Merge branch 'main' into users/mingmingl-llvm/samplefdo-profile-formatusers/mingmingl-llvm/samplefdo-profile-format

author: Mingming Liu <mingmingl@google.com> 2025-09-10 15:25:31 -0700
committer: GitHub <noreply@github.com> 2025-09-10 15:25:31 -0700
commit: 1417dafa1db9cb1b2b09438aa9f53ea5ab6e36e2 (patch)
tree: 57f4b1f313c8cf74eed8819870f39c36ea263c68 /llvm/test/CodeGen/ARM
parent: 898b813bc8a6d0276bf0f4769f5f2f64b34e632d (diff)
parent: b8cefcb601ddaa18482555c4ff363c01a270c2fe (diff)
10 files changed, 2177 insertions, 521 deletions
diff --git a/llvm/test/CodeGen/ARM/atomic-cmpxchg.ll b/llvm/test/CodeGen/ARM/atomic-cmpxchg.ll
index 4bf42d4ac962..c37a21be5ca4 100644
--- a/llvm/test/CodeGen/ARM/atomic-cmpxchg.ll
+++ b/llvm/test/CodeGen/ARM/atomic-cmpxchg.ll
@@ -36,16 +36,18 @@ define zeroext i1 @test_cmpxchg_res_i8(ptr %addr, i8 %desired, i8 zeroext %new)
 ; CHECK-THUMB-NEXT:    bx r1
 ;
 ; CHECK-ARMV6-LABEL: test_cmpxchg_res_i8:
-; CHECK-ARMV6:         uxtb r1, r1
+; CHECK-ARMV6:         .fnstart
+; CHECK-ARMV6-NEXT:    uxtb r12, r1
 ; CHECK-ARMV6-NEXT:  .LBB0_1:
-; CHECK-ARMV6-NEXT:    ldrexb r3, [r0]
-; CHECK-ARMV6-NEXT:    cmp r3, r1
+; CHECK-ARMV6-NEXT:    ldrexb r1, [r0]
+; CHECK-ARMV6-NEXT:    cmp r1, r12
 ; CHECK-ARMV6-NEXT:    movne r0, #0
 ; CHECK-ARMV6-NEXT:    bxne lr
 ; CHECK-ARMV6-NEXT:  .LBB0_2:
 ; CHECK-ARMV6-NEXT:    strexb r3, r2, [r0]
+; CHECK-ARMV6-NEXT:    mov r1, #1
 ; CHECK-ARMV6-NEXT:    cmp r3, #0
-; CHECK-ARMV6-NEXT:    moveq r0, #1
+; CHECK-ARMV6-NEXT:    moveq r0, r1
 ; CHECK-ARMV6-NEXT:    bxeq lr
 ; CHECK-ARMV6-NEXT:    b .LBB0_1
 ;
@@ -61,19 +63,22 @@ define zeroext i1 @test_cmpxchg_res_i8(ptr %addr, i8 %desired, i8 zeroext %new)
 ; CHECK-THUMBV6-NEXT:    pop {r4, pc}
 ;
 ; CHECK-ARMV7-LABEL: test_cmpxchg_res_i8:
-; CHECK-ARMV7:         uxtb r1, r1
+; CHECK-ARMV7:         .fnstart
+; CHECK-ARMV7-NEXT:    uxtb r12, r1
 ; CHECK-ARMV7-NEXT:  .LBB0_1:
-; CHECK-ARMV7-NEXT:    ldrexb r3, [r0]
-; CHECK-ARMV7-NEXT:    cmp r3, r1
-; CHECK-ARMV7-NEXT:    bne .LBB0_3
+; CHECK-ARMV7-NEXT:    ldrexb r1, [r0]
+; CHECK-ARMV7-NEXT:    cmp r1, r12
+; CHECK-ARMV7-NEXT:    bne .LBB0_4
 ; CHECK-ARMV7-NEXT:    strexb r3, r2, [r0]
+; CHECK-ARMV7-NEXT:    mov r1, #1
 ; CHECK-ARMV7-NEXT:    cmp r3, #0
-; CHECK-ARMV7-NEXT:    moveq r0, #1
-; CHECK-ARMV7-NEXT:    bxeq lr
-; CHECK-ARMV7-NEXT:    b .LBB0_1
-; CHECK-ARMV7-NEXT:  .LBB0_3:
-; CHECK-ARMV7-NEXT:    mov r0, #0
+; CHECK-ARMV7-NEXT:    bne .LBB0_1 
+; CHECK-ARMV7-NEXT:    mov r0, r1
+; CHECK-ARMV7-NEXT:    bx lr
+; CHECK-ARMV7-NEXT:  .LBB0_4:
+; CHECK-ARMV7-NEXT:    mov r1, #0
 ; CHECK-ARMV7-NEXT:    clrex
+; CHECK-ARMV7-NEXT:    mov r0, r1
 ; CHECK-ARMV7-NEXT:    bx lr
 ;
 ; CHECK-THUMBV7-LABEL: test_cmpxchg_res_i8:
diff --git a/llvm/test/CodeGen/ARM/bswap-inline-asm.ll b/llvm/test/CodeGen/ARM/bswap-inline-asm.ll
index 31f9d729cf6e..cc92a9710c98 100644
--- a/llvm/test/CodeGen/ARM/bswap-inline-asm.ll
+++ b/llvm/test/CodeGen/ARM/bswap-inline-asm.ll
@@ -1,8 +1,10 @@
 ; RUN: llc < %s -mtriple=arm-apple-darwin -mattr=+v6 | FileCheck %s
 
+; rev inline assembly should be preserved as-is.
+
 define i32 @t1(i32 %x) nounwind {
 ; CHECK-LABEL: t1:
-; CHECK-NOT: InlineAsm
+; CHECK: InlineAsm
 ; CHECK: rev
   %asmtmp = tail call i32 asm "rev $0, $1\0A", "=l,l"(i32 %x) nounwind
   ret i32 %asmtmp
diff --git a/llvm/test/CodeGen/ARM/clz.ll b/llvm/test/CodeGen/ARM/clz.ll
index 0f49fbba1184..9e1e9f6ce6da 100644
--- a/llvm/test/CodeGen/ARM/clz.ll
+++ b/llvm/test/CodeGen/ARM/clz.ll
@@ -1,12 +1,41 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc -mtriple=arm-eabi -mattr=+v5t %s -o - | FileCheck %s -check-prefixes=CHECK,INLINE
 ; RUN: llc -mtriple=arm-eabi %s -o - | FileCheck %s -check-prefixes=CHECK,LIBCALL
 
 declare i32 @llvm.ctlz.i32(i32, i1)
 
-define i32 @test(i32 %x) {
-; CHECK-LABEL: test
-; INLINE: clz r0, r0
-; LIBCALL: b __clzsi2
+define i32 @undef_zero(i32 %x) {
+; INLINE-LABEL: undef_zero:
+; INLINE:       @ %bb.0:
+; INLINE-NEXT:    clz r0, r0
+; INLINE-NEXT:    bx lr
+;
+; LIBCALL-LABEL: undef_zero:
+; LIBCALL:       @ %bb.0:
+; LIBCALL-NEXT:    b __clzsi2
         %tmp.1 = call i32 @llvm.ctlz.i32( i32 %x, i1 true )
         ret i32 %tmp.1
 }
+
+define i32 @no_undef_zero(i32 %x) {
+; INLINE-LABEL: no_undef_zero:
+; INLINE:       @ %bb.0:
+; INLINE-NEXT:    clz r0, r0
+; INLINE-NEXT:    bx lr
+;
+; LIBCALL-LABEL: no_undef_zero:
+; LIBCALL:       @ %bb.0:
+; LIBCALL-NEXT:    cmp r0, #0
+; LIBCALL-NEXT:    moveq r0, #32
+; LIBCALL-NEXT:    moveq pc, lr
+; LIBCALL-NEXT:  .LBB1_1: @ %cond.false
+; LIBCALL-NEXT:    .save {r11, lr}
+; LIBCALL-NEXT:    push {r11, lr}
+; LIBCALL-NEXT:    bl __clzsi2
+; LIBCALL-NEXT:    pop {r11, lr}
+; LIBCALL-NEXT:    mov pc, lr
+        %tmp.1 = call i32 @llvm.ctlz.i32( i32 %x, i1 false )
+        ret i32 %tmp.1
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; CHECK: {{.*}}
diff --git a/llvm/test/CodeGen/ARM/cmpxchg-idioms.ll b/llvm/test/CodeGen/ARM/cmpxchg-idioms.ll
index 4ff71b42d5db..d59f282314b5 100644
--- a/llvm/test/CodeGen/ARM/cmpxchg-idioms.ll
+++ b/llvm/test/CodeGen/ARM/cmpxchg-idioms.ll
@@ -4,14 +4,14 @@ define i32 @test_return(ptr %p, i32 %oldval, i32 %newval) {
 ; CHECK-LABEL: test_return:
 
 ; CHECK: ldrex [[LOADED:r[0-9]+]], [r0]
-; CHECK: cmp [[LOADED]], r1
+; CHECK: cmp [[LOADED]], r1 
 ; CHECK: bne [[FAILED:LBB[0-9]+_[0-9]+]]
 
 ; CHECK: dmb ishst
 
 ; CHECK: [[LOOP:LBB[0-9]+_[0-9]+]]:
 ; CHECK: strex [[STATUS:r[0-9]+]], {{r[0-9]+}}, [r0]
-; CHECK: cbz [[STATUS]], [[SUCCESS:LBB[0-9]+_[0-9]+]]
+; CHECK: cmp r3, #0
 
 ; CHECK: ldrex [[LOADED]], [r0]
 ; CHECK: cmp [[LOADED]], r1
@@ -24,12 +24,6 @@ define i32 @test_return(ptr %p, i32 %oldval, i32 %newval) {
 ; CHECK: dmb ish
 ; CHECK: bx lr
 
-; CHECK: [[SUCCESS]]:
-; CHECK-NOT: cmp {{r[0-9]+}}, {{r[0-9]+}}
-; CHECK: movs r0, #1
-; CHECK: dmb ish
-; CHECK: bx lr
-
   %pair = cmpxchg ptr %p, i32 %oldval, i32 %newval seq_cst seq_cst
   %success = extractvalue { i32, i1 } %pair, 1
   %conv = zext i1 %success to i32
@@ -49,7 +43,7 @@ define i1 @test_return_bool(ptr %value, i8 %oldValue, i8 %newValue) {
 
 ; CHECK: [[LOOP:LBB[0-9]+_[0-9]+]]:
 ; CHECK: strexb [[STATUS:r[0-9]+]], {{r[0-9]+}}, [r0]
-; CHECK: cbz [[STATUS]], [[SUCCESS:LBB[0-9]+_[0-9]+]]
+; CHECK: cmp [[STATUS]], #0
 
 ; CHECK: ldrexb [[LOADED]], [r0]
 ; CHECK: cmp [[LOADED]], [[OLDBYTE]]
@@ -63,12 +57,6 @@ define i1 @test_return_bool(ptr %value, i8 %oldValue, i8 %newValue) {
 ; CHECK: eor r0, [[TMP]], #1
 ; CHECK: bx lr
 
-; CHECK: [[SUCCESS]]:
-; CHECK-NOT: cmp {{r[0-9]+}}, {{r[0-9]+}}
-; CHECK: movs [[TMP:r[0-9]+]], #1
-; CHECK: eor r0, [[TMP]], #1
-; CHECK: bx lr
-
 
   %pair = cmpxchg ptr %value, i8 %oldValue, i8 %newValue acq_rel monotonic
   %success = extractvalue { i8, i1 } %pair, 1
@@ -87,7 +75,7 @@ define void @test_conditional(ptr %p, i32 %oldval, i32 %newval) {
 
 ; CHECK: [[LOOP:LBB[0-9]+_[0-9]+]]:
 ; CHECK: strex [[STATUS:r[0-9]+]], r2, [r0]
-; CHECK: cbz [[STATUS]], [[SUCCESS:LBB[0-9]+_[0-9]+]]
+; CHECK: cmp [[STATUS]], #0 
 
 ; CHECK: ldrex [[LOADED]], [r0]
 ; CHECK: cmp [[LOADED]], r1
@@ -99,11 +87,6 @@ define void @test_conditional(ptr %p, i32 %oldval, i32 %newval) {
 ; CHECK: dmb ish
 ; CHECK: b.w _baz
 
-; CHECK: [[SUCCESS]]:
-; CHECK-NOT: cmp {{r[0-9]+}}, {{r[0-9]+}}
-; CHECK: dmb ish
-; CHECK: b.w _bar
-
   %pair = cmpxchg ptr %p, i32 %oldval, i32 %newval seq_cst seq_cst
   %success = extractvalue { i32, i1 } %pair, 1
   br i1 %success, label %true, label %false
diff --git a/llvm/test/CodeGen/ARM/cmpxchg-weak.ll b/llvm/test/CodeGen/ARM/cmpxchg-weak.ll
index 9963f2d08ba5..b33eea997574 100644
--- a/llvm/test/CodeGen/ARM/cmpxchg-weak.ll
+++ b/llvm/test/CodeGen/ARM/cmpxchg-weak.ll
@@ -10,16 +10,14 @@ define void @test_cmpxchg_weak(ptr %addr, i32 %desired, i32 %new) {
 ; CHECK-NEXT:     dmb ish
 ; CHECK-NEXT:     strex   [[SUCCESS:r[0-9]+]], r2, [r0]
 ; CHECK-NEXT:     cmp     [[SUCCESS]], #0
-; CHECK-NEXT:     beq     [[SUCCESSBB:LBB[0-9]+_[0-9]+]]
+; CHECK-NEXT:     bne     [[FAILBB:LBB[0-9]+_[0-9]+]]
 ; CHECK-NEXT: %bb.2:
+; CHECK-NEXT:     dmb     ish
 ; CHECK-NEXT:     str     r3, [r0]
 ; CHECK-NEXT:     bx      lr
 ; CHECK-NEXT: [[LDFAILBB]]:
 ; CHECK-NEXT:     clrex
-; CHECK-NEXT:     str     r3, [r0]
-; CHECK-NEXT:     bx      lr
-; CHECK-NEXT: [[SUCCESSBB]]:
-; CHECK-NEXT:     dmb     ish
+; CHECK-NEXT: [[FAILBB]]:
 ; CHECK-NEXT:     str     r3, [r0]
 ; CHECK-NEXT:     bx      lr
 ;
@@ -37,19 +35,20 @@ define i1 @test_cmpxchg_weak_to_bool(i32, ptr %addr, i32 %desired, i32 %new) {
 ; CHECK-NEXT:     bne     [[LDFAILBB:LBB[0-9]+_[0-9]+]]
 ; CHECK-NEXT: %bb.1:
 ; CHECK-NEXT:     dmb ish
-; CHECK-NEXT:     mov     r0, #0
 ; CHECK-NEXT:     strex   [[SUCCESS:r[0-9]+]], r3, [r1]
 ; CHECK-NEXT:     cmp     [[SUCCESS]], #0
-; CHECK-NEXT:     bxne    lr
-; CHECK-NEXT: LBB1_2:
+; CHECK-NEXT:     bne     [[FAILBB:LBB[0-9]+_[0-9]+]] 
+; CHECK-NEXT: %bb.2:
 ; CHECK-NEXT:     mov     r0, #1
 ; CHECK-NEXT:     dmb     ish
 ; CHECK-NEXT:     bx      lr
 ; CHECK-NEXT: [[LDFAILBB]]:
-; CHECK-NEXT:     mov     r0, #0
 ; CHECK-NEXT:     clrex
+; CHECK-NEXT: [[FAILBB]]:
+; CHECK-NEXT:     mov     r0, #0
 ; CHECK-NEXT:     bx      lr
 ;
+
   %pair = cmpxchg weak ptr %addr, i32 %desired, i32 %new seq_cst monotonic
   %success = extractvalue { i32, i1 } %pair, 1
   ret i1 %success
diff --git a/llvm/test/CodeGen/ARM/cttz.ll b/llvm/test/CodeGen/ARM/cttz.ll
index 1146ad64ee70..bf42e9f1104b 100644
--- a/llvm/test/CodeGen/ARM/cttz.ll
+++ b/llvm/test/CodeGen/ARM/cttz.ll
@@ -1,4 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple arm-eabi -mattr=+v5t | FileCheck %s --check-prefix=CHECK-5
 ; RUN: llc < %s -mtriple arm-eabi -mattr=+v6t2 | FileCheck %s
 ; RUN: llc < %s -mtriple arm-eabi -mattr=+v6t2 -mattr=+neon | FileCheck %s
 ; RUN: llc < %s -mtriple thumbv6m-none-eabi | FileCheck %s --check-prefix=CHECK-6M
@@ -14,6 +15,15 @@ declare i64 @llvm.cttz.i64(i64, i1)
 ;------------------------------------------------------------------------------
 
 define i8 @test_i8(i8 %a) {
+; CHECK-5-LABEL: test_i8:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    orr r0, r0, #256
+; CHECK-5-NEXT:    sub r1, r0, #1
+; CHECK-5-NEXT:    bic r0, r1, r0
+; CHECK-5-NEXT:    clz r0, r0
+; CHECK-5-NEXT:    rsb r0, r0, #32
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i8:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    orr r0, r0, #256
@@ -81,6 +91,15 @@ define i8 @test_i8(i8 %a) {
 }
 
 define i16 @test_i16(i16 %a) {
+; CHECK-5-LABEL: test_i16:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    orr r0, r0, #65536
+; CHECK-5-NEXT:    sub r1, r0, #1
+; CHECK-5-NEXT:    bic r0, r1, r0
+; CHECK-5-NEXT:    clz r0, r0
+; CHECK-5-NEXT:    rsb r0, r0, #32
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i16:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    orr r0, r0, #65536
@@ -148,6 +167,14 @@ define i16 @test_i16(i16 %a) {
 }
 
 define i32 @test_i32(i32 %a) {
+; CHECK-5-LABEL: test_i32:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    sub r1, r0, #1
+; CHECK-5-NEXT:    bic r0, r1, r0
+; CHECK-5-NEXT:    clz r0, r0
+; CHECK-5-NEXT:    rsb r0, r0, #32
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i32:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    rbit r0, r0
@@ -207,6 +234,21 @@ define i32 @test_i32(i32 %a) {
 }
 
 define i64 @test_i64(i64 %a) {
+; CHECK-5-LABEL: test_i64:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    sub r3, r1, #1
+; CHECK-5-NEXT:    sub r2, r0, #1
+; CHECK-5-NEXT:    bic r1, r3, r1
+; CHECK-5-NEXT:    bic r2, r2, r0
+; CHECK-5-NEXT:    clz r1, r1
+; CHECK-5-NEXT:    clz r2, r2
+; CHECK-5-NEXT:    rsb r1, r1, #64
+; CHECK-5-NEXT:    cmp r0, #0
+; CHECK-5-NEXT:    rsbne r1, r2, #32
+; CHECK-5-NEXT:    mov r0, r1
+; CHECK-5-NEXT:    mov r1, #0
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i64:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    rbit r1, r1
@@ -323,6 +365,14 @@ define i64 @test_i64(i64 %a) {
 ;------------------------------------------------------------------------------
 
 define i8 @test_i8_zero_undef(i8 %a) {
+; CHECK-5-LABEL: test_i8_zero_undef:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    sub r1, r0, #1
+; CHECK-5-NEXT:    bic r0, r1, r0
+; CHECK-5-NEXT:    clz r0, r0
+; CHECK-5-NEXT:    rsb r0, r0, #32
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i8_zero_undef:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    rbit r0, r0
@@ -377,6 +427,14 @@ define i8 @test_i8_zero_undef(i8 %a) {
 }
 
 define i16 @test_i16_zero_undef(i16 %a) {
+; CHECK-5-LABEL: test_i16_zero_undef:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    sub r1, r0, #1
+; CHECK-5-NEXT:    bic r0, r1, r0
+; CHECK-5-NEXT:    clz r0, r0
+; CHECK-5-NEXT:    rsb r0, r0, #32
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i16_zero_undef:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    rbit r0, r0
@@ -432,6 +490,14 @@ define i16 @test_i16_zero_undef(i16 %a) {
 
 
 define i32 @test_i32_zero_undef(i32 %a) {
+; CHECK-5-LABEL: test_i32_zero_undef:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    sub r1, r0, #1
+; CHECK-5-NEXT:    bic r0, r1, r0
+; CHECK-5-NEXT:    clz r0, r0
+; CHECK-5-NEXT:    rsb r0, r0, #32
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i32_zero_undef:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    rbit r0, r0
@@ -486,6 +552,21 @@ define i32 @test_i32_zero_undef(i32 %a) {
 }
 
 define i64 @test_i64_zero_undef(i64 %a) {
+; CHECK-5-LABEL: test_i64_zero_undef:
+; CHECK-5:       @ %bb.0:
+; CHECK-5-NEXT:    sub r3, r1, #1
+; CHECK-5-NEXT:    sub r2, r0, #1
+; CHECK-5-NEXT:    bic r1, r3, r1
+; CHECK-5-NEXT:    bic r2, r2, r0
+; CHECK-5-NEXT:    clz r1, r1
+; CHECK-5-NEXT:    clz r2, r2
+; CHECK-5-NEXT:    rsb r1, r1, #64
+; CHECK-5-NEXT:    cmp r0, #0
+; CHECK-5-NEXT:    rsbne r1, r2, #32
+; CHECK-5-NEXT:    mov r0, r1
+; CHECK-5-NEXT:    mov r1, #0
+; CHECK-5-NEXT:    bx lr
+;
 ; CHECK-LABEL: test_i64_zero_undef:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    rbit r1, r1
diff --git a/llvm/test/CodeGen/ARM/fp16-promote.ll b/llvm/test/CodeGen/ARM/fp16-promote.ll
index 1bd01508808c..800ee87b95ca 100644
--- a/llvm/test/CodeGen/ARM/fp16-promote.ll
+++ b/llvm/test/CodeGen/ARM/fp16-promote.ll
@@ -1,20 +1,59 @@
-; RUN: llc -asm-verbose=false < %s -mattr=+vfp3,+fp16 | FileCheck -allow-deprecated-dag-overlap %s -check-prefix=CHECK-FP16  --check-prefix=CHECK-VFP -check-prefix=CHECK-ALL
-; RUN: llc -asm-verbose=false < %s | FileCheck -allow-deprecated-dag-overlap %s -check-prefix=CHECK-LIBCALL --check-prefix=CHECK-VFP -check-prefix=CHECK-ALL --check-prefix=CHECK-LIBCALL-VFP
-; RUN: llc -asm-verbose=false < %s -mattr=-fpregs | FileCheck -allow-deprecated-dag-overlap %s --check-prefix=CHECK-LIBCALL -check-prefix=CHECK-NOVFP -check-prefix=CHECK-ALL
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -asm-verbose=false < %s -mattr=+vfp3,+fp16 | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-VFP,CHECK-FP16
+; RUN: llc -asm-verbose=false < %s | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-LIBCALL,CHECK-VFP,CHECK-LIBCALL-VFP
+; RUN: llc -asm-verbose=false < %s -mattr=-fpregs | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-LIBCALL,CHECK-NOVFP
 
