[ARM] Regenerate vld2.ll test. NFC
(cherry picked from commit bcdb7d3c8f)
This commit is contained in:
David Green
Cullen Rhodes
parent
5f04300518
commit
d033ffe4ee
1 changed files with 167 additions and 100 deletions
|
|
@ -1,3 +1,4 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
||||
; RUN: llc -mtriple=arm-eabi -mattr=+neon %s -o - | FileCheck %s
|
||||
|
||||
%struct.__neon_int8x8x2_t = type { <8 x i8>, <8 x i8> }
|
||||
|
|
@ -11,137 +12,203 @@
|
|||
%struct.__neon_int32x4x2_t = type { <4 x i32>, <4 x i32> }
|
||||
%struct.__neon_float32x4x2_t = type { <4 x float>, <4 x float> }
|
||||
|
||||
define <8 x i8> @vld2i8(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2i8:
|
||||
;Check the alignment value. Max for this instruction is 128 bits:
|
||||
;CHECK: vld2.8 {d16, d17}, [{{r[0-9]+|lr}}:64]
|
||||
%tmp1 = call %struct.__neon_int8x8x2_t @llvm.arm.neon.vld2.v8i8.p0(ptr %A, i32 8)
|
||||
%tmp2 = extractvalue %struct.__neon_int8x8x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int8x8x2_t %tmp1, 1
|
||||
%tmp4 = add <8 x i8> %tmp2, %tmp3
|
||||
ret <8 x i8> %tmp4
|
||||
define <8 x i8> @vld2i8(ptr %A) nounwind {
|
||||
; CHECK-LABEL: vld2i8:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.8 {d16, d17}, [r0:64]
|
||||
; CHECK-NEXT: vadd.i8 d16, d16, d17
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_int8x8x2_t @llvm.arm.neon.vld2.v8i8.p0(ptr %A, i32 8)
|
||||
%tmp2 = extractvalue %struct.__neon_int8x8x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int8x8x2_t %tmp1, 1
|
||||
%tmp4 = add <8 x i8> %tmp2, %tmp3
|
||||
ret <8 x i8> %tmp4
|
||||
}
|
||||
|
||||
define <4 x i16> @vld2i16(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2i16:
|
||||
;Check the alignment value. Max for this instruction is 128 bits:
|
||||
;CHECK: vld2.16 {d16, d17}, [{{r[0-9]+|lr}}:128]
|
||||
%tmp1 = call %struct.__neon_int16x4x2_t @llvm.arm.neon.vld2.v4i16.p0(ptr %A, i32 32)
|
||||
%tmp2 = extractvalue %struct.__neon_int16x4x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int16x4x2_t %tmp1, 1
|
||||
%tmp4 = add <4 x i16> %tmp2, %tmp3
|
||||
ret <4 x i16> %tmp4
|
||||
define <4 x i16> @vld2i16(ptr %A) nounwind {
|
||||
; CHECK-LABEL: vld2i16:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.16 {d16, d17}, [r0:128]
|
||||
; CHECK-NEXT: vadd.i16 d16, d16, d17
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_int16x4x2_t @llvm.arm.neon.vld2.v4i16.p0(ptr %A, i32 32)
|
||||
%tmp2 = extractvalue %struct.__neon_int16x4x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int16x4x2_t %tmp1, 1
|
||||
%tmp4 = add <4 x i16> %tmp2, %tmp3
|
||||
ret <4 x i16> %tmp4
|
||||
}
|
||||
|
||||
define <2 x i32> @vld2i32(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2i32:
|
||||
;CHECK: vld2.32
|
||||
%tmp1 = call %struct.__neon_int32x2x2_t @llvm.arm.neon.vld2.v2i32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_int32x2x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int32x2x2_t %tmp1, 1
|
||||
%tmp4 = add <2 x i32> %tmp2, %tmp3
|
||||
ret <2 x i32> %tmp4
|
||||
; CHECK-LABEL: vld2i32:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.32 {d16, d17}, [r0]
|
||||
; CHECK-NEXT: vadd.i32 d16, d16, d17
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_int32x2x2_t @llvm.arm.neon.vld2.v2i32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_int32x2x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int32x2x2_t %tmp1, 1
|
||||
%tmp4 = add <2 x i32> %tmp2, %tmp3
|
||||
ret <2 x i32> %tmp4
|
||||
}
|
||||
|
||||
define <2 x float> @vld2f(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2f:
|
||||
;CHECK: vld2.32
|
||||
%tmp1 = call %struct.__neon_float32x2x2_t @llvm.arm.neon.vld2.v2f32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 1
|
||||
%tmp4 = fadd <2 x float> %tmp2, %tmp3
|
||||
ret <2 x float> %tmp4
|
||||
; CHECK-LABEL: vld2f:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.32 {d16, d17}, [r0]
|
||||
; CHECK-NEXT: vadd.f32 d16, d16, d17
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_float32x2x2_t @llvm.arm.neon.vld2.v2f32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 1
|
||||
%tmp4 = fadd <2 x float> %tmp2, %tmp3
|
||||
ret <2 x float> %tmp4
|
||||
}
|
||||
|
||||
;Check for a post-increment updating load.
|
||||
;Check for a post-increment updating load.
|
||||
define <2 x float> @vld2f_update(ptr %ptr) nounwind {
|
||||
;CHECK-LABEL: vld2f_update:
|
||||
;CHECK: vld2.32 {d16, d17}, [{{r[0-9]+|lr}}]!
|
||||
%A = load ptr, ptr %ptr
|
||||
%tmp1 = call %struct.__neon_float32x2x2_t @llvm.arm.neon.vld2.v2f32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 1
|
||||
%tmp4 = fadd <2 x float> %tmp2, %tmp3
|
||||
%tmp5 = getelementptr float, ptr %A, i32 4
|
||||
store ptr %tmp5, ptr %ptr
|
||||
ret <2 x float> %tmp4
|
||||
; CHECK-LABEL: vld2f_update:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: ldr r3, [r0]
|
||||
; CHECK-NEXT: vld2.32 {d16, d17}, [r3]!
|
||||
; CHECK-NEXT: vadd.f32 d16, d16, d17
|
||||
; CHECK-NEXT: str r3, [r0]
|
||||
; CHECK-NEXT: vmov r2, r1, d16
|
||||
; CHECK-NEXT: mov r0, r2
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%A = load ptr, ptr %ptr
|
||||
%tmp1 = call %struct.__neon_float32x2x2_t @llvm.arm.neon.vld2.v2f32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_float32x2x2_t %tmp1, 1
|
||||
%tmp4 = fadd <2 x float> %tmp2, %tmp3
|
||||
%tmp5 = getelementptr float, ptr %A, i32 4
|
||||
store ptr %tmp5, ptr %ptr
|
||||
ret <2 x float> %tmp4
|
||||
}
|
||||
|
||||
define <1 x i64> @vld2i64(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2i64:
|
||||
;Check the alignment value. Max for this instruction is 128 bits:
|
||||
;CHECK: vld1.64 {d16, d17}, [{{r[0-9]+|lr}}:128]
|
||||
%tmp1 = call %struct.__neon_int64x1x2_t @llvm.arm.neon.vld2.v1i64.p0(ptr %A, i32 32)
|
||||
%tmp2 = extractvalue %struct.__neon_int64x1x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int64x1x2_t %tmp1, 1
|
||||
%tmp4 = add <1 x i64> %tmp2, %tmp3
|
||||
ret <1 x i64> %tmp4
|
||||
define <1 x i64> @vld2i64(ptr %A) nounwind {
|
||||
; CHECK-LABEL: vld2i64:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld1.64 {d16, d17}, [r0:128]
|
||||
; CHECK-NEXT: vadd.i64 d16, d16, d17
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_int64x1x2_t @llvm.arm.neon.vld2.v1i64.p0(ptr %A, i32 32)
|
||||
%tmp2 = extractvalue %struct.__neon_int64x1x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int64x1x2_t %tmp1, 1
|
||||
%tmp4 = add <1 x i64> %tmp2, %tmp3
|
||||
ret <1 x i64> %tmp4
|
||||
}
|
||||
|
||||
define <16 x i8> @vld2Qi8(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2Qi8:
|
||||
;Check the alignment value. Max for this instruction is 256 bits:
|
||||
;CHECK: vld2.8 {d16, d17, d18, d19}, [{{r[0-9]+|lr}}:64]
|
||||
%tmp1 = call %struct.__neon_int8x16x2_t @llvm.arm.neon.vld2.v16i8.p0(ptr %A, i32 8)
|
||||
%tmp2 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 1
|
||||
%tmp4 = add <16 x i8> %tmp2, %tmp3
|
||||
ret <16 x i8> %tmp4
|
||||
define <16 x i8> @vld2Qi8(ptr %A) nounwind {
|
||||
; CHECK-LABEL: vld2Qi8:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.8 {d16, d17, d18, d19}, [r0:64]
|
||||
; CHECK-NEXT: vadd.i8 q8, q8, q9
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: vmov r2, r3, d17
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_int8x16x2_t @llvm.arm.neon.vld2.v16i8.p0(ptr %A, i32 8)
|
||||
%tmp2 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 1
|
||||
%tmp4 = add <16 x i8> %tmp2, %tmp3
|
||||
ret <16 x i8> %tmp4
|
||||
}
|
||||
|
||||
;Check for a post-increment updating load with register increment.
|
||||
define <16 x i8> @vld2Qi8_update(ptr %ptr, i32 %inc) nounwind {
|
||||
;CHECK-LABEL: vld2Qi8_update:
|
||||
;CHECK: vld2.8 {d16, d17, d18, d19}, [{{r[0-9]+|lr}}:128], r1
|
||||
%A = load ptr, ptr %ptr
|
||||
%tmp1 = call %struct.__neon_int8x16x2_t @llvm.arm.neon.vld2.v16i8.p0(ptr %A, i32 16)
|
||||
%tmp2 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 1
|
||||
%tmp4 = add <16 x i8> %tmp2, %tmp3
|
||||
%tmp5 = getelementptr i8, ptr %A, i32 %inc
|
||||
store ptr %tmp5, ptr %ptr
|
||||
ret <16 x i8> %tmp4
|
||||
; CHECK-LABEL: vld2Qi8_update:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: .save {r11, lr}
|
||||
; CHECK-NEXT: push {r11, lr}
|
||||
; CHECK-NEXT: ldr lr, [r0]
|
||||
; CHECK-NEXT: vld2.8 {d16, d17, d18, d19}, [lr:128], r1
|
||||
; CHECK-NEXT: vadd.i8 q8, q8, q9
|
||||
; CHECK-NEXT: str lr, [r0]
|
||||
; CHECK-NEXT: vmov r12, r1, d16
|
||||
; CHECK-NEXT: vmov r2, r3, d17
|
||||
; CHECK-NEXT: mov r0, r12
|
||||
; CHECK-NEXT: pop {r11, lr}
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%A = load ptr, ptr %ptr
|
||||
%tmp1 = call %struct.__neon_int8x16x2_t @llvm.arm.neon.vld2.v16i8.p0(ptr %A, i32 16)
|
||||
%tmp2 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int8x16x2_t %tmp1, 1
|
||||
%tmp4 = add <16 x i8> %tmp2, %tmp3
|
||||
%tmp5 = getelementptr i8, ptr %A, i32 %inc
|
||||
store ptr %tmp5, ptr %ptr
|
||||
ret <16 x i8> %tmp4
|
||||
}
|
||||
|
||||
;Check the alignment value. Max for this instruction is 256 bits:
|
||||
define <8 x i16> @vld2Qi16(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2Qi16:
|
||||
;Check the alignment value. Max for this instruction is 256 bits:
|
||||
;CHECK: vld2.16 {d16, d17, d18, d19}, [{{r[0-9]+|lr}}:128]
|
||||
%tmp1 = call %struct.__neon_int16x8x2_t @llvm.arm.neon.vld2.v8i16.p0(ptr %A, i32 16)
|
||||
%tmp2 = extractvalue %struct.__neon_int16x8x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int16x8x2_t %tmp1, 1
|
||||
%tmp4 = add <8 x i16> %tmp2, %tmp3
|
||||
ret <8 x i16> %tmp4
|
||||
; CHECK-LABEL: vld2Qi16:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.16 {d16, d17, d18, d19}, [r0:128]
|
||||
; CHECK-NEXT: vadd.i16 q8, q8, q9
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: vmov r2, r3, d17
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_int16x8x2_t @llvm.arm.neon.vld2.v8i16.p0(ptr %A, i32 16)
|
||||
%tmp2 = extractvalue %struct.__neon_int16x8x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int16x8x2_t %tmp1, 1
|
||||
%tmp4 = add <8 x i16> %tmp2, %tmp3
|
||||
ret <8 x i16> %tmp4
|
||||
}
|
||||
|
||||
define <4 x i32> @vld2Qi32(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2Qi32:
|
||||
;Check the alignment value. Max for this instruction is 256 bits:
|
||||
;CHECK: vld2.32 {d16, d17, d18, d19}, [{{r[0-9]+|lr}}:256]
|
||||
%tmp1 = call %struct.__neon_int32x4x2_t @llvm.arm.neon.vld2.v4i32.p0(ptr %A, i32 64)
|
||||
%tmp2 = extractvalue %struct.__neon_int32x4x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int32x4x2_t %tmp1, 1
|
||||
%tmp4 = add <4 x i32> %tmp2, %tmp3
|
||||
ret <4 x i32> %tmp4
|
||||
define <4 x i32> @vld2Qi32(ptr %A) nounwind {
|
||||
; CHECK-LABEL: vld2Qi32:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.32 {d16, d17, d18, d19}, [r0:256]
|
||||
; CHECK-NEXT: vadd.i32 q8, q8, q9
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: vmov r2, r3, d17
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_int32x4x2_t @llvm.arm.neon.vld2.v4i32.p0(ptr %A, i32 64)
|
||||
%tmp2 = extractvalue %struct.__neon_int32x4x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_int32x4x2_t %tmp1, 1
|
||||
%tmp4 = add <4 x i32> %tmp2, %tmp3
|
||||
ret <4 x i32> %tmp4
|
||||
}
|
||||
|
||||
define <4 x float> @vld2Qf(ptr %A) nounwind {
|
||||
;CHECK-LABEL: vld2Qf:
|
||||
;CHECK: vld2.32
|
||||
%tmp1 = call %struct.__neon_float32x4x2_t @llvm.arm.neon.vld2.v4f32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_float32x4x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_float32x4x2_t %tmp1, 1
|
||||
%tmp4 = fadd <4 x float> %tmp2, %tmp3
|
||||
ret <4 x float> %tmp4
|
||||
; CHECK-LABEL: vld2Qf:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: vld2.32 {d16, d17, d18, d19}, [r0]
|
||||
; CHECK-NEXT: vadd.f32 q8, q8, q9
|
||||
; CHECK-NEXT: vmov r0, r1, d16
|
||||
; CHECK-NEXT: vmov r2, r3, d17
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%tmp1 = call %struct.__neon_float32x4x2_t @llvm.arm.neon.vld2.v4f32.p0(ptr %A, i32 1)
|
||||
%tmp2 = extractvalue %struct.__neon_float32x4x2_t %tmp1, 0
|
||||
%tmp3 = extractvalue %struct.__neon_float32x4x2_t %tmp1, 1
|
||||
%tmp4 = fadd <4 x float> %tmp2, %tmp3
|
||||
ret <4 x float> %tmp4
|
||||
}
|
||||
|
||||
declare %struct.__neon_int8x8x2_t @llvm.arm.neon.vld2.v8i8.p0(ptr, i32) nounwind readonly
|
||||
declare %struct.__neon_int16x4x2_t @llvm.arm.neon.vld2.v4i16.p0(ptr, i32) nounwind readonly
|
||||
declare %struct.__neon_int32x2x2_t @llvm.arm.neon.vld2.v2i32.p0(ptr, i32) nounwind readonly
|
||||
declare %struct.__neon_float32x2x2_t @llvm.arm.neon.vld2.v2f32.p0(ptr, i32) nounwind readonly
|
||||
declare %struct.__neon_int64x1x2_t @llvm.arm.neon.vld2.v1i64.p0(ptr, i32) nounwind readonly
|
||||
|
||||
declare %struct.__neon_int8x16x2_t @llvm.arm.neon.vld2.v16i8.p0(ptr, i32) nounwind readonly
|
||||
declare %struct.__neon_int16x8x2_t @llvm.arm.neon.vld2.v8i16.p0(ptr, i32) nounwind readonly
|
||||
declare %struct.__neon_int32x4x2_t @llvm.arm.neon.vld2.v4i32.p0(ptr, i32) nounwind readonly
|
||||
declare %struct.__neon_float32x4x2_t @llvm.arm.neon.vld2.v4f32.p0(ptr, i32) nounwind readonly
|
||||
; FIXME: Do not update with an invalid increment
|
||||
define ptr @test_or_update(ptr %p, ptr %q) {
|
||||
; CHECK-LABEL: test_or_update:
|
||||
; CHECK: @ %bb.0:
|
||||
; CHECK-NEXT: orr r0, r0, #3
|
||||
; CHECK-NEXT: mov r2, #2
|
||||
; CHECK-NEXT: vld2.32 {d16, d17}, [r0], r2
|
||||
; CHECK-NEXT: vstr d16, [r1]
|
||||
; CHECK-NEXT: mov pc, lr
|
||||
%p1 = ptrtoint ptr %p to i32
|
||||
%p2 = or i32 %p1, 3
|
||||
%p3 = inttoptr i32 %p2 to ptr
|
||||
%l = load <4 x i32>, ptr %p3, align 1
|
||||
%s = shufflevector <4 x i32> %l, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
|
||||
store <2 x i32> %s, ptr %q
|
||||
%q1 = getelementptr i8, ptr %p, i32 2
|
||||
ret ptr %q1
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in a new issue