Merge "Update aosp/master LLVM for rebase to r230699."

1 files changed, 305 insertions, 9 deletions

diff --git a/lib/Target/X86/X86IntrinsicsInfo.h b/lib/Target/X86/X86IntrinsicsInfo.h
index d252f72..e436811 100644
--- a/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/lib/Target/X86/X86IntrinsicsInfo.h
@@ -20,8 +20,9 @@ enum IntrinsicType {
   INTR_NO_TYPE,
   GATHER, SCATTER, PREFETCH, RDSEED, RDRAND, RDPMC, RDTSC, XTEST, ADX,
   INTR_TYPE_1OP, INTR_TYPE_2OP, INTR_TYPE_3OP,
-  CMP_MASK, CMP_MASK_CC, VSHIFT, VSHIFT_MASK, COMI, 
-  INTR_TYPE_1OP_MASK_RM
+  CMP_MASK, CMP_MASK_CC, VSHIFT, VSHIFT_MASK, COMI,
+  INTR_TYPE_1OP_MASK_RM, INTR_TYPE_2OP_MASK, FMA_OP_MASK, INTR_TYPE_SCALAR_MASK_RM,
+  COMPRESS_EXPAND_IN_REG, COMPRESS_TO_MEM, EXPAND_FROM_MEM, BLEND
 };
 
