[ARM] Regenerate vcvt test checks.

llvm-svn: 314160

diff --git a/llvm/test/CodeGen/ARM/vcvt.ll b/llvm/test/CodeGen/ARM/vcvt.ll
index 0d7dc54..be91ba7 100644 (file)
--- a/llvm/test/CodeGen/ARM/vcvt.ll
+++ b/llvm/test/CodeGen/ARM/vcvt.ll
@@ -1,96 +1,149 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=arm-eabi -mattr=+neon,+fp16 %s -o - | FileCheck %s
 
 define <2 x i32> @vcvt_f32tos32(<2 x float>* %A) nounwind {
-;CHECK-LABEL: vcvt_f32tos32:
-;CHECK: vcvt.s32.f32
+; CHECK-LABEL: vcvt_f32tos32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.s32.f32 d16, d16
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x float>, <2 x float>* %A
        %tmp2 = fptosi <2 x float> %tmp1 to <2 x i32>
        ret <2 x i32> %tmp2
 }
 
 define <2 x i32> @vcvt_f32tou32(<2 x float>* %A) nounwind {
-;CHECK-LABEL: vcvt_f32tou32:
-;CHECK: vcvt.u32.f32
+; CHECK-LABEL: vcvt_f32tou32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.u32.f32 d16, d16
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x float>, <2 x float>* %A
        %tmp2 = fptoui <2 x float> %tmp1 to <2 x i32>
        ret <2 x i32> %tmp2
 }
 
 define <2 x float> @vcvt_s32tof32(<2 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvt_s32tof32:
-;CHECK: vcvt.f32.s32
+; CHECK-LABEL: vcvt_s32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.f32.s32 d16, d16
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x i32>, <2 x i32>* %A
        %tmp2 = sitofp <2 x i32> %tmp1 to <2 x float>
        ret <2 x float> %tmp2
 }
 
 define <2 x float> @vcvt_u32tof32(<2 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvt_u32tof32:
-;CHECK: vcvt.f32.u32
+; CHECK-LABEL: vcvt_u32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.f32.u32 d16, d16
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x i32>, <2 x i32>* %A
        %tmp2 = uitofp <2 x i32> %tmp1 to <2 x float>
        ret <2 x float> %tmp2
 }
 
 define <4 x i32> @vcvtQ_f32tos32(<4 x float>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_f32tos32:
-;CHECK: vcvt.s32.f32
+; CHECK-LABEL: vcvtQ_f32tos32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.s32.f32 q8, q8
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x float>, <4 x float>* %A
        %tmp2 = fptosi <4 x float> %tmp1 to <4 x i32>
        ret <4 x i32> %tmp2
 }
 
 define <4 x i32> @vcvtQ_f32tou32(<4 x float>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_f32tou32:
-;CHECK: vcvt.u32.f32
+; CHECK-LABEL: vcvtQ_f32tou32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.u32.f32 q8, q8
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x float>, <4 x float>* %A
        %tmp2 = fptoui <4 x float> %tmp1 to <4 x i32>
        ret <4 x i32> %tmp2
 }
 
 define <4 x float> @vcvtQ_s32tof32(<4 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_s32tof32:
-;CHECK: vcvt.f32.s32
+; CHECK-LABEL: vcvtQ_s32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.f32.s32 q8, q8
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x i32>, <4 x i32>* %A
        %tmp2 = sitofp <4 x i32> %tmp1 to <4 x float>
        ret <4 x float> %tmp2
 }
 
 define <4 x float> @vcvtQ_u32tof32(<4 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_u32tof32:
-;CHECK: vcvt.f32.u32
+; CHECK-LABEL: vcvtQ_u32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.f32.u32 q8, q8
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x i32>, <4 x i32>* %A
        %tmp2 = uitofp <4 x i32> %tmp1 to <4 x float>
        ret <4 x float> %tmp2
 }
 
 define <2 x i32> @vcvt_n_f32tos32(<2 x float>* %A) nounwind {
-;CHECK-LABEL: vcvt_n_f32tos32:
-;CHECK: vcvt.s32.f32
+; CHECK-LABEL: vcvt_n_f32tos32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.s32.f32 d16, d16, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x float>, <2 x float>* %A
        %tmp2 = call <2 x i32> @llvm.arm.neon.vcvtfp2fxs.v2i32.v2f32(<2 x float> %tmp1, i32 1)
        ret <2 x i32> %tmp2
 }
 
 define <2 x i32> @vcvt_n_f32tou32(<2 x float>* %A) nounwind {
-;CHECK-LABEL: vcvt_n_f32tou32:
-;CHECK: vcvt.u32.f32
+; CHECK-LABEL: vcvt_n_f32tou32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.u32.f32 d16, d16, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x float>, <2 x float>* %A
        %tmp2 = call <2 x i32> @llvm.arm.neon.vcvtfp2fxu.v2i32.v2f32(<2 x float> %tmp1, i32 1)
        ret <2 x i32> %tmp2
 }
 
 define <2 x float> @vcvt_n_s32tof32(<2 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvt_n_s32tof32:
-;CHECK: vcvt.f32.s32
+; CHECK-LABEL: vcvt_n_s32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.f32.s32 d16, d16, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x i32>, <2 x i32>* %A
        %tmp2 = call <2 x float> @llvm.arm.neon.vcvtfxs2fp.v2f32.v2i32(<2 x i32> %tmp1, i32 1)
        ret <2 x float> %tmp2
 }
 
 define <2 x float> @vcvt_n_u32tof32(<2 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvt_n_u32tof32:
-;CHECK: vcvt.f32.u32
+; CHECK-LABEL: vcvt_n_u32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.f32.u32 d16, d16, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <2 x i32>, <2 x i32>* %A
        %tmp2 = call <2 x float> @llvm.arm.neon.vcvtfxu2fp.v2f32.v2i32(<2 x i32> %tmp1, i32 1)
        ret <2 x float> %tmp2
@@ -102,32 +155,52 @@ declare <2 x float> @llvm.arm.neon.vcvtfxs2fp.v2f32.v2i32(<2 x i32>, i32) nounwi
 declare <2 x float> @llvm.arm.neon.vcvtfxu2fp.v2f32.v2i32(<2 x i32>, i32) nounwind readnone
 
 define <4 x i32> @vcvtQ_n_f32tos32(<4 x float>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_n_f32tos32:
-;CHECK: vcvt.s32.f32
+; CHECK-LABEL: vcvtQ_n_f32tos32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.s32.f32 q8, q8, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x float>, <4 x float>* %A
        %tmp2 = call <4 x i32> @llvm.arm.neon.vcvtfp2fxs.v4i32.v4f32(<4 x float> %tmp1, i32 1)
        ret <4 x i32> %tmp2
 }
 
 define <4 x i32> @vcvtQ_n_f32tou32(<4 x float>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_n_f32tou32:
-;CHECK: vcvt.u32.f32
+; CHECK-LABEL: vcvtQ_n_f32tou32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.u32.f32 q8, q8, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x float>, <4 x float>* %A
        %tmp2 = call <4 x i32> @llvm.arm.neon.vcvtfp2fxu.v4i32.v4f32(<4 x float> %tmp1, i32 1)
        ret <4 x i32> %tmp2
 }
 
 define <4 x float> @vcvtQ_n_s32tof32(<4 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_n_s32tof32:
-;CHECK: vcvt.f32.s32
+; CHECK-LABEL: vcvtQ_n_s32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.f32.s32 q8, q8, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x i32>, <4 x i32>* %A
        %tmp2 = call <4 x float> @llvm.arm.neon.vcvtfxs2fp.v4f32.v4i32(<4 x i32> %tmp1, i32 1)
        ret <4 x float> %tmp2
 }
 
 define <4 x float> @vcvtQ_n_u32tof32(<4 x i32>* %A) nounwind {
-;CHECK-LABEL: vcvtQ_n_u32tof32:
-;CHECK: vcvt.f32.u32
+; CHECK-LABEL: vcvtQ_n_u32tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.f32.u32 q8, q8, #1
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x i32>, <4 x i32>* %A
        %tmp2 = call <4 x float> @llvm.arm.neon.vcvtfxu2fp.v4f32.v4i32(<4 x i32> %tmp1, i32 1)
        ret <4 x float> %tmp2
@@ -139,16 +212,25 @@ declare <4 x float> @llvm.arm.neon.vcvtfxs2fp.v4f32.v4i32(<4 x i32>, i32) nounwi
 declare <4 x float> @llvm.arm.neon.vcvtfxu2fp.v4f32.v4i32(<4 x i32>, i32) nounwind readnone
 
 define <4 x float> @vcvt_f16tof32(<4 x i16>* %A) nounwind {
-;CHECK-LABEL: vcvt_f16tof32:
-;CHECK: vcvt.f32.f16
+; CHECK-LABEL: vcvt_f16tof32:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vldr d16, [r0]
+; CHECK-NEXT:    vcvt.f32.f16 q8, d16
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x i16>, <4 x i16>* %A
        %tmp2 = call <4 x float> @llvm.arm.neon.vcvthf2fp(<4 x i16> %tmp1)
        ret <4 x float> %tmp2
 }
 
 define <4 x i16> @vcvt_f32tof16(<4 x float>* %A) nounwind {
-;CHECK-LABEL: vcvt_f32tof16:
-;CHECK: vcvt.f16.f32
+; CHECK-LABEL: vcvt_f32tof16:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r0]
+; CHECK-NEXT:    vcvt.f16.f32 d16, q8
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
        %tmp1 = load <4 x float>, <4 x float>* %A
        %tmp2 = call <4 x i16> @llvm.arm.neon.vcvtfp2hf(<4 x float> %tmp1)
        ret <4 x i16> %tmp2
@@ -160,8 +242,13 @@ declare <4 x i16> @llvm.arm.neon.vcvtfp2hf(<4 x float>) nounwind readnone
 
 define <4 x i16> @fix_float_to_i16(<4 x float> %in) {
 ; CHECK-LABEL: fix_float_to_i16:
-; CHECK: vcvt.u32.f32 [[TMP:q[0-9]+]], {{q[0-9]+}}, #1
-; CHECK: vmovn.i32 {{d[0-9]+}}, [[TMP]]
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vmov d17, r2, r3
+; CHECK-NEXT:    vmov d16, r0, r1
+; CHECK-NEXT:    vcvt.u32.f32 q8, q8, #1
+; CHECK-NEXT:    vmovn.i32 d16, q8
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
 
   %scale = fmul <4 x float> %in, <float 2.0, float 2.0, float 2.0, float 2.0>
   %conv = fptoui <4 x float> %scale to <4 x i16>
@@ -170,8 +257,28 @@ define <4 x i16> @fix_float_to_i16(<4 x float> %in) {
 
 define <2 x i64> @fix_float_to_i64(<2 x float> %in) {
 ; CHECK-LABEL: fix_float_to_i64:
-; CHECK: bl
-; CHECK: bl
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    .save {r4, lr}
+; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov d16, r0, r1
+; CHECK-NEXT:    vadd.f32 d8, d16, d16
+; CHECK-NEXT:    vmov r0, s17
+; CHECK-NEXT:    bl __aeabi_f2ulz
+; CHECK-NEXT:    mov r4, r1
+; CHECK-NEXT:    vmov r1, s16
+; CHECK-NEXT:    vmov.32 d9[0], r0
+; CHECK-NEXT:    mov r0, r1
+; CHECK-NEXT:    bl __aeabi_f2ulz
+; CHECK-NEXT:    vmov.32 d8[0], r0
+; CHECK-NEXT:    vmov.32 d9[1], r4
+; CHECK-NEXT:    vmov.32 d8[1], r1
+; CHECK-NEXT:    vmov r2, r3, d9
+; CHECK-NEXT:    vmov r0, r1, d8
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    pop {r4, lr}
+; CHECK-NEXT:    mov pc, lr
 
   %scale = fmul <2 x float> %in, <float 2.0, float 2.0>
   %conv = fptoui <2 x float> %scale to <2 x i64>
@@ -180,8 +287,30 @@ define <2 x i64> @fix_float_to_i64(<2 x float> %in) {
 
 define <4 x i16> @fix_double_to_i16(<4 x double> %in) {
 ; CHECK-LABEL: fix_double_to_i16:
-; CHECK: vcvt.u32.f64
-; CHECK: vcvt.u32.f64
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vmov d18, r0, r1
+; CHECK-NEXT:    mov r12, sp
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r12]
+; CHECK-NEXT:    vmov d19, r2, r3
+; CHECK-NEXT:    vadd.f64 d18, d18, d18
+; CHECK-NEXT:    vcvt.u32.f64 s0, d18
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    vadd.f64 d20, d16, d16
+; CHECK-NEXT:    vadd.f64 d19, d19, d19
+; CHECK-NEXT:    vadd.f64 d16, d17, d17
+; CHECK-NEXT:    vcvt.u32.f64 s2, d20
+; CHECK-NEXT:    vcvt.u32.f64 s4, d19
+; CHECK-NEXT:    vcvt.u32.f64 s6, d16
+; CHECK-NEXT:    vmov.32 d16[0], r0
+; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    vmov.32 d17[0], r0
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    vmov.32 d16[1], r0
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov.32 d17[1], r0
+; CHECK-NEXT:    vmovn.i32 d16, q8
+; CHECK-NEXT:    vmov r0, r1, d16
+; CHECK-NEXT:    mov pc, lr
 
   %scale = fmul <4 x double> %in, <double 2.0, double 2.0, double 2.0, double 2.0>
   %conv = fptoui <4 x double> %scale to <4 x i16>
@@ -190,29 +319,63 @@ define <4 x i16> @fix_double_to_i16(<4 x double> %in) {
 
 define <2 x i64> @fix_double_to_i64(<2 x double> %in) {
 ; CHECK-LABEL: fix_double_to_i64:
-; CHECK: bl
-; CHECK: bl
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    .save {r4, lr}
+; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov d16, r2, r3
+; CHECK-NEXT:    vadd.f64 d16, d16, d16
+; CHECK-NEXT:    vmov r2, r3, d16
+; CHECK-NEXT:    vmov d16, r0, r1
+; CHECK-NEXT:    vadd.f64 d8, d16, d16
+; CHECK-NEXT:    mov r0, r2
+; CHECK-NEXT:    mov r1, r3
+; CHECK-NEXT:    bl __aeabi_d2ulz
+; CHECK-NEXT:    mov r4, r1
+; CHECK-NEXT:    vmov r2, r1, d8
+; CHECK-NEXT:    vmov.32 d9[0], r0
+; CHECK-NEXT:    mov r0, r2
+; CHECK-NEXT:    bl __aeabi_d2ulz
+; CHECK-NEXT:    vmov.32 d8[0], r0
+; CHECK-NEXT:    vmov.32 d9[1], r4
+; CHECK-NEXT:    vmov.32 d8[1], r1
+; CHECK-NEXT:    vmov r2, r3, d9
+; CHECK-NEXT:    vmov r0, r1, d8
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    pop {r4, lr}
+; CHECK-NEXT:    mov pc, lr
   %scale = fmul <2 x double> %in, <double 2.0, double 2.0>
   %conv = fptoui <2 x double> %scale to <2 x i64>
   ret <2 x i64> %conv
 }
 
 define i32 @multi_sint(double %c, i32* nocapture %p, i32* nocapture %q) {
+; CHECK-LABEL: multi_sint:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vmov d16, r0, r1
+; CHECK-NEXT:    vcvt.s32.f64 s0, d16
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    str r0, [r2]
+; CHECK-NEXT:    str r0, [r3]
+; CHECK-NEXT:    mov pc, lr
   %conv = fptosi double %c to i32
   store i32 %conv, i32* %p, align 4
   store i32 %conv, i32* %q, align 4
   ret i32 %conv
-; CHECK-LABEL: multi_sint:
-; CHECK: vcvt.s32.f64
-; CHECK-NOT: vcvt
 }
 
 define i32 @multi_uint(double %c, i32* nocapture %p, i32* nocapture %q) {
+; CHECK-LABEL: multi_uint:
+; CHECK:       @ BB#0:
+; CHECK-NEXT:    vmov d16, r0, r1
+; CHECK-NEXT:    vcvt.u32.f64 s0, d16
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    str r0, [r2]
+; CHECK-NEXT:    str r0, [r3]
+; CHECK-NEXT:    mov pc, lr
   %conv = fptoui double %c to i32
   store i32 %conv, i32* %p, align 4
   store i32 %conv, i32* %q, align 4
   ret i32 %conv
-; CHECK-LABEL: multi_uint:
-; CHECK: vcvt.u32.f64
-; CHECK-NOT: vcvt
 }