1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve,+f64mm < %s | FileCheck %s
8 define <vscale x 16 x i8> @ld1rob_i8(<vscale x 16 x i1> %pg, ptr %a, i64 %index) {
9 ; CHECK-LABEL: ld1rob_i8:
11 ; CHECK-NEXT: ld1rob { z0.b }, p0/z, [x0, x1]
13 %base = getelementptr i8, ptr %a, i64 %index
14 %load = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1ro.nxv16i8(<vscale x 16 x i1> %pg, ptr %base)
15 ret <vscale x 16 x i8> %load
22 define <vscale x 8 x i16> @ld1roh_i16(<vscale x 8 x i1> %pg, ptr %a, i64 %index) {
23 ; CHECK-LABEL: ld1roh_i16:
25 ; CHECK-NEXT: ld1roh { z0.h }, p0/z, [x0, x1, lsl #1]
27 %base = getelementptr i16, ptr %a, i64 %index
28 %load = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1ro.nxv8i16(<vscale x 8 x i1> %pg, ptr %base)
29 ret <vscale x 8 x i16> %load
32 define <vscale x 8 x half> @ld1roh_f16(<vscale x 8 x i1> %pg, ptr %a, i64 %index) {
33 ; CHECK-LABEL: ld1roh_f16:
35 ; CHECK-NEXT: ld1roh { z0.h }, p0/z, [x0, x1, lsl #1]
37 %base = getelementptr half, ptr %a, i64 %index
38 %load = call <vscale x 8 x half> @llvm.aarch64.sve.ld1ro.nxv8f16(<vscale x 8 x i1> %pg, ptr %base)
39 ret <vscale x 8 x half> %load
42 define <vscale x 8 x bfloat> @ld1roh_bf16(<vscale x 8 x i1> %pg, ptr %a, i64 %index) #0 {
43 ; CHECK-LABEL: ld1roh_bf16:
45 ; CHECK-NEXT: ld1roh { z0.h }, p0/z, [x0, x1, lsl #1]
47 %base = getelementptr bfloat, ptr %a, i64 %index
48 %load = call <vscale x 8 x bfloat> @llvm.aarch64.sve.ld1ro.nxv8bf16(<vscale x 8 x i1> %pg, ptr %base)
49 ret <vscale x 8 x bfloat> %load
56 define <vscale x 4 x i32> @ld1row_i32(<vscale x 4 x i1> %pg, ptr %a, i64 %index) {
57 ; CHECK-LABEL: ld1row_i32:
59 ; CHECK-NEXT: ld1row { z0.s }, p0/z, [x0, x1, lsl #2]
61 %base = getelementptr i32, ptr %a, i64 %index
62 %load = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1ro.nxv4i32(<vscale x 4 x i1> %pg, ptr %base)
63 ret <vscale x 4 x i32> %load
66 define <vscale x 4 x float> @ld1row_f32(<vscale x 4 x i1> %pg, ptr %a, i64 %index) {
67 ; CHECK-LABEL: ld1row_f32:
69 ; CHECK-NEXT: ld1row { z0.s }, p0/z, [x0, x1, lsl #2]
71 %base = getelementptr float, ptr %a, i64 %index
72 %load = call <vscale x 4 x float> @llvm.aarch64.sve.ld1ro.nxv4f32(<vscale x 4 x i1> %pg, ptr %base)
73 ret <vscale x 4 x float> %load
80 define <vscale x 2 x i64> @ld1rod_i64(<vscale x 2 x i1> %pg, ptr %a, i64 %index) {
81 ; CHECK-LABEL: ld1rod_i64:
83 ; CHECK-NEXT: ld1rod { z0.d }, p0/z, [x0, x1, lsl #3]
85 %base = getelementptr i64, ptr %a, i64 %index
86 %load = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1ro.nxv2i64(<vscale x 2 x i1> %pg, ptr %base)
87 ret <vscale x 2 x i64> %load
90 define <vscale x 2 x double> @ld1rod_f64(<vscale x 2 x i1> %pg, ptr %a, i64 %index) {
91 ; CHECK-LABEL: ld1rod_f64:
93 ; CHECK-NEXT: ld1rod { z0.d }, p0/z, [x0, x1, lsl #3]
95 %base = getelementptr double, ptr %a, i64 %index
96 %load = call <vscale x 2 x double> @llvm.aarch64.sve.ld1ro.nxv2f64(<vscale x 2 x i1> %pg, ptr %base)
97 ret <vscale x 2 x double> %load
100 declare <vscale x 16 x i8> @llvm.aarch64.sve.ld1ro.nxv16i8(<vscale x 16 x i1>, ptr)
102 declare <vscale x 8 x i16> @llvm.aarch64.sve.ld1ro.nxv8i16(<vscale x 8 x i1>, ptr)
103 declare <vscale x 8 x half> @llvm.aarch64.sve.ld1ro.nxv8f16(<vscale x 8 x i1>, ptr)
104 declare <vscale x 8 x bfloat> @llvm.aarch64.sve.ld1ro.nxv8bf16(<vscale x 8 x i1>, ptr)
106 declare <vscale x 4 x i32> @llvm.aarch64.sve.ld1ro.nxv4i32(<vscale x 4 x i1>, ptr)
107 declare <vscale x 4 x float> @llvm.aarch64.sve.ld1ro.nxv4f32(<vscale x 4 x i1>, ptr)
109 declare <vscale x 2 x i64> @llvm.aarch64.sve.ld1ro.nxv2i64(<vscale x 2 x i1>, ptr)
110 declare <vscale x 2 x double> @llvm.aarch64.sve.ld1ro.nxv2f64(<vscale x 2 x i1>, ptr)
112 ; +bf16 is required for the bfloat version.
113 attributes #0 = { "target-features"="+sve,+f64mm,+bf16" }