[llvm] 1fae86d - [ARM] Improve fp16-promote.ll test (NFC) (#156341)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Sep 1 09:10:15 PDT 2025
Author: beetrees
Date: 2025-09-01T16:10:10Z
New Revision: 1fae86d5d361328b0598dd51dfb32f620c7d4158
URL: https://github.com/llvm/llvm-project/commit/1fae86d5d361328b0598dd51dfb32f620c7d4158
DIFF: https://github.com/llvm/llvm-project/commit/1fae86d5d361328b0598dd51dfb32f620c7d4158.diff
LOG: [ARM] Improve fp16-promote.ll test (NFC) (#156341)
Update the test to use `utils/update_llc_test_checks.py`, and add a
check for `fneg`. Prerequisite to #156343.
Added:
Modified:
llvm/test/CodeGen/ARM/fp16-promote.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/ARM/fp16-promote.ll b/llvm/test/CodeGen/ARM/fp16-promote.ll
index 1bd01508808c8..800ee87b95ca8 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 }
More information about the llvm-commits
mailing list