; 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
; NOTE: invalid, upper and lower bound immediate values of the reg+imm
; addressing mode are checked only for the byte version of each
; instruction (`st<N>b`), as the code for detecting the immediate is
; common to all instructions, and varies only for the number of
; elements of the structured store, which is <N> = 2, 3, 4.
;
; ST2B
;
define void @st2b_i8_valid_imm(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2b_i8_valid_imm:
; 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, #2, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 2, i64 0
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 %base)
ret void
}
define void @st2b_i8_invalid_imm_not_multiple_of_2(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2b_i8_invalid_imm_not_multiple_of_2:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: rdvl x8, #3
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2b { z0.b, z1.b }, p0, [x0, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 3, i64 0
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 %base)
ret void
}
define void @st2b_i8_invalid_imm_out_of_lower_bound(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2b_i8_invalid_imm_out_of_lower_bound:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: rdvl x8, #-18
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2b { z0.b, z1.b }, p0, [x0, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 -18, i64 0
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 %base)
ret void
}
define void @st2b_i8_invalid_imm_out_of_upper_bound(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2b_i8_invalid_imm_out_of_upper_bound:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: rdvl x8, #16
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: st2b { z0.b, z1.b }, p0, [x0, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 16, i64 0
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 %base)
ret void
}
define void @st2b_i8_valid_imm_lower_bound(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2b_i8_valid_imm_lower_bound:
; 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, #-16, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 -16, i64 0
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 %base)
ret void
}
define void @st2b_i8_valid_imm_upper_bound(<vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1, <vscale x 16 x i1> %pred, ptr %addr) {
; CHECK-LABEL: st2b_i8_valid_imm_upper_bound:
; 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, #14, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 14, i64 0
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 %base)
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, #2, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 8 x i16>, ptr %addr, i64 2, i64 0
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 %base)
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, #2, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 8 x half>, ptr %addr, i64 2, i64 0
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 %base)
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, #4, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 4 x i32>, ptr %addr, i64 4, i64 0
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 %base)
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, #6, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 4 x float>, ptr %addr, i64 6, i64 0
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 %base)
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, #8, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 2 x i64>, ptr %addr, i64 8, i64 0
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 %base)
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, #10, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 2 x double>, ptr %addr, i64 10, i64 0
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 %base)
ret void
}
;
; ST3B
;
define void @st3b_i8_valid_imm(<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_valid_imm:
; 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, #3, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 3, i64 0
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 %base)
ret void
}
define void @st3b_i8_invalid_imm_not_multiple_of_3_01(<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_invalid_imm_not_multiple_of_3_01:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: rdvl x8, #4
; 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, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 4, i64 0
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 %base)
ret void
}
define void @st3b_i8_invalid_imm_not_multiple_of_3_02(<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_invalid_imm_not_multiple_of_3_02:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: rdvl x8, #5
; 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, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 5, i64 0
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 %base)
ret void
}
define void @st3b_i8_invalid_imm_out_of_lower_bound(<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_invalid_imm_out_of_lower_bound:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: rdvl x8, #-27
; 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, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 -27, i64 0
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 %base)
ret void
}
define void @st3b_i8_invalid_imm_out_of_upper_bound(<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_invalid_imm_out_of_upper_bound:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z2 killed $z2 killed $z0_z1_z2 def $z0_z1_z2
; CHECK-NEXT: rdvl x8, #24
; 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, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 24, i64 0
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 %base)
ret void
}
define void @st3b_i8_valid_imm_lower_bound(<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_valid_imm_lower_bound:
; 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, #-24, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 -24, i64 0
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 %base)
ret void
}
define void @st3b_i8_valid_imm_upper_bound(<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_valid_imm_upper_bound:
; 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, #21, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 21, i64 0
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 %base)
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, #6, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 8 x i16>, ptr %addr, i64 6, i64 0
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 %base)
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, #9, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 8 x half>, ptr %addr, i64 9, i64 0
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 %base)
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, #12, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 4 x i32>, ptr %addr, i64 12, i64 0
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 %base)
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, #15, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 4 x float>, ptr %addr, i64 15, i64 0
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 %base)
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, #18, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 2 x i64>, ptr %addr, i64 18, i64 0
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 %base)
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, #-3, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 2 x double>, ptr %addr, i64 -3, i64 0
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 %base)
ret void
}
;
; ST4B
;
define void @st4b_i8_valid_imm(<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_valid_imm:
; 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, #4, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 4, i64 0
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 %base)
ret void
}
define void @st4b_i8_invalid_imm_not_multiple_of_4_01(<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_invalid_imm_not_multiple_of_4_01:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: rdvl x8, #5
; 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, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 5, i64 0
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 %base)
ret void
}
define void @st4b_i8_invalid_imm_not_multiple_of_4_02(<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_invalid_imm_not_multiple_of_4_02:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: rdvl x8, #6
; 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, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 6, i64 0
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 %base)
ret void
}
define void @st4b_i8_invalid_imm_not_multiple_of_4_03(<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_invalid_imm_not_multiple_of_4_03:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: rdvl x8, #7
; 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, x8]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 7, i64 0
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 %base)
ret void
}
define void @st4b_i8_invalid_imm_out_of_lower_bound(<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_invalid_imm_out_of_lower_bound:
; CHECK: // %bb.0:
; CHECK-NEXT: rdvl x8, #1
; CHECK-NEXT: mov x9, #-576 // =0xfffffffffffffdc0
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: lsr x8, x8, #4
; 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: mul x8, x8, x9
; 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, x8]
; CHECK-NEXT: ret
; FIXME: optimize OFFSET computation so that xOFFSET = (mul (RDVL #4) #9)
; xM = -9 * 2^6
; xP = RDVL * 2^-4
; xBASE = RDVL * 2^-4 * -9 * 2^6 = RDVL * -36
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 -36, i64 0
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 %base)
ret void
}
define void @st4b_i8_invalid_imm_out_of_upper_bound(<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_invalid_imm_out_of_upper_bound:
; CHECK: // %bb.0:
; CHECK-NEXT: rdvl x8, #1
; CHECK-NEXT: mov w9, #512 // =0x200
; CHECK-NEXT: // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT: lsr x8, x8, #4
; 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: mul x8, x8, x9
; 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, x8]
; CHECK-NEXT: ret
; FIXME: optimize OFFSET computation so that xOFFSET = (shl (RDVL #16) #1)
; xM = 2^9
; xP = RDVL * 2^-4
; xOFFSET = RDVL * 2^-4 * 2^9 = RDVL * 32
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 32, i64 0
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 %base)
ret void
}
define void @st4b_i8_valid_imm_lower_bound(<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_valid_imm_lower_bound:
; 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, #-32, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 -32, i64 0
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 %base)
ret void
}
define void @st4b_i8_valid_imm_upper_bound(<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_valid_imm_upper_bound:
; 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, #28, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 16 x i8>, ptr %addr, i64 28, i64 0
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 %base)
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, #8, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 8 x i16>, ptr %addr, i64 8, i64 0
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 %base)
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, #12, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 8 x half>, ptr %addr, i64 12, i64 0
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 %base)
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, #16, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 4 x i32>, ptr %addr, i64 16, i64 0
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 %base)
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, #20, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 4 x float>, ptr %addr, i64 20, i64 0
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 %base)
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, #24, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 2 x i64>, ptr %addr, i64 24, i64 0
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 %base)
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, #28, mul vl]
; CHECK-NEXT: ret
%base = getelementptr <vscale x 2 x double>, ptr %addr, i64 28, i64 0
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 %base)
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.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.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.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.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.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)