diff options
Diffstat (limited to 'lib/Target/AArch64/AArch64InstrFormats.td')
| -rw-r--r-- | lib/Target/AArch64/AArch64InstrFormats.td | 215 |
1 files changed, 212 insertions, 3 deletions
diff --git a/lib/Target/AArch64/AArch64InstrFormats.td b/lib/Target/AArch64/AArch64InstrFormats.td index d295c02..0c0efaf 100644 --- a/lib/Target/AArch64/AArch64InstrFormats.td +++ b/lib/Target/AArch64/AArch64InstrFormats.td @@ -1637,10 +1637,16 @@ multiclass AddSub<bit isSub, string mnemonic, SDPatternOperator OpNode = null_frag> { let hasSideEffects = 0, isReMaterializable = 1, isAsCheapAsAMove = 1 in { // Add/Subtract immediate + // Increase the weight of the immediate variant to try to match it before + // the extended register variant. + // We used to match the register variant before the immediate when the + // register argument could be implicitly zero-extended. + let AddedComplexity = 6 in def Wri : BaseAddSubImm<isSub, 0, GPR32sp, GPR32sp, addsub_shifted_imm32, mnemonic, OpNode> { let Inst{31} = 0; } + let AddedComplexity = 6 in def Xri : BaseAddSubImm<isSub, 0, GPR64sp, GPR64sp, addsub_shifted_imm64, mnemonic, OpNode> { let Inst{31} = 1; @@ -3282,6 +3288,10 @@ class LoadStoreExclusiveSimple<bits<2> sz, bit o2, bit L, bit o1, bit o0, : BaseLoadStoreExclusive<sz, o2, L, o1, o0, oops, iops, asm, operands> { bits<5> Rt; bits<5> Rn; + let Inst{20-16} = 0b11111; + let Unpredictable{20-16} = 0b11111; + let Inst{14-10} = 0b11111; + let Unpredictable{14-10} = 0b11111; let Inst{9-5} = Rn; let Inst{4-0} = Rt; @@ -5298,6 +5308,27 @@ class BaseSIMDThreeScalar<bit U, bits<2> size, bits<5> opcode, let Inst{4-0} = Rd; } +let mayStore = 0, mayLoad = 0, hasSideEffects = 0 in +class BaseSIMDThreeScalarTied<bit U, bits<2> size, bit R, bits<5> opcode, + dag oops, dag iops, string asm, + list<dag> pattern> + : I<oops, iops, asm, "\t$Rd, $Rn, $Rm", "$Rd = $dst", pattern>, + Sched<[WriteV]> { + bits<5> Rd; + bits<5> Rn; + bits<5> Rm; + let Inst{31-30} = 0b01; + let Inst{29} = U; + let Inst{28-24} = 0b11110; + let Inst{23-22} = size; + let Inst{21} = R; + let Inst{20-16} = Rm; + let Inst{15-11} = opcode; + let Inst{10} = 1; + let Inst{9-5} = Rn; + let Inst{4-0} = Rd; +} + multiclass SIMDThreeScalarD<bit U, bits<5> opc, string asm, SDPatternOperator OpNode> { def v1i64 : BaseSIMDThreeScalar<U, 0b11, opc, FPR64, asm, @@ -5325,6 +5356,16 @@ multiclass SIMDThreeScalarHS<bit U, bits<5> opc, string asm, def v1i16 : BaseSIMDThreeScalar<U, 0b01, opc, FPR16, asm, []>; } +multiclass SIMDThreeScalarHSTied<bit U, bit R, bits<5> opc, string asm, + SDPatternOperator OpNode = null_frag> { + def v1i32: BaseSIMDThreeScalarTied<U, 0b10, R, opc, (outs FPR32:$dst), + (ins FPR32:$Rd, FPR32:$Rn, FPR32:$Rm), + asm, []>; + def v1i16: BaseSIMDThreeScalarTied<U, 0b01, R, opc, (outs FPR16:$dst), + (ins FPR16:$Rd, FPR16:$Rn, FPR16:$Rm), + asm, []>; +} + multiclass SIMDThreeScalarSD<bit U, bit S, bits<5> opc, string asm, SDPatternOperator OpNode = null_frag> { let mayLoad = 0, mayStore = 0, hasSideEffects = 0 in { @@ -5885,7 +5926,7 @@ multiclass SIMDIns { let Inst{20-18} = idx; let Inst{17-16} = 0b10; let Inst{14-12} = idx2; - let Inst{11} = 0; + let Inst{11} = {?