/external/llvm-project/llvm/test/CodeGen/AArch64/ |
D | aarch64-bf16-ldst-intrinsics.ll | 88 …%vld1xN = tail call { <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld1x2.v8bf16.p0bf16(bfloat* … 97 declare { <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld1x2.v8bf16.p0bf16(bfloat*) nounwind 124 …%vld1xN = tail call { <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld1x3.v8bf16.p… 135 declare { <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld1x3.v8bf16.p0bf16(bfloat*… 164 …at>, <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld1x4.v8bf16.p0bf16(bfloat* %pt… 177 declare { <8 x bfloat>, <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld1x4.v8bf16.… 216 …%vld2 = tail call { <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld2.v8bf16.p0v8bf16(<8 x bfloa… 225 declare { <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld2.v8bf16.p0v8bf16(<8 x bfloat>*) nounwi… 255 …%vld2_lane = tail call { <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld2lane.v8bf16.p0i8(<8 x … 264 declare { <8 x bfloat>, <8 x bfloat> } @llvm.aarch64.neon.ld2lane.v8bf16.p0i8(<8 x bfloat>, <8 x bf… [all …]
|
D | aarch64-bf16-dotprod-intrinsics.ll | 20 …%vbfdot3.i = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> %r, <8 x bfloat> %… 47 …%vbfdot3.i = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> %r, <8 x bfloat> %… 74 …%vbfdot3.i = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> %r, <8 x bfloat> %… 155 declare <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float>, <8 x bfloat>, <8 x bfloat>)
|
/external/llvm-project/llvm/test/CodeGen/ARM/ |
D | bf16-intrinsics-ld-st.ll | 84 …%vld1xN = tail call { <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld1x2.v8bf16.p0bf16(bfloat* %pt… 120 …%vld1xN = tail call { <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld1x3.v8bf16.p0bf… 162 …float>, <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld1x4.v8bf16.p0bf16(bfloat* %pt… 214 %vld2q_v = tail call { <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld2.v8bf16.p0i8(i8* %0, i32 2) 260 …%vld2q_lane_v = tail call { <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld2lane.v8bf16.p0i8(i8* %… 298 …%vld3q_v = tail call { <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld3.v8bf16.p0i8(… 356 …all { <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld3lane.v8bf16.p0i8(i8* %3, <8 x … 400 … bfloat>, <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld4.v8bf16.p0i8(i8* %0, i32 2) 470 …oat>, <8 x bfloat>, <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld4lane.v8bf16.p0i8(i8* %4, <8 x … 511 …%vld2q_dup_v = tail call { <8 x bfloat>, <8 x bfloat> } @llvm.arm.neon.vld2dup.v8bf16.p0i8(i8* %0,… [all …]
|
D | arm-bf16-dotprod-intrinsics.ll | 20 …%vbfdot3.i = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> %r, <8 x bfloat> %a, <… 47 …%vbfdot3.i = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> %r, <8 x bfloat> %a, <… 74 …%vbfdot3.i = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> %r, <8 x bfloat> %a, <… 167 declare <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float>, <8 x bfloat>, <8 x bfloat>)
|
/external/llvm-project/llvm/lib/Target/ARM/ |
D | ARMCallingConv.td | 34 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 60 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 75 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 95 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 112 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 169 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 187 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 214 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>, 237 CCIfType<[v2i64, v4i32, v8i16, v8f16, v8bf16, v16i8, v4f32], CCBitConvertToType<v2f64>>,
|
D | ARMInstrNEON.td | 1066 def : Pat<(vector_insert (v8bf16 QPR:$src), 6388 defm : ExtractEltOddF16VMOVH<v4bf16, v8bf16>; 6392 defm : ExtractEltEvenF16<v4bf16, v8bf16>; 6401 def : Pat<(extractelt (v8bf16 QPR:$src), imm_odd:$lane), 6519 defm : InsertEltF16<bf16, v4bf16, v8bf16>; 6643 def : Pat<(v8bf16 (ARMvduplane (v8bf16 QPR:$src), imm:$lane)), 6644 (v8bf16 (VDUPLN16q (v4bf16 (EXTRACT_SUBREG QPR:$src, 6651 def : Pat<(v8bf16 (ARMvdup (bf16 HPR:$src))), 6652 (v8bf16 (VDUPLN16q (INSERT_SUBREG (v4bf16 (IMPLICIT_DEF)), 7412 def : Pat<(v8i16 (bitconvert (v8bf16 QPR:$src))), (v8i16 QPR:$src)>; [all …]
|
D | ARMCallingConv.cpp | 222 case MVT::v8bf16: in CC_ARM_AAPCS_Custom_Aggregate()
|
D | ARMRegisterInfo.td | 449 def QPR : RegisterClass<"ARM", [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64, v8f16, v8bf16], 128,
|
D | ARMISelDAGToDAG.cpp | 2089 case MVT::v8bf16: in SelectVLD() 2234 case MVT::v8bf16: in SelectVST() 2402 case MVT::v8bf16: in SelectVLDSTLane() 2933 case MVT::v8bf16: in SelectVLDDup()
|
D | ARMISelLowering.cpp | 809 addQRTypeForNEON(MVT::v8bf16); in ARMTargetLowering() 4352 RegVT == MVT::v8bf16) in LowerFormalArguments()
|
/external/llvm-project/llvm/lib/Target/AArch64/ |
D | AArch64ISelDAGToDAG.cpp | 3546 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3573 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3600 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3627 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3654 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3681 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3708 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3735 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3762 } else if (VT == MVT::v8i16 || VT == MVT::v8f16 || VT == MVT::v8bf16) { in Select() 3784 VT == MVT::v8f16 || VT == MVT::v4bf16 || VT == MVT::v8bf16) { in Select() [all …]
|
D | AArch64CallingConvention.td | 35 CCIfBigEndian<CCIfType<[v2i64, v2f64, v4i32, v4f32, v8i16, v8f16, v8bf16, v16i8], 111 CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16, v8bf16], 119 CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16, v8bf16], 136 CCIfBigEndian<CCIfType<[v2i64, v2f64, v4i32, v4f32, v8i16, v8f16, v8bf16, v16i8], 155 CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16, v8bf16], 232 CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16, v8bf16], 248 CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16, v8bf16], 269 CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16, v8bf16], 291 CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16, v8bf16],
|
D | AArch64InstrInfo.td | 818 (v8bf16 (insert_subvector undef, 2327 defm : VecROLoadPat<ro128, v8bf16, LDRQroW, LDRQroX>; 2488 def : Pat<(v8bf16 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))), 3004 defm : VecROStorePat<ro128, v8bf16, FPR128, STRQroW, STRQroX>; 3130 def : Pat<(store (v8bf16 FPR128:$Rt), 3275 def : Pat<(store (v8bf16 FPR128:$Rt), 3986 def : Pat<(v8bf16 (AArch64rev32 V128:$Rn)), (REV32v8i16 V128:$Rn)>; 3987 def : Pat<(v8bf16 (AArch64rev64 V128:$Rn)), (REV64v8i16 V128:$Rn)>; 4861 defm : ExtPat<v4bf16, v8bf16, 4>; 4991 def : Pat<(v8bf16 (AArch64dup (bf16 FPR16:$Rn))), [all …]
|
D | AArch64RegisterInfo.td | 449 v8f16, v8bf16], 456 v8bf16],
|
D | AArch64InstrFormats.td | 7852 def v8bf16 : BaseSIMDThreeSameVectorBFDot<1, U, asm, ".4s", ".8h", V128, 7853 v4f32, v8bf16>; 7882 def v8bf16 : BaseSIMDThreeSameVectorBF16DotI<1, U, asm, ".4s", ".8h", 7883 ".2h", V128, v4f32, v8bf16>; 7889 (v8bf16 V128:$Rn), 7890 (v8bf16 V128:$Rm)))]> { 7900 (v8bf16 V128:$Rn), 7901 (v8bf16 7902 (AArch64duplane16 (v8bf16 V128_lo:$Rm), 7927 (v8bf16 V128:$Rn), [all …]
|
D | AArch64ISelLowering.cpp | 253 addQRTypeForNEON(MVT::v8bf16); in AArch64TargetLowering() 9691 VT == MVT::v8f16 || VT == MVT::v8bf16) in LowerINSERT_VECTOR_ELT() 9726 VT == MVT::v8f16 || VT == MVT::v8bf16) in LowerEXTRACT_VECTOR_ELT()
|
/external/llvm-project/llvm/include/llvm/Support/ |
D | MachineValueType.h | 134 v8bf16 = 79, // 8 x bf16 enumerator 384 SimpleTy == MVT::v8f16 || SimpleTy == MVT::v8bf16 || in is128BitVector() 597 case v8bf16: in getVectorElementType() 715 case v8bf16: in getVectorNumElements() 897 case v8bf16: in getSizeInBits() 1188 if (NumElements == 8) return MVT::v8bf16; in getVectorVT()
|
/external/llvm-project/llvm/test/Bitcode/ |
D | arm-bf16-upgrade.ll | 25 …; CHECK-NEXT: %vbfdot1.i = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> %r, <8 x… 70 ; CHECK: declare <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float>, <8 x bfloat>, <8 x bflo…
|
D | aarch64-bf16-upgrade.ll | 24 …; CHECK-NEXT: %vbfdot1.i = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> %r, … 70 ; CHECK: declare <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float>, <8 x bfloat>, <8 x …
|
/external/llvm-project/llvm/include/llvm/CodeGen/ |
D | ValueTypes.td | 107 def v8bf16 : ValueType<128, 79>; // 8 x bf16 vector value
|
/external/llvm-project/llvm/lib/CodeGen/ |
D | ValueTypes.cpp | 326 case MVT::v8bf16: in getTypeForEVT()
|
/external/llvm-project/llvm/utils/TableGen/ |
D | CodeGenTarget.cpp | 145 case MVT::v8bf16: return "MVT::v8bf16"; in getEnumName()
|
/external/llvm-project/llvm/include/llvm/IR/ |
D | Intrinsics.td | 310 def llvm_v8bf16_ty : LLVMType<v8bf16>; // 8 x bfloat (__bf16)
|