diff options
Diffstat (limited to 'lib/Target/X86/X86IntrinsicsInfo.h')
-rw-r--r-- | lib/Target/X86/X86IntrinsicsInfo.h | 928 |
1 files changed, 374 insertions, 554 deletions
diff --git a/lib/Target/X86/X86IntrinsicsInfo.h b/lib/Target/X86/X86IntrinsicsInfo.h index b525d5eb60a7c..b647d11e38663 100644 --- a/lib/Target/X86/X86IntrinsicsInfo.h +++ b/lib/Target/X86/X86IntrinsicsInfo.h @@ -14,31 +14,36 @@ #ifndef LLVM_LIB_TARGET_X86_X86INTRINSICSINFO_H #define LLVM_LIB_TARGET_X86_X86INTRINSICSINFO_H +#include "X86ISelLowering.h" +#include "X86InstrInfo.h" + namespace llvm { -enum IntrinsicType { +enum IntrinsicType : uint16_t { INTR_NO_TYPE, GATHER, SCATTER, PREFETCH, RDSEED, RDRAND, RDPMC, RDTSC, XTEST, ADX, FPCLASS, FPCLASSS, INTR_TYPE_1OP, INTR_TYPE_2OP, INTR_TYPE_2OP_IMM8, INTR_TYPE_3OP, INTR_TYPE_4OP, - CMP_MASK, CMP_MASK_CC,CMP_MASK_SCALAR_CC, VSHIFT, VSHIFT_MASK, COMI, COMI_RM, + CMP_MASK, CMP_MASK_CC,CMP_MASK_SCALAR_CC, VSHIFT, COMI, COMI_RM, INTR_TYPE_1OP_MASK, INTR_TYPE_1OP_MASK_RM, INTR_TYPE_2OP_MASK, INTR_TYPE_2OP_MASK_RM, INTR_TYPE_2OP_IMM8_MASK, INTR_TYPE_3OP_MASK, INTR_TYPE_3OP_MASK_RM, INTR_TYPE_3OP_IMM8_MASK, - FMA_OP_MASK, FMA_OP_MASKZ, FMA_OP_MASK3, VPERM_3OP_MASK, - VPERM_3OP_MASKZ, INTR_TYPE_SCALAR_MASK, + FMA_OP_MASK, FMA_OP_MASKZ, FMA_OP_MASK3, + FMA_OP_SCALAR_MASK, FMA_OP_SCALAR_MASKZ, FMA_OP_SCALAR_MASK3, + VPERM_2OP_MASK, VPERM_3OP_MASK, VPERM_3OP_MASKZ, INTR_TYPE_SCALAR_MASK, INTR_TYPE_SCALAR_MASK_RM, INTR_TYPE_3OP_SCALAR_MASK_RM, - COMPRESS_EXPAND_IN_REG, COMPRESS_TO_MEM, BRCST_SUBVEC_TO_VEC, + COMPRESS_EXPAND_IN_REG, COMPRESS_TO_MEM, BRCST_SUBVEC_TO_VEC, BRCST32x2_TO_VEC, TRUNCATE_TO_MEM_VI8, TRUNCATE_TO_MEM_VI16, TRUNCATE_TO_MEM_VI32, - EXPAND_FROM_MEM, LOADA, LOADU, BLEND, INSERT_SUBVEC, - TERLOG_OP_MASK, TERLOG_OP_MASKZ, BROADCASTM, KUNPCK, CONVERT_MASK_TO_VEC, CONVERT_TO_MASK + EXPAND_FROM_MEM, INSERT_SUBVEC, + TERLOG_OP_MASK, TERLOG_OP_MASKZ, BROADCASTM, KUNPCK, FIXUPIMM, FIXUPIMM_MASKZ, FIXUPIMMS, + FIXUPIMMS_MASKZ, CONVERT_MASK_TO_VEC, CONVERT_TO_MASK }; struct IntrinsicData { - unsigned Id; + uint16_t Id; IntrinsicType Type; - unsigned Opc0; - unsigned Opc1; + uint16_t Opc0; + uint16_t Opc1; bool operator<(const IntrinsicData &RHS) const { return Id < RHS.Id; @@ -61,6 +66,14 @@ static const IntrinsicData IntrinsicsWithChain[] = { X86_INTRINSIC_DATA(addcarryx_u32, ADX, X86ISD::ADC, 0), X86_INTRINSIC_DATA(addcarryx_u64, ADX, X86ISD::ADC, 0), + X86_INTRINSIC_DATA(avx512_gather_dpd_512, GATHER, X86::VGATHERDPDZrm, 0), + X86_INTRINSIC_DATA(avx512_gather_dpi_512, GATHER, X86::VPGATHERDDZrm, 0), + X86_INTRINSIC_DATA(avx512_gather_dpq_512, GATHER, X86::VPGATHERDQZrm, 0), + X86_INTRINSIC_DATA(avx512_gather_dps_512, GATHER, X86::VGATHERDPSZrm, 0), + X86_INTRINSIC_DATA(avx512_gather_qpd_512, GATHER, X86::VGATHERQPDZrm, 0), + X86_INTRINSIC_DATA(avx512_gather_qpi_512, GATHER, X86::VPGATHERQDZrm, 0), + X86_INTRINSIC_DATA(avx512_gather_qpq_512, GATHER, X86::VPGATHERQQZrm, 0), + X86_INTRINSIC_DATA(avx512_gather_qps_512, GATHER, X86::VGATHERQPSZrm, 0), X86_INTRINSIC_DATA(avx512_gather3div2_df, GATHER, X86::VGATHERQPDZ128rm, 0), X86_INTRINSIC_DATA(avx512_gather3div2_di, GATHER, X86::VPGATHERQQZ128rm, 0), X86_INTRINSIC_DATA(avx512_gather3div4_df, GATHER, X86::VGATHERQPDZ256rm, 0), @@ -77,14 +90,6 @@ static const IntrinsicData IntrinsicsWithChain[] = { X86_INTRINSIC_DATA(avx512_gather3siv4_si, GATHER, X86::VPGATHERDDZ128rm, 0), X86_INTRINSIC_DATA(avx512_gather3siv8_sf, GATHER, X86::VGATHERDPSZ256rm, 0), X86_INTRINSIC_DATA(avx512_gather3siv8_si, GATHER, X86::VPGATHERDDZ256rm, 0), - X86_INTRINSIC_DATA(avx512_gather_dpd_512, GATHER, X86::VGATHERDPDZrm, 0), - X86_INTRINSIC_DATA(avx512_gather_dpi_512, GATHER, X86::VPGATHERDDZrm, 0), - X86_INTRINSIC_DATA(avx512_gather_dpq_512, GATHER, X86::VPGATHERDQZrm, 0), - X86_INTRINSIC_DATA(avx512_gather_dps_512, GATHER, X86::VGATHERDPSZrm, 0), - X86_INTRINSIC_DATA(avx512_gather_qpd_512, GATHER, X86::VGATHERQPDZrm, 0), - X86_INTRINSIC_DATA(avx512_gather_qpi_512, GATHER, X86::VPGATHERQDZrm, 0), - X86_INTRINSIC_DATA(avx512_gather_qpq_512, GATHER, X86::VPGATHERQQZrm, 0), - X86_INTRINSIC_DATA(avx512_gather_qps_512, GATHER, X86::VGATHERQPSZrm, 0), X86_INTRINSIC_DATA(avx512_gatherpf_dpd_512, PREFETCH, X86::VGATHERPF0DPDm, X86::VGATHERPF1DPDm), @@ -143,18 +148,6 @@ static const IntrinsicData IntrinsicsWithChain[] = { EXPAND_FROM_MEM, X86ISD::EXPAND, 0), X86_INTRINSIC_DATA(avx512_mask_expand_load_q_512, EXPAND_FROM_MEM, X86ISD::EXPAND, 0), - X86_INTRINSIC_DATA(avx512_mask_load_pd_128, LOADA, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_load_pd_256, LOADA, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_load_pd_512, LOADA, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_load_ps_128, LOADA, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_load_ps_256, LOADA, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_load_ps_512, LOADA, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_loadu_pd_128, LOADU, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_loadu_pd_256, LOADU, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_loadu_pd_512, LOADU, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_loadu_ps_128, LOADU, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_loadu_ps_256, LOADU, ISD::DELETED_NODE, 0), - X86_INTRINSIC_DATA(avx512_mask_loadu_ps_512, LOADU, ISD::DELETED_NODE, 0), X86_INTRINSIC_DATA(avx512_mask_pmov_db_mem_128, TRUNCATE_TO_MEM_VI8, X86ISD::VTRUNC, 0), X86_INTRINSIC_DATA(avx512_mask_pmov_db_mem_256, TRUNCATE_TO_MEM_VI8, @@ -223,7 +216,6 @@ static const IntrinsicData IntrinsicsWithChain[] = { X86_INTRINSIC_DATA(avx512_scattersiv4_si, SCATTER, X86::VPSCATTERDDZ128mr, 0), X86_INTRINSIC_DATA(avx512_scattersiv8_sf, SCATTER, X86::VSCATTERDPSZ256mr, 0), X86_INTRINSIC_DATA(avx512_scattersiv8_si, SCATTER, X86::VPSCATTERDDZ256mr, 0), - X86_INTRINSIC_DATA(rdpmc, RDPMC, X86ISD::RDPMC_DAG, 0), X86_INTRINSIC_DATA(rdrand_16, RDRAND, X86ISD::RDRAND, 0), X86_INTRINSIC_DATA(rdrand_32, RDRAND, X86ISD::RDRAND, 0), @@ -242,7 +234,7 @@ static const IntrinsicData IntrinsicsWithChain[] = { /* * Find Intrinsic data by intrinsic ID */ -static const IntrinsicData* getIntrinsicWithChain(unsigned IntNo) { +static const IntrinsicData* getIntrinsicWithChain(uint16_t IntNo) { IntrinsicData IntrinsicToFind = {IntNo, INTR_NO_TYPE, 0, 0 }; const IntrinsicData *Data = std::lower_bound(std::begin(IntrinsicsWithChain), @@ -258,49 +250,51 @@ static const IntrinsicData* getIntrinsicWithChain(unsigned IntNo) { * the alphabetical order. */ static const IntrinsicData IntrinsicsWithoutChain[] = { + X86_INTRINSIC_DATA(avx_hadd_pd_256, INTR_TYPE_2OP, X86ISD::FHADD, 0), + X86_INTRINSIC_DATA(avx_hadd_ps_256, INTR_TYPE_2OP, X86ISD::FHADD, 0), + X86_INTRINSIC_DATA(avx_hsub_pd_256, INTR_TYPE_2OP, X86ISD::FHSUB, 0), + X86_INTRINSIC_DATA(avx_hsub_ps_256, INTR_TYPE_2OP, X86ISD::FHSUB, 0), + X86_INTRINSIC_DATA(avx_max_pd_256, INTR_TYPE_2OP, X86ISD::FMAX, 0), + X86_INTRINSIC_DATA(avx_max_ps_256, INTR_TYPE_2OP, X86ISD::FMAX, 0), + X86_INTRINSIC_DATA(avx_min_pd_256, INTR_TYPE_2OP, X86ISD::FMIN, 0), + X86_INTRINSIC_DATA(avx_min_ps_256, INTR_TYPE_2OP, X86ISD::FMIN, 0), + X86_INTRINSIC_DATA(avx_movmsk_pd_256, INTR_TYPE_1OP, X86ISD::MOVMSK, 0), + X86_INTRINSIC_DATA(avx_movmsk_ps_256, INTR_TYPE_1OP, X86ISD::MOVMSK, 0), + X86_INTRINSIC_DATA(avx_rcp_ps_256, INTR_TYPE_1OP, X86ISD::FRCP, 0), + X86_INTRINSIC_DATA(avx_rsqrt_ps_256, INTR_TYPE_1OP, X86ISD::FRSQRT, 0), + X86_INTRINSIC_DATA(avx_sqrt_pd_256, INTR_TYPE_1OP, ISD::FSQRT, 0), + X86_INTRINSIC_DATA(avx_sqrt_ps_256, INTR_TYPE_1OP, ISD::FSQRT, 0), + X86_INTRINSIC_DATA(avx_vperm2f128_pd_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0), + X86_INTRINSIC_DATA(avx_vperm2f128_ps_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0), + X86_INTRINSIC_DATA(avx_vperm2f128_si_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0), + X86_INTRINSIC_DATA(avx_vpermilvar_pd, INTR_TYPE_2OP, X86ISD::VPERMILPV, 0), + X86_INTRINSIC_DATA(avx_vpermilvar_pd_256, INTR_TYPE_2OP, X86ISD::VPERMILPV, 0), + X86_INTRINSIC_DATA(avx_vpermilvar_ps, INTR_TYPE_2OP, X86ISD::VPERMILPV, 0), + X86_INTRINSIC_DATA(avx_vpermilvar_ps_256, INTR_TYPE_2OP, X86ISD::VPERMILPV, 0), + X86_INTRINSIC_DATA(avx2_pabs_b, INTR_TYPE_1OP, X86ISD::ABS, 0), + X86_INTRINSIC_DATA(avx2_pabs_d, INTR_TYPE_1OP, X86ISD::ABS, 0), + X86_INTRINSIC_DATA(avx2_pabs_w, INTR_TYPE_1OP, X86ISD::ABS, 0), X86_INTRINSIC_DATA(avx2_packssdw, INTR_TYPE_2OP, X86ISD::PACKSS, 0), X86_INTRINSIC_DATA(avx2_packsswb, INTR_TYPE_2OP, X86ISD::PACKSS, 0), X86_INTRINSIC_DATA(avx2_packusdw, INTR_TYPE_2OP, X86ISD::PACKUS, 0), X86_INTRINSIC_DATA(avx2_packuswb, INTR_TYPE_2OP, X86ISD::PACKUS, 0), + X86_INTRINSIC_DATA(avx2_padds_b, INTR_TYPE_2OP, X86ISD::ADDS, 0), + X86_INTRINSIC_DATA(avx2_padds_w, INTR_TYPE_2OP, X86ISD::ADDS, 0), + X86_INTRINSIC_DATA(avx2_paddus_b, INTR_TYPE_2OP, X86ISD::ADDUS, 0), + X86_INTRINSIC_DATA(avx2_paddus_w, INTR_TYPE_2OP, X86ISD::ADDUS, 0), X86_INTRINSIC_DATA(avx2_pavg_b, INTR_TYPE_2OP, X86ISD::AVG, 0), X86_INTRINSIC_DATA(avx2_pavg_w, INTR_TYPE_2OP, X86ISD::AVG, 0), X86_INTRINSIC_DATA(avx2_phadd_d, INTR_TYPE_2OP, X86ISD::HADD, 0), X86_INTRINSIC_DATA(avx2_phadd_w, INTR_TYPE_2OP, X86ISD::HADD, 0), X86_INTRINSIC_DATA(avx2_phsub_d, INTR_TYPE_2OP, X86ISD::HSUB, 0), X86_INTRINSIC_DATA(avx2_phsub_w, INTR_TYPE_2OP, X86ISD::HSUB, 0), - X86_INTRINSIC_DATA(avx2_pmaxs_b, INTR_TYPE_2OP, ISD::SMAX, 0), - X86_INTRINSIC_DATA(avx2_pmaxs_d, INTR_TYPE_2OP, ISD::SMAX, 0), - X86_INTRINSIC_DATA(avx2_pmaxs_w, INTR_TYPE_2OP, ISD::SMAX, 0), - X86_INTRINSIC_DATA(avx2_pmaxu_b, INTR_TYPE_2OP, ISD::UMAX, 0), - X86_INTRINSIC_DATA(avx2_pmaxu_d, INTR_TYPE_2OP, ISD::UMAX, 0), - X86_INTRINSIC_DATA(avx2_pmaxu_w, INTR_TYPE_2OP, ISD::UMAX, 0), - X86_INTRINSIC_DATA(avx2_pmins_b, INTR_TYPE_2OP, ISD::SMIN, 0), - X86_INTRINSIC_DATA(avx2_pmins_d, INTR_TYPE_2OP, ISD::SMIN, 0), - X86_INTRINSIC_DATA(avx2_pmins_w, INTR_TYPE_2OP, ISD::SMIN, 0), - X86_INTRINSIC_DATA(avx2_pminu_b, INTR_TYPE_2OP, ISD::UMIN, 0), - X86_INTRINSIC_DATA(avx2_pminu_d, INTR_TYPE_2OP, ISD::UMIN, 0), - X86_INTRINSIC_DATA(avx2_pminu_w, INTR_TYPE_2OP, ISD::UMIN, 0), - X86_INTRINSIC_DATA(avx2_pmovsxbd, INTR_TYPE_1OP, X86ISD::VSEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovsxbq, INTR_TYPE_1OP, X86ISD::VSEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovsxbw, INTR_TYPE_1OP, X86ISD::VSEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovsxdq, INTR_TYPE_1OP, X86ISD::VSEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovsxwd, INTR_TYPE_1OP, X86ISD::VSEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovsxwq, INTR_TYPE_1OP, X86ISD::VSEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovzxbd, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovzxbq, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovzxbw, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovzxdq, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovzxwd, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(avx2_pmovzxwq, INTR_TYPE_1OP, X86ISD::VZEXT, 0), + X86_INTRINSIC_DATA(avx2_pmovmskb, INTR_TYPE_1OP, X86ISD::MOVMSK, 0), X86_INTRINSIC_DATA(avx2_pmul_dq, INTR_TYPE_2OP, X86ISD::PMULDQ, 0), X86_INTRINSIC_DATA(avx2_pmulh_w, INTR_TYPE_2OP, ISD::MULHS, 0), X86_INTRINSIC_DATA(avx2_pmulhu_w, INTR_TYPE_2OP, ISD::MULHU, 0), X86_INTRINSIC_DATA(avx2_pmulu_dq, INTR_TYPE_2OP, X86ISD::PMULUDQ, 0), X86_INTRINSIC_DATA(avx2_psad_bw, INTR_TYPE_2OP, X86ISD::PSADBW, 0), X86_INTRINSIC_DATA(avx2_pshuf_b, INTR_TYPE_2OP, X86ISD::PSHUFB, 0), - X86_INTRINSIC_DATA(avx2_psign_b, INTR_TYPE_2OP, X86ISD::PSIGN, 0), - X86_INTRINSIC_DATA(avx2_psign_d, INTR_TYPE_2OP, X86ISD::PSIGN, 0), - X86_INTRINSIC_DATA(avx2_psign_w, INTR_TYPE_2OP, X86ISD::PSIGN, 0), X86_INTRINSIC_DATA(avx2_psll_d, INTR_TYPE_2OP, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx2_psll_q, INTR_TYPE_2OP, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx2_psll_w, INTR_TYPE_2OP, X86ISD::VSHL, 0), @@ -315,8 +309,8 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx2_psra_w, INTR_TYPE_2OP, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx2_psrai_d, VSHIFT, X86ISD::VSRAI, 0), X86_INTRINSIC_DATA(avx2_psrai_w, VSHIFT, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx2_psrav_d, INTR_TYPE_2OP, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx2_psrav_d_256, INTR_TYPE_2OP, ISD::SRA, 0), + X86_INTRINSIC_DATA(avx2_psrav_d, INTR_TYPE_2OP, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx2_psrav_d_256, INTR_TYPE_2OP, X86ISD::VSRAV, 0), X86_INTRINSIC_DATA(avx2_psrl_d, INTR_TYPE_2OP, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx2_psrl_q, INTR_TYPE_2OP, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx2_psrl_w, INTR_TYPE_2OP, X86ISD::VSRL, 0), @@ -327,6 +321,8 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx2_psrlv_d_256, INTR_TYPE_2OP, ISD::SRL, 0), X86_INTRINSIC_DATA(avx2_psrlv_q, INTR_TYPE_2OP, ISD::SRL, 0), X86_INTRINSIC_DATA(avx2_psrlv_q_256, INTR_TYPE_2OP, ISD::SRL, 0), + X86_INTRINSIC_DATA(avx2_psubs_b, INTR_TYPE_2OP, X86ISD::SUBS, 0), + X86_INTRINSIC_DATA(avx2_psubs_w, INTR_TYPE_2OP, X86ISD::SUBS, 0), X86_INTRINSIC_DATA(avx2_psubus_b, INTR_TYPE_2OP, X86ISD::SUBUS, 0), X86_INTRINSIC_DATA(avx2_psubus_w, INTR_TYPE_2OP, X86ISD::SUBUS, 0), X86_INTRINSIC_DATA(avx2_vperm2i128, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0), @@ -380,50 +376,6 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_kunpck_bw, KUNPCK, ISD::CONCAT_VECTORS, 0), X86_INTRINSIC_DATA(avx512_kunpck_dq, KUNPCK, ISD::CONCAT_VECTORS, 0), X86_INTRINSIC_DATA(avx512_kunpck_wd, KUNPCK, ISD::CONCAT_VECTORS, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_128, FMA_OP_MASK3, X86ISD::FMADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_256, FMA_OP_MASK3, X86ISD::FMADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_512, FMA_OP_MASK3, X86ISD::FMADD, - X86ISD::FMADD_RND), - X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_128, FMA_OP_MASK3, X86ISD::FMADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_256, FMA_OP_MASK3, X86ISD::FMADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_512, FMA_OP_MASK3, X86ISD::FMADD, - X86ISD::FMADD_RND), - - X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_pd_128, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_pd_256, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_pd_512, FMA_OP_MASK3, X86ISD::FMADDSUB, - X86ISD::FMADDSUB_RND), - X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_ps_128, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_ps_256, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_ps_512, FMA_OP_MASK3, X86ISD::FMADDSUB, - X86ISD::FMADDSUB_RND), - - X86_INTRINSIC_DATA(avx512_mask3_vfmsub_pd_128, FMA_OP_MASK3, X86ISD::FMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsub_pd_256, FMA_OP_MASK3, X86ISD::FMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsub_pd_512, FMA_OP_MASK3, X86ISD::FMSUB, - X86ISD::FMSUB_RND), - X86_INTRINSIC_DATA(avx512_mask3_vfmsub_ps_128, FMA_OP_MASK3, X86ISD::FMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsub_ps_256, FMA_OP_MASK3, X86ISD::FMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsub_ps_512, FMA_OP_MASK3, X86ISD::FMSUB, - X86ISD::FMSUB_RND), - - X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_pd_128, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_pd_256, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_pd_512, FMA_OP_MASK3, X86ISD::FMSUBADD, - X86ISD::FMSUBADD_RND), - X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_ps_128, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_ps_256, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_ps_512, FMA_OP_MASK3, X86ISD::FMSUBADD, - X86ISD::FMSUBADD_RND), - - X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_pd_128, FMA_OP_MASK3, X86ISD::FNMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_pd_256, FMA_OP_MASK3, X86ISD::FNMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_pd_512, FMA_OP_MASK3, X86ISD::FNMSUB, - X86ISD::FNMSUB_RND), - X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_ps_128, FMA_OP_MASK3, X86ISD::FNMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_ps_256, FMA_OP_MASK3, X86ISD::FNMSUB, 0), - X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_ps_512, FMA_OP_MASK3, X86ISD::FNMSUB, - X86ISD::FNMSUB_RND), X86_INTRINSIC_DATA(avx512_mask_add_pd_128, INTR_TYPE_2OP_MASK, ISD::FADD, 0), X86_INTRINSIC_DATA(avx512_mask_add_pd_256, INTR_TYPE_2OP_MASK, ISD::FADD, 0), @@ -449,38 +401,10 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_andn_ps_128, INTR_TYPE_2OP_MASK, X86ISD::FANDN, 0), X86_INTRINSIC_DATA(avx512_mask_andn_ps_256, INTR_TYPE_2OP_MASK, X86ISD::FANDN, 0), X86_INTRINSIC_DATA(avx512_mask_andn_ps_512, INTR_TYPE_2OP_MASK, X86ISD::FANDN, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_b_128, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_b_256, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_b_512, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_d_128, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_d_256, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_d_512, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_pd_128, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_pd_256, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_pd_512, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_ps_128, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_ps_256, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_ps_512, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_q_128, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_q_256, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_q_512, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_w_128, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_w_256, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_blend_w_512, BLEND, X86ISD::SELECT, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcast_sd_pd_256, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcast_sd_pd_512, INTR_TYPE_1OP_MASK, + X86_INTRINSIC_DATA(avx512_mask_broadcastf32x2_256, BRCST32x2_TO_VEC, X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcast_ss_ps_128, INTR_TYPE_1OP_MASK, + X86_INTRINSIC_DATA(avx512_mask_broadcastf32x2_512, BRCST32x2_TO_VEC, X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcast_ss_ps_256, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcast_ss_ps_512, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcastf32x2_256, INTR_TYPE_1OP_MASK, - X86ISD::SUBV_BROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcastf32x2_512, INTR_TYPE_1OP_MASK, - X86ISD::SUBV_BROADCAST, 0), X86_INTRINSIC_DATA(avx512_mask_broadcastf32x4_256, BRCST_SUBVEC_TO_VEC, X86ISD::SHUF128, 0), X86_INTRINSIC_DATA(avx512_mask_broadcastf32x4_512, BRCST_SUBVEC_TO_VEC, @@ -493,12 +417,12 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::SHUF128, 0), X86_INTRINSIC_DATA(avx512_mask_broadcastf64x4_512, BRCST_SUBVEC_TO_VEC, X86ISD::SHUF128, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcasti32x2_128, INTR_TYPE_1OP_MASK, - X86ISD::SUBV_BROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcasti32x2_256, INTR_TYPE_1OP_MASK, - X86ISD::SUBV_BROADCAST, 0), - X86_INTRINSIC_DATA(avx512_mask_broadcasti32x2_512, INTR_TYPE_1OP_MASK, - X86ISD::SUBV_BROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_broadcasti32x2_128, BRCST32x2_TO_VEC, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_broadcasti32x2_256, BRCST32x2_TO_VEC, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_broadcasti32x2_512, BRCST32x2_TO_VEC, + X86ISD::VBROADCAST, 0), X86_INTRINSIC_DATA(avx512_mask_broadcasti32x4_256, BRCST_SUBVEC_TO_VEC, X86ISD::SHUF128, 0), X86_INTRINSIC_DATA(avx512_mask_broadcasti32x4_512, BRCST_SUBVEC_TO_VEC, @@ -773,6 +697,14 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::EXPAND, 0), X86_INTRINSIC_DATA(avx512_mask_expand_q_512, COMPRESS_EXPAND_IN_REG, X86ISD::EXPAND, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_pd_128, FIXUPIMM, X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_pd_256, FIXUPIMM, X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_pd_512, FIXUPIMM, X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_ps_128, FIXUPIMM, X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_ps_256, FIXUPIMM, X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_ps_512, FIXUPIMM, X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_sd, FIXUPIMMS, X86ISD::VFIXUPIMMS, 0), + X86_INTRINSIC_DATA(avx512_mask_fixupimm_ss, FIXUPIMMS, X86ISD::VFIXUPIMMS, 0), X86_INTRINSIC_DATA(avx512_mask_fpclass_pd_128, FPCLASS, X86ISD::VFPCLASS, 0), X86_INTRINSIC_DATA(avx512_mask_fpclass_pd_256, FPCLASS, X86ISD::VFPCLASS, 0), X86_INTRINSIC_DATA(avx512_mask_fpclass_pd_512, FPCLASS, X86ISD::VFPCLASS, 0), @@ -873,28 +805,10 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::FMIN, X86ISD::FMIN_RND), X86_INTRINSIC_DATA(avx512_mask_min_ss_round, INTR_TYPE_SCALAR_MASK_RM, X86ISD::FMIN, X86ISD::FMIN_RND), - X86_INTRINSIC_DATA(avx512_mask_movddup_128, INTR_TYPE_1OP_MASK, - X86ISD::MOVDDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_movddup_256, INTR_TYPE_1OP_MASK, - X86ISD::MOVDDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_movddup_512, INTR_TYPE_1OP_MASK, - X86ISD::MOVDDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_move_sd, INTR_TYPE_SCALAR_MASK, + X86_INTRINSIC_DATA(avx512_mask_move_sd, INTR_TYPE_SCALAR_MASK, X86ISD::MOVSD, 0), - X86_INTRINSIC_DATA(avx512_mask_move_ss, INTR_TYPE_SCALAR_MASK, + X86_INTRINSIC_DATA(avx512_mask_move_ss, INTR_TYPE_SCALAR_MASK, X86ISD::MOVSS, 0), - X86_INTRINSIC_DATA(avx512_mask_movshdup_128, INTR_TYPE_1OP_MASK, - X86ISD::MOVSHDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_movshdup_256, INTR_TYPE_1OP_MASK, - X86ISD::MOVSHDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_movshdup_512, INTR_TYPE_1OP_MASK, - X86ISD::MOVSHDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_movsldup_128, INTR_TYPE_1OP_MASK, - X86ISD::MOVSLDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_movsldup_256, INTR_TYPE_1OP_MASK, - X86ISD::MOVSLDUP, 0), - X86_INTRINSIC_DATA(avx512_mask_movsldup_512, INTR_TYPE_1OP_MASK, - X86ISD::MOVSLDUP, 0), X86_INTRINSIC_DATA(avx512_mask_mul_pd_128, INTR_TYPE_2OP_MASK, ISD::FMUL, 0), X86_INTRINSIC_DATA(avx512_mask_mul_pd_256, INTR_TYPE_2OP_MASK, ISD::FMUL, 0), X86_INTRINSIC_DATA(avx512_mask_mul_pd_512, INTR_TYPE_2OP_MASK, ISD::FMUL, @@ -961,54 +875,64 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_paddus_w_128, INTR_TYPE_2OP_MASK, X86ISD::ADDUS, 0), X86_INTRINSIC_DATA(avx512_mask_paddus_w_256, INTR_TYPE_2OP_MASK, X86ISD::ADDUS, 0), X86_INTRINSIC_DATA(avx512_mask_paddus_w_512, INTR_TYPE_2OP_MASK, X86ISD::ADDUS, 0), - X86_INTRINSIC_DATA(avx512_mask_palignr_128, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::PALIGNR, 0), - X86_INTRINSIC_DATA(avx512_mask_palignr_256, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::PALIGNR, 0), - X86_INTRINSIC_DATA(avx512_mask_palignr_512, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::PALIGNR, 0), - X86_INTRINSIC_DATA(avx512_mask_pand_d_128, INTR_TYPE_2OP_MASK, ISD::AND, 0), - X86_INTRINSIC_DATA(avx512_mask_pand_d_256, INTR_TYPE_2OP_MASK, ISD::AND, 0), - X86_INTRINSIC_DATA(avx512_mask_pand_d_512, INTR_TYPE_2OP_MASK, ISD::AND, 0), - X86_INTRINSIC_DATA(avx512_mask_pand_q_128, INTR_TYPE_2OP_MASK, ISD::AND, 0), - X86_INTRINSIC_DATA(avx512_mask_pand_q_256, INTR_TYPE_2OP_MASK, ISD::AND, 0), - X86_INTRINSIC_DATA(avx512_mask_pand_q_512, INTR_TYPE_2OP_MASK, ISD::AND, 0), - X86_INTRINSIC_DATA(avx512_mask_pandn_d_128, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0), - X86_INTRINSIC_DATA(avx512_mask_pandn_d_256, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0), - X86_INTRINSIC_DATA(avx512_mask_pandn_d_512, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0), - X86_INTRINSIC_DATA(avx512_mask_pandn_q_128, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0), - X86_INTRINSIC_DATA(avx512_mask_pandn_q_256, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0), - X86_INTRINSIC_DATA(avx512_mask_pandn_q_512, INTR_TYPE_2OP_MASK, X86ISD::ANDNP, 0), X86_INTRINSIC_DATA(avx512_mask_pavg_b_128, INTR_TYPE_2OP_MASK, X86ISD::AVG, 0), X86_INTRINSIC_DATA(avx512_mask_pavg_b_256, INTR_TYPE_2OP_MASK, X86ISD::AVG, 0), X86_INTRINSIC_DATA(avx512_mask_pavg_b_512, INTR_TYPE_2OP_MASK, X86ISD::AVG, 0), X86_INTRINSIC_DATA(avx512_mask_pavg_w_128, INTR_TYPE_2OP_MASK, X86ISD::AVG, 0), X86_INTRINSIC_DATA(avx512_mask_pavg_w_256, INTR_TYPE_2OP_MASK, X86ISD::AVG, 0), X86_INTRINSIC_DATA(avx512_mask_pavg_w_512, INTR_TYPE_2OP_MASK, X86ISD::AVG, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_b_128, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_b_256, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_b_512, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_d_128, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_d_256, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_d_512, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_q_128, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_q_256, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_q_512, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_w_128, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_w_256, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpeq_w_512, CMP_MASK, X86ISD::PCMPEQM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_b_128, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_b_256, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_b_512, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_d_128, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_d_256, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_d_512, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_q_128, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_q_256, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_q_512, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_w_128, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_w_256, CMP_MASK, X86ISD::PCMPGTM, 0), - X86_INTRINSIC_DATA(avx512_mask_pcmpgt_w_512, CMP_MASK, X86ISD::PCMPGTM, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_b_gpr_128, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_b_gpr_256, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_b_gpr_512, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_d_gpr_128, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_d_gpr_256, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_d_gpr_512, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_q_gpr_128, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_q_gpr_256, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_q_gpr_512, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_w_gpr_128, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_w_gpr_256, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_pbroadcast_w_gpr_512, INTR_TYPE_1OP_MASK, + X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_df_256, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_df_512, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_di_256, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_di_512, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_hi_128, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_hi_256, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_hi_512, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_qi_128, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_qi_256, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_qi_512, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_sf_256, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_sf_512, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_si_256, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), + X86_INTRINSIC_DATA(avx512_mask_permvar_si_512, VPERM_2OP_MASK, + X86ISD::VPERMV, 0), X86_INTRINSIC_DATA(avx512_mask_pmaddubs_w_128, INTR_TYPE_2OP_MASK, X86ISD::VPMADDUBSW, 0), X86_INTRINSIC_DATA(avx512_mask_pmaddubs_w_256, INTR_TYPE_2OP_MASK, @@ -1273,36 +1197,36 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_pmull_w_128, INTR_TYPE_2OP_MASK, ISD::MUL, 0), X86_INTRINSIC_DATA(avx512_mask_pmull_w_256, INTR_TYPE_2OP_MASK, ISD::MUL, 0), X86_INTRINSIC_DATA(avx512_mask_pmull_w_512, INTR_TYPE_2OP_MASK, ISD::MUL, 0), + X86_INTRINSIC_DATA(avx512_mask_pmultishift_qb_128, INTR_TYPE_2OP_MASK, + X86ISD::MULTISHIFT, 0), + X86_INTRINSIC_DATA(avx512_mask_pmultishift_qb_256, INTR_TYPE_2OP_MASK, + X86ISD::MULTISHIFT, 0), + X86_INTRINSIC_DATA(avx512_mask_pmultishift_qb_512, INTR_TYPE_2OP_MASK, + X86ISD::MULTISHIFT, 0), X86_INTRINSIC_DATA(avx512_mask_pmulu_dq_128, INTR_TYPE_2OP_MASK, X86ISD::PMULUDQ, 0), X86_INTRINSIC_DATA(avx512_mask_pmulu_dq_256, INTR_TYPE_2OP_MASK, X86ISD::PMULUDQ, 0), X86_INTRINSIC_DATA(avx512_mask_pmulu_dq_512, INTR_TYPE_2OP_MASK, X86ISD::PMULUDQ, 0), - X86_INTRINSIC_DATA(avx512_mask_por_d_128, INTR_TYPE_2OP_MASK, ISD::OR, 0), - X86_INTRINSIC_DATA(avx512_mask_por_d_256, INTR_TYPE_2OP_MASK, ISD::OR, 0), - X86_INTRINSIC_DATA(avx512_mask_por_d_512, INTR_TYPE_2OP_MASK, ISD::OR, 0), - X86_INTRINSIC_DATA(avx512_mask_por_q_128, INTR_TYPE_2OP_MASK, ISD::OR, 0), - X86_INTRINSIC_DATA(avx512_mask_por_q_256, INTR_TYPE_2OP_MASK, ISD::OR, 0), - X86_INTRINSIC_DATA(avx512_mask_por_q_512, INTR_TYPE_2OP_MASK, ISD::OR, 0), - X86_INTRINSIC_DATA(avx512_mask_prol_d_128, INTR_TYPE_2OP_MASK, X86ISD::VROTLI, 0), - X86_INTRINSIC_DATA(avx512_mask_prol_d_256, INTR_TYPE_2OP_MASK, X86ISD::VROTLI, 0), - X86_INTRINSIC_DATA(avx512_mask_prol_d_512, INTR_TYPE_2OP_MASK, X86ISD::VROTLI, 0), - X86_INTRINSIC_DATA(avx512_mask_prol_q_128, INTR_TYPE_2OP_MASK, X86ISD::VROTLI, 0), - X86_INTRINSIC_DATA(avx512_mask_prol_q_256, INTR_TYPE_2OP_MASK, X86ISD::VROTLI, 0), - X86_INTRINSIC_DATA(avx512_mask_prol_q_512, INTR_TYPE_2OP_MASK, X86ISD::VROTLI, 0), + X86_INTRINSIC_DATA(avx512_mask_prol_d_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTLI, 0), + X86_INTRINSIC_DATA(avx512_mask_prol_d_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTLI, 0), + X86_INTRINSIC_DATA(avx512_mask_prol_d_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTLI, 0), + X86_INTRINSIC_DATA(avx512_mask_prol_q_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTLI, 0), + X86_INTRINSIC_DATA(avx512_mask_prol_q_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTLI, 0), + X86_INTRINSIC_DATA(avx512_mask_prol_q_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTLI, 0), X86_INTRINSIC_DATA(avx512_mask_prolv_d_128, INTR_TYPE_2OP_MASK, ISD::ROTL, 0), X86_INTRINSIC_DATA(avx512_mask_prolv_d_256, INTR_TYPE_2OP_MASK, ISD::ROTL, 0), X86_INTRINSIC_DATA(avx512_mask_prolv_d_512, INTR_TYPE_2OP_MASK, ISD::ROTL, 0), X86_INTRINSIC_DATA(avx512_mask_prolv_q_128, INTR_TYPE_2OP_MASK, ISD::ROTL, 0), X86_INTRINSIC_DATA(avx512_mask_prolv_q_256, INTR_TYPE_2OP_MASK, ISD::ROTL, 0), X86_INTRINSIC_DATA(avx512_mask_prolv_q_512, INTR_TYPE_2OP_MASK, ISD::ROTL, 0), - X86_INTRINSIC_DATA(avx512_mask_pror_d_128, INTR_TYPE_2OP_MASK, X86ISD::VROTRI, 0), - X86_INTRINSIC_DATA(avx512_mask_pror_d_256, INTR_TYPE_2OP_MASK, X86ISD::VROTRI, 0), - X86_INTRINSIC_DATA(avx512_mask_pror_d_512, INTR_TYPE_2OP_MASK, X86ISD::VROTRI, 0), - X86_INTRINSIC_DATA(avx512_mask_pror_q_128, INTR_TYPE_2OP_MASK, X86ISD::VROTRI, 0), - X86_INTRINSIC_DATA(avx512_mask_pror_q_256, INTR_TYPE_2OP_MASK, X86ISD::VROTRI, 0), - X86_INTRINSIC_DATA(avx512_mask_pror_q_512, INTR_TYPE_2OP_MASK, X86ISD::VROTRI, 0), + X86_INTRINSIC_DATA(avx512_mask_pror_d_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTRI, 0), + X86_INTRINSIC_DATA(avx512_mask_pror_d_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTRI, 0), + X86_INTRINSIC_DATA(avx512_mask_pror_d_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTRI, 0), + X86_INTRINSIC_DATA(avx512_mask_pror_q_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTRI, 0), + X86_INTRINSIC_DATA(avx512_mask_pror_q_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTRI, 0), + X86_INTRINSIC_DATA(avx512_mask_pror_q_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VROTRI, 0), X86_INTRINSIC_DATA(avx512_mask_prorv_d_128, INTR_TYPE_2OP_MASK, ISD::ROTR, 0), X86_INTRINSIC_DATA(avx512_mask_prorv_d_256, INTR_TYPE_2OP_MASK, ISD::ROTR, 0), X86_INTRINSIC_DATA(avx512_mask_prorv_d_512, INTR_TYPE_2OP_MASK, ISD::ROTR, 0), @@ -1315,44 +1239,26 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::PSHUFB, 0), X86_INTRINSIC_DATA(avx512_mask_pshuf_b_512, INTR_TYPE_2OP_MASK, X86ISD::PSHUFB, 0), - X86_INTRINSIC_DATA(avx512_mask_pshuf_d_128, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFD, 0), - X86_INTRINSIC_DATA(avx512_mask_pshuf_d_256, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFD, 0), - X86_INTRINSIC_DATA(avx512_mask_pshuf_d_512, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFD, 0), - X86_INTRINSIC_DATA(avx512_mask_pshufh_w_128, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFHW, 0), - X86_INTRINSIC_DATA(avx512_mask_pshufh_w_256, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFHW, 0), - X86_INTRINSIC_DATA(avx512_mask_pshufh_w_512, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFHW, 0), - X86_INTRINSIC_DATA(avx512_mask_pshufl_w_128, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFLW, 0), - X86_INTRINSIC_DATA(avx512_mask_pshufl_w_256, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFLW, 0), - X86_INTRINSIC_DATA(avx512_mask_pshufl_w_512, INTR_TYPE_2OP_MASK, - X86ISD::PSHUFLW, 0), X86_INTRINSIC_DATA(avx512_mask_psll_d, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx512_mask_psll_d_128, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx512_mask_psll_d_256, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_di_128, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_di_256, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_di_512, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_di_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_di_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_di_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), X86_INTRINSIC_DATA(avx512_mask_psll_q, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx512_mask_psll_q_128, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx512_mask_psll_q_256, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_qi_128, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_qi_256, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_qi_512, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_qi_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_qi_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_qi_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), X86_INTRINSIC_DATA(avx512_mask_psll_w_128, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx512_mask_psll_w_256, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(avx512_mask_psll_w_512, INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_wi_128, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_wi_256, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psll_wi_512, INTR_TYPE_2OP_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_pslli_d, VSHIFT_MASK, X86ISD::VSHLI, 0), - X86_INTRINSIC_DATA(avx512_mask_pslli_q, VSHIFT_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_wi_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_wi_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psll_wi_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSHLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psllv_d, INTR_TYPE_2OP_MASK, ISD::SHL, 0), + X86_INTRINSIC_DATA(avx512_mask_psllv_q, INTR_TYPE_2OP_MASK, ISD::SHL, 0), X86_INTRINSIC_DATA(avx512_mask_psllv16_hi, INTR_TYPE_2OP_MASK, ISD::SHL, 0), X86_INTRINSIC_DATA(avx512_mask_psllv2_di, INTR_TYPE_2OP_MASK, ISD::SHL, 0), X86_INTRINSIC_DATA(avx512_mask_psllv32hi, INTR_TYPE_2OP_MASK, ISD::SHL, 0), @@ -1360,57 +1266,53 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_psllv4_si, INTR_TYPE_2OP_MASK, ISD::SHL, 0), X86_INTRINSIC_DATA(avx512_mask_psllv8_hi, INTR_TYPE_2OP_MASK, ISD::SHL, 0), X86_INTRINSIC_DATA(avx512_mask_psllv8_si, INTR_TYPE_2OP_MASK, ISD::SHL, 0), - X86_INTRINSIC_DATA(avx512_mask_psllv_d, INTR_TYPE_2OP_MASK, ISD::SHL, 0), - X86_INTRINSIC_DATA(avx512_mask_psllv_q, INTR_TYPE_2OP_MASK, ISD::SHL, 0), X86_INTRINSIC_DATA(avx512_mask_psra_d, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx512_mask_psra_d_128, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx512_mask_psra_d_256, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_di_128, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_di_256, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_di_512, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_di_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_di_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_di_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), X86_INTRINSIC_DATA(avx512_mask_psra_q, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx512_mask_psra_q_128, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx512_mask_psra_q_256, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_qi_128, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_qi_256, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_qi_512, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_qi_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_qi_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_qi_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), X86_INTRINSIC_DATA(avx512_mask_psra_w_128, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx512_mask_psra_w_256, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx512_mask_psra_w_512, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_wi_128, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_wi_256, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psra_wi_512, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrai_d, VSHIFT_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrai_q, VSHIFT_MASK, X86ISD::VSRAI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav16_hi, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav32_hi, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav4_si, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav8_hi, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav8_si, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav_d, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav_q, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav_q_128, INTR_TYPE_2OP_MASK, ISD::SRA, 