[llvm] r361719 - [ARM] Select fp16 fma

[David Green via llvm-commits]

Author: dmgreen
Date: Sun May 26 04:34:30 2019
New Revision: 361719

URL: http://llvm.org/viewvc/llvm-project?rev=361719&view=rev
Log:
[ARM] Select fp16 fma

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

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

Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrVFP.td
    llvm/trunk/test/CodeGen/ARM/fp16-fullfp16.ll

Modified: llvm/trunk/lib/Target/ARM/ARMInstrVFP.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrVFP.td?rev=361719&r1=361718&r2=361719&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrVFP.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrVFP.td Sun May 26 04:34:30 2019
@@ -2073,6 +2073,9 @@ def : Pat<(f64 (fma DPR:$Dn, DPR:$Dm, DP
 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),

Modified: llvm/trunk/test/CodeGen/ARM/fp16-fullfp16.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/fp16-fullfp16.ll?rev=361719&r1=361718&r2=361719&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/ARM/fp16-fullfp16.ll (original)
+++ llvm/trunk/test/CodeGen/ARM/fp16-fullfp16.ll Sun May 26 04:34:30 2019
@@ -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: