <32 x i8> %b to <32 x i16>
|
|
store <32 x i16> %c, <32 x i16>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; sext i8 -> i32
|
|
;
|
|
|
|
define void @sext_v8i8_v8i32(<8 x i8> %a, <8 x i32>* %out) #0 {
|
|
; CHECK-LABEL: sext_v8i8_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: sunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <8 x i8> %a to <8 x i32>
|
|
store <8 x i32>%b, <8 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v16i8_v16i32(<16 x i8> %a, <16 x i32>* %out) #0 {
|
|
; CHECK-LABEL: sext_v16i8_v16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sunpklo z1.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z2.s, z1.h
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z3.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q3, q0, [x0, #32]
|
|
; CHECK-NEXT: sunpklo z0.s, z1.h
|
|
; CHECK-NEXT: stp q2, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <16 x i8> %a to <16 x i32>
|
|
store <16 x i32> %b, <16 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v32i8_v32i32(<32 x i8>* %in, <32 x i32>* %out) #0 {
|
|
; CHECK-LABEL: sext_v32i8_v32i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.b, z0.b, z0.b
|
|
; CHECK-NEXT: sunpklo z2.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.b, z1.b, z1.b
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: sunpklo z3.h, z1.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z5.s, z3.h
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: sunpklo z1.h, z1.b
|
|
; CHECK-NEXT: sunpklo z3.s, z3.h
|
|
; CHECK-NEXT: sunpklo z4.s, z2.h
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: sunpklo z6.s, z0.h
|
|
; CHECK-NEXT: stp q5, q3, [x1, #64]
|
|
; CHECK-NEXT: sunpklo z5.s, z1.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z2.s, z2.h
|
|
; CHECK-NEXT: sunpklo z1.s, z1.h
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q4, q2, [x1]
|
|
; CHECK-NEXT: stp q6, q0, [x1, #32]
|
|
; CHECK-NEXT: stp q5, q1, [x1, #96]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i8>, <32 x i8>* %in
|
|
%b = add <32 x i8> %a, %a
|
|
%c = sext <32 x i8> %b to <32 x i32>
|
|
store <32 x i32> %c, <32 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; sext i8 -> i64
|
|
;
|
|
|
|
; NOTE: v4i8 is an unpacked typed stored within a v4i16 container. The sign
|
|
; extend is a two step process where the container is any_extend'd with the
|
|
; result feeding an inreg sign extend.
|
|
define void @sext_v4i8_v4i64(<4 x i8> %a, <4 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v4i8_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI7_0
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z2.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: ldr q1, [x8, :lo12:.LCPI7_0]
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: lsl z2.d, p0/m, z2.d, z1.d
|
|
; CHECK-NEXT: lsl z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: asr z2.d, p0/m, z2.d, z1.d
|
|
; CHECK-NEXT: asr z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: stp q2, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <4 x i8> %a to <4 x i64>
|
|
store <4 x i64>%b, <4 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v8i8_v8i64(<8 x i8> %a, <8 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v8i8_v8i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: sunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: sunpklo z2.d, z1.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z3.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z1.d, z1.s
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q2, q1, [x0]
|
|
; CHECK-NEXT: stp q3, q0, [x0, #32]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <8 x i8> %a to <8 x i64>
|
|
store <8 x i64>%b, <8 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v16i8_v16i64(<16 x i8> %a, <16 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v16i8_v16i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sunpklo z1.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z2.s, z1.h
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z3.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: sunpklo z1.s, z1.h
|
|
; CHECK-NEXT: sunpklo z6.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z4.d, z2.s
|
|
; CHECK-NEXT: sunpklo z5.d, z3.s
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: sunpklo z7.d, z1.s
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z3.d, z3.s
|
|
; CHECK-NEXT: sunpklo z2.d, z2.s
|
|
; CHECK-NEXT: stp q5, q3, [x0, #64]
|
|
; CHECK-NEXT: stp q4, q2, [x0]
|
|
; CHECK-NEXT: stp q6, q0, [x0, #96]
|
|
; CHECK-NEXT: sunpklo z0.d, z1.s
|
|
; CHECK-NEXT: stp q7, q0, [x0, #32]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <16 x i8> %a to <16 x i64>
|
|
store <16 x i64> %b, <16 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v32i8_v32i64(<32 x i8>* %in, <32 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v32i8_v32i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.b, z0.b, z0.b
|
|
; CHECK-NEXT: sunpklo z2.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.b, z1.b, z1.b
|
|
; CHECK-NEXT: sunpklo z0.h, z0.b
|
|
; CHECK-NEXT: sunpklo z3.h, z1.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z5.s, z3.h
|
|
; CHECK-NEXT: sunpklo z1.h, z1.b
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: sunpklo z4.s, z2.h
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: sunpklo z6.s, z0.h
|
|
; CHECK-NEXT: sunpklo z7.s, z1.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z3.s, z3.h
|
|
; CHECK-NEXT: sunpklo z2.s, z2.h
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: sunpklo z1.s, z1.h
|
|
; CHECK-NEXT: sunpklo z20.d, z3.s
|
|
; CHECK-NEXT: sunpklo z22.d, z4.s
|
|
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: sunpklo z16.d, z2.s
|
|
; CHECK-NEXT: sunpklo z17.d, z6.s
|
|
; CHECK-NEXT: sunpklo z18.d, z0.s
|
|
; CHECK-NEXT: sunpklo z19.d, z1.s
|
|
; CHECK-NEXT: sunpklo z21.d, z7.s
|
|
; CHECK-NEXT: sunpklo z23.d, z5.s
|
|
; CHECK-NEXT: ext z5.b, z5.b, z5.b, #8
|
|
; CHECK-NEXT: sunpklo z4.d, z4.s
|
|
; CHECK-NEXT: ext z7.b, z7.b, z7.b, #8
|
|
; CHECK-NEXT: sunpklo z3.d, z3.s
|
|
; CHECK-NEXT: ext z6.b, z6.b, z6.b, #8
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z5.d, z5.s
|
|
; CHECK-NEXT: stp q22, q4, [x1]
|
|
; CHECK-NEXT: sunpklo z4.d, z7.s
|
|
; CHECK-NEXT: stp q23, q5, [x1, #128]
|
|
; CHECK-NEXT: sunpklo z2.d, z2.s
|
|
; CHECK-NEXT: stp q20, q3, [x1, #160]
|
|
; CHECK-NEXT: sunpklo z3.d, z6.s
|
|
; CHECK-NEXT: sunpklo z1.d, z1.s
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q16, q2, [x1, #32]
|
|
; CHECK-NEXT: stp q17, q3, [x1, #64]
|
|
; CHECK-NEXT: stp q18, q0, [x1, #96]
|
|
; CHECK-NEXT: stp q21, q4, [x1, #192]
|
|
; CHECK-NEXT: stp q19, q1, [x1, #224]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i8>, <32 x i8>* %in
|
|
%b = add <32 x i8> %a, %a
|
|
%c = sext <32 x i8> %b to <32 x i64>
|
|
store <32 x i64> %c, <32 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; sext i16 -> i32
|
|
;
|
|
|
|
define void @sext_v8i16_v8i32(<8 x i16> %a, <8 x i32>* %out) #0 {
|
|
; CHECK-LABEL: sext_v8i16_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <8 x i16> %a to <8 x i32>
|
|
store <8 x i32>%b, <8 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v16i16_v16i32(<16 x i16>* %in, <16 x i32>* %out) #0 {
|
|
; CHECK-LABEL: sext_v16i16_v16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.h, z0.h, z0.h
|
|
; CHECK-NEXT: sunpklo z2.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.h, z1.h, z1.h
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: sunpklo z3.s, z1.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z1.s, z1.h
|
|
; CHECK-NEXT: stp q2, q0, [x1]
|
|
; CHECK-NEXT: stp q3, q1, [x1, #32]
|
|
; CHECK-NEXT: ret
|
|
%a = load <16 x i16>, <16 x i16>* %in
|
|
%b = add <16 x i16> %a, %a
|
|
%c = sext <16 x i16> %b to <16 x i32>
|
|
store <16 x i32> %c, <16 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; sext i16 -> i64
|
|
;
|
|
|
|
define void @sext_v4i16_v4i64(<4 x i16> %a, <4 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v4i16_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: sunpklo z1.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <4 x i16> %a to <4 x i64>
|
|
store <4 x i64>%b, <4 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v8i16_v8i64(<8 x i16> %a, <8 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v8i16_v8i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z2.d, z1.s
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z3.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q3, q0, [x0, #32]
|
|
; CHECK-NEXT: sunpklo z0.d, z1.s
|
|
; CHECK-NEXT: stp q2, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <8 x i16> %a to <8 x i64>
|
|
store <8 x i64>%b, <8 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v16i16_v16i64(<16 x i16>* %in, <16 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v16i16_v16i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.h, z0.h, z0.h
|
|
; CHECK-NEXT: sunpklo z2.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.h, z1.h, z1.h
|
|
; CHECK-NEXT: sunpklo z0.s, z0.h
|
|
; CHECK-NEXT: sunpklo z3.s, z1.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z5.d, z3.s
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: sunpklo z1.s, z1.h
|
|
; CHECK-NEXT: sunpklo z3.d, z3.s
|
|
; CHECK-NEXT: sunpklo z4.d, z2.s
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: sunpklo z6.d, z0.s
|
|
; CHECK-NEXT: stp q5, q3, [x1, #64]
|
|
; CHECK-NEXT: sunpklo z5.d, z1.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z2.d, z2.s
|
|
; CHECK-NEXT: sunpklo z1.d, z1.s
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q4, q2, [x1]
|
|
; CHECK-NEXT: stp q6, q0, [x1, #32]
|
|
; CHECK-NEXT: stp q5, q1, [x1, #96]
|
|
; CHECK-NEXT: ret
|
|
%a = load <16 x i16>, <16 x i16>* %in
|
|
%b = add <16 x i16> %a, %a
|
|
%c = sext <16 x i16> %b to <16 x i64>
|
|
store <16 x i64> %c, <16 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; sext i32 -> i64
|
|
;
|
|
|
|
define void @sext_v4i32_v4i64(<4 x i32> %a, <4 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v4i32_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sunpklo z1.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = sext <4 x i32> %a to <4 x i64>
|
|
store <4 x i64>%b, <4 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @sext_v8i32_v8i64(<8 x i32>* %in, <8 x i64>* %out) #0 {
|
|
; CHECK-LABEL: sext_v8i32_v8i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.s, z0.s, z0.s
|
|
; CHECK-NEXT: sunpklo z2.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.s, z1.s, z1.s
|
|
; CHECK-NEXT: sunpklo z0.d, z0.s
|
|
; CHECK-NEXT: sunpklo z3.d, z1.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: sunpklo z1.d, z1.s
|
|
; CHECK-NEXT: stp q2, q0, [x1]
|
|
; CHECK-NEXT: stp q3, q1, [x1, #32]
|
|
; CHECK-NEXT: ret
|
|
%a = load <8 x i32>, <8 x i32>* %in
|
|
%b = add <8 x i32> %a, %a
|
|
%c = sext <8 x i32> %b to <8 x i64>
|
|
store <8 x i64> %c, <8 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; zext i8 -> i16
|
|
;
|
|
|
|
define void @zext_v16i8_v16i16(<16 x i8> %a, <16 x i16>* %out) #0 {
|
|
; CHECK-LABEL: zext_v16i8_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uunpklo z1.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <16 x i8> %a to <16 x i16>
|
|
store <16 x i16>%b, <16 x i16>* %out
|
|
ret void
|
|
}
|
|
|
|
; NOTE: Extra 'add' is to prevent the extend being combined with the load.
|
|
define void @zext_v32i8_v32i16(<32 x i8>* %in, <32 x i16>* %out) #0 {
|
|
; CHECK-LABEL: zext_v32i8_v32i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.b, z0.b, z0.b
|
|
; CHECK-NEXT: uunpklo z2.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.b, z1.b, z1.b
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: uunpklo z3.h, z1.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z1.h, z1.b
|
|
; CHECK-NEXT: stp q2, q0, [x1]
|
|
; CHECK-NEXT: stp q3, q1, [x1, #32]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i8>, <32 x i8>* %in
|
|
%b = add <32 x i8> %a, %a
|
|
%c = zext <32 x i8> %b to <32 x i16>
|
|
store <32 x i16> %c, <32 x i16>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; zext i8 -> i32
|
|
;
|
|
|
|
define void @zext_v8i8_v8i32(<8 x i8> %a, <8 x i32>* %out) #0 {
|
|
; CHECK-LABEL: zext_v8i8_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: uunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <8 x i8> %a to <8 x i32>
|
|
store <8 x i32>%b, <8 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v16i8_v16i32(<16 x i8> %a, <16 x i32>* %out) #0 {
|
|
; CHECK-LABEL: zext_v16i8_v16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uunpklo z1.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z2.s, z1.h
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z3.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q3, q0, [x0, #32]
|
|
; CHECK-NEXT: uunpklo z0.s, z1.h
|
|
; CHECK-NEXT: stp q2, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <16 x i8> %a to <16 x i32>
|
|
store <16 x i32> %b, <16 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v32i8_v32i32(<32 x i8>* %in, <32 x i32>* %out) #0 {
|
|
; CHECK-LABEL: zext_v32i8_v32i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.b, z0.b, z0.b
|
|
; CHECK-NEXT: uunpklo z2.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.b, z1.b, z1.b
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: uunpklo z3.h, z1.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z5.s, z3.h
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: uunpklo z1.h, z1.b
|
|
; CHECK-NEXT: uunpklo z3.s, z3.h
|
|
; CHECK-NEXT: uunpklo z4.s, z2.h
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: uunpklo z6.s, z0.h
|
|
; CHECK-NEXT: stp q5, q3, [x1, #64]
|
|
; CHECK-NEXT: uunpklo z5.s, z1.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z2.s, z2.h
|
|
; CHECK-NEXT: uunpklo z1.s, z1.h
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q4, q2, [x1]
|
|
; CHECK-NEXT: stp q6, q0, [x1, #32]
|
|
; CHECK-NEXT: stp q5, q1, [x1, #96]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i8>, <32 x i8>* %in
|
|
%b = add <32 x i8> %a, %a
|
|
%c = zext <32 x i8> %b to <32 x i32>
|
|
store <32 x i32> %c, <32 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; zext i8 -> i64
|
|
;
|
|
|
|
; NOTE: v4i8 is an unpacked typed stored within a v4i16 container. The zero
|
|
; extend is a two step process where the container is zero_extend_inreg'd with
|
|
; the result feeding a normal zero extend from halfs to doublewords.
|
|
define void @zext_v4i8_v4i64(<4 x i8> %a, <4 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v4i8_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: adrp x8, .LCPI23_0
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: ldr d1, [x8, :lo12:.LCPI23_0]
|
|
; CHECK-NEXT: and z0.d, z0.d, z1.d
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z1.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <4 x i8> %a to <4 x i64>
|
|
store <4 x i64>%b, <4 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v8i8_v8i64(<8 x i8> %a, <8 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v8i8_v8i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: uunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z2.d, z1.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z3.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z1.d, z1.s
