Google Git
Sign in
llvm / llvm-project / 15d8b3cae9debc2bd7d27ca92ff599ba9fb30da5 / . / llvm / test / CodeGen / AArch64 / sve-intrinsics-stores.ll
blob: b09baa6bf7e0ac2249b2aa83e699d28163126ade [file] [log] [blame]
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme -force-streaming < %s | FileCheck %s
;
; ST2B
;
define void @st2b_i8(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2b_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2b { z0.b, z1.b }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> %v0,
<vscale x 16 x i8> %v1,
<vscale x 16 x i1> %pred,
ptr %addr)
ret void
}
;
; ST2H
;
define void @st2h_i16(<vscale x 8 x i16> %v0, <vscale x 8 x i16> %v1, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2h_i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2h { z0.h, z1.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> %v0,
<vscale x 8 x i16> %v1,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @st2h_f16(<vscale x 8 x half> %v0, <vscale x 8 x half> %v1, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2h_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2h { z0.h, z1.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> %v0,
<vscale x 8 x half> %v1,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @st2h_bf16(<vscale x 8 x bfloat> %v0, <vscale x 8 x bfloat> %v1, <vscale x 8 x i1> %pred, ptr %addr) #0 {
; CHECK-LABEL: st2h_bf16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2h { z0.h, z1.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> %v0,
<vscale x 8 x bfloat> %v1,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
;
; ST2W
;
define void @st2w_i32(<vscale x 4 x i32> %v0, <vscale x 4 x i32> %v1, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2w_i32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2w { z0.s, z1.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> %v0,
<vscale x 4 x i32> %v1,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
define void @st2w_f32(<vscale x 4 x float> %v0, <vscale x 4 x float> %v1, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2w_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2w { z0.s, z1.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> %v0,
<vscale x 4 x float> %v1,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
;
; ST2D
;
define void @st2d_i64(<vscale x 2 x i64> %v0, <vscale x 2 x i64> %v1, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2d_i64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2d { z0.d, z1.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> %v0,
<vscale x 2 x i64> %v1,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
define void @st2d_f64(<vscale x 2 x double> %v0, <vscale x 2 x double> %v1, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2d_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2d { z0.d, z1.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> %v0,
<vscale x 2 x double> %v1,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
define void @st2d_ptr(<vscale x 2 x ptr> %v0, <vscale x 2 x ptr> %v1, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2d_ptr:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2d { z0.d, z1.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st2.nxv2p0(<vscale x 2 x ptr> %v0,
<vscale x 2 x ptr> %v1,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
;
; ST3B
;
define void @st3b_i8(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i8> %v2, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3b_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3b { z0.b - z2.b }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> %v0,
<vscale x 16 x i8> %v1,
<vscale x 16 x i8> %v2,
<vscale x 16 x i1> %pred,
ptr %addr)
ret void
}
;
; ST3H
;
define void @st3h_i16(<vscale x 8 x i16> %v0, <vscale x 8 x i16> %v1, <vscale x 8 x i16> %v2, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3h_i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3h { z0.h - z2.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> %v0,
<vscale x 8 x i16> %v1,
<vscale x 8 x i16> %v2,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @st3h_f16(<vscale x 8 x half> %v0, <vscale x 8 x half> %v1, <vscale x 8 x half> %v2, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3h_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3h { z0.h - z2.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> %v0,
<vscale x 8 x half> %v1,
<vscale x 8 x half> %v2,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @st3h_bf16(<vscale x 8 x bfloat> %v0, <vscale x 8 x bfloat> %v1, <vscale x 8 x bfloat> %v2, <vscale x 8 x i1> %pred, ptr %addr) #0 {
; CHECK-LABEL: st3h_bf16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3h { z0.h - z2.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> %v0,
<vscale x 8 x bfloat> %v1,
<vscale x 8 x bfloat> %v2,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
;
; ST3W
;
define void @st3w_i32(<vscale x 4 x i32> %v0, <vscale x 4 x i32> %v1, <vscale x 4 x i32> %v2, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3w_i32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3w { z0.s - z2.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> %v0,
<vscale x 4 x i32> %v1,
<vscale x 4 x i32> %v2,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
define void @st3w_f32(<vscale x 4 x float> %v0, <vscale x 4 x float> %v1, <vscale x 4 x float> %v2, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3w_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3w { z0.s - z2.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> %v0,
<vscale x 4 x float> %v1,
<vscale x 4 x float> %v2,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
;
; ST3D
;
define void @st3d_i64(<vscale x 2 x i64> %v0, <vscale x 2 x i64> %v1, <vscale x 2 x i64> %v2, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3d_i64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3d { z0.d - z2.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> %v0,
<vscale x 2 x i64> %v1,
<vscale x 2 x i64> %v2,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
define void @st3d_f64(<vscale x 2 x double> %v0, <vscale x 2 x double> %v1, <vscale x 2 x double> %v2, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3d_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3d { z0.d - z2.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> %v0,
<vscale x 2 x double> %v1,
<vscale x 2 x double> %v2,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
define void @st3d_ptr(<vscale x 2 x ptr> %v0, <vscale x 2 x ptr> %v1, <vscale x 2 x ptr> %v2, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st3d_ptr:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: st3d { z0.d - z2.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st3.nxv2p0(<vscale x 2 x ptr> %v0,
<vscale x 2 x ptr> %v1,
<vscale x 2 x ptr> %v2,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
;
; ST4B
;
define void @st4b_i8(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i8> %v2, <vscale x 16 x i8> %v3, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4b_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4b { z0.b - z3.b }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> %v0,
<vscale x 16 x i8> %v1,
<vscale x 16 x i8> %v2,
<vscale x 16 x i8> %v3,
<vscale x 16 x i1> %pred,
ptr %addr)
ret void
}
;
; ST4H
;
define void @st4h_i16(<vscale x 8 x i16> %v0, <vscale x 8 x i16> %v1, <vscale x 8 x i16> %v2, <vscale x 8 x i16> %v3, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4h_i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4h { z0.h - z3.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> %v0,
<vscale x 8 x i16> %v1,
<vscale x 8 x i16> %v2,
<vscale x 8 x i16> %v3,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @st4h_f16(<vscale x 8 x half> %v0, <vscale x 8 x half> %v1, <vscale x 8 x half> %v2, <vscale x 8 x half> %v3, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4h_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4h { z0.h - z3.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> %v0,
<vscale x 8 x half> %v1,
<vscale x 8 x half> %v2,
<vscale x 8 x half> %v3,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @st4h_bf16(<vscale x 8 x bfloat> %v0, <vscale x 8 x bfloat> %v1, <vscale x 8 x bfloat> %v2, <vscale x 8 x bfloat> %v3, <vscale x 8 x i1> %pred, ptr %addr) #0 {
; CHECK-LABEL: st4h_bf16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4h { z0.h - z3.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> %v0,
<vscale x 8 x bfloat> %v1,
<vscale x 8 x bfloat> %v2,
<vscale x 8 x bfloat> %v3,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
;
; ST4W
;
define void @st4w_i32(<vscale x 4 x i32> %v0, <vscale x 4 x i32> %v1, <vscale x 4 x i32> %v2, <vscale x 4 x i32> %v3, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4w_i32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4w { z0.s - z3.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> %v0,
<vscale x 4 x i32> %v1,
<vscale x 4 x i32> %v2,
<vscale x 4 x i32> %v3,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
define void @st4w_f32(<vscale x 4 x float> %v0, <vscale x 4 x float> %v1, <vscale x 4 x float> %v2, <vscale x 4 x float> %v3, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4w_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4w { z0.s - z3.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> %v0,
<vscale x 4 x float> %v1,
<vscale x 4 x float> %v2,
<vscale x 4 x float> %v3,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
;
; ST4D
;
define void @st4d_i64(<vscale x 2 x i64> %v0, <vscale x 2 x i64> %v1, <vscale x 2 x i64> %v2, <vscale x 2 x i64> %v3, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4d_i64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4d { z0.d - z3.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> %v0,
<vscale x 2 x i64> %v1,
<vscale x 2 x i64> %v2,