0), - X86_INTRINSIC_DATA(avx512_mask_psrav_q_256, INTR_TYPE_2OP_MASK, ISD::SRA, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_wi_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_wi_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_wi_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav_d, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav_q, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav_q_128, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav_q_256, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav16_hi, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav32_hi, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav4_si, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav8_hi, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), + X86_INTRINSIC_DATA(avx512_mask_psrav8_si, INTR_TYPE_2OP_MASK, X86ISD::VSRAV, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_d, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_d_128, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_d_256, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_di_128, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_di_256, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_di_512, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_di_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_di_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_di_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_q, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_q_128, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_q_256, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_qi_128, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_qi_256, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_qi_512, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_qi_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_qi_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_qi_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_w_128, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_w_256, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrl_w_512, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_wi_128, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_wi_256, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrl_wi_512, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrli_d, VSHIFT_MASK, X86ISD::VSRLI, 0), - X86_INTRINSIC_DATA(avx512_mask_psrli_q, VSHIFT_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_wi_128, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_wi_256, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrl_wi_512, INTR_TYPE_2OP_IMM8_MASK, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(avx512_mask_psrlv_d, INTR_TYPE_2OP_MASK, ISD::SRL, 0), + X86_INTRINSIC_DATA(avx512_mask_psrlv_q, INTR_TYPE_2OP_MASK, ISD::SRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrlv16_hi, INTR_TYPE_2OP_MASK, ISD::SRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrlv2_di, INTR_TYPE_2OP_MASK, ISD::SRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrlv32hi, INTR_TYPE_2OP_MASK, ISD::SRL, 0), @@ -1418,8 +1320,6 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_psrlv4_si, INTR_TYPE_2OP_MASK, ISD::SRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrlv8_hi, INTR_TYPE_2OP_MASK, ISD::SRL, 0), X86_INTRINSIC_DATA(avx512_mask_psrlv8_si, INTR_TYPE_2OP_MASK, ISD::SRL, 0), - X86_INTRINSIC_DATA(avx512_mask_psrlv_d, INTR_TYPE_2OP_MASK, ISD::SRL, 0), - X86_INTRINSIC_DATA(avx512_mask_psrlv_q, INTR_TYPE_2OP_MASK, ISD::SRL, 0), X86_INTRINSIC_DATA(avx512_mask_psub_b_128, INTR_TYPE_2OP_MASK, ISD::SUB, 0), X86_INTRINSIC_DATA(avx512_mask_psub_b_256, INTR_TYPE_2OP_MASK, ISD::SUB, 0), X86_INTRINSIC_DATA(avx512_mask_psub_b_512, INTR_TYPE_2OP_MASK, ISD::SUB, 0), @@ -1456,60 +1356,6 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::VPTERNLOG, 0), X86_INTRINSIC_DATA(avx512_mask_pternlog_q_512, TERLOG_OP_MASK, X86ISD::VPTERNLOG, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhb_w_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhb_w_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhb_w_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhd_q_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhd_q_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhd_q_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhqd_q_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhqd_q_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhqd_q_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhw_d_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhw_d_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckhw_d_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklb_w_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklb_w_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklb_w_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckld_q_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckld_q_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpckld_q_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklqd_q_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklqd_q_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklqd_q_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklw_d_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklw_d_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_punpcklw_d_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_pxor_d_128, INTR_TYPE_2OP_MASK, ISD::XOR, 0), - X86_INTRINSIC_DATA(avx512_mask_pxor_d_256, INTR_TYPE_2OP_MASK, ISD::XOR, 0), - X86_INTRINSIC_DATA(avx512_mask_pxor_d_512, INTR_TYPE_2OP_MASK, ISD::XOR, 0), - X86_INTRINSIC_DATA(avx512_mask_pxor_q_128, INTR_TYPE_2OP_MASK, ISD::XOR, 0), - X86_INTRINSIC_DATA(avx512_mask_pxor_q_256, INTR_TYPE_2OP_MASK, ISD::XOR, 0), - X86_INTRINSIC_DATA(avx512_mask_pxor_q_512, INTR_TYPE_2OP_MASK, ISD::XOR, 0), X86_INTRINSIC_DATA(avx512_mask_range_pd_128, INTR_TYPE_3OP_MASK_RM, X86ISD::VRANGE, 0), X86_INTRINSIC_DATA(avx512_mask_range_pd_256, INTR_TYPE_3OP_MASK_RM, X86ISD::VRANGE, 0), X86_INTRINSIC_DATA(avx512_mask_range_pd_512, INTR_TYPE_3OP_MASK_RM, X86ISD::VRANGE, 0), @@ -1549,9 +1395,9 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_scalef_ps_512, INTR_TYPE_2OP_MASK_RM, X86ISD::SCALEF, 0), X86_INTRINSIC_DATA(avx512_mask_scalef_sd, INTR_TYPE_SCALAR_MASK_RM, - X86ISD::SCALEF, 0), + X86ISD::SCALEFS, 0), X86_INTRINSIC_DATA(avx512_mask_scalef_ss, INTR_TYPE_SCALAR_MASK_RM, - X86ISD::SCALEF, 0), + X86ISD::SCALEFS, 0), X86_INTRINSIC_DATA(avx512_mask_shuf_f32x4, INTR_TYPE_3OP_IMM8_MASK, X86ISD::SHUF128, 0), X86_INTRINSIC_DATA(avx512_mask_shuf_f32x4_256, INTR_TYPE_3OP_IMM8_MASK, @@ -1616,30 +1462,6 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_ucmp_w_128, CMP_MASK_CC, X86ISD::CMPMU, 0), X86_INTRINSIC_DATA(avx512_mask_ucmp_w_256, CMP_MASK_CC, X86ISD::CMPMU, 0), X86_INTRINSIC_DATA(avx512_mask_ucmp_w_512, CMP_MASK_CC, X86ISD::CMPMU, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckh_pd_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckh_pd_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckh_pd_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckh_ps_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckh_ps_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckh_ps_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKH, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckl_pd_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckl_pd_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckl_pd_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckl_ps_128, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckl_ps_256, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), - X86_INTRINSIC_DATA(avx512_mask_unpckl_ps_512, INTR_TYPE_2OP_MASK, - X86ISD::UNPCKL, 0), X86_INTRINSIC_DATA(avx512_mask_valign_d_128, INTR_TYPE_3OP_IMM8_MASK, X86ISD::VALIGN, 0), X86_INTRINSIC_DATA(avx512_mask_valign_d_256, INTR_TYPE_3OP_IMM8_MASK, @@ -1673,6 +1495,8 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_vfmadd_ps_512, FMA_OP_MASK, X86ISD::FMADD, X86ISD::FMADD_RND), + X86_INTRINSIC_DATA(avx512_mask_vfmadd_sd, FMA_OP_SCALAR_MASK, X86ISD::FMADD_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_vfmadd_ss, FMA_OP_SCALAR_MASK, X86ISD::FMADD_RND, 0), X86_INTRINSIC_DATA(avx512_mask_vfmaddsub_pd_128, FMA_OP_MASK, X86ISD::FMADDSUB, 0), X86_INTRINSIC_DATA(avx512_mask_vfmaddsub_pd_256, FMA_OP_MASK, X86ISD::FMADDSUB, 0), X86_INTRINSIC_DATA(avx512_mask_vfmaddsub_pd_512, FMA_OP_MASK, X86ISD::FMADDSUB, @@ -1730,18 +1554,12 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::VPERMIV3, 0), X86_INTRINSIC_DATA(avx512_mask_vpermi2var_q_512, VPERM_3OP_MASK, X86ISD::VPERMIV3, 0), - X86_INTRINSIC_DATA(avx512_mask_vpermil_pd_128, INTR_TYPE_2OP_IMM8_MASK, - X86ISD::VPERMILPI, 0), - X86_INTRINSIC_DATA(avx512_mask_vpermil_pd_256, INTR_TYPE_2OP_IMM8_MASK, - X86ISD::VPERMILPI, 0), - X86_INTRINSIC_DATA(avx512_mask_vpermil_pd_512, INTR_TYPE_2OP_IMM8_MASK, - X86ISD::VPERMILPI, 0), - X86_INTRINSIC_DATA(avx512_mask_vpermil_ps_128, INTR_TYPE_2OP_IMM8_MASK, - X86ISD::VPERMILPI, 0), - X86_INTRINSIC_DATA(avx512_mask_vpermil_ps_256, INTR_TYPE_2OP_IMM8_MASK, - X86ISD::VPERMILPI, 0), - X86_INTRINSIC_DATA(avx512_mask_vpermil_ps_512, INTR_TYPE_2OP_IMM8_MASK, - X86ISD::VPERMILPI, 