From 0532e48bcd896a98e94a52c8d272394b998f67f6 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Sun, 26 May 2019 11:34:30 +0000
Subject: [PATCH] [ARM] Select fp16 fma

This adds a pattern for fma, similar to the float and double patterns.

Differential Revision: https://reviews.llvm.org/D62330

llvm-svn: 361719
---
 lib/Target/ARM/ARMInstrVFP.td     |  3 +++
 test/CodeGen/ARM/fp16-fullfp16.ll | 24 +++++++++++++++---------
 2 files changed, 18 insertions(+), 9 deletions(-)

diff --git a/lib/Target/ARM/ARMInstrVFP.td b/lib/Target/ARM/ARMInstrVFP.td
index e3d2a947788..d1b32f531c0 100644
--- a/lib/Target/ARM/ARMInstrVFP.td
+++ b/lib/Target/ARM/ARMInstrVFP.td
@@ -2073,6 +2073,9 @@ def : Pat<(f64 (fma DPR:$Dn, DPR:$Dm, DPR:$Ddin)),
 def : Pat<(f32 (fma SPR:$Sn, SPR:$Sm, SPR:$Sdin)),
           (VFMAS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
+def : Pat<(f16 (fma HPR:$Sn, HPR:$Sm, HPR:$Sdin)),
+          (VFMAH HPR:$Sdin, HPR:$Sn, HPR:$Sm)>,
+      Requires<[HasFullFP16]>;
 
 def VFMSD : ADbI<0b11101, 0b10, 1, 0,
                  (outs DPR:$Dd), (ins DPR:$Ddin, DPR:$Dn, DPR:$Dm),
diff --git a/test/CodeGen/ARM/fp16-fullfp16.ll b/test/CodeGen/ARM/fp16-fullfp16.ll
index 19afba05db6..a30b62acbac 100644
--- a/test/CodeGen/ARM/fp16-fullfp16.ll
+++ b/test/CodeGen/ARM/fp16-fullfp16.ll
@@ -386,15 +386,21 @@ define void @test_log2(half* %p) {
   ret void
 }
 
-; FIXME
-;define void @test_fma(half* %p, half* %q, half* %r) {
-;  %a = load half, half* %p, align 2
-;  %b = load half, half* %q, align 2
-;  %c = load half, half* %r, align 2
-;  %v = call half @llvm.fma.f16(half %a, half %b, half %c)
-;  store half %v, half* %p
-;  ret void
-;}
+define void @test_fma(half* %p, half* %q, half* %r) {
+; CHECK-LABEL: test_fma:
+; CHECK:         vldr.16 s0, [r1]
+; CHECK-NEXT:    vldr.16 s2, [r0]
+; CHECK-NEXT:    vldr.16 s4, [r2]
+; CHECK-NEXT:    vfma.f16 s4, s2, s0
+; CHECK-NEXT:    vstr.16 s4, [r0]
+; CHECK-NEXT:    bx lr
+  %a = load half, half* %p, align 2
+  %b = load half, half* %q, align 2
+  %c = load half, half* %r, align 2
+  %v = call half @llvm.fma.f16(half %a, half %b, half %c)
+  store half %v, half* %p
+  ret void
+}
 
 define void @test_fabs(half* %p) {
 ; CHECK-LABEL: test_fabs: