[AArch64][SME]: Add streaming-compatible testing files.

Testing files:
 - limit-duplane.ll
 - optimize-ptrue.ll
 - ptest.ll

Reviewed By: david-arm, sdesmalen

Differential Revision: https://reviews.llvm.org/D138768

diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-limit-duplane.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-limit-duplane.ll
new file mode 100644 (file)
index 0000000..c752b1a
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-limit-duplane.ll
@@ -0,0 +1,52 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mattr=+sve -force-streaming-compatible-sve < %s | FileCheck %s
+
+target triple = "aarch64-unknown-linux-gnu"
+
+define <4 x i32> @test(ptr %arg1, ptr %arg2) {
+; CHECK-LABEL: test:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ldp q2, q1, [x0, #32]
+; CHECK-NEXT:    add z2.s, z2.s, z2.s
+; CHECK-NEXT:    ldp q3, q4, [x0]
+; CHECK-NEXT:    mov z0.s, z1.s[2]
+; CHECK-NEXT:    add z1.s, z1.s, z1.s
+; CHECK-NEXT:    stp q2, q1, [x0, #32]
+; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
+; CHECK-NEXT:    add z2.s, z3.s, z3.s
+; CHECK-NEXT:    add z1.s, z4.s, z4.s
+; CHECK-NEXT:    stp q2, q1, [x0]
+; CHECK-NEXT:    ret
+entry:
+  %0 = load <16 x i32>, ptr %arg1, align 256
+  %1 = load <16 x i32>, ptr %arg2, align 256
+  %shvec = shufflevector <16 x i32> %0, <16 x i32> %1, <4 x i32> <i32 14, i32 14, i32 14, i32 14>
+  %2 = add <16 x i32> %0, %0
+  store <16 x i32> %2, ptr %arg1, align 256
+  ret <4 x i32> %shvec
+}
+
+define <2 x i32> @test2(ptr %arg1, ptr %arg2) {
+; CHECK-LABEL: test2:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ldp q2, q0, [x0, #32]
+; CHECK-NEXT:    ldp q4, q5, [x0]
+; CHECK-NEXT:    mov z1.d, z0.d
+; CHECK-NEXT:    add z3.s, z0.s, z0.s
+; CHECK-NEXT:    ext z1.b, z1.b, z0.b, #8
+; CHECK-NEXT:    mov z0.s, s1
+; CHECK-NEXT:    add z1.s, z2.s, z2.s
+; CHECK-NEXT:    stp q1, q3, [x0, #32]
+; CHECK-NEXT:    add z1.s, z4.s, z4.s
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
+; CHECK-NEXT:    add z2.s, z5.s, z5.s
+; CHECK-NEXT:    stp q1, q2, [x0]
+; CHECK-NEXT:    ret
+entry:
+  %0 = load <16 x i32>, ptr %arg1, align 256
+  %1 = load <16 x i32>, ptr %arg2, align 256
+  %shvec = shufflevector <16 x i32> %0, <16 x i32> %1, <2 x i32> <i32 14, i32 14>
+  %2 = add <16 x i32> %0, %0
+  store <16 x i32> %2, ptr %arg1, align 256
+  ret <2 x i32> %shvec
+}

diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-optimize-ptrue.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-optimize-ptrue.ll
new file mode 100644 (file)
index 0000000..020d44b
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-optimize-ptrue.ll
@@ -0,0 +1,368 @@
+; 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"
+
+define void @add_v4i8(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: add_v4i8:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr s0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl4
+; CHECK-NEXT:    ldr s1, [x1]
+; CHECK-NEXT:    uunpklo z0.h, z0.b
+; CHECK-NEXT:    uunpklo z1.h, z1.b
+; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    st1b { z0.h }, p0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <4 x i8>, ptr %a
+  %op2 = load <4 x i8>, ptr %b
+  %res = add <4 x i8> %op1, %op2
+  store <4 x i8> %res, ptr %a
+  ret void
+}
+
+define void @add_v8i8(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: add_v8i8:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr d0, [x0]
+; CHECK-NEXT:    ldr d1, [x1]
+; CHECK-NEXT:    add z0.b, z0.b, z1.b