}; } def vi32lane : SIMDInsFromElement<".s", v4i32, i32, VectorIndexS> { bits<2> idx; @@ -5893,7 +5934,7 @@ multiclass SIMDIns { let Inst{20-19} = idx; let Inst{18-16} = 0b100; let Inst{14-13} = idx2; - let Inst{12-11} = 0; + let Inst{12-11} = {?,?}; } def vi64lane : SIMDInsFromElement<".d", v2i64, i64, VectorIndexD> { bits<1> idx; @@ -5901,7 +5942,7 @@ multiclass SIMDIns { let Inst{20} = idx; let Inst{19-16} = 0b1000; let Inst{14} = idx2; - let Inst{13-11} = 0; + let Inst{13-11} = {?,?,?}; } // For all forms of the INS instruction, the "mov" mnemonic is the @@ -8517,6 +8558,174 @@ multiclass SIMDLdSt4SingleAliases<string asm> { } // end of 'let Predicates = [HasNEON]' //---------------------------------------------------------------------------- +// AdvSIMD v8.1 Rounding Double Multiply Add/Subtract +//---------------------------------------------------------------------------- + +let Predicates = [HasNEON, HasV8_1a] in { + +class BaseSIMDThreeSameVectorTiedR0<bit Q, bit U, bits<2> size, bits<5> opcode, + RegisterOperand regtype, string asm, + string kind, list<dag> pattern> + : BaseSIMDThreeSameVectorTied<Q, U, size, opcode, regtype, asm, kind, + pattern> { + let Inst{21}=0; +} +multiclass SIMDThreeSameVectorSQRDMLxHTiedHS<bit U, bits<5> opc, string asm, + SDPatternOperator Accum> { + def v4i16 : BaseSIMDThreeSameVectorTiedR0<0, U, 0b01, opc, V64, asm, ".4h", + [(set (v4i16 V64:$dst), + (Accum (v4i16 V64:$Rd), + (v4i16 (int_aarch64_neon_sqrdmulh (v4i16 V64:$Rn), + (v4i16 V64:$Rm)))))]>; + def v8i16 : BaseSIMDThreeSameVectorTiedR0<1, U, 0b01, opc, V128, asm, ".8h", + [(set (v8i16 V128:$dst), + (Accum (v8i16 V128:$Rd), + (v8i16 (int_aarch64_neon_sqrdmulh (v8i16 V128:$Rn), + (v8i16 V128:$Rm)))))]>; + def v2i32 : BaseSIMDThreeSameVectorTiedR0<0, U, 0b10, opc, V64, asm, ".2s", + [(set (v2i32 V64:$dst), + (Accum (v2i32 V64:$Rd), + (v2i32 (int_aarch64_neon_sqrdmulh (v2i32 V64:$Rn), + (v2i32 V64:$Rm)))))]>; + def v4i32 : BaseSIMDThreeSameVectorTiedR0<1, U, 0b10, opc, V128, asm, ".4s", + [(set (v4i32 V128:$dst), + (Accum (v4i32 V128:$Rd), + (v4i32 (int_aarch64_neon_sqrdmulh (v4i32 V128:$Rn), + (v4i32 V128:$Rm)))))]>; +} + +multiclass SIMDIndexedSQRDMLxHSDTied<bit U, bits<4> opc, string asm, + SDPatternOperator Accum> { + def v4i16_indexed : BaseSIMDIndexedTied<0, U, 0, 0b01, opc, + V64, V64, V128_lo, VectorIndexH, + asm, ".4h", ".4h", ".4h", ".h", + [(set (v4i16 V64:$dst), + (Accum (v4i16 V64:$Rd), + (v4i16 (int_aarch64_neon_sqrdmulh + (v4i16 V64:$Rn), + (v4i16 (AArch64duplane16 (v8i16 V128_lo:$Rm), + VectorIndexH:$idx))))))]> { + bits<3> idx; + let Inst{11} = idx{2}; + let Inst{21} = idx{1}; + let Inst{20} = idx{0}; + } + + def v8i16_indexed : BaseSIMDIndexedTied<1, U, 0, 0b01, opc, + V128, V128, V128_lo, VectorIndexH, + asm, ".8h", ".8h", ".8h", ".h", + [(set (v8i16 V128:$dst), + (Accum (v8i16 V128:$Rd), + (v8i16 (int_aarch64_neon_sqrdmulh + (v8i16 V128:$Rn), + (v8i16 (AArch64duplane16 (v8i16 