<vscale x 2 x i64> %v3,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
define void @st4d_f64(<vscale x 2 x double> %v0, <vscale x 2 x double> %v1, <vscale x 2 x double> %v2, <vscale x 2 x double> %v3, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4d_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4d { z0.d - z3.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> %v0,
<vscale x 2 x double> %v1,
<vscale x 2 x double> %v2,
<vscale x 2 x double> %v3,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
define void @st4d_ptr(<vscale x 2 x ptr> %v0, <vscale x 2 x ptr> %v1, <vscale x 2 x ptr> %v2, <vscale x 2 x ptr> %v3, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st4d_ptr:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: st4d { z0.d - z3.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.st4.nxv2p0(<vscale x 2 x ptr> %v0,
<vscale x 2 x ptr> %v1,
<vscale x 2 x ptr> %v2,
<vscale x 2 x ptr> %v3,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
;
; STNT1B
;
define void @stnt1b_i8(<vscale x 16 x i8> %data, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: stnt1b_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1b { z0.b }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv16i8(<vscale x 16 x i8> %data,
<vscale x 16 x i1> %pred,
ptr %addr)
ret void
}
;
; STNT1H
;
define void @stnt1h_i16(<vscale x 8 x i16> %data, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: stnt1h_i16:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1h { z0.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv8i16(<vscale x 8 x i16> %data,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @stnt1h_f16(<vscale x 8 x half> %data, <vscale x 8 x i1> %pred, ptr %addr) {
; CHECK-LABEL: stnt1h_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1h { z0.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv8f16(<vscale x 8 x half> %data,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
define void @stnt1h_bf16(<vscale x 8 x bfloat> %data, <vscale x 8 x i1> %pred, ptr %addr) #0 {
; CHECK-LABEL: stnt1h_bf16:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1h { z0.h }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv8bf16(<vscale x 8 x bfloat> %data,
<vscale x 8 x i1> %pred,
ptr %addr)
ret void
}
;
; STNT1W
;
define void @stnt1w_i32(<vscale x 4 x i32> %data, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: stnt1w_i32:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1w { z0.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv4i32(<vscale x 4 x i32> %data,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
define void @stnt1w_f32(<vscale x 4 x float> %data, <vscale x 4 x i1> %pred, ptr %addr) {
; CHECK-LABEL: stnt1w_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1w { z0.s }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv4f32(<vscale x 4 x float> %data,
<vscale x 4 x i1> %pred,
ptr %addr)
ret void
}
;
; STNT1D
;
define void @stnt1d_i64(<vscale x 2 x i64> %data, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: stnt1d_i64:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1d { z0.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv2i64(<vscale x 2 x i64> %data,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
define void @stnt1d_f64(<vscale x 2 x double> %data, <vscale x 2 x i1> %pred, ptr %addr) {
; CHECK-LABEL: stnt1d_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: stnt1d { z0.d }, p0, [x0]
; CHECK-NEXT: ret
call void @llvm.aarch64.sve.stnt1.nxv2f64(<vscale x 2 x double> %data,
<vscale x 2 x i1> %pred,
ptr %addr)
ret void
}
declare void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, ptr)
declare void @llvm.aarch64.sve.st2.nxv2p0(<vscale x 2 x ptr>, <vscale x 2 x ptr>, <vscale x 2 x i1>, ptr nocapture)
declare void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, ptr)
declare void @llvm.aarch64.sve.st3.nxv2p0(<vscale x 2 x ptr>, <vscale x 2 x ptr>, <vscale x 2 x ptr>, <vscale x 2 x i1>, ptr nocapture)
declare void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i1>, ptr)
declare void @llvm.aarch64.sve.st4.nxv2p0(<vscale x 2 x ptr>, <vscale x 2 x ptr>, <vscale x 2 x ptr>, <vscale x 2 x ptr>, <vscale x 2 x i1>, ptr nocapture)
declare void @llvm.aarch64.sve.stnt1.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i1>, ptr)
declare void @llvm.aarch64.sve.stnt1.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.stnt1.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.stnt1.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, ptr)
declare void @llvm.aarch64.sve.stnt1.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.stnt1.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x i1>, ptr)
declare void @llvm.aarch64.sve.stnt1.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, ptr)
declare void @llvm.aarch64.sve.stnt1.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, ptr)
; +bf16 is required for the bfloat version.
attributes #0 = { "target-features"="+bf16" }