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q2, q1, [x0]
|
|
; CHECK-NEXT: stp q3, q0, [x0, #32]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <8 x i8> %a to <8 x i64>
|
|
store <8 x i64>%b, <8 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v16i8_v16i64(<16 x i8> %a, <16 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v16i8_v16i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uunpklo z1.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z2.s, z1.h
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z3.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z1.s, z1.h
|
|
; CHECK-NEXT: uunpklo z6.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z4.d, z2.s
|
|
; CHECK-NEXT: uunpklo z5.d, z3.s
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: uunpklo z7.d, z1.s
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z3.d, z3.s
|
|
; CHECK-NEXT: uunpklo z2.d, z2.s
|
|
; CHECK-NEXT: stp q5, q3, [x0, #64]
|
|
; CHECK-NEXT: stp q4, q2, [x0]
|
|
; CHECK-NEXT: stp q6, q0, [x0, #96]
|
|
; CHECK-NEXT: uunpklo z0.d, z1.s
|
|
; CHECK-NEXT: stp q7, q0, [x0, #32]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <16 x i8> %a to <16 x i64>
|
|
store <16 x i64> %b, <16 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v32i8_v32i64(<32 x i8>* %in, <32 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v32i8_v32i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.b, z0.b, z0.b
|
|
; CHECK-NEXT: uunpklo z2.h, z0.b
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.b, z1.b, z1.b
|
|
; CHECK-NEXT: uunpklo z0.h, z0.b
|
|
; CHECK-NEXT: uunpklo z3.h, z1.b
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z5.s, z3.h
|
|
; CHECK-NEXT: uunpklo z1.h, z1.b
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: uunpklo z4.s, z2.h
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: uunpklo z6.s, z0.h
|
|
; CHECK-NEXT: uunpklo z7.s, z1.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z3.s, z3.h
|
|
; CHECK-NEXT: uunpklo z2.s, z2.h
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z1.s, z1.h
|
|
; CHECK-NEXT: uunpklo z20.d, z3.s
|
|
; CHECK-NEXT: uunpklo z22.d, z4.s
|
|
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: uunpklo z16.d, z2.s
|
|
; CHECK-NEXT: uunpklo z17.d, z6.s
|
|
; CHECK-NEXT: uunpklo z18.d, z0.s
|
|
; CHECK-NEXT: uunpklo z19.d, z1.s
|
|
; CHECK-NEXT: uunpklo z21.d, z7.s
|
|
; CHECK-NEXT: uunpklo z23.d, z5.s
|
|
; CHECK-NEXT: ext z5.b, z5.b, z5.b, #8
|
|
; CHECK-NEXT: uunpklo z4.d, z4.s
|
|
; CHECK-NEXT: ext z7.b, z7.b, z7.b, #8
|
|
; CHECK-NEXT: uunpklo z3.d, z3.s
|
|
; CHECK-NEXT: ext z6.b, z6.b, z6.b, #8
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z5.d, z5.s
|
|
; CHECK-NEXT: stp q22, q4, [x1]
|
|
; CHECK-NEXT: uunpklo z4.d, z7.s
|
|
; CHECK-NEXT: stp q23, q5, [x1, #128]
|
|
; CHECK-NEXT: uunpklo z2.d, z2.s
|
|
; CHECK-NEXT: stp q20, q3, [x1, #160]
|
|
; CHECK-NEXT: uunpklo z3.d, z6.s
|
|
; CHECK-NEXT: uunpklo z1.d, z1.s
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q16, q2, [x1, #32]
|
|
; CHECK-NEXT: stp q17, q3, [x1, #64]
|
|
; CHECK-NEXT: stp q18, q0, [x1, #96]
|
|
; CHECK-NEXT: stp q21, q4, [x1, #192]
|
|
; CHECK-NEXT: stp q19, q1, [x1, #224]
|
|
; CHECK-NEXT: ret
|
|
%a = load <32 x i8>, <32 x i8>* %in
|
|
%b = add <32 x i8> %a, %a
|
|
%c = zext <32 x i8> %b to <32 x i64>
|
|
store <32 x i64> %c, <32 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; zext i16 -> i32
|
|
;
|
|
|
|
define void @zext_v8i16_v8i32(<8 x i16> %a, <8 x i32>* %out) #0 {
|
|
; CHECK-LABEL: zext_v8i16_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <8 x i16> %a to <8 x i32>
|
|
store <8 x i32>%b, <8 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v16i16_v16i32(<16 x i16>* %in, <16 x i32>* %out) #0 {
|
|
; CHECK-LABEL: zext_v16i16_v16i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.h, z0.h, z0.h
|
|
; CHECK-NEXT: uunpklo z2.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.h, z1.h, z1.h
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z3.s, z1.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z1.s, z1.h
|
|
; CHECK-NEXT: stp q2, q0, [x1]
|
|
; CHECK-NEXT: stp q3, q1, [x1, #32]
|
|
; CHECK-NEXT: ret
|
|
%a = load <16 x i16>, <16 x i16>* %in
|
|
%b = add <16 x i16> %a, %a
|
|
%c = zext <16 x i16> %b to <16 x i32>
|
|
store <16 x i32> %c, <16 x i32>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; zext i16 -> i64
|
|
;
|
|
|
|
define void @zext_v4i16_v4i64(<4 x i16> %a, <4 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v4i16_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z1.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <4 x i16> %a to <4 x i64>
|
|
store <4 x i64>%b, <4 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v8i16_v8i64(<8 x i16> %a, <8 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v8i16_v8i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uunpklo z1.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z2.d, z1.s
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z3.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q3, q0, [x0, #32]
|
|
; CHECK-NEXT: uunpklo z0.d, z1.s
|
|
; CHECK-NEXT: stp q2, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <8 x i16> %a to <8 x i64>
|
|
store <8 x i64>%b, <8 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v16i16_v16i64(<16 x i16>* %in, <16 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v16i16_v16i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.h, z0.h, z0.h
|
|
; CHECK-NEXT: uunpklo z2.s, z0.h
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.h, z1.h, z1.h
|
|
; CHECK-NEXT: uunpklo z0.s, z0.h
|
|
; CHECK-NEXT: uunpklo z3.s, z1.h
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z5.d, z3.s
|
|
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
|
|
; CHECK-NEXT: uunpklo z1.s, z1.h
|
|
; CHECK-NEXT: uunpklo z3.d, z3.s
|
|
; CHECK-NEXT: uunpklo z4.d, z2.s
|
|
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
|
|
; CHECK-NEXT: uunpklo z6.d, z0.s
|
|
; CHECK-NEXT: stp q5, q3, [x1, #64]
|
|
; CHECK-NEXT: uunpklo z5.d, z1.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z2.d, z2.s
|
|
; CHECK-NEXT: uunpklo z1.d, z1.s
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q4, q2, [x1]
|
|
; CHECK-NEXT: stp q6, q0, [x1, #32]
|
|
; CHECK-NEXT: stp q5, q1, [x1, #96]
|
|
; CHECK-NEXT: ret
|
|
%a = load <16 x i16>, <16 x i16>* %in
|
|
%b = add <16 x i16> %a, %a
|
|
%c = zext <16 x i16> %b to <16 x i64>
|
|
store <16 x i64> %c, <16 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; zext i32 -> i64
|
|
;
|
|
|
|
define void @zext_v4i32_v4i64(<4 x i32> %a, <4 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v4i32_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uunpklo z1.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: stp q1, q0, [x0]
|
|
; CHECK-NEXT: ret
|
|
%b = zext <4 x i32> %a to <4 x i64>
|
|
store <4 x i64>%b, <4 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
define void @zext_v8i32_v8i64(<8 x i32>* %in, <8 x i64>* %out) #0 {
|
|
; CHECK-LABEL: zext_v8i32_v8i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: add z0.s, z0.s, z0.s
|
|
; CHECK-NEXT: uunpklo z2.d, z0.s
|
|
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
|
|
; CHECK-NEXT: add z1.s, z1.s, z1.s
|
|
; CHECK-NEXT: uunpklo z0.d, z0.s
|
|
; CHECK-NEXT: uunpklo z3.d, z1.s
|
|
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
|
|
; CHECK-NEXT: uunpklo z1.d, z1.s
|
|
; CHECK-NEXT: stp q2, q0, [x1]
|
|
; CHECK-NEXT: stp q3, q1, [x1, #32]
|
|
; CHECK-NEXT: ret
|
|
%a = load <8 x i32>, <8 x i32>* %in
|
|
%b = add <8 x i32> %a, %a
|
|
%c = zext <8 x i32> %b to <8 x i64>
|
|
store <8 x i64> %c, <8 x i64>* %out
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind "target-features"="+sve" }
|