+; CHECK-NEXT:    str d0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <8 x i8>, ptr %a
+  %op2 = load <8 x i8>, ptr %b
+  %res = add <8 x i8> %op1, %op2
+  store <8 x i8> %res, ptr %a
+  ret void
+}
+
+define void @add_v16i8(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: add_v16i8:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    add z0.b, z0.b, z1.b
+; CHECK-NEXT:    str q0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <16 x i8>, ptr %a
+  %op2 = load <16 x i8>, ptr %b
+  %res = add <16 x i8> %op1, %op2
+  store <16 x i8> %res, ptr %a
+  ret void
+}
+
+define void @add_v32i8(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: add_v32i8:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ldp q2, q3, [x1]
+; CHECK-NEXT:    add z0.b, z0.b, z2.b
+; CHECK-NEXT:    add z1.b, z1.b, z3.b
+; CHECK-NEXT:    stp q0, q1, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <32 x i8>, ptr %a
+  %op2 = load <32 x i8>, ptr %b
+  %res = add <32 x i8> %op1, %op2
+  store <32 x i8> %res, ptr %a
+  ret void
+}
+
+define void @add_v2i16(ptr %a, ptr %b, ptr %c) #0 {
+; CHECK-LABEL: add_v2i16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub sp, sp, #16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    ldrh w8, [x0, #2]
+; CHECK-NEXT:    ptrue p0.s, vl2
+; CHECK-NEXT:    str w8, [sp, #4]
+; CHECK-NEXT:    ldrh w8, [x0]
+; CHECK-NEXT:    str w8, [sp]
+; CHECK-NEXT:    ldrh w8, [x1, #2]
+; CHECK-NEXT:    str w8, [sp, #12]
+; CHECK-NEXT:    ldrh w8, [x1]
+; CHECK-NEXT:    str w8, [sp, #8]
+; CHECK-NEXT:    ldp d0, d1, [sp]
+; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    st1h { z0.s }, p0, [x0]
+; CHECK-NEXT:    add sp, sp, #16
+; CHECK-NEXT:    ret
+  %op1 = load <2 x i16>, ptr %a
+  %op2 = load <2 x i16>, ptr %b
+  %res = add <2 x i16> %op1, %op2
+  store <2 x i16> %res, ptr %a
+  ret void
+}
+
+define void @add_v4i16(ptr %a, ptr %b, ptr %c) #0 {
+; CHECK-LABEL: add_v4i16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr d0, [x0]
+; CHECK-NEXT:    ldr d1, [x1]
+; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    str d0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <4 x i16>, ptr %a
+  %op2 = load <4 x i16>, ptr %b
+  %res = add <4 x i16> %op1, %op2
+  store <4 x i16> %res, ptr %a
+  ret void
+}
+
+define void @add_v8i16(ptr %a, ptr %b, ptr %c) #0 {
+; CHECK-LABEL: add_v8i16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    str q0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <8 x i16>, ptr %a
+  %op2 = load <8 x i16>, ptr %b
+  %res = add <8 x i16> %op1, %op2
+  store <8 x i16> %res, ptr %a
+  ret void
+}
+
+define void @add_v16i16(ptr %a, ptr %b, ptr %c) #0 {
+; CHECK-LABEL: add_v16i16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ldp q2, q3, [x1]
+; CHECK-NEXT:    add z0.h, z0.h, z2.h
+; CHECK-NEXT:    add z1.h, z1.h, z3.h
+; CHECK-NEXT:    stp q0, q1, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <16 x i16>, ptr %a
+  %op2 = load <16 x i16>, ptr %b
+  %res = add <16 x i16> %op1, %op2
+  store <16 x i16> %res, ptr %a
+  ret void
+}
+
+define void @abs_v2i32(ptr %a) #0 {
+; CHECK-LABEL: abs_v2i32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr d0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl2
+; CHECK-NEXT:    abs z0.s, p0/m, z0.s
+; CHECK-NEXT:    str d0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <2 x i32>, ptr %a
+  %res = call <2 x i32> @llvm.abs.v2i32(<2 x i32> %op1, i1 false)
+  store <2 x i32> %res, ptr %a
+  ret void
+}
+
+define void @abs_v4i32(ptr %a) #0 {