V128_lo:$Rm), + VectorIndexH:$idx))))))]> { + bits<3> idx; + let Inst{11} = idx{2}; + let Inst{21} = idx{1}; + let Inst{20} = idx{0}; + } + + def v2i32_indexed : BaseSIMDIndexedTied<0, U, 0, 0b10, opc, + V64, V64, V128, VectorIndexS, + asm, ".2s", ".2s", ".2s", ".s", + [(set (v2i32 V64:$dst), + (Accum (v2i32 V64:$Rd), + (v2i32 (int_aarch64_neon_sqrdmulh + (v2i32 V64:$Rn), + (v2i32 (AArch64duplane32 (v4i32 V128:$Rm), + VectorIndexS:$idx))))))]> { + bits<2> idx; + let Inst{11} = idx{1}; + let Inst{21} = idx{0}; + } + + // FIXME: it would be nice to use the scalar (v1i32) instruction here, but + // an intermediate EXTRACT_SUBREG would be untyped. + // FIXME: direct EXTRACT_SUBREG from v2i32 to i32 is illegal, that's why we + // got it lowered here as (i32 vector_extract (v4i32 insert_subvector(..))) + def : Pat<(i32 (Accum (i32 FPR32Op:$Rd), + (i32 (vector_extract + (v4i32 (insert_subvector + (undef), + (v2i32 (int_aarch64_neon_sqrdmulh + (v2i32 V64:$Rn), + (v2i32 (AArch64duplane32 + (v4i32 V128:$Rm), + VectorIndexS:$idx)))), + (i32 0))), + (i64 0))))), + (EXTRACT_SUBREG + (v2i32 (!cast<Instruction>(NAME # v2i32_indexed) + (v2i32 (INSERT_SUBREG (v2i32 (IMPLICIT_DEF)), + FPR32Op:$Rd, + ssub)), + V64:$Rn, + V128:$Rm, + VectorIndexS:$idx)), + ssub)>; + + def v4i32_indexed : BaseSIMDIndexedTied<1, U, 0, 0b10, opc, + V128, V128, V128, VectorIndexS, + asm, ".4s", ".4s", ".4s", ".s", + [(set (v4i32 V128:$dst), + (Accum (v4i32 V128:$Rd), + (v4i32 (int_aarch64_neon_sqrdmulh + (v4i32 V128:$Rn), + (v4i32 (AArch64duplane32 (v4i32 V128:$Rm), + VectorIndexS:$idx))))))]> { + bits<2> idx; + let Inst{11} = idx{1}; + let Inst{21} = idx{0}; + } + + // FIXME: it would be nice to use the scalar (v1i32) instruction here, but + // an intermediate EXTRACT_SUBREG would be untyped. + def : Pat<(i32 (Accum (i32 FPR32Op:$Rd), + (i32 (vector_extract + (v4i32 (int_aarch64_neon_sqrdmulh + (v4i32 V128:$Rn), + (v4i32 (AArch64duplane32 + (v4i32 V128:$Rm), + VectorIndexS:$idx)))), + (i64 0))))), + (EXTRACT_SUBREG + (v4i32 (!cast<Instruction>(NAME # v4i32_indexed) + (v4i32 (INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), + FPR32Op:$Rd, + ssub)), + V128:$Rn, + V128:$Rm, + VectorIndexS:$idx)), + ssub)>; + + def i16_indexed : BaseSIMDIndexedTied<1, U, 1, 0b01, opc, + FPR16Op, FPR16Op, V128_lo, + VectorIndexH, asm, ".h", "", "", ".h", + []> { + bits<3> idx; + let Inst{11} = idx{2}; + let Inst{21} = idx{1}; + let Inst{20} = idx{0}; + } + + def i32_indexed : BaseSIMDIndexedTied<1, U, 1, 0b10, opc, + FPR32Op, FPR32Op, V128, VectorIndexS, + asm, ".s", "", "", ".s", + [(set (i32 FPR32Op:$dst), + (Accum (i32 FPR32Op:$Rd), + (i32 (int_aarch64_neon_sqrdmulh + (i32 FPR32Op:$Rn), + (i32 (vector_extract (v4i32 V128:$Rm), + VectorIndexS:$idx))))))]> { + bits<2> idx; + let Inst{11} = idx{1}; + let Inst{21} = idx{0}; + } +} +} // let Predicates = [HasNeon, HasV8_1a] + +//---------------------------------------------------------------------------- // Crypto extensions //---------------------------------------------------------------------------- |