llvm-project/llvm/test/CodeGen/Thumb2/mve-gather-ind16-unscaled.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve.fp -enable-arm-maskedgatscat %s -o - | FileCheck %s

define arm_aapcs_vfpcc <8 x i16> @zext_unscaled_i8_i16(i8* %base, <8 x i16>* %offptr) {
; CHECK-LABEL: zext_unscaled_i8_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vldrb.u16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %gather = call <8 x i8> @llvm.masked.gather.v8i8.v8p0i8(<8 x i8*> %ptrs, i32 1, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i8> undef)
  %gather.zext = zext <8 x i8> %gather to <8 x i16>
  ret <8 x i16> %gather.zext
}

define arm_aapcs_vfpcc <8 x i16> @sext_unscaled_i8_i16(i8* %base, <8 x i16>* %offptr) {
; CHECK-LABEL: sext_unscaled_i8_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vldrb.s16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %gather = call <8 x i8> @llvm.masked.gather.v8i8.v8p0i8(<8 x i8*> %ptrs, i32 1, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i8> undef)
  %gather.sext = sext <8 x i8> %gather to <8 x i16>
  ret <8 x i16> %gather.sext
}

define arm_aapcs_vfpcc <8 x i16> @unscaled_i16_i16(i8* %base, <8 x i16>* %offptr) {
; CHECK-LABEL: unscaled_i16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vldrh.u16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %byte_ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %ptrs = bitcast <8 x i8*> %byte_ptrs to <8 x i16*>
  %gather = call <8 x i16> @llvm.masked.gather.v8i16.v8p0i16(<8 x i16*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i16> undef)
  ret <8 x i16> %gather
}

define arm_aapcs_vfpcc <8 x half> @unscaled_f16_i16(i8* %base, <8 x i16>* %offptr) {
; CHECK-LABEL: unscaled_f16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vldrh.u16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %byte_ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %ptrs = bitcast <8 x i8*> %byte_ptrs to <8 x half*>
  %gather = call <8 x half> @llvm.masked.gather.v8f16.v8p0f16(<8 x half*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x half> undef)
  ret <8 x half> %gather
}

define arm_aapcs_vfpcc <8 x i16> @zext_unsigned_unscaled_i8_i8(i8* %base, <8 x i8>* %offptr) {
; CHECK-LABEL: zext_unsigned_unscaled_i8_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u16 q1, [r1]
; CHECK-NEXT:    vldrb.u16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i8>, <8 x i8>* %offptr, align 1
  %offs.zext = zext <8 x i8> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %gather = call <8 x i8> @llvm.masked.gather.v8i8.v8p0i8(<8 x i8*> %ptrs, i32 1, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i8> undef)
  %gather.zext = zext <8 x i8> %gather to <8 x i16>
  ret <8 x i16> %gather.zext
}

define arm_aapcs_vfpcc <8 x i16> @sext_unsigned_unscaled_i8_i8(i8* %base, <8 x i8>* %offptr) {
; CHECK-LABEL: sext_unsigned_unscaled_i8_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u16 q1, [r1]
; CHECK-NEXT:    vldrb.s16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i8>, <8 x i8>* %offptr, align 1
  %offs.zext = zext <8 x i8> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %gather = call <8 x i8> @llvm.masked.gather.v8i8.v8p0i8(<8 x i8*> %ptrs, i32 1, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i8> undef)
  %gather.sext = sext <8 x i8> %gather to <8 x i16>
  ret <8 x i16> %gather.sext
}

define arm_aapcs_vfpcc <8 x i16> @unsigned_unscaled_i16_i8(i8* %base, <8 x i8>* %offptr) {
; CHECK-LABEL: unsigned_unscaled_i16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u16 q1, [r1]
; CHECK-NEXT:    vldrh.u16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i8>, <8 x i8>* %offptr, align 1
  %offs.zext = zext <8 x i8> %offs to <8 x i32>
  %byte_ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %ptrs = bitcast <8 x i8*> %byte_ptrs to <8 x i16*>
  %gather = call <8 x i16> @llvm.masked.gather.v8i16.v8p0i16(<8 x i16*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i16> undef)
  ret <8 x i16> %gather
}

define arm_aapcs_vfpcc <8 x half> @unsigned_unscaled_f16_i8(i8* %base, <8 x i8>* %offptr) {
; CHECK-LABEL: unsigned_unscaled_f16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u16 q1, [r1]
; CHECK-NEXT:    vldrh.u16 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i8>, <8 x i8>* %offptr, align 1
  %offs.zext = zext <8 x i8> %offs to <8 x i32>
  %byte_ptrs = getelementptr inbounds i8, i8* %base, <8 x i32> %offs.zext
  %ptrs = bitcast <8 x i8*> %byte_ptrs to <8 x half*>
  %gather = call <8 x half> @llvm.masked.gather.v8f16.v8p0f16(<8 x half*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x half> undef)
  ret <8 x half> %gather
}

declare <8 x i8> @llvm.masked.gather.v8i8.v8p0i8(<8 x i8*>, i32, <8 x i1>, <8 x i8>) #1
declare <8 x i16> @llvm.masked.gather.v8i16.v8p0i16(<8 x i16*>, i32, <8 x i1>, <8 x i16>) #1
declare <8 x half> @llvm.masked.gather.v8f16.v8p0f16(<8 x half*>, i32, <8 x i1>, <8 x half>) #1