0), + X86_INTRINSIC_DATA(avx512_mask_vpermi2var_qi_128, VPERM_3OP_MASK, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_mask_vpermi2var_qi_256, VPERM_3OP_MASK, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_mask_vpermi2var_qi_512, VPERM_3OP_MASK, + X86ISD::VPERMV3, 0), X86_INTRINSIC_DATA(avx512_mask_vpermilvar_pd_128, INTR_TYPE_2OP_MASK, X86ISD::VPERMILPV, 0), X86_INTRINSIC_DATA(avx512_mask_vpermilvar_pd_256, INTR_TYPE_2OP_MASK, @@ -1784,12 +1602,92 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::VPERMV3, 0), X86_INTRINSIC_DATA(avx512_mask_vpermt2var_q_512, VPERM_3OP_MASK, X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_mask_vpermt2var_qi_128, VPERM_3OP_MASK, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_mask_vpermt2var_qi_256, VPERM_3OP_MASK, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_mask_vpermt2var_qi_512, VPERM_3OP_MASK, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_mask_vpmadd52h_uq_128 , FMA_OP_MASK, + X86ISD::VPMADD52H, 0), + X86_INTRINSIC_DATA(avx512_mask_vpmadd52h_uq_256 , FMA_OP_MASK, + X86ISD::VPMADD52H, 0), + X86_INTRINSIC_DATA(avx512_mask_vpmadd52h_uq_512 , FMA_OP_MASK, + X86ISD::VPMADD52H, 0), + X86_INTRINSIC_DATA(avx512_mask_vpmadd52l_uq_128 , FMA_OP_MASK, + X86ISD::VPMADD52L, 0), + X86_INTRINSIC_DATA(avx512_mask_vpmadd52l_uq_256 , FMA_OP_MASK, + X86ISD::VPMADD52L, 0), + X86_INTRINSIC_DATA(avx512_mask_vpmadd52l_uq_512 , FMA_OP_MASK, + X86ISD::VPMADD52L, 0), X86_INTRINSIC_DATA(avx512_mask_xor_pd_128, INTR_TYPE_2OP_MASK, X86ISD::FXOR, 0), X86_INTRINSIC_DATA(avx512_mask_xor_pd_256, INTR_TYPE_2OP_MASK, X86ISD::FXOR, 0), X86_INTRINSIC_DATA(avx512_mask_xor_pd_512, INTR_TYPE_2OP_MASK, X86ISD::FXOR, 0), X86_INTRINSIC_DATA(avx512_mask_xor_ps_128, INTR_TYPE_2OP_MASK, X86ISD::FXOR, 0), X86_INTRINSIC_DATA(avx512_mask_xor_ps_256, INTR_TYPE_2OP_MASK, X86ISD::FXOR, 0), X86_INTRINSIC_DATA(avx512_mask_xor_ps_512, INTR_TYPE_2OP_MASK, X86ISD::FXOR, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_128, FMA_OP_MASK3, X86ISD::FMADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_256, FMA_OP_MASK3, X86ISD::FMADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_512, FMA_OP_MASK3, X86ISD::FMADD, + X86ISD::FMADD_RND), + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_128, FMA_OP_MASK3, X86ISD::FMADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_256, FMA_OP_MASK3, X86ISD::FMADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_512, FMA_OP_MASK3, X86ISD::FMADD, + X86ISD::FMADD_RND), + + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_sd, FMA_OP_SCALAR_MASK3, X86ISD::FMADD_RND, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ss, FMA_OP_SCALAR_MASK3, X86ISD::FMADD_RND, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_pd_128, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_pd_256, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_pd_512, FMA_OP_MASK3, X86ISD::FMADDSUB, + X86ISD::FMADDSUB_RND), + X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_ps_128, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_ps_256, FMA_OP_MASK3, X86ISD::FMADDSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmaddsub_ps_512, FMA_OP_MASK3, X86ISD::FMADDSUB, + X86ISD::FMADDSUB_RND), + + X86_INTRINSIC_DATA(avx512_mask3_vfmsub_pd_128, FMA_OP_MASK3, X86ISD::FMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsub_pd_256, FMA_OP_MASK3, X86ISD::FMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsub_pd_512, FMA_OP_MASK3, X86ISD::FMSUB, + X86ISD::FMSUB_RND), + X86_INTRINSIC_DATA(avx512_mask3_vfmsub_ps_128, FMA_OP_MASK3, X86ISD::FMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsub_ps_256, FMA_OP_MASK3, X86ISD::FMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsub_ps_512, FMA_OP_MASK3, X86ISD::FMSUB, + X86ISD::FMSUB_RND), + + X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_pd_128, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_pd_256, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_pd_512, FMA_OP_MASK3, X86ISD::FMSUBADD, + X86ISD::FMSUBADD_RND), + X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_ps_128, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_ps_256, FMA_OP_MASK3, X86ISD::FMSUBADD, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfmsubadd_ps_512, FMA_OP_MASK3, X86ISD::FMSUBADD, + X86ISD::FMSUBADD_RND), + + X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_pd_128, FMA_OP_MASK3, X86ISD::FNMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_pd_256, FMA_OP_MASK3, X86ISD::FNMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_pd_512, FMA_OP_MASK3, X86ISD::FNMSUB, + X86ISD::FNMSUB_RND), + X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_ps_128, FMA_OP_MASK3, X86ISD::FNMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_ps_256, FMA_OP_MASK3, X86ISD::FNMSUB, 0), + X86_INTRINSIC_DATA(avx512_mask3_vfnmsub_ps_512, FMA_OP_MASK3, X86ISD::FNMSUB, + X86ISD::FNMSUB_RND), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_pd_128, FIXUPIMM_MASKZ, + X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_pd_256, FIXUPIMM_MASKZ, + X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_pd_512, FIXUPIMM_MASKZ, + X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_ps_128, FIXUPIMM_MASKZ, + X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_ps_256, FIXUPIMM_MASKZ, + X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_ps_512, FIXUPIMM_MASKZ, + X86ISD::VFIXUPIMM, 0), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_sd, FIXUPIMMS_MASKZ, + X86ISD::VFIXUPIMMS, 0), + X86_INTRINSIC_DATA(avx512_maskz_fixupimm_ss, FIXUPIMMS_MASKZ, + X86ISD::VFIXUPIMMS, 0), X86_INTRINSIC_DATA(avx512_maskz_pternlog_d_128, TERLOG_OP_MASKZ, X86ISD::VPTERNLOG, 0), X86_INTRINSIC_DATA(avx512_maskz_pternlog_d_256, TERLOG_OP_MASKZ, @@ -1811,6 +1709,8 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ps_512, FMA_OP_MASKZ, X86ISD::FMADD, X86ISD::FMADD_RND), + X86_INTRINSIC_DATA(avx512_maskz_vfmadd_sd, FMA_OP_SCALAR_MASKZ, X86ISD::FMADD_RND, 0), + X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ss, FMA_OP_SCALAR_MASKZ, X86ISD::FMADD_RND, 0), X86_INTRINSIC_DATA(avx512_maskz_vfmaddsub_pd_128, FMA_OP_MASKZ, X86ISD::FMADDSUB, 0), X86_INTRINSIC_DATA(avx512_maskz_vfmaddsub_pd_256, FMA_OP_MASKZ, X86ISD::FMADDSUB, 0), X86_INTRINSIC_DATA(avx512_maskz_vfmaddsub_pd_512, FMA_OP_MASKZ, X86ISD::FMADDSUB, @@ -1850,41 +1750,57 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::VPERMV3, 0), X86_INTRINSIC_DATA(avx512_maskz_vpermt2var_q_512, VPERM_3OP_MASKZ, X86ISD::VPERMV3, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastb_128, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastb_256, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastb_512, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastd_128, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastd_256, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastd_512, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastq_128, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastq_256, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastq_512, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastw_128, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastw_256, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), - X86_INTRINSIC_DATA(avx512_pbroadcastw_512, INTR_TYPE_1OP_MASK, - X86ISD::VBROADCAST, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpermt2var_qi_128, VPERM_3OP_MASKZ, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpermt2var_qi_256, VPERM_3OP_MASKZ, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpermt2var_qi_512, VPERM_3OP_MASKZ, + X86ISD::VPERMV3, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpmadd52h_uq_128, FMA_OP_MASKZ, + X86ISD::VPMADD52H, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpmadd52h_uq_256, FMA_OP_MASKZ, + X86ISD::VPMADD52H, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpmadd52h_uq_512, FMA_OP_MASKZ, + X86ISD::VPMADD52H, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpmadd52l_uq_128, FMA_OP_MASKZ, + X86ISD::VPMADD52L, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpmadd52l_uq_256, FMA_OP_MASKZ, + X86ISD::VPMADD52L, 0), + X86_INTRINSIC_DATA(avx512_maskz_vpmadd52l_uq_512, FMA_OP_MASKZ, + X86ISD::VPMADD52L, 0), X86_INTRINSIC_DATA(avx512_psad_bw_512, INTR_TYPE_2OP, X86ISD::PSADBW, 0), - X86_INTRINSIC_DATA(avx512_psll_dq_512, INTR_TYPE_2OP_IMM8, X86ISD::VSHLDQ, 0), - X86_INTRINSIC_DATA(avx512_psrl_dq_512, INTR_TYPE_2OP_IMM8, X86ISD::VSRLDQ, 0), + X86_INTRINSIC_DATA(avx512_ptestm_b_128, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_b_256, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_b_512, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_d_128, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_d_256, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_d_512, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_q_128, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_q_256, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_q_512, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_w_128, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_w_256, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestm_w_512, CMP_MASK, X86ISD::TESTM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_b_128, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_b_256, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_b_512, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_d_128, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_d_256, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_d_512, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_q_128, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_q_256, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_q_512, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_w_128, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_w_256, CMP_MASK, X86ISD::TESTNM, 0), + X86_INTRINSIC_DATA(avx512_ptestnm_w_512, CMP_MASK, X86ISD::TESTNM, 0), X86_INTRINSIC_DATA(avx512_rcp14_pd_128, INTR_TYPE_1OP_MASK, X86ISD::FRCP, 0), X86_INTRINSIC_DATA(avx512_rcp14_pd_256, INTR_TYPE_1OP_MASK, X86ISD::FRCP, 0), X86_INTRINSIC_DATA(avx512_rcp14_pd_512, INTR_TYPE_1OP_MASK, X86ISD::FRCP, 0), X86_INTRINSIC_DATA(avx512_rcp14_ps_128, INTR_TYPE_1OP_MASK, X86ISD::FRCP, 0), X86_INTRINSIC_DATA(avx512_rcp14_ps_256, INTR_TYPE_1OP_MASK, X86ISD::FRCP, 0), X86_INTRINSIC_DATA(avx512_rcp14_ps_512, INTR_TYPE_1OP_MASK, X86ISD::FRCP, 0), - X86_INTRINSIC_DATA(avx512_rcp14_sd, INTR_TYPE_SCALAR_MASK, X86ISD::FRCP, 0), - X86_INTRINSIC_DATA(avx512_rcp14_ss, INTR_TYPE_SCALAR_MASK, X86ISD::FRCP, 0), + X86_INTRINSIC_DATA(avx512_rcp14_sd, INTR_TYPE_SCALAR_MASK, X86ISD::FRCPS, 0), + X86_INTRINSIC_DATA(avx512_rcp14_ss, INTR_TYPE_SCALAR_MASK, X86ISD::FRCPS, 0), X86_INTRINSIC_DATA(avx512_rcp28_pd, INTR_TYPE_1OP_MASK_RM, X86ISD::RCP28, 0), X86_INTRINSIC_DATA(avx512_rcp28_ps, INTR_TYPE_1OP_MASK_RM, X86ISD::RCP28, 0), X86_INTRINSIC_DATA(avx512_rcp28_sd, INTR_TYPE_SCALAR_MASK_RM, X86ISD::RCP28, 0), @@ -1895,29 +1811,30 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_rsqrt14_ps_128, INTR_TYPE_1OP_MASK, X86ISD::FRSQRT, 0), X86_INTRINSIC_DATA(avx512_rsqrt14_ps_256, INTR_TYPE_1OP_MASK, X86ISD::FRSQRT, 0), X86_INTRINSIC_DATA(avx512_rsqrt14_ps_512, INTR_TYPE_1OP_MASK, X86ISD::FRSQRT, 0), - X86_INTRINSIC_DATA(avx512_rsqrt14_sd, INTR_TYPE_SCALAR_MASK, X86ISD::FRSQRT, 0), - X86_INTRINSIC_DATA(avx512_rsqrt14_ss, INTR_TYPE_SCALAR_MASK, X86ISD::FRSQRT, 0), + X86_INTRINSIC_DATA(avx512_rsqrt14_sd, INTR_TYPE_SCALAR_MASK, X86ISD::FRSQRTS, 0), + X86_INTRINSIC_DATA(avx512_rsqrt14_ss, INTR_TYPE_SCALAR_MASK, X86ISD::FRSQRTS, 0), X86_INTRINSIC_DATA(avx512_rsqrt28_pd, INTR_TYPE_1OP_MASK_RM,X86ISD::RSQRT28, 0), X86_INTRINSIC_DATA(avx512_rsqrt28_ps, INTR_TYPE_1OP_MASK_RM,X86ISD::RSQRT28, 0), X86_INTRINSIC_DATA(avx512_rsqrt28_sd, INTR_TYPE_SCALAR_MASK_RM,X86ISD::RSQRT28, 0), X86_INTRINSIC_DATA(avx512_rsqrt28_ss, INTR_TYPE_SCALAR_MASK_RM,X86ISD::RSQRT28, 0), X86_INTRINSIC_DATA(avx512_vcomi_sd, COMI_RM, X86ISD::COMI, X86ISD::UCOMI), X86_INTRINSIC_DATA(avx512_vcomi_ss, COMI_RM, X86ISD::COMI, X86ISD::UCOMI), - X86_INTRINSIC_DATA(avx_hadd_pd_256, INTR_TYPE_2OP, X86ISD::FHADD, 0), - X86_INTRINSIC_DATA(avx_hadd_ps_256, INTR_TYPE_2OP, X86ISD::FHADD, 0), - X86_INTRINSIC_DATA(avx_hsub_pd_256, INTR_TYPE_2OP, X86ISD::FHSUB, 0), - X86_INTRINSIC_DATA(avx_hsub_ps_256, INTR_TYPE_2OP, X86ISD::FHSUB, 0), - X86_INTRINSIC_DATA(avx_max_pd_256, INTR_TYPE_2OP, X86ISD::FMAX, 0), - X86_INTRINSIC_DATA(avx_max_ps_256, INTR_TYPE_2OP, X86ISD::FMAX, 0), - X86_INTRINSIC_DATA(avx_min_pd_256, INTR_TYPE_2OP, X86ISD::FMIN, 0), - X86_INTRINSIC_DATA(avx_min_ps_256, INTR_TYPE_2OP, X86ISD::FMIN, 0), - X86_INTRINSIC_DATA(avx_rcp_ps_256, INTR_TYPE_1OP, X86ISD::FRCP, 0), - X86_INTRINSIC_DATA(avx_rsqrt_ps_256, INTR_TYPE_1OP, X86ISD::FRSQRT, 0), - X86_INTRINSIC_DATA(avx_sqrt_pd_256, INTR_TYPE_1OP, ISD::FSQRT, 0), - X86_INTRINSIC_DATA(avx_sqrt_ps_256, INTR_TYPE_1OP, ISD::FSQRT, 0), - X86_INTRINSIC_DATA(avx_vperm2f128_pd_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0), - X86_INTRINSIC_DATA(avx_vperm2f128_ps_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0), - X86_INTRINSIC_DATA(avx_vperm2f128_si_256, INTR_TYPE_3OP, X86ISD::VPERM2X128, 0), + X86_INTRINSIC_DATA(avx512_vcvtsd2si32, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_vcvtsd2si64, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_vcvtsd2usi32, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_vcvtsd2usi64, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_vcvtss2si32, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_vcvtss2si64, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_vcvtss2usi32, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_vcvtss2usi64, INTR_TYPE_2OP, + X86ISD::SCALAR_FP_TO_UINT_RND, 0), X86_INTRINSIC_DATA(fma_vfmadd_pd, INTR_TYPE_3OP, X86ISD::FMADD, 0), X86_INTRINSIC_DATA(fma_vfmadd_pd_256, INTR_TYPE_3OP, X86ISD::FMADD, 0), X86_INTRINSIC_DATA(fma_vfmadd_ps, INTR_TYPE_3OP, X86ISD::FMADD, 0), @@ -1942,6 +1859,24 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(fma_vfnmsub_pd_256, INTR_TYPE_3OP, X86ISD::FNMSUB, 0), X86_INTRINSIC_DATA(fma_vfnmsub_ps, INTR_TYPE_3OP, X86ISD::FNMSUB, 0), X86_INTRINSIC_DATA(fma_vfnmsub_ps_256, INTR_TYPE_3OP, X86ISD::FNMSUB, 0), + X86_INTRINSIC_DATA(sse_comieq_ss, COMI, X86ISD::COMI, ISD::SETEQ), + X86_INTRINSIC_DATA(sse_comige_ss, COMI, X86ISD::COMI, ISD::SETGE), + X86_INTRINSIC_DATA(sse_comigt_ss, COMI, X86ISD::COMI, ISD::SETGT), + X86_INTRINSIC_DATA(sse_comile_ss, COMI, X86ISD::COMI, ISD::SETLE), + X86_INTRINSIC_DATA(sse_comilt_ss, COMI, X86ISD::COMI, ISD::SETLT), + X86_INTRINSIC_DATA(sse_comineq_ss, COMI, X86ISD::COMI, ISD::SETNE), + X86_INTRINSIC_DATA(sse_max_ps, INTR_TYPE_2OP, X86ISD::FMAX, 0), + X86_INTRINSIC_DATA(sse_min_ps, INTR_TYPE_2OP, X86ISD::FMIN, 0), + X86_INTRINSIC_DATA(sse_movmsk_ps, INTR_TYPE_1OP, X86ISD::MOVMSK, 0), + X86_INTRINSIC_DATA(sse_rcp_ps, INTR_TYPE_1OP, X86ISD::FRCP, 0), + X86_INTRINSIC_DATA(sse_rsqrt_ps, INTR_TYPE_1OP, X86ISD::FRSQRT, 0), + X86_INTRINSIC_DATA(sse_sqrt_ps, INTR_TYPE_1OP, ISD::FSQRT, 0), + X86_INTRINSIC_DATA(sse_ucomieq_ss, COMI, X86ISD::UCOMI, ISD::SETEQ), + X86_INTRINSIC_DATA(sse_ucomige_ss, COMI, X86ISD::UCOMI, ISD::SETGE), + X86_INTRINSIC_DATA(sse_ucomigt_ss, COMI, X86ISD::UCOMI, ISD::SETGT), + X86_INTRINSIC_DATA(sse_ucomile_ss, COMI, X86ISD::UCOMI, ISD::SETLE), + X86_INTRINSIC_DATA(sse_ucomilt_ss, COMI, X86ISD::UCOMI, ISD::SETLT), + X86_INTRINSIC_DATA(sse_ucomineq_ss, COMI, X86ISD::UCOMI, ISD::SETNE), X86_INTRINSIC_DATA(sse2_comieq_sd, COMI, X86ISD::COMI, ISD::SETEQ), X86_INTRINSIC_DATA(sse2_comige_sd, COMI, X86ISD::COMI, ISD::SETGE), X86_INTRINSIC_DATA(sse2_comigt_sd, COMI, X86ISD::COMI, ISD::SETGT), @@ -1950,22 +1885,21 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(sse2_comineq_sd, COMI, X86ISD::COMI, ISD::SETNE), X86_INTRINSIC_DATA(sse2_max_pd, INTR_TYPE_2OP, X86ISD::FMAX, 0), X86_INTRINSIC_DATA(sse2_min_pd, INTR_TYPE_2OP, X86ISD::FMIN, 0), + X86_INTRINSIC_DATA(sse2_movmsk_pd, INTR_TYPE_1OP, X86ISD::MOVMSK, 0), X86_INTRINSIC_DATA(sse2_packssdw_128, INTR_TYPE_2OP, X86ISD::PACKSS, 0), X86_INTRINSIC_DATA(sse2_packsswb_128, INTR_TYPE_2OP, X86ISD::PACKSS, 0), X86_INTRINSIC_DATA(sse2_packuswb_128, INTR_TYPE_2OP, X86ISD::PACKUS, 0), + X86_INTRINSIC_DATA(sse2_padds_b, INTR_TYPE_2OP, X86ISD::ADDS, 0), + X86_INTRINSIC_DATA(sse2_padds_w, INTR_TYPE_2OP, X86ISD::ADDS, 0), + X86_INTRINSIC_DATA(sse2_paddus_b, INTR_TYPE_2OP, X86ISD::ADDUS, 0), + X86_INTRINSIC_DATA(sse2_paddus_w, INTR_TYPE_2OP, X86ISD::ADDUS, 0), X86_INTRINSIC_DATA(sse2_pavg_b, INTR_TYPE_2OP, X86ISD::AVG, 0), X86_INTRINSIC_DATA(sse2_pavg_w, INTR_TYPE_2OP, X86ISD::AVG, 0), - X86_INTRINSIC_DATA(sse2_pmaxs_w, INTR_TYPE_2OP, ISD::SMAX, 0), - X86_INTRINSIC_DATA(sse2_pmaxu_b, INTR_TYPE_2OP, ISD::UMAX, 0), - X86_INTRINSIC_DATA(sse2_pmins_w, INTR_TYPE_2OP, ISD::SMIN, 0), - X86_INTRINSIC_DATA(sse2_pminu_b, INTR_TYPE_2OP, ISD::UMIN, 0), + X86_INTRINSIC_DATA(sse2_pmovmskb_128, INTR_TYPE_1OP, X86ISD::MOVMSK, 0), X86_INTRINSIC_DATA(sse2_pmulh_w, INTR_TYPE_2OP, ISD::MULHS, 0), X86_INTRINSIC_DATA(sse2_pmulhu_w, INTR_TYPE_2OP, ISD::MULHU, 0), X86_INTRINSIC_DATA(sse2_pmulu_dq, INTR_TYPE_2OP, X86ISD::PMULUDQ, 0), X86_INTRINSIC_DATA(sse2_psad_bw, INTR_TYPE_2OP, X86ISD::PSADBW, 0), - X86_INTRINSIC_DATA(sse2_pshuf_d, INTR_TYPE_2OP, X86ISD::PSHUFD, 0), - X86_INTRINSIC_DATA(sse2_pshufh_w, INTR_TYPE_2OP, X86ISD::PSHUFHW, 0), - X86_INTRINSIC_DATA(sse2_pshufl_w, INTR_TYPE_2OP, X86ISD::PSHUFLW, 0), X86_INTRINSIC_DATA(sse2_psll_d, INTR_TYPE_2OP, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(sse2_psll_q, INTR_TYPE_2OP, X86ISD::VSHL, 0), X86_INTRINSIC_DATA(sse2_psll_w, INTR_TYPE_2OP, X86ISD::VSHL, 0), @@ -1982,6 +1916,8 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(sse2_psrli_d, VSHIFT, X86ISD::VSRLI, 0), X86_INTRINSIC_DATA(sse2_psrli_q, VSHIFT, X86ISD::VSRLI, 0), X86_INTRINSIC_DATA(sse2_psrli_w, VSHIFT, X86ISD::VSRLI, 0), + X86_INTRINSIC_DATA(sse2_psubs_b, INTR_TYPE_2OP, X86ISD::SUBS, 0), + X86_INTRINSIC_DATA(sse2_psubs_w, INTR_TYPE_2OP, X86ISD::SUBS, 0), X86_INTRINSIC_DATA(sse2_psubus_b, INTR_TYPE_2OP, X86ISD::SUBUS, 0), X86_INTRINSIC_DATA(sse2_psubus_w, INTR_TYPE_2OP, X86ISD::SUBUS, 0), X86_INTRINSIC_DATA(sse2_sqrt_pd, INTR_TYPE_1OP, ISD::FSQRT, 0), @@ -1997,48 +1933,17 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(sse3_hsub_ps, INTR_TYPE_2OP, X86ISD::FHSUB, 0), X86_INTRINSIC_DATA(sse41_insertps, INTR_TYPE_3OP, X86ISD::INSERTPS, 0), X86_INTRINSIC_DATA(sse41_packusdw, INTR_TYPE_2OP, X86ISD::PACKUS, 0), - X86_INTRINSIC_DATA(sse41_pmaxsb, INTR_TYPE_2OP, ISD::SMAX, 0), - X86_INTRINSIC_DATA(sse41_pmaxsd, INTR_TYPE_2OP, ISD::SMAX, 0), - X86_INTRINSIC_DATA(sse41_pmaxud, INTR_TYPE_2OP, ISD::UMAX, 0), - X86_INTRINSIC_DATA(sse41_pmaxuw, INTR_TYPE_2OP, ISD::UMAX, 0), - X86_INTRINSIC_DATA(sse41_pminsb, INTR_TYPE_2OP, ISD::SMIN, 0), - X86_INTRINSIC_DATA(sse41_pminsd, INTR_TYPE_2OP, ISD::SMIN, 0), - X86_INTRINSIC_DATA(sse41_pminud, INTR_TYPE_2OP, ISD::UMIN, 0), - X86_INTRINSIC_DATA(sse41_pminuw, INTR_TYPE_2OP, ISD::UMIN, 0), - X86_INTRINSIC_DATA(sse41_pmovzxbd, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(sse41_pmovzxbq, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(sse41_pmovzxbw, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(sse41_pmovzxdq, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(sse41_pmovzxwd, INTR_TYPE_1OP, X86ISD::VZEXT, 0), - X86_INTRINSIC_DATA(sse41_pmovzxwq, INTR_TYPE_1OP, X86ISD::VZEXT, 0), X86_INTRINSIC_DATA(sse41_pmuldq, INTR_TYPE_2OP, X86ISD::PMULDQ, 0), X86_INTRINSIC_DATA(sse4a_extrqi, INTR_TYPE_3OP, X86ISD::EXTRQI, 0), X86_INTRINSIC_DATA(sse4a_insertqi, INTR_TYPE_4OP, X86ISD::INSERTQI, 0), - X86_INTRINSIC_DATA(sse_comieq_ss, COMI, X86ISD::COMI, ISD::SETEQ), - X86_INTRINSIC_DATA(sse_comige_ss, COMI, X86ISD::COMI, ISD::SETGE), - X86_INTRINSIC_DATA(sse_comigt_ss, COMI, X86ISD::COMI, ISD::SETGT), - X86_INTRINSIC_DATA(sse_comile_ss, COMI, X86ISD::COMI, ISD::SETLE), - X86_INTRINSIC_DATA(sse_comilt_ss, COMI, X86ISD::COMI, ISD::SETLT), - X86_INTRINSIC_DATA(sse_comineq_ss, COMI, X86ISD::COMI, ISD::SETNE), - X86_INTRINSIC_DATA(sse_max_ps, INTR_TYPE_2OP, X86ISD::FMAX, 0), - X86_INTRINSIC_DATA(sse_min_ps, INTR_TYPE_2OP, X86ISD::FMIN, 0), - X86_INTRINSIC_DATA(sse_rcp_ps, INTR_TYPE_1OP, X86ISD::FRCP, 0), - X86_INTRINSIC_DATA(sse_rsqrt_ps, INTR_TYPE_1OP, X86ISD::FRSQRT, 0), - X86_INTRINSIC_DATA(sse_sqrt_ps, INTR_TYPE_1OP, ISD::FSQRT, 0), - X86_INTRINSIC_DATA(sse_ucomieq_ss, COMI, X86ISD::UCOMI, ISD::SETEQ), - X86_INTRINSIC_DATA(sse_ucomige_ss, COMI, X86ISD::UCOMI, ISD::SETGE), - X86_INTRINSIC_DATA(sse_ucomigt_ss, COMI, X86ISD::UCOMI, ISD::SETGT), - X86_INTRINSIC_DATA(sse_ucomile_ss, COMI, X86ISD::UCOMI, ISD::SETLE), - X86_INTRINSIC_DATA(sse_ucomilt_ss, COMI, X86ISD::UCOMI, ISD::SETLT), - X86_INTRINSIC_DATA(sse_ucomineq_ss, COMI, X86ISD::UCOMI, ISD::SETNE), + X86_INTRINSIC_DATA(ssse3_pabs_b_128, INTR_TYPE_1OP, X86ISD::ABS, 0), + X86_INTRINSIC_DATA(ssse3_pabs_d_128, INTR_TYPE_1OP, X86ISD::ABS, 0), + X86_INTRINSIC_DATA(ssse3_pabs_w_128, INTR_TYPE_1OP, X86ISD::ABS, 0), X86_INTRINSIC_DATA(ssse3_phadd_d_128, INTR_TYPE_2OP, X86ISD::HADD, 0), X86_INTRINSIC_DATA(ssse3_phadd_w_128, INTR_TYPE_2OP, X86ISD::HADD, 0), X86_INTRINSIC_DATA(ssse3_phsub_d_128, INTR_TYPE_2OP, X86ISD::HSUB, 0), X86_INTRINSIC_DATA(ssse3_phsub_w_128, INTR_TYPE_2OP, X86ISD::HSUB, 0), X86_INTRINSIC_DATA(ssse3_pshuf_b_128, INTR_TYPE_2OP, X86ISD::PSHUFB, 0), - X86_INTRINSIC_DATA(ssse3_psign_b_128, INTR_TYPE_2OP, X86ISD::PSIGN, 0), - X86_INTRINSIC_DATA(ssse3_psign_d_128, INTR_TYPE_2OP, X86ISD::PSIGN, 0), - X86_INTRINSIC_DATA(ssse3_psign_w_128, INTR_TYPE_2OP, X86ISD::PSIGN, 0), X86_INTRINSIC_DATA(xop_vpcomb, INTR_TYPE_3OP, X86ISD::VPCOM, 0), X86_INTRINSIC_DATA(xop_vpcomd, INTR_TYPE_3OP, X86ISD::VPCOM, 0), X86_INTRINSIC_DATA(xop_vpcomq, INTR_TYPE_3OP, X86ISD::VPCOM, 0), @@ -2047,6 +1952,11 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(xop_vpcomuq, INTR_TYPE_3OP, X86ISD::VPCOMU, 0), X86_INTRINSIC_DATA(xop_vpcomuw, INTR_TYPE_3OP, X86ISD::VPCOMU, 0), X86_INTRINSIC_DATA(xop_vpcomw, INTR_TYPE_3OP, X86ISD::VPCOM, 0), + X86_INTRINSIC_DATA(xop_vpermil2pd, INTR_TYPE_4OP, X86ISD::VPERMIL2, 0), + X86_INTRINSIC_DATA(xop_vpermil2pd_256, INTR_TYPE_4OP, X86ISD::VPERMIL2, 0), + X86_INTRINSIC_DATA(xop_vpermil2ps, INTR_TYPE_4OP, X86ISD::VPERMIL2, 0), + X86_INTRINSIC_DATA(xop_vpermil2ps_256, INTR_TYPE_4OP, X86ISD::VPERMIL2, 0), + X86_INTRINSIC_DATA(xop_vpperm, INTR_TYPE_3OP, X86ISD::VPPERM, 0), X86_INTRINSIC_DATA(xop_vprotb, INTR_TYPE_2OP, X86ISD::VPROT, 0), X86_INTRINSIC_DATA(xop_vprotbi, INTR_TYPE_2OP, X86ISD::VPROTI, 0), X86_INTRINSIC_DATA(xop_vprotd, INTR_TYPE_2OP, X86ISD::VPROT, 0), @@ -2069,7 +1979,7 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { * Retrieve data for Intrinsic without chain. * Return nullptr if intrinsic is not defined in the table. */ -static const IntrinsicData* getIntrinsicWithoutChain(unsigned IntNo) { +static const IntrinsicData* getIntrinsicWithoutChain(uint16_t IntNo) { IntrinsicData IntrinsicToFind = { IntNo, INTR_NO_TYPE, 0, 0 }; const IntrinsicData *Data = std::lower_bound(std::begin(IntrinsicsWithoutChain), std::end(IntrinsicsWithoutChain), @@ -2093,96 +2003,6 @@ static void verifyIntrinsicTables() { std::end(IntrinsicsWithChain)) && "Intrinsic data tables should have unique entries"); } - -// X86 specific compare constants. -// They must be kept in synch with avxintrin.h -#define _X86_CMP_EQ_OQ 0x00 /* Equal (ordered, non-signaling) */ -#define _X86_CMP_LT_OS 0x01 /* Less-than (ordered, signaling) */ -#define _X86_CMP_LE_OS 0x02 /* Less-than-or-equal (ordered, signaling) */ -#define _X86_CMP_UNORD_Q 0x03 /* Unordered (non-signaling) */ -#define _X86_CMP_NEQ_UQ 0x04 /* Not-equal (unordered, non-signaling) */ -#define _X86_CMP_NLT_US 0x05 /* Not-less-than (unordered, signaling) */ -#define _X86_CMP_NLE_US 0x06 /* Not-less-than-or-equal (unordered, signaling) */ -#define _X86_CMP_ORD_Q 0x07 /* Ordered (nonsignaling) */ -#define _X86_CMP_EQ_UQ 0x08 /* Equal (unordered, non-signaling) */ -#define _X86_CMP_NGE_US 0x09 /* Not-greater-than-or-equal (unord, signaling) */ -#define _X86_CMP_NGT_US 0x0a /* Not-greater-than (unordered, signaling) */ -#define _X86_CMP_FALSE_OQ 0x0b /* False (ordered, non-signaling) */ -#define _X86_CMP_NEQ_OQ 0x0c /* Not-equal (ordered, non-signaling) */ -#define _X86_CMP_GE_OS 0x0d /* Greater-than-or-equal (ordered, signaling) */ -#define _X86_CMP_GT_OS 0x0e /* Greater-than (ordered, signaling) */ -#define _X86_CMP_TRUE_UQ 0x0f /* True (unordered, non-signaling) */ -#define _X86_CMP_EQ_OS 0x10 /* Equal (ordered, signaling) */ -#define _X86_CMP_LT_OQ 0x11 /* Less-than (ordered, non-signaling) */ -#define _X86_CMP_LE_OQ 0x12 /* Less-than-or-equal (ordered, non-signaling) */ -#define _X86_CMP_UNORD_S 0x13 /* Unordered (signaling) */ -#define _X86_CMP_NEQ_US 0x14 /* Not-equal (unordered, signaling) */ -#define _X86_CMP_NLT_UQ 0x15 /* Not-less-than (unordered, non-signaling) */ -#define _X86_CMP_NLE_UQ 0x16 /* Not-less-than-or-equal (unord, non-signaling) */ -#define _X86_CMP_ORD_S 0x17 /* Ordered (signaling) */ -#define _X86_CMP_EQ_US 0x18 /* Equal (unordered, signaling) */ -#define _X86_CMP_NGE_UQ 0x19 /* Not-greater-than-or-equal (unord, non-sign) */ -#define _X86_CMP_NGT_UQ 0x1a /* Not-greater-than (unordered, non-signaling) */ -#define _X86_CMP_FALSE_OS 0x1b /* False (ordered, signaling) */ -#define _X86_CMP_NEQ_OS 0x1c /* Not-equal (ordered, signaling) */ -#define _X86_CMP_GE_OQ 0x1d /* Greater-than-or-equal (ordered, non-signaling) */ -#define _X86_CMP_GT_OQ 0x1e /* Greater-than (ordered, non-signaling) */ -#define _X86_CMP_TRUE_US 0x1f /* True (unordered, signaling) */ - -/* -* Get comparison modifier from _mm_comi_round_sd/ss intrinsic -* Return tuple <isOrdered, X86 condcode> -*/ -static std::tuple<bool,unsigned> TranslateX86ConstCondToX86CC(SDValue &imm) { - ConstantSDNode *CImm = dyn_cast<ConstantSDNode>(imm); - unsigned IntImm = CImm->getZExtValue(); - // On a floating point condition, the flags are set as follows: - // ZF PF CF op - // 0 | 0 | 0 | X > Y - // 0 | 0 | 1 | X < Y - // 1 | 0 | 0 | X == Y - // 1 | 1 | 1 | unordered - switch (IntImm) { - default: llvm_unreachable("Invalid floating point compare value for Comi!"); - case _X86_CMP_EQ_OQ: // 0x00 - Equal (ordered, nonsignaling) - case _X86_CMP_EQ_OS: // 0x10 - Equal (ordered, signaling) - return std::make_tuple(true, X86::COND_E); - case _X86_CMP_EQ_UQ: // 0x08 - Equal (unordered, non-signaling) - case _X86_CMP_EQ_US: // 0x18 - Equal (unordered, signaling) - return std::make_tuple(false , X86::COND_E); - case _X86_CMP_LT_OS: // 0x01 - Less-than (ordered, signaling) - case _X86_CMP_LT_OQ: // 0x11 - Less-than (ordered, nonsignaling) - return std::make_tuple(true, X86::COND_B); - case _X86_CMP_NGE_US: // 0x09 - Not-greater-than-or-equal (unordered, signaling) - case _X86_CMP_NGE_UQ: // 0x19 - Not-greater-than-or-equal (unordered, nonsignaling) - return std::make_tuple(false , X86::COND_B); - case _X86_CMP_LE_OS: // 0x02 - Less-than-or-equal (ordered, signaling) - case _X86_CMP_LE_OQ: // 0x12 - Less-than-or-equal (ordered, nonsignaling) - return std::make_tuple(true, X86::COND_BE); - case _X86_CMP_NGT_US: // 0x0A - Not-greater-than (unordered, signaling) - case _X86_CMP_NGT_UQ: // 0x1A - Not-greater-than (unordered, nonsignaling) - return std::make_tuple(false, X86::COND_BE); - case _X86_CMP_GT_OS: // 0x0E - Greater-than (ordered, signaling) - case _X86_CMP_GT_OQ: // 0x1E - Greater-than (ordered, nonsignaling) - return std::make_tuple(true, X86::COND_A); - case _X86_CMP_NLE_US: // 0x06 - Not-less-than-or-equal (unordered,signaling) - case _X86_CMP_NLE_UQ: // 0x16 - Not-less-than-or-equal (unordered, nonsignaling) - return std::make_tuple(false, X86::COND_A); - case _X86_CMP_GE_OS: // 0x0D - Greater-than-or-equal (ordered, signaling) - case _X86_CMP_GE_OQ: // 0x1D - Greater-than-or-equal (ordered, nonsignaling) - return std::make_tuple(true, X86::COND_AE); - case _X86_CMP_NLT_US: // 0x05 - Not-less-than (unordered, signaling) - case _X86_CMP_NLT_UQ: // 0x15 - Not-less-than (unordered, nonsignaling) - return std::make_tuple(false, X86::COND_AE); - case _X86_CMP_NEQ_OQ: // 0x0C - Not-equal (ordered, non-signaling) - case _X86_CMP_NEQ_OS: // 0x1C - Not-equal (ordered, signaling) - return std::make_tuple(true, X86::COND_NE); - case _X86_CMP_NEQ_UQ: // 0x04 - Not-equal (unordered, nonsignaling) - case _X86_CMP_NEQ_US: // 0x14 - Not-equal (unordered, signaling) - return std::make_tuple(false, X86::COND_NE); - } -} - } // End llvm namespace #endif |