diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp --- a/llvm/lib/Target/ARM/ARMISelLowering.cpp +++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp @@ -1156,6 +1156,8 @@ setOperationAction(ISD::FLOG, MVT::f16, Promote); setOperationAction(ISD::FLOG10, MVT::f16, Promote); setOperationAction(ISD::FLOG2, MVT::f16, Promote); + + setOperationAction(ISD::FROUND, MVT::f16, Legal); } if (Subtarget->hasNEON()) { diff --git a/llvm/lib/Target/ARM/ARMInstrVFP.td b/llvm/lib/Target/ARM/ARMInstrVFP.td --- a/llvm/lib/Target/ARM/ARMInstrVFP.td +++ b/llvm/lib/Target/ARM/ARMInstrVFP.td @@ -930,9 +930,9 @@ multiclass vrint_inst_zrx { def H : AHuI<0b11101, 0b11, 0b0110, 0b11, 0, - (outs SPR:$Sd), (ins SPR:$Sm), + (outs HPR:$Sd), (ins HPR:$Sm), NoItinerary, !strconcat("vrint", opc), ".f16\t$Sd, $Sm", - []>, + [(set (f16 HPR:$Sd), (node (f16 HPR:$Sm)))]>, Requires<[HasFullFP16]> { let Inst{7} = op2; let Inst{16} = op; @@ -975,9 +975,9 @@ let PostEncoderMethod = "", DecoderNamespace = "VFPV8", isUnpredicable = 1 in { def H : AHuInp<0b11101, 0b11, 0b1000, 0b01, 0, - (outs SPR:$Sd), (ins SPR:$Sm), + (outs HPR:$Sd), (ins HPR:$Sm), NoItinerary, !strconcat("vrint", opc, ".f16\t$Sd, $Sm"), - []>, + [(set (f16 HPR:$Sd), (node (f16 HPR:$Sm)))]>, Requires<[HasFullFP16]> { let Inst{17-16} = rm; } diff --git a/llvm/test/CodeGen/ARM/fp16-fullfp16.ll b/llvm/test/CodeGen/ARM/fp16-fullfp16.ll --- a/llvm/test/CodeGen/ARM/fp16-fullfp16.ll +++ b/llvm/test/CodeGen/ARM/fp16-fullfp16.ll @@ -488,53 +488,77 @@ ret void } -; FIXME -;define void @test_floor(half* %p) { -; %a = load half, half* %p, align 2 -; %r = call half @llvm.floor.f16(half %a) -; store half %r, half* %p -; ret void -;} +define void @test_floor(half* %p) { +; CHECK-LABEL: test_floor: +; CHECK: vldr.16 s0, [r0] +; CHECK-NEXT: vrintm.f16 s0, s0 +; CHECK-NEXT: vstr.16 s0, [r0] +; CHECK-NEXT: bx lr + %a = load half, half* %p, align 2 + %r = call half @llvm.floor.f16(half %a) + store half %r, half* %p + ret void +} -; FIXME -;define void @test_ceil(half* %p) { -; %a = load half, half* %p, align 2 -; %r = call half @llvm.ceil.f16(half %a) -; store half %r, half* %p -; ret void -;} +define void @test_ceil(half* %p) { +; CHECK-LABEL: test_ceil: +; CHECK: vldr.16 s0, [r0] +; CHECK-NEXT: vrintp.f16 s0, s0 +; CHECK-NEXT: vstr.16 s0, [r0] +; CHECK-NEXT: bx lr + %a = load half, half* %p, align 2 + %r = call half @llvm.ceil.f16(half %a) + store half %r, half* %p + ret void +} -; FIXME -;define void @test_trunc(half* %p) { -; %a = load half, half* %p, align 2 -; %r = call half @llvm.trunc.f16(half %a) -; store half %r, half* %p -; ret void -;} +define void @test_trunc(half* %p) { +; CHECK-LABEL: test_trunc: +; CHECK: vldr.16 s0, [r0] +; CHECK-NEXT: vrintz.f16 s0, s0 +; CHECK-NEXT: vstr.16 s0, [r0] +; CHECK-NEXT: bx lr + %a = load half, half* %p, align 2 + %r = call half @llvm.trunc.f16(half %a) + store half %r, half* %p + ret void +} -; FIXME -;define void @test_rint(half* %p) { -; %a = load half, half* %p, align 2 -; %r = call half @llvm.rint.f16(half %a) -; store half %r, half* %p -; ret void -;} +define void @test_rint(half* %p) { +; CHECK-LABEL: test_rint: +; CHECK: vldr.16 s0, [r0] +; CHECK-NEXT: vrintx.f16 s0, s0 +; CHECK-NEXT: vstr.16 s0, [r0] +; CHECK-NEXT: bx lr + %a = load half, half* %p, align 2 + %r = call half @llvm.rint.f16(half %a) + store half %r, half* %p + ret void +} -; FIXME -;define void @test_nearbyint(half* %p) { -; %a = load half, half* %p, align 2 -; %r = call half @llvm.nearbyint.f16(half %a) -; store half %r, half* %p -; ret void -;} +define void @test_nearbyint(half* %p) { +; CHECK-LABEL: test_nearbyint: +; CHECK: vldr.16 s0, [r0] +; CHECK-NEXT: vrintr.f16 s0, s0 +; CHECK-NEXT: vstr.16 s0, [r0] +; CHECK-NEXT: bx lr + %a = load half, half* %p, align 2 + %r = call half @llvm.nearbyint.f16(half %a) + store half %r, half* %p + ret void +} -; FIXME -;define void @test_round(half* %p) { -; %a = load half, half* %p, align 2 -; %r = call half @llvm.round.f16(half %a) -; store half %r, half* %p -; ret void -;} +define void @test_round(half* %p) { +; CHECK-LABEL: test_round: +; CHECK: vldr.16 s0, [r0] +; CHECK-NEXT: vrinta.f16 s0, s0 +; CHECK-NEXT: vstr.16 s0, [r0] +; CHECK-NEXT: bx lr + %a = load half, half* %p, align 2 + %r = call half @llvm.round.f16(half %a) + store half %r, half* %p + ret void +} define void @test_fmuladd(half* %p, half* %q, half* %r) { ; CHECK-LABEL: test_fmuladd: diff --git a/llvm/test/CodeGen/ARM/fp16-instructions.ll b/llvm/test/CodeGen/ARM/fp16-instructions.ll --- a/llvm/test/CodeGen/ARM/fp16-instructions.ll +++ b/llvm/test/CodeGen/ARM/fp16-instructions.ll @@ -43,8 +43,6 @@ ; CHECK-HARDFP-FULLFP16: {{.*}} lr } -; 1. VABS: TODO - ; 2. VADD define float @Add(float %a.coerce, float %b.coerce) { entry: @@ -691,15 +689,6 @@ ; CHECK-HARDFP-FULLFP16: vnmul.f16 s0, s0, s1 } -; TODO: -; 28. VRINTA -; 29. VRINTM -; 30. VRINTN -; 31. VRINTP -; 32. VRINTR -; 33. VRINTX -; 34. VRINTZ - ; 35. VSELEQ define half @select_cc1(half* %a0) { %1 = load half, half* %a0 @@ -955,8 +944,6 @@ ; CHECK-SOFTFP-FP16-T32-NEXT: vcvtb.f16.f32 s0, [[S4]] } -; 39. VSQRT - TODO - ; 40. VSUB define float @Sub(float %a.coerce, float %b.coerce) { entry: