[ARM] FP16: codegen support for VACGT

Differential Revision: https://reviews.llvm.org/D50236 llvm-svn: 339148
llvm · Aug 7, 2018 · b39cd88 · b39cd88
1 parent 1bfadb0
commit b39cd88
Show file tree

Hide file tree

Showing 2 changed files with 18 additions and 28 deletions.
diff --git a/llvm/lib/Target/ARM/ARMInstrNEON.td b/llvm/lib/Target/ARM/ARMInstrNEON.td
@@ -5072,7 +5072,7 @@ def  VACGThd   : N3VDInt<1, 0, 0b11, 0b1110, 1, N3RegFrm, IIC_VBIND, "vacgt",
                         "f16", v4i16, v4f16, int_arm_neon_vacgt, 0>,
                  Requires<[HasNEON, HasFullFP16]>;
 def  VACGThq   : N3VQInt<1, 0, 0b11, 0b1110, 1, N3RegFrm, IIC_VBINQ, "vacgt",
-                        "f16", v8f16, v8f16, int_arm_neon_vacgt, 0>,
+                        "f16", v8i16, v8f16, int_arm_neon_vacgt, 0>,
                  Requires<[HasNEON, HasFullFP16]>;
 //   VTST     : Vector Test Bits
 defm VTST     : N3V_QHS<0, 0, 0b1000, 1, IIC_VBINi4D, IIC_VBINi4D, IIC_VBINi4Q,

diff --git a/llvm/test/CodeGen/ARM/armv8.2a-fp16-vector-intrinsics.ll b/llvm/test/CodeGen/ARM/armv8.2a-fp16-vector-intrinsics.ll
@@ -522,19 +522,23 @@ entry:
   ret <8 x i16> %vcageq_v2.i
 }
 
-; FIXME (PR38404)
-;
-;define dso_local <4 x i16> @test_vcagt_f16(<4 x half> %a, <4 x half> %b) {
-;entry:
-;  %vcagt_v2.i = tail call <4 x i16> @llvm.arm.neon.vacgt.v4i16.v4f16(<4 x half> %a, <4 x half> %b)
-;  ret <4 x i16> %vcagt_v2.i
-;}
-;
-;define dso_local <8 x i16> @test_vcagtq_f16(<8 x half> %a, <8 x half> %b) {
-;entry:
-;  %vcagtq_v2.i = tail call <8 x i16> @llvm.arm.neon.vacgt.v8i16.v8f16(<8 x half> %a, <8 x half> %b)
-;  ret <8 x i16> %vcagtq_v2.i
-;}
+define dso_local <4 x i16> @test_vcagt_f16(<4 x half> %a, <4 x half> %b) {
+; CHECK-LABEL: test_vcagt_f16:
+; CHECK:         vacgt.f16 d0, d0, d1
+; CHECK-NEXT:    bx lr
+entry:
+  %vcagt_v2.i = tail call <4 x i16> @llvm.arm.neon.vacgt.v4i16.v4f16(<4 x half> %a, <4 x half> %b)
+  ret <4 x i16> %vcagt_v2.i
+}
+
+define dso_local <8 x i16> @test_vcagtq_f16(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: test_vcagtq_f16:
+; CHECK:         vacgt.f16 q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %vcagtq_v2.i = tail call <8 x i16> @llvm.arm.neon.vacgt.v8i16.v8f16(<8 x half> %a, <8 x half> %b)
+  ret <8 x i16> %vcagtq_v2.i
+}
 
 define dso_local <4 x i16> @test_vcale_f16(<4 x half> %a, <4 x half> %b) {
 ; CHECKLABEL: test_vcale_f16:
@@ -554,20 +558,6 @@ entry:
   ret <8 x i16> %vcaleq_v2.i
 }
 
-; FIXME (PR38404)
-;
-;define dso_local <4 x i16> @test_vcalt_f16(<4 x half> %a, <4 x half> %b) {
-;entry:
-;  %vcalt_v2.i = tail call <4 x i16> @llvm.arm.neon.vacgt.v4i16.v4f16(<4 x half> %b, <4 x half> %a)
-;  ret <4 x i16> %vcalt_v2.i
-;}
-
-;define dso_local <8 x i16> @test_vcaltq_f16(<8 x half> %a, <8 x half> %b) {
-;entry:
-;  %vcaltq_v2.i = tail call <8 x i16> @llvm.arm.neon.vacgt.v8i16.v8f16(<8 x half> %b, <8 x half> %a)
-;  ret <8 x i16> %vcaltq_v2.i
-;}
-
 define dso_local <4 x i16> @test_vceq_f16(<4 x half> %a, <4 x half> %b) {
 ; CHECKLABEL: test_vceq_f16:
 ; CHECK:         vceq.f16 d0, d0, d1