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-n32"
 target triple = "armv7---eabihf"
 
 define void @test_fadd(ptr %p, ptr %q) #0 {
-; CHECK-ALL-LABEL: test_fadd:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vadd.f32
-; CHECK-NOVFP: bl __aeabi_fadd
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_fadd:
+; CHECK-FP16:         ldrh r2, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r2
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vadd.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fadd:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vadd.f32 s0, s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fadd:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_fadd
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = fadd half %a, %b
@@ -23,15 +62,53 @@ define void @test_fadd(ptr %p, ptr %q) #0 {
 }
 
 define void @test_fsub(ptr %p, ptr %q) #0 {
-; CHECK-ALL-LABEL: test_fsub:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vsub.f32
-; CHECK-NOVFP: bl __aeabi_fsub
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_fsub:
+; CHECK-FP16:         ldrh r2, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r2
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vsub.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fsub:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vsub.f32 s0, s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fsub:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_fsub
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = fsub half %a, %b
@@ -40,15 +117,53 @@ define void @test_fsub(ptr %p, ptr %q) #0 {
 }
 
 define void @test_fmul(ptr %p, ptr %q) #0 {
-; CHECK-ALL-LABEL: test_fmul:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vmul.f32
-; CHECK-NOVFP: bl __aeabi_fmul
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_fmul:
+; CHECK-FP16:         ldrh r2, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r2
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vmul.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fmul:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmul.f32 s0, s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fmul:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_fmul
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = fmul half %a, %b
@@ -57,15 +172,53 @@ define void @test_fmul(ptr %p, ptr %q) #0 {
 }
 
 define void @test_fdiv(ptr %p, ptr %q) #0 {
-; CHECK-ALL-LABEL: test_fdiv:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vdiv.f32
-; CHECK-NOVFP: bl __aeabi_fdiv
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_fdiv:
+; CHECK-FP16:         ldrh r2, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r2
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vdiv.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fdiv:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vdiv.f32 s0, s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fdiv:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_fdiv
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = fdiv half %a, %b
@@ -74,14 +227,53 @@ define void @test_fdiv(ptr %p, ptr %q) #0 {
 }
 
 define void @test_frem(ptr %p, ptr %q) #0 {
-; CHECK-ALL-LABEL: test_frem:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl fmodf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_frem:
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s1, s2
+; CHECK-FP16-NEXT:    bl fmodf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_frem:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s1, r5
+; CHECK-LIBCALL-VFP-NEXT:    bl fmodf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_frem:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl fmodf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = frem half %a, %b
@@ -91,9 +283,9 @@ define void @test_frem(ptr %p, ptr %q) #0 {
 
 define void @test_load_store(ptr %p, ptr %q) #0 {
 ; CHECK-ALL-LABEL: test_load_store:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL: ldrh {{r[0-9]+}}, [{{r[0-9]+}}]
-; CHECK-ALL: strh {{r[0-9]+}}, [{{r[0-9]+}}]
+; CHECK-ALL:         ldrh r0, [r0]
+; CHECK-ALL-NEXT:    strh r0, [r1]
+; CHECK-ALL-NEXT:    bx lr
   %a = load half, ptr %p, align 2
   store half %a, ptr %q
   ret void
@@ -106,42 +298,48 @@ declare half @test_callee(half %a, half %b) #0
 
 define half @test_call(half %a, half %b) #0 {
 ; CHECK-ALL-LABEL: test_call:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: .save {r11, lr}
-; CHECK-ALL-NEXT: push {r11, lr}
-; CHECK-ALL-NEXT: bl test_callee
-; CHECK-ALL-NEXT: pop {r11, pc}
+; CHECK-ALL:         .save {r11, lr}
+; CHECK-ALL-NEXT:    push {r11, lr}
+; CHECK-ALL-NEXT:    bl test_callee
+; CHECK-ALL-NEXT:    pop {r11, pc}
   %r = call half @test_callee(half %a, half %b)
   ret half %r
 }
 
 define half @test_call_flipped(half %a, half %b) #0 {
-; CHECK-ALL-LABEL: test_call_flipped:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: .save {r11, lr}
-; CHECK-ALL-NEXT: push {r11, lr}
-; CHECK-VFP-NEXT: vmov.f32 s2, s0
-; CHECK-VFP-NEXT: vmov.f32 s0, s1
-; CHECK-VFP-NEXT: vmov.f32 s1, s2
-; CHECK-NOVFP-NEXT: mov r2, r0
-; CHECK-NOVFP-NEXT: mov r0, r1
-; CHECK-NOVFP-NEXT: mov r1, r2
-; CHECK-ALL-NEXT: bl test_callee
-; CHECK-ALL-NEXT: pop {r11, pc}
+; CHECK-VFP-LABEL: test_call_flipped:
+; CHECK-VFP:         .save {r11, lr}
+; CHECK-VFP-NEXT:    push {r11, lr}
+; CHECK-VFP-NEXT:    vmov.f32 s2, s0
+; CHECK-VFP-NEXT:    vmov.f32 s0, s1
+; CHECK-VFP-NEXT:    vmov.f32 s1, s2
+; CHECK-VFP-NEXT:    bl test_callee
+; CHECK-VFP-NEXT:    pop {r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_call_flipped:
+; CHECK-NOVFP:         .save {r11, lr}
+; CHECK-NOVFP-NEXT:    push {r11, lr}
+; CHECK-NOVFP-NEXT:    mov r2, r0
+; CHECK-NOVFP-NEXT:    mov r0, r1
+; CHECK-NOVFP-NEXT:    mov r1, r2
+; CHECK-NOVFP-NEXT:    bl test_callee
+; CHECK-NOVFP-NEXT:    pop {r11, pc}
   %r = call half @test_callee(half %b, half %a)
   ret half %r
 }
 
 define half @test_tailcall_flipped(half %a, half %b) #0 {
-; CHECK-ALL-LABEL: test_tailcall_flipped:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-VFP-NEXT: vmov.f32 s2, s0
-; CHECK-VFP-NEXT: vmov.f32 s0, s1
-; CHECK-VFP-NEXT: vmov.f32 s1, s2
-; CHECK-NOVFP-NEXT: mov r2, r0
-; CHECK-NOVFP-NEXT: mov r0, r1
-; CHECK-NOVFP-NEXT: mov r1, r2
-; CHECK-ALL-NEXT: b test_callee
+; CHECK-VFP-LABEL: test_tailcall_flipped:
+; CHECK-VFP:         vmov.f32 s2, s0
+; CHECK-VFP-NEXT:    vmov.f32 s0, s1
+; CHECK-VFP-NEXT:    vmov.f32 s1, s2
+; CHECK-VFP-NEXT:    b test_callee
+;
+; CHECK-NOVFP-LABEL: test_tailcall_flipped:
+; CHECK-NOVFP:         mov r2, r0
+; CHECK-NOVFP-NEXT:    mov r0, r1
+; CHECK-NOVFP-NEXT:    mov r1, r2
+; CHECK-NOVFP-NEXT:    b test_callee
   %r = tail call half @test_callee(half %b, half %a)
   ret half %r
 }
@@ -150,10 +348,11 @@ define half @test_tailcall_flipped(half %a, half %b) #0 {
 ; No conversion is needed
 define void @test_select(ptr %p, ptr %q, i1 zeroext %c) #0 {
 ; CHECK-ALL-LABEL: test_select:
-; CHECK-ALL: cmp {{r[0-9]+}}, #0
-; CHECK-ALL: movne {{r[0-9]+}}, {{r[0-9]+}}
-; CHECK-ALL: ldrh {{r[0-9]+}}, [{{r[0-9]+}}]
-; CHECK-ALL: strh {{r[0-9]+}}, [{{r[0-9]+}}]
+; CHECK-ALL:         cmp r2, #0
+; CHECK-ALL-NEXT:    movne r1, r0
+; CHECK-ALL-NEXT:    ldrh r1, [r1]
+; CHECK-ALL-NEXT:    strh r1, [r0]
+; CHECK-ALL-NEXT:    bx lr
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = select i1 %c, half %a, half %b
@@ -164,17 +363,53 @@ define void @test_select(ptr %p, ptr %q, i1 zeroext %c) #0 {
 ; Test only two variants of fcmp.  These get translated to f32 vcmp
 ; instructions anyway.
 define i1 @test_fcmp_une(ptr %p, ptr %q) #0 {
-; CHECK-ALL-LABEL: test_fcmp_une:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vcmp.f32
-; CHECK-NOVFP: bl __aeabi_fcmpeq
-; CHECK-VFP-NEXT: vmrs APSR_nzcv, fpscr
-; CHECK-VFP-NEXT: movwne
-; CHECK-NOVFP-NEXT: clz r0, r0
-; CHECK-NOVFP-NEXT: lsr r0, r0, #5
+; CHECK-FP16-LABEL: test_fcmp_une:
+; CHECK-FP16:         ldrh r2, [r0]
+; CHECK-FP16-NEXT:    mov r0, #0
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r2
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vcmp.f32 s0, s2
+; CHECK-FP16-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-FP16-NEXT:    movwne r0, #1
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fcmp_une:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, #0
+; CHECK-LIBCALL-VFP-NEXT:    vcmp.f32 s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-LIBCALL-VFP-NEXT:    movwne r0, #1
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fcmp_une:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_fcmpeq
+; CHECK-NOVFP-NEXT:    clz r0, r0
+; CHECK-NOVFP-NEXT:    lsr r0, r0, #5
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = fcmp une half %a, %b
@@ -182,15 +417,60 @@ define i1 @test_fcmp_une(ptr %p, ptr %q) #0 {
 }
 
 define i1 @test_fcmp_ueq(ptr %p, ptr %q) #0 {
-; CHECK-ALL-LABEL: test_fcmp_ueq:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vcmp.f32
-; CHECK-NOVFP: bl __aeabi_fcmpeq
-; CHECK-FP16: vmrs APSR_nzcv, fpscr
-; CHECK-LIBCALL: movw{{ne|eq}}
+; CHECK-FP16-LABEL: test_fcmp_ueq:
+; CHECK-FP16:         ldrh r2, [r0]
+; CHECK-FP16-NEXT:    mov r0, #0
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r2
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vcmp.f32 s0, s2
+; CHECK-FP16-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-FP16-NEXT:    movweq r0, #1
+; CHECK-FP16-NEXT:    movwvs r0, #1
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fcmp_ueq:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, #0
+; CHECK-LIBCALL-VFP-NEXT:    vcmp.f32 s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-LIBCALL-VFP-NEXT:    movweq r0, #1
+; CHECK-LIBCALL-VFP-NEXT:    movwvs r0, #1
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fcmp_ueq:
+; CHECK-NOVFP:         .save {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    bl __aeabi_fcmpeq
+; CHECK-NOVFP-NEXT:    mov r6, r0
+; CHECK-NOVFP-NEXT:    mov r0, r4
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_fcmpun
+; CHECK-NOVFP-NEXT:    orrs r0, r0, r6
+; CHECK-NOVFP-NEXT:    movwne r0, #1
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r6, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = fcmp ueq half %a, %b
@@ -198,17 +478,61 @@ define i1 @test_fcmp_ueq(ptr %p, ptr %q) #0 {
 }
 
 define void @test_br_cc(ptr %p, ptr %q, ptr %p1, ptr %p2) #0 {
-; CHECK-ALL-LABEL: test_br_cc:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vcmp.f32
-; CHECK-NOVFP: bl __aeabi_fcmplt
-; CHECK-FP16: vmrs APSR_nzcv, fpscr
-; CHECK-VFP: movmi
-; CHECK-VFP: str
-; CHECK-NOVFP: str
+; CHECK-FP16-LABEL: test_br_cc:
+; CHECK-FP16:         ldrh r0, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    mov r0, #0
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vcmp.f32 s0, s2
+; CHECK-FP16-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-FP16-NEXT:    movmi r2, r3
+; CHECK-FP16-NEXT:    str r0, [r2]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_br_cc:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r6, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r6, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r3
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r2
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r6]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, #0
+; CHECK-LIBCALL-VFP-NEXT:    vcmp.f32 s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-LIBCALL-VFP-NEXT:    movmi r5, r4
+; CHECK-LIBCALL-VFP-NEXT:    str r0, [r5]
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r6, pc}
+;
+; CHECK-NOVFP-LABEL: test_br_cc:
+; CHECK-NOVFP:         .save {r4, r5, r6, r7, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r6, r7, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r6, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    mov r4, r3
+; CHECK-NOVFP-NEXT:    mov r5, r2
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r7, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r6]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r7
+; CHECK-NOVFP-NEXT:    bl __aeabi_fcmplt
+; CHECK-NOVFP-NEXT:    cmp r0, #0
+; CHECK-NOVFP-NEXT:    mov r0, #0
+; CHECK-NOVFP-NEXT:    movne r5, r4
+; CHECK-NOVFP-NEXT:    str r0, [r5]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r6, r7, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %c = fcmp uge half %a, %b
@@ -222,14 +546,37 @@ else:
 }
 
 declare i1 @test_dummy(ptr %p) #0
-; CHECK-ALL-LABEL: test_phi:
-; CHECK-FP16: [[LOOP:.LBB[0-9_]+]]:
-; CHECK-FP16: bl      test_dummy
-; CHECK-FP16: bne     [[LOOP]]
-; CHECK-LIBCALL: [[LOOP:.LBB[0-9_]+]]:
-; CHECK-LIBCALL: bl test_dummy
-; CHECK-LIBCALL: bne     [[LOOP]]
+
 define void @test_phi(ptr %p) #0 {
+; CHECK-VFP-LABEL: test_phi:
+; CHECK-VFP:         .save {r4, r5, r6, lr}
+; CHECK-VFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-VFP-NEXT:    ldrh r6, [r0]
+; CHECK-VFP-NEXT:    mov r4, r0
+; CHECK-VFP-NEXT:  .LBB13_1:
+; CHECK-VFP-NEXT:    mov r0, r4
+; CHECK-VFP-NEXT:    mov r5, r6
+; CHECK-VFP-NEXT:    ldrh r6, [r4]
+; CHECK-VFP-NEXT:    bl test_dummy
+; CHECK-VFP-NEXT:    tst r0, #1
+; CHECK-VFP-NEXT:    bne .LBB13_1
+; CHECK-VFP-NEXT:    strh r5, [r4]
+; CHECK-VFP-NEXT:    pop {r4, r5, r6, pc}
+;
+; CHECK-NOVFP-LABEL: test_phi:
+; CHECK-NOVFP:         .save {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    ldrh r5, [r0]
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:  .LBB13_1:
+; CHECK-NOVFP-NEXT:    mov r0, r4
+; CHECK-NOVFP-NEXT:    mov r6, r5
+; CHECK-NOVFP-NEXT:    ldrh r5, [r4]
+; CHECK-NOVFP-NEXT:    bl test_dummy
+; CHECK-NOVFP-NEXT:    tst r0, #1
+; CHECK-NOVFP-NEXT:    bne .LBB13_1
+; CHECK-NOVFP-NEXT:    strh r6, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %a = load half, ptr %p
   br label %loop
@@ -244,84 +591,224 @@ return:
 }
 
 define i32 @test_fptosi_i32(ptr %p) #0 {
-; CHECK-ALL-LABEL: test_fptosi_i32:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vcvt.s32.f32
-; CHECK-NOVFP: bl __aeabi_f2iz
+; CHECK-FP16-LABEL: test_fptosi_i32:
+; CHECK-FP16:         ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvt.s32.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fptosi_i32:
+; CHECK-LIBCALL-VFP:         .save {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vcvt.s32.f32 s0, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    pop {r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_fptosi_i32:
+; CHECK-NOVFP:         .save {r11, lr}
+; CHECK-NOVFP-NEXT:    push {r11, lr}
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2iz
+; CHECK-NOVFP-NEXT:    pop {r11, pc}
   %a = load half, ptr %p, align 2
   %r = fptosi half %a to i32
   ret i32 %r
 }
 
 define i64 @test_fptosi_i64(ptr %p) #0 {
-; CHECK-ALL-LABEL: test_fptosi_i64:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-ALL: bl __aeabi_f2lz
+; CHECK-FP16-LABEL: test_fptosi_i64:
+; CHECK-FP16:         .save {r11, lr}
+; CHECK-FP16-NEXT:    push {r11, lr}
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    bl __aeabi_f2lz
+; CHECK-FP16-NEXT:    pop {r11, pc}
+;
+; CHECK-LIBCALL-LABEL: test_fptosi_i64:
+; CHECK-LIBCALL:         .save {r11, lr}
+; CHECK-LIBCALL-NEXT:    push {r11, lr}
+; CHECK-LIBCALL-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-NEXT:    bl __aeabi_f2lz
+; CHECK-LIBCALL-NEXT:    pop {r11, pc}
   %a = load half, ptr %p, align 2
   %r = fptosi half %a to i64
   ret i64 %r
 }
 
 define i32 @test_fptoui_i32(ptr %p) #0 {
-; CHECK-ALL-LABEL: test_fptoui_i32:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vcvt.u32.f32
-; CHECK-NOVFP: bl __aeabi_f2uiz
+; CHECK-FP16-LABEL: test_fptoui_i32:
+; CHECK-FP16:         ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvt.u32.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fptoui_i32:
+; CHECK-LIBCALL-VFP:         .save {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vcvt.u32.f32 s0, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    pop {r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_fptoui_i32:
+; CHECK-NOVFP:         .save {r11, lr}
+; CHECK-NOVFP-NEXT:    push {r11, lr}
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2uiz
+; CHECK-NOVFP-NEXT:    pop {r11, pc}
   %a = load half, ptr %p, align 2
   %r = fptoui half %a to i32
   ret i32 %r
 }
 
 define i64 @test_fptoui_i64(ptr %p) #0 {
-; CHECK-ALL-LABEL: test_fptoui_i64:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-ALL: bl __aeabi_f2ulz
+; CHECK-FP16-LABEL: test_fptoui_i64:
+; CHECK-FP16:         .save {r11, lr}
+; CHECK-FP16-NEXT:    push {r11, lr}
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    bl __aeabi_f2ulz
+; CHECK-FP16-NEXT:    pop {r11, pc}
+;
+; CHECK-LIBCALL-LABEL: test_fptoui_i64:
+; CHECK-LIBCALL:         .save {r11, lr}
+; CHECK-LIBCALL-NEXT:    push {r11, lr}
+; CHECK-LIBCALL-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-NEXT:    bl __aeabi_f2ulz
+; CHECK-LIBCALL-NEXT:    pop {r11, pc}
   %a = load half, ptr %p, align 2
   %r = fptoui half %a to i64
   ret i64 %r
 }
 
 define void @test_sitofp_i32(i32 %a, ptr %p) #0 {
-; CHECK-ALL-LABEL: test_sitofp_i32:
-; CHECK-VFP: vcvt.f32.s32
-; CHECK-NOVFP: bl __aeabi_i2f
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_sitofp_i32:
+; CHECK-FP16:         vmov s0, r0
+; CHECK-FP16-NEXT:    vcvt.f32.s32 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r1]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_sitofp_i32:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r1
+; CHECK-LIBCALL-VFP-NEXT:    vcvt.f32.s32 s0, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_sitofp_i32:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r1
+; CHECK-NOVFP-NEXT:    bl __aeabi_i2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %r = sitofp i32 %a to half
   store half %r, ptr %p
   ret void
 }
 
 define void @test_uitofp_i32(i32 %a, ptr %p) #0 {
-; CHECK-ALL-LABEL: test_uitofp_i32:
-; CHECK-VFP: vcvt.f32.u32
-; CHECK-NOVFP: bl __aeabi_ui2f
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_uitofp_i32:
+; CHECK-FP16:         vmov s0, r0
+; CHECK-FP16-NEXT:    vcvt.f32.u32 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r1]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_uitofp_i32:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r1
+; CHECK-LIBCALL-VFP-NEXT:    vcvt.f32.u32 s0, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_uitofp_i32:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r1
+; CHECK-NOVFP-NEXT:    bl __aeabi_ui2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %r = uitofp i32 %a to half
   store half %r, ptr %p
   ret void
 }
 
 define void @test_sitofp_i64(i64 %a, ptr %p) #0 {
-; CHECK-ALL-LABEL: test_sitofp_i64:
-; CHECK-ALL: bl __aeabi_l2f
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_sitofp_i64:
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r2
+; CHECK-FP16-NEXT:    bl __aeabi_l2f
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-LABEL: test_sitofp_i64:
+; CHECK-LIBCALL:         .save {r4, lr}
+; CHECK-LIBCALL-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-NEXT:    mov r4, r2
+; CHECK-LIBCALL-NEXT:    bl __aeabi_l2f
+; CHECK-LIBCALL-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-NEXT:    pop {r4, pc}
   %r = sitofp i64 %a to half
   store half %r, ptr %p
   ret void
 }
 
 define void @test_uitofp_i64(i64 %a, ptr %p) #0 {
-; CHECK-ALL-LABEL: test_uitofp_i64:
-; CHECK-ALL: bl __aeabi_ul2f
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_uitofp_i64:
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r2
+; CHECK-FP16-NEXT:    bl __aeabi_ul2f
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-LABEL: test_uitofp_i64:
+; CHECK-LIBCALL:         .save {r4, lr}
+; CHECK-LIBCALL-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-NEXT:    mov r4, r2
+; CHECK-LIBCALL-NEXT:    bl __aeabi_ul2f
+; CHECK-LIBCALL-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-NEXT:    pop {r4, pc}
   %r = uitofp i64 %a to half
   store half %r, ptr %p
   ret void
@@ -329,19 +816,49 @@ define void @test_uitofp_i64(i64 %a, ptr %p) #0 {
 
 define void @test_fptrunc_float(float %f, ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_fptrunc_float:
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fptrunc_float:
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fptrunc_float:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fptrunc_float:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r1
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = fptrunc float %f to half
   store half %a, ptr %p
   ret void
 }
 
 define void @test_fptrunc_double(double %d, ptr %p) #0 {
-; CHECK-FP16-LABEL: test_fptrunc_double:
-; CHECK-FP16: bl __aeabi_d2h
-; CHECK-LIBCALL-LABEL: test_fptrunc_double:
-; CHECK-LIBCALL: bl __aeabi_d2h
+; CHECK-VFP-LABEL: test_fptrunc_double:
+; CHECK-VFP:         .save {r4, lr}
+; CHECK-VFP-NEXT:    push {r4, lr}
+; CHECK-VFP-NEXT:    mov r4, r0
+; CHECK-VFP-NEXT:    vmov r0, r1, d0
+; CHECK-VFP-NEXT:    bl __aeabi_d2h
+; CHECK-VFP-NEXT:    strh r0, [r4]
+; CHECK-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_fptrunc_double:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r2
+; CHECK-NOVFP-NEXT:    bl __aeabi_d2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = fptrunc double %d to half
   store half %a, ptr %p
   ret void
@@ -349,9 +866,25 @@ define void @test_fptrunc_double(double %d, ptr %p) #0 {
 
 define float @test_fpextend_float(ptr %p) {
 ; CHECK-FP16-LABEL: test_fpextend_float:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL-LABEL: test_fpextend_float:
-; CHECK-LIBCALL: bl __aeabi_h2f
+; CHECK-FP16:         ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fpextend_float:
+; CHECK-LIBCALL-VFP:         .save {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    pop {r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_fpextend_float:
+; CHECK-NOVFP:         .save {r11, lr}
+; CHECK-NOVFP-NEXT:    push {r11, lr}
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    pop {r11, pc}
   %a = load half, ptr %p, align 2
   %r = fpext half %a to float
   ret float %r
@@ -359,11 +892,28 @@ define float @test_fpextend_float(ptr %p) {
 
 define double @test_fpextend_double(ptr %p) {
 ; CHECK-FP16-LABEL: test_fpextend_double:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL-LABEL: test_fpextend_double:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-VFP: vcvt.f64.f32
-; CHECK-NOVFP: bl __aeabi_f2d
+; CHECK-FP16:         ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvt.f64.f32 d0, s0
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fpextend_double:
+; CHECK-LIBCALL-VFP:         .save {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vcvt.f64.f32 d0, s0
+; CHECK-LIBCALL-VFP-NEXT:    pop {r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_fpextend_double:
+; CHECK-NOVFP:         .save {r11, lr}
+; CHECK-NOVFP-NEXT:    push {r11, lr}
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2d
+; CHECK-NOVFP-NEXT:    pop {r11, pc}
   %a = load half, ptr %p, align 2
   %r = fpext half %a to double
   ret double %r
@@ -371,9 +921,8 @@ define double @test_fpextend_double(ptr %p) {
 
 define i16 @test_bitcast_halftoi16(ptr %p) #0 {
 ; CHECK-ALL-LABEL: test_bitcast_halftoi16:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: ldrh r0, [r0]
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL:         ldrh r0, [r0]
+; CHECK-ALL-NEXT:    bx lr
   %a = load half, ptr %p, align 2
   %r = bitcast half %a to i16
   ret i16 %r
@@ -381,9 +930,8 @@ define i16 @test_bitcast_halftoi16(ptr %p) #0 {
 
 define void @test_bitcast_i16tohalf(i16 %a, ptr %p) #0 {
 ; CHECK-ALL-LABEL: test_bitcast_i16tohalf:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: strh r0, [r1]
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL:         strh r0, [r1]
+; CHECK-ALL-NEXT:    bx lr
   %r = bitcast i16 %a to half
   store half %r, ptr %p
   ret void
@@ -415,14 +963,39 @@ declare half @llvm.roundeven.f16(half %a) #0
 declare half @llvm.fmuladd.f16(half %a, half %b, half %c) #0
 
 define void @test_sqrt(ptr %p) #0 {
-; CHECK-ALL-LABEL: test_sqrt:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vsqrt.f32
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL-VFP: vsqrt.f32
-; CHECK-NOVFP: bl sqrtf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16-LABEL: test_sqrt:
+; CHECK-FP16:         ldrh r1, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vsqrt.f32 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_sqrt:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vsqrt.f32 s0, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_sqrt:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl sqrtf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.sqrt.f16(half %a)
   store half %r, ptr %p
@@ -431,13 +1004,46 @@ define void @test_sqrt(ptr %p) #0 {
 
 define void @test_fpowi(ptr %p, i32 %b) #0 {
 ; CHECK-FP16-LABEL: test_fpowi:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl __powisf2
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fpowi:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __powisf2
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    mov r0, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl __powisf2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fpowi:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r4
+; CHECK-LIBCALL-VFP-NEXT:    bl __powisf2
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r5]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_fpowi:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    mov r4, r1
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r4
+; CHECK-NOVFP-NEXT:    bl __powisf2
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r5]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.powi.f16.i32(half %a, i32 %b)
   store half %r, ptr %p
@@ -446,13 +1052,41 @@ define void @test_fpowi(ptr %p, i32 %b) #0 {
 
 define void @test_sin(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_sin:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl sinf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_sin:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl sinf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl sinf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_sin:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl sinf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_sin:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl sinf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.sin.f16(half %a)
   store half %r, ptr %p
@@ -461,13 +1095,41 @@ define void @test_sin(ptr %p) #0 {
 
 define void @test_cos(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_cos:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl cosf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_cos:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl cosf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl cosf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_cos:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl cosf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_cos:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl cosf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.cos.f16(half %a)
   store half %r, ptr %p
@@ -476,13 +1138,41 @@ define void @test_cos(ptr %p) #0 {
 
 define void @test_tan(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_tan:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl tanf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_tan:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl tanf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl tanf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_tan:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl tanf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_tan:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl tanf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.tan.f16(half %a)
   store half %r, ptr %p
@@ -491,15 +1181,52 @@ define void @test_tan(ptr %p) #0 {
 
 define void @test_pow(ptr %p, ptr %q) #0 {
 ; CHECK-FP16-LABEL: test_pow:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl powf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_pow:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl powf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s1, s2
+; CHECK-FP16-NEXT:    bl powf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_pow:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s1, r5
+; CHECK-LIBCALL-VFP-NEXT:    bl powf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_pow:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl powf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = call half @llvm.pow.f16(half %a, half %b)
@@ -509,13 +1236,51 @@ define void @test_pow(ptr %p, ptr %q) #0 {
 
 define void @test_cbrt(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_cbrt:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl powf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_cbrt:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl powf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vldr s1, .LCPI34_0
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl powf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+; CHECK-FP16-NEXT:    .p2align 2
+; CHECK-FP16-NEXT:  .LCPI34_0:
+; CHECK-FP16-NEXT:    .long 0x3eaaa000
+;
+; CHECK-LIBCALL-VFP-LABEL: test_cbrt:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vldr s1, .LCPI34_0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl powf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+; CHECK-LIBCALL-VFP-NEXT:    .p2align 2
+; CHECK-LIBCALL-VFP-NEXT:  .LCPI34_0:
+; CHECK-LIBCALL-VFP-NEXT:    .long 0x3eaaa000
+;
+; CHECK-NOVFP-LABEL: test_cbrt:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    movw r1, #40960
+; CHECK-NOVFP-NEXT:    movt r1, #16042
+; CHECK-NOVFP-NEXT:    bl powf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.pow.f16(half %a, half 0x3FD5540000000000)
   store half %r, ptr %p
@@ -524,13 +1289,41 @@ define void @test_cbrt(ptr %p) #0 {
 
 define void @test_exp(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_exp:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl expf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_exp:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl expf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl expf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_exp:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl expf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_exp:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl expf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.exp.f16(half %a)
   store half %r, ptr %p
@@ -539,13 +1332,41 @@ define void @test_exp(ptr %p) #0 {
 
 define void @test_exp2(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_exp2:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl exp2f
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_exp2:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl exp2f
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl exp2f
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_exp2:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl exp2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_exp2:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl exp2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.exp2.f16(half %a)
   store half %r, ptr %p
@@ -554,13 +1375,41 @@ define void @test_exp2(ptr %p) #0 {
 
 define void @test_log(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_log:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl logf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_log:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl logf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl logf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_log:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl logf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_log:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl logf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.log.f16(half %a)
   store half %r, ptr %p
@@ -569,13 +1418,41 @@ define void @test_log(ptr %p) #0 {
 
 define void @test_log10(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_log10:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl log10f
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_log10:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl log10f
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl log10f
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_log10:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl log10f
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_log10:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl log10f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.log10.f16(half %a)
   store half %r, ptr %p
@@ -584,13 +1461,41 @@ define void @test_log10(ptr %p) #0 {
 
 define void @test_log2(ptr %p) #0 {
 ; CHECK-FP16-LABEL: test_log2:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl log2f
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_log2:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl log2f
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl log2f
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_log2:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl log2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_log2:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl log2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.log2.f16(half %a)
   store half %r, ptr %p
@@ -599,17 +1504,65 @@ define void @test_log2(ptr %p) #0 {
 
 define void @test_fma(ptr %p, ptr %q, ptr %r) #0 {
 ; CHECK-FP16-LABEL: test_fma:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl fmaf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fma:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl fmaf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r1]
+; CHECK-FP16-NEXT:    ldrh r1, [r4]
+; CHECK-FP16-NEXT:    ldrh r2, [r2]
+; CHECK-FP16-NEXT:    vmov s2, r0
+; CHECK-FP16-NEXT:    vmov s0, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s1, s2
+; CHECK-FP16-NEXT:    vmov s2, r2
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    bl fmaf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fma:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r6, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r2]
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r6, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r5]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s1, r5
+; CHECK-LIBCALL-VFP-NEXT:    vmov s2, r6
+; CHECK-LIBCALL-VFP-NEXT:    bl fmaf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r6, pc}
+;
+; CHECK-NOVFP-LABEL: test_fma:
+; CHECK-NOVFP:         .save {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    mov r5, r2
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r6, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r5]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r6
+; CHECK-NOVFP-NEXT:    mov r2, r5
+; CHECK-NOVFP-NEXT:    bl fmaf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r6, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %c = load half, ptr %r, align 2
@@ -620,13 +1573,25 @@ define void @test_fma(ptr %p, ptr %q, ptr %r) #0 {
 
 define void @test_fabs(ptr %p) {
 ; CHECK-FP16-LABEL: test_fabs:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vabs.f32
-; CHECK-FP16: vcvtb.f16.f32
+; CHECK-FP16:         ldrh r1, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vabs.f32 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
 ; CHECK-LIBCALL-LABEL: test_fabs:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bic
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-LIBCALL:         .save {r4, lr}
+; CHECK-LIBCALL-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-NEXT:    mov r4, r0
+; CHECK-LIBCALL-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-NEXT:    bic r0, r0, #-2147483648
+; CHECK-LIBCALL-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.fabs.f16(half %a)
   store half %r, ptr %p
@@ -635,15 +1600,52 @@ define void @test_fabs(ptr %p) {
 
 define void @test_minnum(ptr %p, ptr %q) #0 {
 ; CHECK-FP16-LABEL: test_minnum:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl fminf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_minnum:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl fminf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s1, s2
+; CHECK-FP16-NEXT:    bl fminf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_minnum:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s1, r5
+; CHECK-LIBCALL-VFP-NEXT:    bl fminf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_minnum:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl fminf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = call half @llvm.minnum.f16(half %a, half %b)
@@ -653,15 +1655,52 @@ define void @test_minnum(ptr %p, ptr %q) #0 {
 
 define void @test_maxnum(ptr %p, ptr %q) #0 {
 ; CHECK-FP16-LABEL: test_maxnum:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl fmaxf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_maxnum:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl fmaxf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s1, s2
+; CHECK-FP16-NEXT:    bl fmaxf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_maxnum:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s1, r5
+; CHECK-LIBCALL-VFP-NEXT:    bl fmaxf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_maxnum:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl fmaxf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %r = call half @llvm.maxnum.f16(half %a, half %b)
@@ -670,16 +1709,45 @@ define void @test_maxnum(ptr %p, ptr %q) #0 {
 }
 
 define void @test_minimum(ptr %p) #0 {
-; CHECK-ALL-LABEL: test_minimum:
-; CHECK-FP16: vmov.f32 s0, #1.000000e+00
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL-VFP: vmov.f32 s{{[0-9]+}}, #1.000000e+00
-; CHECK-NOVFP: mov r{{[0-9]+}}, #1065353216
-; CHECK-VFP: vcmp.f32
-; CHECK-VFP: vmrs
-; CHECK-VFP: movge
-; CHECK-NOVFP: bl __aeabi_fcmpge
+; CHECK-FP16-LABEL: test_minimum:
+; CHECK-FP16:         vmov.f32 s0, #1.000000e+00
+; CHECK-FP16-NEXT:    ldrh r1, [r0]
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vcmp.f32 s2, s0
+; CHECK-FP16-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-FP16-NEXT:    movge r1, #15360
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_minimum:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r5, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r5
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov.f32 s0, #1.000000e+00
+; CHECK-LIBCALL-VFP-NEXT:    vmov s2, r0
+; CHECK-LIBCALL-VFP-NEXT:    vcmp.f32 s2, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-LIBCALL-VFP-NEXT:    movge r5, #15360
+; CHECK-LIBCALL-VFP-NEXT:    strh r5, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_minimum:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    ldrh r5, [r0]
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    mov r0, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, #1065353216
+; CHECK-NOVFP-NEXT:    bl __aeabi_fcmpge
+; CHECK-NOVFP-NEXT:    cmp r0, #0
+; CHECK-NOVFP-NEXT:    movne r5, #15360
+; CHECK-NOVFP-NEXT:    strh r5, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %c = fcmp ult half %a, 1.0
   %r = select i1 %c, half %a, half 1.0
@@ -688,16 +1756,45 @@ define void @test_minimum(ptr %p) #0 {
 }
 
 define void @test_maximum(ptr %p) #0 {
-; CHECK-ALL-LABEL: test_maximum:
-; CHECK-FP16: vmov.f32 s0, #1.000000e+00
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL-VFP: vmov.f32 s0, #1.000000e+00
-; CHECK-NOVFP: mov r{{[0-9]+}}, #1065353216
-; CHECK-VFP: vcmp.f32
-; CHECK-VFP: vmrs
-; CHECK-VFP: movls
-; CHECK-NOVFP: bl __aeabi_fcmple
+; CHECK-FP16-LABEL: test_maximum:
+; CHECK-FP16:         vmov.f32 s0, #1.000000e+00
+; CHECK-FP16-NEXT:    ldrh r1, [r0]
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vcmp.f32 s2, s0
+; CHECK-FP16-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-FP16-NEXT:    movls r1, #15360
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_maximum:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r5, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r5
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov.f32 s0, #1.000000e+00
+; CHECK-LIBCALL-VFP-NEXT:    vmov s2, r0
+; CHECK-LIBCALL-VFP-NEXT:    vcmp.f32 s2, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmrs APSR_nzcv, fpscr
+; CHECK-LIBCALL-VFP-NEXT:    movls r5, #15360
+; CHECK-LIBCALL-VFP-NEXT:    strh r5, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_maximum:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    ldrh r5, [r0]
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    mov r0, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, #1065353216
+; CHECK-NOVFP-NEXT:    bl __aeabi_fcmple
+; CHECK-NOVFP-NEXT:    cmp r0, #0
+; CHECK-NOVFP-NEXT:    movne r5, #15360
+; CHECK-NOVFP-NEXT:    strh r5, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
   %a = load half, ptr %p, align 2
   %c = fcmp ugt half %a, 1.0
   %r = select i1 %c, half %a, half 1.0
@@ -723,13 +1820,41 @@ define void @test_copysign(ptr %p, ptr %q) #0 {
 
 define void @test_floor(ptr %p) {
 ; CHECK-FP16-LABEL: test_floor:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl floorf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_floor:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl floorf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl floorf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_floor:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl floorf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_floor:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl floorf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.floor.f16(half %a)
   store half %r, ptr %p
@@ -738,13 +1863,41 @@ define void @test_floor(ptr %p) {
 
 define void @test_ceil(ptr %p) {
 ; CHECK-FP16-LABEL: test_ceil:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl ceilf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_ceil:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl ceilf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl ceilf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_ceil:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl ceilf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_ceil:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl ceilf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.ceil.f16(half %a)
   store half %r, ptr %p
@@ -753,13 +1906,41 @@ define void @test_ceil(ptr %p) {
 
 define void @test_trunc(ptr %p) {
 ; CHECK-FP16-LABEL: test_trunc:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl truncf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_trunc:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl truncf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl truncf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_trunc:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl truncf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_trunc:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl truncf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.trunc.f16(half %a)
   store half %r, ptr %p
@@ -768,13 +1949,41 @@ define void @test_trunc(ptr %p) {
 
 define void @test_rint(ptr %p) {
 ; CHECK-FP16-LABEL: test_rint:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl rintf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_rint:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl rintf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl rintf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_rint:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl rintf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_rint:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl rintf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.rint.f16(half %a)
   store half %r, ptr %p
@@ -783,13 +1992,41 @@ define void @test_rint(ptr %p) {
 
 define void @test_nearbyint(ptr %p) {
 ; CHECK-FP16-LABEL: test_nearbyint:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl nearbyintf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_nearbyint:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl nearbyintf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl nearbyintf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_nearbyint:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl nearbyintf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_nearbyint:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl nearbyintf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.nearbyint.f16(half %a)
   store half %r, ptr %p
@@ -798,13 +2035,41 @@ define void @test_nearbyint(ptr %p) {
 
 define void @test_round(ptr %p) {
 ; CHECK-FP16-LABEL: test_round:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl roundf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_round:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl roundf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl roundf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_round:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl roundf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_round:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl roundf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.round.f16(half %a)
   store half %r, ptr %p
@@ -813,13 +2078,41 @@ define void @test_round(ptr %p) {
 
 define void @test_roundeven(ptr %p) {
 ; CHECK-FP16-LABEL: test_roundeven:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl roundevenf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_roundeven:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl roundevenf
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         .save {r4, lr}
+; CHECK-FP16-NEXT:    push {r4, lr}
+; CHECK-FP16-NEXT:    mov r4, r0
+; CHECK-FP16-NEXT:    ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    bl roundevenf
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r4]
+; CHECK-FP16-NEXT:    pop {r4, pc}
+;
+; CHECK-LIBCALL-VFP-LABEL: test_roundeven:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    bl roundevenf
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_roundeven:
+; CHECK-NOVFP:         .save {r4, lr}
+; CHECK-NOVFP-NEXT:    push {r4, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    bl roundevenf
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, pc}
   %a = load half, ptr %p, align 2
   %r = call half @llvm.roundeven.f16(half %a)
   store half %r, ptr %p
@@ -828,25 +2121,78 @@ define void @test_roundeven(ptr %p) {
 
 define void @test_fmuladd(ptr %p, ptr %q, ptr %r) #0 {
 ; CHECK-FP16-LABEL: test_fmuladd:
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vmul.f32
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vadd.f32
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fmuladd:
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL-VFP: vmul.f32
-; CHECK-NOVFP: bl __aeabi_fmul
-; CHECK-LIBCALL: bl __aeabi_f2h
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL-VFP: vadd.f32
-; CHECK-NOVFP: bl __aeabi_fadd
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-FP16:         ldrh r3, [r0]
+; CHECK-FP16-NEXT:    ldrh r1, [r1]
+; CHECK-FP16-NEXT:    ldrh r2, [r2]
+; CHECK-FP16-NEXT:    vmov s0, r3
+; CHECK-FP16-NEXT:    vmov s2, r1
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vmul.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vmov s2, r2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vadd.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r1, s0
+; CHECK-FP16-NEXT:    strh r1, [r0]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_fmuladd:
+; CHECK-LIBCALL-VFP:         .save {r4, r5, r6, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r1]
+; CHECK-LIBCALL-VFP-NEXT:    mov r5, r2
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r1, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmul.f32 s0, s0, s16
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r6, r0
+; CHECK-LIBCALL-VFP-NEXT:    ldrh r0, [r5]
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s2, r6
+; CHECK-LIBCALL-VFP-NEXT:    vadd.f32 s0, s2, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, r5, r6, pc}
+;
+; CHECK-NOVFP-LABEL: test_fmuladd:
+; CHECK-NOVFP:         .save {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r6, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r1]
+; CHECK-NOVFP-NEXT:    mov r5, r2
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r6, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r4]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r6
+; CHECK-NOVFP-NEXT:    bl __aeabi_fmul
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    mov r6, r0
+; CHECK-NOVFP-NEXT:    ldrh r0, [r5]
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    uxth r0, r6
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_fadd
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    strh r0, [r4]
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r6, pc}
   %a = load half, ptr %p, align 2
   %b = load half, ptr %q, align 2
   %c = load half, ptr %r, align 2
@@ -859,26 +2205,53 @@ define void @test_fmuladd(ptr %p, ptr %q, ptr %r) #0 {
 ; to the register, but are stored in the stack instead.  Hence insertelement
 ; and extractelement have these extra loads and stores.
 define void @test_insertelement(ptr %p, ptr %q, i32 %i) #0 {
-; CHECK-ALL-LABEL: test_insertelement:
-; CHECK-ALL: sub sp, sp, #8
-
-; CHECK-ALL-DAG: and
-; CHECK-ALL-DAG: mov
-; CHECK-ALL-DAG: ldrd
-; CHECK-ALL-DAG: orr
-; CHECK-ALL-DAG: ldrh
-; CHECK-ALL-DAG: stm
-; CHECK-ALL: ldrh
-; CHECK-ALL-DAG: ldrh
-; CHECK-ALL-DAG: ldrh
-; CHECK-ALL-DAG: ldrh
-; CHECK-ALL-DAG: strh
-; CHECK-ALL-DAG: strh
-; CHECK-ALL-DAG: strh
-; CHECK-ALL-DAG: strh
-; CHECK-ALL: strh
-
-; CHECK-ALL: add sp, sp, #8
+; CHECK-VFP-LABEL: test_insertelement:
+; CHECK-VFP:         .save {r4, r5, r11, lr}
+; CHECK-VFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-VFP-NEXT:    .pad #8
+; CHECK-VFP-NEXT:    sub sp, sp, #8
+; CHECK-VFP-NEXT:    and r2, r2, #3
+; CHECK-VFP-NEXT:    mov r3, sp
+; CHECK-VFP-NEXT:    ldrd r4, r5, [r1]
+; CHECK-VFP-NEXT:    orr r2, r3, r2, lsl #1
+; CHECK-VFP-NEXT:    ldrh r0, [r0]
+; CHECK-VFP-NEXT:    stm sp, {r4, r5}
+; CHECK-VFP-NEXT:    strh r0, [r2]
+; CHECK-VFP-NEXT:    ldrh r0, [sp, #6]
+; CHECK-VFP-NEXT:    ldrh r2, [sp, #4]
+; CHECK-VFP-NEXT:    ldrh r3, [sp, #2]
+; CHECK-VFP-NEXT:    ldrh r5, [sp]
+; CHECK-VFP-NEXT:    strh r0, [r1, #6]
+; CHECK-VFP-NEXT:    strh r2, [r1, #4]
+; CHECK-VFP-NEXT:    strh r3, [r1, #2]
+; CHECK-VFP-NEXT:    strh r5, [r1]
+; CHECK-VFP-NEXT:    add sp, sp, #8
+; CHECK-VFP-NEXT:    pop {r4, r5, r11, pc}
+;
+; CHECK-NOVFP-LABEL: test_insertelement:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    .pad #8
+; CHECK-NOVFP-NEXT:    sub sp, sp, #8
+; CHECK-NOVFP-NEXT:    ldrd r4, r5, [r1]
+; CHECK-NOVFP-NEXT:    and r2, r2, #3
+; CHECK-NOVFP-NEXT:    mov r3, sp
+; CHECK-NOVFP-NEXT:    stm sp, {r4, r5}
+; CHECK-NOVFP-NEXT:    orr r2, r3, r2, lsl #1
+; CHECK-NOVFP-NEXT:    ldrh r0, [r0]
+; CHECK-NOVFP-NEXT:    strh r0, [r2]
+; CHECK-NOVFP-NEXT:    ldrh r0, [sp, #6]
+; CHECK-NOVFP-NEXT:    strh r0, [r1, #6]
+; CHECK-NOVFP-NEXT:    ldrh r0, [sp, #4]
+; CHECK-NOVFP-NEXT:    strh r0, [r1, #4]
+; CHECK-NOVFP-NEXT:    ldrh r0, [sp, #2]
+; CHECK-NOVFP-NEXT:    strh r0, [r1, #2]
+; CHECK-NOVFP-NEXT:    ldrh r0, [sp]
+; CHECK-NOVFP-NEXT:    strh r0, [r1]
+; CHECK-NOVFP-NEXT:    add sp, sp, #8
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
+
+
   %a = load half, ptr %p, align 2
   %b = load <4 x half>, ptr %q, align 8
   %c = insertelement <4 x half> %b, half %a, i32 %i
@@ -888,15 +2261,19 @@ define void @test_insertelement(ptr %p, ptr %q, i32 %i) #0 {
 
 define void @test_extractelement(ptr %p, ptr %q, i32 %i) #0 {
 ; CHECK-ALL-LABEL: test_extractelement:
-; CHECK-ALL: push {{{.*}}, lr}
-; CHECK-ALL: sub sp, sp, #8
-; CHECK-ALL: ldrd
-; CHECK-ALL: mov
-; CHECK-ALL: orr
-; CHECK-ALL: ldrh
-; CHECK-ALL: strh
-; CHECK-ALL: add sp, sp, #8
-; CHECK-ALL: pop {{{.*}}, pc}
+; CHECK-ALL:         .save {r4, r5, r11, lr}
+; CHECK-ALL-NEXT:    push {r4, r5, r11, lr}
+; CHECK-ALL-NEXT:    .pad #8
+; CHECK-ALL-NEXT:    sub sp, sp, #8
+; CHECK-ALL-NEXT:    ldrd r4, r5, [r1]
+; CHECK-ALL-NEXT:    and r1, r2, #3
+; CHECK-ALL-NEXT:    mov r2, sp
+; CHECK-ALL-NEXT:    orr r1, r2, r1, lsl #1
+; CHECK-ALL-NEXT:    stm sp, {r4, r5}
+; CHECK-ALL-NEXT:    ldrh r1, [r1]
+; CHECK-ALL-NEXT:    strh r1, [r0]
+; CHECK-ALL-NEXT:    add sp, sp, #8
+; CHECK-ALL-NEXT:    pop {r4, r5, r11, pc}
   %a = load <4 x half>, ptr %q, align 8
   %b = extractelement <4 x half> %a, i32 %i
   store half %b, ptr %p
@@ -909,10 +2286,9 @@ define void @test_extractelement(ptr %p, ptr %q, i32 %i) #0 {
 
 define void @test_insertvalue(ptr %p, ptr %q) {
 ; CHECK-ALL-LABEL: test_insertvalue:
-; CHECK-ALL-DAG: ldr
-; CHECK-ALL-DAG: ldrh
-; CHECK-ALL-DAG: strh
-; CHECK-ALL-DAG: str
+; CHECK-ALL:         ldrh r1, [r1]
+; CHECK-ALL-NEXT:    strh r1, [r0, #4]
+; CHECK-ALL-NEXT:    bx lr
   %a = load %struct.dummy, ptr %p
   %b = load half, ptr %q
   %c = insertvalue %struct.dummy %a, half %b, 1
@@ -922,9 +2298,9 @@ define void @test_insertvalue(ptr %p, ptr %q) {
 
 define void @test_extractvalue(ptr %p, ptr %q) {
 ; CHECK-ALL-LABEL: test_extractvalue:
-; CHECK-ALL: .fnstart
-; CHECK-ALL: ldrh
-; CHECK-ALL: strh
+; CHECK-ALL:         ldrh r0, [r0, #4]
+; CHECK-ALL-NEXT:    strh r0, [r1]
+; CHECK-ALL-NEXT:    bx lr
   %a = load %struct.dummy, ptr %p
   %b = extractvalue %struct.dummy %a, 1
   store half %b, ptr %q
@@ -932,61 +2308,176 @@ define void @test_extractvalue(ptr %p, ptr %q) {
 }
 
 define %struct.dummy @test_struct_return(ptr %p) {
-; CHECK-ALL-LABEL: test_struct_return:
+; CHECK-VFP-LABEL: test_struct_return:
+; CHECK-VFP:         ldrh r1, [r0, #4]
+; CHECK-VFP-NEXT:    ldr r0, [r0]
+; CHECK-VFP-NEXT:    vmov s0, r1
+; CHECK-VFP-NEXT:    bx lr
+;
+; CHECK-NOVFP-LABEL: test_struct_return:
+; CHECK-NOVFP:         ldr r2, [r0]
+; CHECK-NOVFP-NEXT:    ldrh r1, [r0, #4]
+; CHECK-NOVFP-NEXT:    mov r0, r2
+; CHECK-NOVFP-NEXT:    bx lr
 ; CHECK-VFP-LIBCALL: bl __aeabi_h2f
-; CHECK-NOVFP-DAG: ldr
-; CHECK-NOVFP-DAG: ldrh
   %a = load %struct.dummy, ptr %p
   ret %struct.dummy %a
 }
 
 define half @test_struct_arg(%struct.dummy %p) {
-; CHECK-ALL-LABEL: test_struct_arg:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-NOVFP-NEXT: mov r0, r1
-; CHECK-ALL-NEXT: bx lr
+; CHECK-VFP-LABEL: test_struct_arg:
+; CHECK-VFP:         bx lr
+;
+; CHECK-NOVFP-LABEL: test_struct_arg:
+; CHECK-NOVFP:         mov r0, r1
+; CHECK-NOVFP-NEXT:    bx lr
   %a = extractvalue %struct.dummy %p, 1
   ret half %a
 }
 
 define half @test_uitofp_i32_fadd(i32 %a, half %b) #0 {
-; CHECK-LABEL: test_uitofp_i32_fadd:
-; CHECK-VFP-DAG: vcvt.f32.u32
-; CHECK-NOVFP-DAG: bl __aeabi_ui2f
+; CHECK-FP16-LABEL: test_uitofp_i32_fadd:
+; CHECK-FP16:         vmov s2, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvt.f32.u32 s2, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s2, s2
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vadd.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_uitofp_i32_fadd:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vmov r1, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vcvt.f32.u32 s16, s0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s16
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r4
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vadd.f32 s0, s16, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_uitofp_i32_fadd:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r1
+; CHECK-NOVFP-NEXT:    bl __aeabi_ui2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    uxth r0, r4
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    uxth r0, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r0
+; CHECK-NOVFP-NEXT:    mov r0, r4
+; CHECK-NOVFP-NEXT:    bl __aeabi_fadd
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
 
-; CHECK-FP16-DAG: vcvtb.f16.f32
-; CHECK-FP16-DAG: vcvtb.f32.f16
-; CHECK-LIBCALL-DAG: bl __aeabi_h2f
-; CHECK-LIBCALL-DAG: bl __aeabi_h2f
 
-; CHECK-VFP-DAG: vadd.f32
-; CHECK-NOVFP-DAG: bl __aeabi_fadd
 
-; CHECK-FP16-DAG: vcvtb.f16.f32
-; CHECK-LIBCALL-DAG: bl __aeabi_f2h
   %c = uitofp i32 %a to half
   %r = fadd half %b, %c
   ret half %r
 }
 
 define half @test_sitofp_i32_fadd(i32 %a, half %b) #0 {
-; CHECK-LABEL: test_sitofp_i32_fadd:
-; CHECK-VFP-DAG: vcvt.f32.s32
-; CHECK-NOVFP-DAG: bl __aeabi_i2f
+; CHECK-FP16-LABEL: test_sitofp_i32_fadd:
+; CHECK-FP16:         vmov s2, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vcvt.f32.s32 s2, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s2, s2
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s2, s2
+; CHECK-FP16-NEXT:    vadd.f32 s0, s0, s2
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-VFP-LABEL: test_sitofp_i32_fadd:
+; CHECK-LIBCALL-VFP:         .save {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-VFP-NEXT:    .vsave {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vpush {d8}
+; CHECK-LIBCALL-VFP-NEXT:    vmov r1, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vcvt.f32.s32 s16, s0
+; CHECK-LIBCALL-VFP-NEXT:    mov r0, r1
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    mov r4, r0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s16
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    vmov s16, r4
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vadd.f32 s0, s16, s0
+; CHECK-LIBCALL-VFP-NEXT:    vmov r0, s0
+; CHECK-LIBCALL-VFP-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-VFP-NEXT:    vmov s0, r0
+; CHECK-LIBCALL-VFP-NEXT:    vpop {d8}
+; CHECK-LIBCALL-VFP-NEXT:    pop {r4, pc}
+;
+; CHECK-NOVFP-LABEL: test_sitofp_i32_fadd:
+; CHECK-NOVFP:         .save {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    push {r4, r5, r11, lr}
+; CHECK-NOVFP-NEXT:    mov r4, r1
+; CHECK-NOVFP-NEXT:    bl __aeabi_i2f
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    mov r5, r0
+; CHECK-NOVFP-NEXT:    uxth r0, r4
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r4, r0
+; CHECK-NOVFP-NEXT:    uxth r0, r5
+; CHECK-NOVFP-NEXT:    bl __aeabi_h2f
+; CHECK-NOVFP-NEXT:    mov r1, r0
+; CHECK-NOVFP-NEXT:    mov r0, r4
+; CHECK-NOVFP-NEXT:    bl __aeabi_fadd
+; CHECK-NOVFP-NEXT:    bl __aeabi_f2h
+; CHECK-NOVFP-NEXT:    pop {r4, r5, r11, pc}
 
-; CHECK-FP16-DAG: vcvtb.f16.f32
-; CHECK-FP16-DAG: vcvtb.f32.f16
-; CHECK-LIBCALL-DAG: bl __aeabi_h2f
-; CHECK-LIBCALL-DAG: bl __aeabi_h2f
 
-; CHECK-VFP-DAG: vadd.f32
-; CHECK-NOVFP-DAG: bl __aeabi_fadd
 
-; CHECK-FP16-DAG: vcvtb.f16.f32
-; CHECK-LIBCALL-DAG: bl __aeabi_f2h
   %c = sitofp i32 %a to half
   %r = fadd half %b, %c
   ret half %r
 }
 
+define void @test_fneg(ptr %p1, ptr %p2) #0 {
+; CHECK-FP16-LABEL: test_fneg:
+; CHECK-FP16:         ldrh r0, [r0]
+; CHECK-FP16-NEXT:    vmov s0, r0
+; CHECK-FP16-NEXT:    vcvtb.f32.f16 s0, s0
+; CHECK-FP16-NEXT:    vneg.f32 s0, s0
+; CHECK-FP16-NEXT:    vcvtb.f16.f32 s0, s0
+; CHECK-FP16-NEXT:    vmov r0, s0
+; CHECK-FP16-NEXT:    strh r0, [r1]
+; CHECK-FP16-NEXT:    bx lr
+;
+; CHECK-LIBCALL-LABEL: test_fneg:
+; CHECK-LIBCALL:         .save {r4, lr}
+; CHECK-LIBCALL-NEXT:    push {r4, lr}
+; CHECK-LIBCALL-NEXT:    ldrh r0, [r0]
+; CHECK-LIBCALL-NEXT:    mov r4, r1
+; CHECK-LIBCALL-NEXT:    bl __aeabi_h2f
+; CHECK-LIBCALL-NEXT:    eor r0, r0, #-2147483648
+; CHECK-LIBCALL-NEXT:    bl __aeabi_f2h
+; CHECK-LIBCALL-NEXT:    strh r0, [r4]
+; CHECK-LIBCALL-NEXT:    pop {r4, pc}
+  %v = load half, ptr %p1, align 2
+  %res = fneg half %v
+  store half %res, ptr %p2, align 2
+  ret void
+}
+
 attributes #0 = { nounwind }
diff --git a/llvm/test/CodeGen/ARM/fpcmp-opt.ll b/llvm/test/CodeGen/ARM/fpcmp-opt.ll
index 447e470b2363..a40fd4244af1 100644
--- a/llvm/test/CodeGen/ARM/fpcmp-opt.ll
+++ b/llvm/test/CodeGen/ARM/fpcmp-opt.ll
@@ -1,4 +1,4 @@
-; RUN: llc -mtriple=arm-eabi -mcpu=cortex-a8 -mattr=+vfp2 -enable-unsafe-fp-math %s -o - \
+; RUN: llc -mtriple=arm-eabi -mcpu=cortex-a8 -mattr=+vfp2 %s -o - \
 ; RUN:  | FileCheck %s
 
 ; rdar://7461510
@@ -42,7 +42,7 @@ entry:
 ; CHECK-NOT: vmrs
 ; CHECK: bne
   %0 = load double, ptr %a
-  %1 = fcmp oeq double %0, 0.000000e+00
+  %1 = fcmp nnan oeq double %0, 0.000000e+00
   br i1 %1, label %bb1, label %bb2
 
 bb1:
@@ -65,7 +65,7 @@ entry:
 ; CHECK-NOT: vmrs
 ; CHECK: bne
   %0 = load float, ptr %a
-  %1 = fcmp oeq float %0, 0.000000e+00
+  %1 = fcmp nnan oeq float %0, 0.000000e+00
   br i1 %1, label %bb1, label %bb2
 
 bb1:
diff --git a/llvm/test/CodeGen/ARM/frame-chain.ll b/llvm/test/CodeGen/ARM/frame-chain.ll
index 7b722cd5fcef..51de31aa393f 100644
--- a/llvm/test/CodeGen/ARM/frame-chain.ll
+++ b/llvm/test/CodeGen/ARM/frame-chain.ll
@@ -223,3 +223,33 @@ define dso_local void @required_fp(i32 %0, i32 %1) {
   store i64 %8, ptr %6, align 8
   ret void
 }
+
+define dso_local noundef i32 @leaf_tailcall(i32 noundef %0) {
+; LEAF-FP-LABEL: leaf_tailcall:
+; LEAF-FP:       @ %bb.0:
+; LEAF-FP-NEXT:    .save {r11, lr}
+; LEAF-FP-NEXT:    push {r11, lr}
+; LEAF-FP-NEXT:    .setfp r11, sp
+; LEAF-FP-NEXT:    mov r11, sp
+; LEAF-FP-NEXT:    pop {r11, lr}
+; LEAF-FP-NEXT:    b leaf
+;
+; LEAF-FP-AAPCS-LABEL: leaf_tailcall:
+; LEAF-FP-AAPCS:       @ %bb.0:
+; LEAF-FP-AAPCS-NEXT:    .save {r11, lr}
+; LEAF-FP-AAPCS-NEXT:    push {r11, lr}
+; LEAF-FP-AAPCS-NEXT:    .setfp r11, sp
+; LEAF-FP-AAPCS-NEXT:    mov r11, sp
+; LEAF-FP-AAPCS-NEXT:    pop {r11, lr}
+; LEAF-FP-AAPCS-NEXT:    b leaf
+;
+; LEAF-NOFP-LABEL: leaf_tailcall:
+; LEAF-NOFP:       @ %bb.0:
+; LEAF-NOFP-NEXT:    b leaf
+;
+; LEAF-NOFP-AAPCS-LABEL: leaf_tailcall:
+; LEAF-NOFP-AAPCS:       @ %bb.0:
+; LEAF-NOFP-AAPCS-NEXT:    b leaf
+  %a = tail call noundef i32 @leaf(i32 noundef %0)
+  ret i32 %a
+}
diff --git a/llvm/test/CodeGen/ARM/stack-guard-nomovt.ll b/llvm/test/CodeGen/ARM/stack-guard-nomovt.ll
index 6802dabfda87..2003fd747f43 100644
--- a/llvm/test/CodeGen/ARM/stack-guard-nomovt.ll
+++ b/llvm/test/CodeGen/ARM/stack-guard-nomovt.ll
@@ -1,27 +1,63 @@
-; RUN: llc -relocation-model=static -mattr=+no-movt < %s | FileCheck %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -relocation-model=static -mattr=+no-movt < %s | FileCheck %s --check-prefix=STATIC
+; RUN: llc -relocation-model=pic -mattr=+no-movt < %s | FileCheck %s --check-prefix=PIC
 
 target triple = "thumbv7a-linux-gnueabi"
 
 define i32 @test1() #0 {
-; CHECK-LABEL: test1:
-; CHECK:       @ %bb.0:
-; CHECK-NEXT:    push	{r7, lr}
-; CHECK-NEXT:    sub.w	sp, sp, #1032
-; CHECK-NEXT:    ldr	r0, .LCPI0_0
-; CHECK-NEXT:    ldr	r0, [r0]
-; CHECK-NEXT:    str.w	r0, [sp, #1028]
-; CHECK-NEXT:    add	r0, sp, #4
-; CHECK-NEXT:    bl	foo
-; CHECK-NEXT:    ldr.w	r0, [sp, #1028]
-; CHECK-NEXT:    ldr	r1, .LCPI0_0
-; CHECK-NEXT:    ldr	r1, [r1]
-; CHECK-NEXT:    cmp	r1, r0
-; CHECK-NEXT:    ittt	eq
-; CHECK-NEXT:    moveq	r0, #0
-; CHECK-NEXT:    addeq.w	sp, sp, #1032
-; CHECK-NEXT:    popeq	{r7, pc}
-; CHECK-NEXT:  .LBB0_1:
-; CHECK-NEXT:    bl __stack_chk_fail
+; STATIC-LABEL: test1:
+; STATIC:       @ %bb.0:
+; STATIC-NEXT:    push {r7, lr}
+; STATIC-NEXT:    sub.w sp, sp, #1032
+; STATIC-NEXT:    ldr r0, .LCPI0_0
+; STATIC-NEXT:    ldr r0, [r0]
+; STATIC-NEXT:    str.w r0, [sp, #1028]
+; STATIC-NEXT:    add r0, sp, #4
+; STATIC-NEXT:    bl foo
+; STATIC-NEXT:    ldr.w r0, [sp, #1028]
+; STATIC-NEXT:    ldr r1, .LCPI0_0
+; STATIC-NEXT:    ldr r1, [r1]
+; STATIC-NEXT:    cmp r1, r0
+; STATIC-NEXT:    ittt eq
+; STATIC-NEXT:    moveq r0, #0
+; STATIC-NEXT:    addeq.w sp, sp, #1032
+; STATIC-NEXT:    popeq {r7, pc}
+; STATIC-NEXT:  .LBB0_1:
+; STATIC-NEXT:    bl __stack_chk_fail
+; STATIC-NEXT:    .p2align 2
+; STATIC-NEXT:  @ %bb.2:
+; STATIC-NEXT:  .LCPI0_0:
+; STATIC-NEXT:    .long __stack_chk_guard
+;
+; PIC-LABEL: test1:
+; PIC:       @ %bb.0:
+; PIC-NEXT:    push {r7, lr}
+; PIC-NEXT:    sub.w sp, sp, #1032
+; PIC-NEXT:    ldr r0, .LCPI0_0
+; PIC-NEXT:  .LPC0_0:
+; PIC-NEXT:    add r0, pc
+; PIC-NEXT:    ldr r0, [r0]
+; PIC-NEXT:    str.w r0, [sp, #1028]
+; PIC-NEXT:    add r0, sp, #4
+; PIC-NEXT:    bl foo
+; PIC-NEXT:    ldr.w r0, [sp, #1028]
+; PIC-NEXT:    ldr r1, .LCPI0_1
+; PIC-NEXT:  .LPC0_1:
+; PIC-NEXT:    add r1, pc
+; PIC-NEXT:    ldr r1, [r1]
+; PIC-NEXT:    cmp r1, r0
+; PIC-NEXT:    ittt eq
+; PIC-NEXT:    moveq r0, #0
+; PIC-NEXT:    addeq.w sp, sp, #1032
+; PIC-NEXT:    popeq {r7, pc}
+; PIC-NEXT:  .LBB0_1:
+; PIC-NEXT:    bl __stack_chk_fail
+; PIC-NEXT:    .p2align 2
+; PIC-NEXT:  @ %bb.2:
+; PIC-NEXT:  .LCPI0_0:
+; PIC-NEXT:    .long __stack_chk_guard-(.LPC0_0+4)
+; PIC-NEXT:  .LCPI0_1:
+; PIC-NEXT:    .long __stack_chk_guard-(.LPC0_1+4)
   %a1 = alloca [256 x i32], align 4
   call void @foo(ptr %a1) #3
   ret i32 0
author	Mingming Liu <mingmingl@google.com>	2025-09-10 15:25:31 -0700
committer	GitHub <noreply@github.com>	2025-09-10 15:25:31 -0700
commit	1417dafa1db9cb1b2b09438aa9f53ea5ab6e36e2 (patch)
tree	57f4b1f313c8cf74eed8819870f39c36ea263c68 /llvm/test/CodeGen/ARM
parent	898b813bc8a6d0276bf0f4769f5f2f64b34e632d (diff)
parent	b8cefcb601ddaa18482555c4ff363c01a270c2fe (diff)