+; CHECK-LABEL: abs_v4i32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    abs z0.s, p0/m, z0.s
+; CHECK-NEXT:    str q0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <4 x i32>, ptr %a
+  %res = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %op1, i1 false)
+  store <4 x i32> %res, ptr %a
+  ret void
+}
+
+define void @abs_v8i32(ptr %a) #0 {
+; CHECK-LABEL: abs_v8i32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    abs z0.s, p0/m, z0.s
+; CHECK-NEXT:    abs z1.s, p0/m, z1.s
+; CHECK-NEXT:    stp q0, q1, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <8 x i32>, ptr %a
+  %res = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %op1, i1 false)
+  store <8 x i32> %res, ptr %a
+  ret void
+}
+
+define void @abs_v2i64(ptr %a) #0 {
+; CHECK-LABEL: abs_v2i64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    abs z0.d, p0/m, z0.d
+; CHECK-NEXT:    str q0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <2 x i64>, ptr %a
+  %res = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %op1, i1 false)
+  store <2 x i64> %res, ptr %a
+  ret void
+}
+
+define void @abs_v4i64(ptr %a) #0 {
+; CHECK-LABEL: abs_v4i64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    abs z0.d, p0/m, z0.d
+; CHECK-NEXT:    abs z1.d, p0/m, z1.d
+; CHECK-NEXT:    stp q0, q1, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <4 x i64>, ptr %a
+  %res = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %op1, i1 false)
+  store <4 x i64> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v2f16(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v2f16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr s0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl4
+; CHECK-NEXT:    ldr s1, [x1]
+; CHECK-NEXT:    fadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    str w8, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <2 x half>, ptr %a
+  %op2 = load <2 x half>, ptr %b
+  %res = fadd <2 x half> %op1, %op2
+  store <2 x half> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v4f16(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v4f16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr d0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl4
+; CHECK-NEXT:    ldr d1, [x1]
+; CHECK-NEXT:    fadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    str d0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <4 x half>, ptr %a
+  %op2 = load <4 x half>, ptr %b
+  %res = fadd <4 x half> %op1, %op2
+  store <4 x half> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v8f16(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v8f16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
+; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    fadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    str q0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <8 x half>, ptr %a
+  %op2 = load <8 x half>, ptr %b
+  %res = fadd <8 x half> %op1, %op2
+  store <8 x half> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v16f16(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v16f16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
+; CHECK-NEXT:    ldp q2, q3, [x1]
+; CHECK-NEXT:    fadd z0.h, p0/m, z0.h, z2.h
+; CHECK-NEXT:    fadd z1.h, p0/m, z1.h, z3.h
+; CHECK-NEXT:    stp q0, q1, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <16 x half>, ptr %a
+  %op2 = load <16 x half>, ptr %b
+  %res = fadd <16 x half> %op1, %op2
+  store <16 x half> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v2f32(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v2f32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr d0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl2
+; CHECK-NEXT:    ldr d1, [x1]
+; CHECK-NEXT:    fadd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    str d0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <2 x float>, ptr %a
+  %op2 = load <2 x float>, ptr %b
+  %res = fadd <2 x float> %op1, %op2
+  store <2 x float> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v4f32(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v4f32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    fadd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    str q0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <4 x float>, ptr %a
+  %op2 = load <4 x float>, ptr %b
+  %res = fadd <4 x float> %op1, %op2
+  store <4 x float> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v8f32(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v8f32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    ldp q2, q3, [x1]
+; CHECK-NEXT:    fadd z0.s, p0/m, z0.s, z2.s
+; CHECK-NEXT:    fadd z1.s, p0/m, z1.s, z3.s
+; CHECK-NEXT:    stp q0, q1, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <8 x float>, ptr %a
+  %op2 = load <8 x float>, ptr %b
+  %res = fadd <8 x float> %op1, %op2
+  store <8 x float> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v2f64(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v2f64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    fadd z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    str q0, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <2 x double>, ptr %a
+  %op2 = load <2 x double>, ptr %b
+  %res = fadd <2 x double> %op1, %op2
+  store <2 x double> %res, ptr %a
+  ret void
+}
+
+define void @fadd_v4f64(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: fadd_v4f64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    ldp q2, q3, [x1]
+; CHECK-NEXT:    fadd z0.d, p0/m, z0.d, z2.d
+; CHECK-NEXT:    fadd z1.d, p0/m, z1.d, z3.d
+; CHECK-NEXT:    stp q0, q1, [x0]
+; CHECK-NEXT:    ret
+  %op1 = load <4 x double>, ptr %a
+  %op2 = load <4 x double>, ptr %b
+  %res = fadd <4 x double> %op1, %op2
+  store <4 x double> %res, ptr %a
+  ret void
+}
+
+declare <2 x i32> @llvm.abs.v2i32(<2 x i32>, i1)
+declare <4 x i32> @llvm.abs.v4i32(<4 x i32>, i1)
+declare <8 x i32> @llvm.abs.v8i32(<8 x i32>, i1)
+declare <2 x i64> @llvm.abs.v2i64(<2 x i64>, i1)
+declare <4 x i64> @llvm.abs.v4i64(<4 x i64>, i1)
+
+attributes #0 = { "target-features"="+sve" }

diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ptest.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ptest.ll
new file mode 100644 (file)
index 0000000..6004130
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ptest.ll
@@ -0,0 +1,363 @@
+; 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"
+
+define i1 @ptest_v16i1(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: ptest_v16i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub sp, sp, #16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    adrp x8, .LCPI0_0
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    ldp q2, q0, [x0, #32]
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI0_0]
+; CHECK-NEXT:    ldp q4, q3, [x0]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z0.s, z1.s
+; CHECK-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p1.s, p0/z, z2.s, z1.s
+; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    mov z5.s, z0.s[3]
+; CHECK-NEXT:    mov z6.s, z0.s[2]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z3.s, z1.s
+; CHECK-NEXT:    fcmne p0.s, p0/z, z4.s, z1.s
+; CHECK-NEXT:    mov z3.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    mov z7.s, z0.s[1]
+; CHECK-NEXT:    mov z0.s, z2.s[3]
+; CHECK-NEXT:    mov z16.s, z2.s[2]
+; CHECK-NEXT:    mov z17.s, z2.s[1]
+; CHECK-NEXT:    fmov w9, s2
+; CHECK-NEXT:    strb w8, [sp, #12]
+; CHECK-NEXT:    mov z1.s, z3.s[3]
+; CHECK-NEXT:    mov z2.s, z3.s[2]
+; CHECK-NEXT:    mov z4.s, z3.s[1]
+; CHECK-NEXT:    fmov w8, s3
+; CHECK-NEXT:    mov z3.s, p0/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fmov w10, s3
+; CHECK-NEXT:    strb w9, [sp, #8]
+; CHECK-NEXT:    fmov w9, s5
+; CHECK-NEXT:    mov z18.s, z3.s[3]
+; CHECK-NEXT:    strb w8, [sp, #4]
+; CHECK-NEXT:    fmov w8, s6
+; CHECK-NEXT:    strb w10, [sp]
+; CHECK-NEXT:    fmov w10, s7
+; CHECK-NEXT:    strb w9, [sp, #15]
+; CHECK-NEXT:    fmov w9, s0
+; CHECK-NEXT:    strb w8, [sp, #14]
+; CHECK-NEXT:    fmov w8, s16
+; CHECK-NEXT:    strb w10, [sp, #13]
+; CHECK-NEXT:    fmov w10, s17
+; CHECK-NEXT:    strb w9, [sp, #11]
+; CHECK-NEXT:    fmov w9, s1
+; CHECK-NEXT:    strb w8, [sp, #10]
+; CHECK-NEXT:    fmov w8, s2
+; CHECK-NEXT:    strb w10, [sp, #9]
+; CHECK-NEXT:    fmov w10, s4
+; CHECK-NEXT:    mov z19.s, z3.s[2]
+; CHECK-NEXT:    mov z20.s, z3.s[1]
+; CHECK-NEXT:    strb w9, [sp, #7]
+; CHECK-NEXT:    fmov w9, s18
+; CHECK-NEXT:    strb w8, [sp, #6]
+; CHECK-NEXT:    fmov w8, s19
+; CHECK-NEXT:    strb w10, [sp, #5]
+; CHECK-NEXT:    fmov w10, s20
+; CHECK-NEXT:    strb w9, [sp, #3]
+; CHECK-NEXT:    strb w8, [sp, #2]
+; CHECK-NEXT:    strb w10, [sp, #1]
+; CHECK-NEXT:    ldr q0, [sp], #16
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    orv b0, p0, z0.b
+; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    and w0, w8, #0x1
+; CHECK-NEXT:    ret
+  %v0 = bitcast ptr %a to <16 x float>*
+  %v1 = load <16 x float>, <16 x float>* %v0, align 4
+  %v2 = fcmp une <16 x float> %v1, zeroinitializer
+  %v3 = call i1 @llvm.vector.reduce.or.i1.v16i1 (<16 x i1> %v2)
+  ret i1 %v3
+}
+
+define i1 @ptest_or_v16i1(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: ptest_or_v16i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub sp, sp, #32
+; CHECK-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NEXT:    adrp x8, .LCPI1_0
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    ldp q2, q1, [x0, #32]
+; CHECK-NEXT:    ldr q0, [x8, :lo12:.LCPI1_0]
+; CHECK-NEXT:    ldp q4, q3, [x0]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z1.s, z0.s
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p1.s, p0/z, z2.s, z0.s
+; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    mov z5.s, z1.s[2]
+; CHECK-NEXT:    mov z6.s, z1.s[1]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z3.s, z0.s
+; CHECK-NEXT:    mov z3.s, z1.s[3]
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p1.s, p0/z, z4.s, z0.s
+; CHECK-NEXT:    mov z7.s, z2.s[3]
+; CHECK-NEXT:    mov z16.s, z2.s[2]
+; CHECK-NEXT:    mov z17.s, z2.s[1]
+; CHECK-NEXT:    fmov w9, s2
+; CHECK-NEXT:    strb w8, [sp, #12]
+; CHECK-NEXT:    mov z2.s, z1.s[3]
+; CHECK-NEXT:    mov z4.s, z1.s[2]
+; CHECK-NEXT:    mov z18.s, z1.s[1]
+; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fmov w10, s1
+; CHECK-NEXT:    strb w9, [sp, #8]
+; CHECK-NEXT:    fmov w9, s3
+; CHECK-NEXT:    mov z19.s, z1.s[3]
+; CHECK-NEXT:    strb w8, [sp, #4]
+; CHECK-NEXT:    fmov w8, s5
+; CHECK-NEXT:    strb w10, [sp]
+; CHECK-NEXT:    fmov w10, s6
+; CHECK-NEXT:    strb w9, [sp, #15]
+; CHECK-NEXT:    fmov w9, s7
+; CHECK-NEXT:    strb w8, [sp, #14]
+; CHECK-NEXT:    fmov w8, s16
+; CHECK-NEXT:    strb w10, [sp, #13]
+; CHECK-NEXT:    fmov w10, s17
+; CHECK-NEXT:    strb w9, [sp, #11]
+; CHECK-NEXT:    fmov w9, s2
+; CHECK-NEXT:    strb w8, [sp, #10]
+; CHECK-NEXT:    fmov w8, s4
+; CHECK-NEXT:    strb w10, [sp, #9]
+; CHECK-NEXT:    fmov w10, s18
+; CHECK-NEXT:    mov z20.s, z1.s[2]
+; CHECK-NEXT:    mov z21.s, z1.s[1]
+; CHECK-NEXT:    strb w9, [sp, #7]
+; CHECK-NEXT:    fmov w9, s19
+; CHECK-NEXT:    strb w8, [sp, #6]
+; CHECK-NEXT:    fmov w8, s20
+; CHECK-NEXT:    strb w10, [sp, #5]
+; CHECK-NEXT:    fmov w10, s21
+; CHECK-NEXT:    strb w9, [sp, #3]
+; CHECK-NEXT:    strb w8, [sp, #2]
+; CHECK-NEXT:    strb w10, [sp, #1]
+; CHECK-NEXT:    ldr q1, [x1, #48]
+; CHECK-NEXT:    ldr q5, [x1]
+; CHECK-NEXT:    ldr q7, [sp]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z1.s, z0.s
+; CHECK-NEXT:    ldp q6, q1, [x1, #16]
+; CHECK-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fmov w8, s2
+; CHECK-NEXT:    mov z3.s, z2.s[3]
+; CHECK-NEXT:    mov z4.s, z2.s[2]
+; CHECK-NEXT:    mov z2.s, z2.s[1]
+; CHECK-NEXT:    strb w8, [sp, #28]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z1.s, z0.s
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p1.s, p0/z, z6.s, z0.s
+; CHECK-NEXT:    mov z6.s, z1.s[3]
+; CHECK-NEXT:    mov z16.s, z1.s[2]
+; CHECK-NEXT:    mov z17.s, z1.s[1]
+; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p0.s, p0/z, z5.s, z0.s
+; CHECK-NEXT:    mov z0.s, z1.s[3]
+; CHECK-NEXT:    mov z5.s, z1.s[2]
+; CHECK-NEXT:    mov z18.s, z1.s[1]
+; CHECK-NEXT:    fmov w9, s1
+; CHECK-NEXT:    mov z1.s, p0/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    strb w8, [sp, #24]
+; CHECK-NEXT:    fmov w10, s1
+; CHECK-NEXT:    fmov w8, s3
+; CHECK-NEXT:    strb w9, [sp, #20]
+; CHECK-NEXT:    fmov w9, s4
+; CHECK-NEXT:    mov z19.s, z1.s[3]
+; CHECK-NEXT:    mov z20.s, z1.s[2]
+; CHECK-NEXT:    strb w10, [sp, #16]
+; CHECK-NEXT:    fmov w10, s2
+; CHECK-NEXT:    strb w8, [sp, #31]
+; CHECK-NEXT:    fmov w8, s6
+; CHECK-NEXT:    strb w9, [sp, #30]
+; CHECK-NEXT:    fmov w9, s16
+; CHECK-NEXT:    strb w10, [sp, #29]
+; CHECK-NEXT:    fmov w10, s17
+; CHECK-NEXT:    strb w8, [sp, #27]
+; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    strb w9, [sp, #26]
+; CHECK-NEXT:    fmov w9, s5
+; CHECK-NEXT:    strb w10, [sp, #25]
+; CHECK-NEXT:    fmov w10, s18
+; CHECK-NEXT:    mov z21.s, z1.s[1]
+; CHECK-NEXT:    strb w8, [sp, #23]
+; CHECK-NEXT:    fmov w8, s19
+; CHECK-NEXT:    strb w9, [sp, #22]
+; CHECK-NEXT:    fmov w9, s20
+; CHECK-NEXT:    strb w10, [sp, #21]
+; CHECK-NEXT:    fmov w10, s21
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    strb w8, [sp, #19]
+; CHECK-NEXT:    strb w9, [sp, #18]
+; CHECK-NEXT:    strb w10, [sp, #17]
+; CHECK-NEXT:    ldr q0, [sp, #16]
+; CHECK-NEXT:    orr z0.d, z7.d, z0.d
+; CHECK-NEXT:    orv b0, p0, z0.b
+; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    and w0, w8, #0x1
+; CHECK-NEXT:    add sp, sp, #32
+; CHECK-NEXT:    ret
+  %v0 = bitcast ptr %a to <16 x float>*
+  %v1 = load <16 x float>, <16 x float>* %v0, align 4
+  %v2 = fcmp une <16 x float> %v1, zeroinitializer
+  %v3 = bitcast float* %b to <16 x float>*
+  %v4 = load <16 x float>, <16 x float>* %v3, align 4
+  %v5 = fcmp une <16 x float> %v4, zeroinitializer
+  %v6 = or <16 x i1> %v2, %v5
+  %v7 = call i1 @llvm.vector.reduce.or.i1.v16i1 (<16 x i1> %v6)
+  ret i1 %v7
+}
+
+declare i1 @llvm.vector.reduce.or.i1.v16i1(<16 x i1>)
+
+;
+; AND reduction.
+;
+
+define i1 @ptest_and_v16i1(ptr %a, ptr %b) #0 {
+; CHECK-LABEL: ptest_and_v16i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub sp, sp, #32
+; CHECK-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NEXT:    adrp x8, .LCPI2_0
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    ldp q2, q1, [x0, #32]
+; CHECK-NEXT:    ldr q0, [x8, :lo12:.LCPI2_0]
+; CHECK-NEXT:    ldp q4, q3, [x0]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z1.s, z0.s
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p1.s, p0/z, z2.s, z0.s
+; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    mov z5.s, z1.s[2]
+; CHECK-NEXT:    mov z6.s, z1.s[1]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z3.s, z0.s
+; CHECK-NEXT:    mov z3.s, z1.s[3]
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p1.s, p0/z, z4.s, z0.s
+; CHECK-NEXT:    mov z7.s, z2.s[3]
+; CHECK-NEXT:    mov z16.s, z2.s[2]
+; CHECK-NEXT:    mov z17.s, z2.s[1]
+; CHECK-NEXT:    fmov w9, s2
+; CHECK-NEXT:    strb w8, [sp, #12]
+; CHECK-NEXT:    mov z2.s, z1.s[3]
+; CHECK-NEXT:    mov z4.s, z1.s[2]
+; CHECK-NEXT:    mov z18.s, z1.s[1]
+; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fmov w10, s1
+; CHECK-NEXT:    strb w9, [sp, #8]
+; CHECK-NEXT:    fmov w9, s3
+; CHECK-NEXT:    mov z19.s, z1.s[3]
+; CHECK-NEXT:    strb w8, [sp, #4]
+; CHECK-NEXT:    fmov w8, s5
+; CHECK-NEXT:    strb w10, [sp]
+; CHECK-NEXT:    fmov w10, s6
+; CHECK-NEXT:    strb w9, [sp, #15]
+; CHECK-NEXT:    fmov w9, s7
+; CHECK-NEXT:    strb w8, [sp, #14]
+; CHECK-NEXT:    fmov w8, s16
+; CHECK-NEXT:    strb w10, [sp, #13]
+; CHECK-NEXT:    fmov w10, s17
+; CHECK-NEXT:    strb w9, [sp, #11]
+; CHECK-NEXT:    fmov w9, s2
+; CHECK-NEXT:    strb w8, [sp, #10]
+; CHECK-NEXT:    fmov w8, s4
+; CHECK-NEXT:    strb w10, [sp, #9]
+; CHECK-NEXT:    fmov w10, s18
+; CHECK-NEXT:    mov z20.s, z1.s[2]
+; CHECK-NEXT:    mov z21.s, z1.s[1]
+; CHECK-NEXT:    strb w9, [sp, #7]
+; CHECK-NEXT:    fmov w9, s19
+; CHECK-NEXT:    strb w8, [sp, #6]
+; CHECK-NEXT:    fmov w8, s20
+; CHECK-NEXT:    strb w10, [sp, #5]
+; CHECK-NEXT:    fmov w10, s21
+; CHECK-NEXT:    strb w9, [sp, #3]
+; CHECK-NEXT:    strb w8, [sp, #2]
+; CHECK-NEXT:    strb w10, [sp, #1]
+; CHECK-NEXT:    ldr q1, [x1, #48]
+; CHECK-NEXT:    ldr q5, [x1]
+; CHECK-NEXT:    ldr q7, [sp]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z1.s, z0.s
+; CHECK-NEXT:    ldp q6, q1, [x1, #16]
+; CHECK-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fmov w8, s2
+; CHECK-NEXT:    mov z3.s, z2.s[3]
+; CHECK-NEXT:    mov z4.s, z2.s[2]
+; CHECK-NEXT:    mov z2.s, z2.s[1]
+; CHECK-NEXT:    strb w8, [sp, #28]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z1.s, z0.s
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p1.s, p0/z, z6.s, z0.s
+; CHECK-NEXT:    mov z6.s, z1.s[3]
+; CHECK-NEXT:    mov z16.s, z1.s[2]
+; CHECK-NEXT:    mov z17.s, z1.s[1]
+; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    fcmne p0.s, p0/z, z5.s, z0.s
+; CHECK-NEXT:    mov z0.s, z1.s[3]
+; CHECK-NEXT:    mov z5.s, z1.s[2]
+; CHECK-NEXT:    mov z18.s, z1.s[1]
+; CHECK-NEXT:    fmov w9, s1
+; CHECK-NEXT:    mov z1.s, p0/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    strb w8, [sp, #24]
+; CHECK-NEXT:    fmov w10, s1
+; CHECK-NEXT:    fmov w8, s3
+; CHECK-NEXT:    strb w9, [sp, #20]
+; CHECK-NEXT:    fmov w9, s4
+; CHECK-NEXT:    mov z19.s, z1.s[3]
+; CHECK-NEXT:    mov z20.s, z1.s[2]
+; CHECK-NEXT:    strb w10, [sp, #16]
+; CHECK-NEXT:    fmov w10, s2
+; CHECK-NEXT:    strb w8, [sp, #31]
+; CHECK-NEXT:    fmov w8, s6
+; CHECK-NEXT:    strb w9, [sp, #30]
+; CHECK-NEXT:    fmov w9, s16
+; CHECK-NEXT:    strb w10, [sp, #29]
+; CHECK-NEXT:    fmov w10, s17
+; CHECK-NEXT:    strb w8, [sp, #27]
+; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    strb w9, [sp, #26]
+; CHECK-NEXT:    fmov w9, s5
+; CHECK-NEXT:    strb w10, [sp, #25]
+; CHECK-NEXT:    fmov w10, s18
+; CHECK-NEXT:    mov z21.s, z1.s[1]
+; CHECK-NEXT:    strb w8, [sp, #23]
+; CHECK-NEXT:    fmov w8, s19
+; CHECK-NEXT:    strb w9, [sp, #22]
+; CHECK-NEXT:    fmov w9, s20
+; CHECK-NEXT:    strb w10, [sp, #21]
+; CHECK-NEXT:    fmov w10, s21
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    strb w8, [sp, #19]
+; CHECK-NEXT:    strb w9, [sp, #18]
+; CHECK-NEXT:    strb w10, [sp, #17]
+; CHECK-NEXT:    ldr q0, [sp, #16]
+; CHECK-NEXT:    and z0.d, z7.d, z0.d
+; CHECK-NEXT:    andv b0, p0, z0.b
+; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    and w0, w8, #0x1
+; CHECK-NEXT:    add sp, sp, #32
+; CHECK-NEXT:    ret
+  %v0 = bitcast ptr %a to <16 x float>*
+  %v1 = load <16 x float>, <16 x float>* %v0, align 4
+  %v2 = fcmp une <16 x float> %v1, zeroinitializer
+  %v3 = bitcast float* %b to <16 x float>*
+  %v4 = load <16 x float>, <16 x float>* %v3, align 4
+  %v5 = fcmp une <16 x float> %v4, zeroinitializer
+  %v6 = and <16 x i1> %v2, %v5
+  %v7 = call i1 @llvm.vector.reduce.and.i1.v16i1 (<16 x i1> %v6)
+  ret i1 %v7
+}
+
+attributes #0 = { "target-features"="+sve" }
+
+declare i1 @llvm.vector.reduce.and.i1.v16i1(<16 x i1>)