 struct IntrinsicData {
@@ -51,7 +52,7 @@ static const IntrinsicData IntrinsicsWithChain[] = {
   X86_INTRINSIC_DATA(addcarry_u64,  ADX, X86ISD::ADC, 0),
   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),
@@ -60,7 +61,7 @@ static const IntrinsicData IntrinsicsWithChain[] = {
   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),
   X86_INTRINSIC_DATA(avx512_gatherpf_dps_512, PREFETCH,
@@ -69,7 +70,55 @@ static const IntrinsicData IntrinsicsWithChain[] = {
                      X86::VGATHERPF0QPDm, X86::VGATHERPF1QPDm),
   X86_INTRINSIC_DATA(avx512_gatherpf_qps_512, PREFETCH,
                      X86::VGATHERPF0QPSm, X86::VGATHERPF1QPSm),
-  
+
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_d_128,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_d_256,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_d_512,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_pd_128,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_pd_256,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_pd_512,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_ps_128,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_ps_256,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_ps_512,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_q_128,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_q_256,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_store_q_512,
+                     COMPRESS_TO_MEM, X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_d_128,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_d_256,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_d_512,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_pd_128,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_pd_256,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_pd_512,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_ps_128,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_ps_256,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_ps_512,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_q_128,
+                     EXPAND_FROM_MEM, X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_load_q_256,
+                     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_scatter_dpd_512, SCATTER, X86::VSCATTERDPDZmr, 0),
   X86_INTRINSIC_DATA(avx512_scatter_dpi_512, SCATTER, X86::VPSCATTERDDZmr, 0),
   X86_INTRINSIC_DATA(avx512_scatter_dpq_512, SCATTER, X86::VPSCATTERDQZmr, 0),
@@ -78,7 +127,7 @@ static const IntrinsicData IntrinsicsWithChain[] = {
   X86_INTRINSIC_DATA(avx512_scatter_qpi_512, SCATTER, X86::VPSCATTERQDZmr, 0),
   X86_INTRINSIC_DATA(avx512_scatter_qpq_512, SCATTER, X86::VPSCATTERQQZmr, 0),
   X86_INTRINSIC_DATA(avx512_scatter_qps_512, SCATTER, X86::VSCATTERQPSZmr, 0),
-  
+
   X86_INTRINSIC_DATA(avx512_scatterpf_dpd_512, PREFETCH,
                      X86::VSCATTERPF0DPDm, X86::VSCATTERPF1DPDm),
   X86_INTRINSIC_DATA(avx512_scatterpf_dps_512, PREFETCH,
@@ -87,7 +136,7 @@ static const IntrinsicData IntrinsicsWithChain[] = {
                      X86::VSCATTERPF0QPDm, X86::VSCATTERPF1QPDm),
   X86_INTRINSIC_DATA(avx512_scatterpf_qps_512, PREFETCH,
                      X86::VSCATTERPF0QPSm, X86::VSCATTERPF1QPSm),
-  
+
   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),
@@ -97,7 +146,7 @@ static const IntrinsicData IntrinsicsWithChain[] = {
   X86_INTRINSIC_DATA(rdseed_64, RDSEED, X86ISD::RDSEED, 0),
   X86_INTRINSIC_DATA(rdtsc,     RDTSC,  X86ISD::RDTSC_DAG, 0),
   X86_INTRINSIC_DATA(rdtscp,    RDTSC,  X86ISD::RDTSCP_DAG, 0),
-  
+
   X86_INTRINSIC_DATA(subborrow_u32, ADX, X86ISD::SBB, 0),
   X86_INTRINSIC_DATA(subborrow_u64, ADX, X86ISD::SBB, 0),
   X86_INTRINSIC_DATA(xtest,     XTEST,  X86ISD::XTEST,  0),
@@ -122,6 +171,12 @@ static const IntrinsicData* getIntrinsicWithChain(unsigned IntNo) {
  * the alphabetical order.
  */
 static const IntrinsicData  IntrinsicsWithoutChain[] = {
+  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_permd,        INTR_TYPE_2OP, X86ISD::VPERMV, 0),
+  X86_INTRINSIC_DATA(avx2_permps,       INTR_TYPE_2OP, X86ISD::VPERMV, 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),
@@ -138,27 +193,79 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx2_pminu_b,      INTR_TYPE_2OP, X86ISD::UMIN, 0),
   X86_INTRINSIC_DATA(avx2_pminu_d,      INTR_TYPE_2OP, X86ISD::UMIN, 0),
   X86_INTRINSIC_DATA(avx2_pminu_w,      INTR_TYPE_2OP, X86ISD::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_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_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),
   X86_INTRINSIC_DATA(avx2_pslli_d,      VSHIFT, X86ISD::VSHLI, 0),
   X86_INTRINSIC_DATA(avx2_pslli_q,      VSHIFT, X86ISD::VSHLI, 0),
   X86_INTRINSIC_DATA(avx2_pslli_w,      VSHIFT, X86ISD::VSHLI, 0),
+  X86_INTRINSIC_DATA(avx2_psllv_d,      INTR_TYPE_2OP, ISD::SHL, 0),
+  X86_INTRINSIC_DATA(avx2_psllv_d_256,  INTR_TYPE_2OP, ISD::SHL, 0),
+  X86_INTRINSIC_DATA(avx2_psllv_q,      INTR_TYPE_2OP, ISD::SHL, 0),
+  X86_INTRINSIC_DATA(avx2_psllv_q_256,  INTR_TYPE_2OP, ISD::SHL, 0),
   X86_INTRINSIC_DATA(avx2_psra_d,       INTR_TYPE_2OP, X86ISD::VSRA, 0),
   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_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),
   X86_INTRINSIC_DATA(avx2_psrli_d,      VSHIFT, X86ISD::VSRLI, 0),
   X86_INTRINSIC_DATA(avx2_psrli_q,      VSHIFT, X86ISD::VSRLI, 0),
   X86_INTRINSIC_DATA(avx2_psrli_w,      VSHIFT, X86ISD::VSRLI, 0),
+  X86_INTRINSIC_DATA(avx2_psrlv_d,      INTR_TYPE_2OP, ISD::SRL, 0),
+  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_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),
   X86_INTRINSIC_DATA(avx512_exp2_pd,    INTR_TYPE_1OP_MASK_RM,X86ISD::EXP2, 0),
   X86_INTRINSIC_DATA(avx512_exp2_ps,    INTR_TYPE_1OP_MASK_RM,X86ISD::EXP2, 0),
+  X86_INTRINSIC_DATA(avx512_mask_add_pd_512, INTR_TYPE_2OP_MASK, ISD::FADD,
+                     X86ISD::FADD_RND),
+  X86_INTRINSIC_DATA(avx512_mask_add_ps_512, INTR_TYPE_2OP_MASK, ISD::FADD,
+                     X86ISD::FADD_RND),
+  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_cmp_b_128,     CMP_MASK_CC,  X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_b_256,     CMP_MASK_CC,  X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_b_512,     CMP_MASK_CC,  X86ISD::CMPM, 0),
@@ -171,6 +278,64 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_cmp_w_128,     CMP_MASK_CC,  X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_w_256,     CMP_MASK_CC,  X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_w_512,     CMP_MASK_CC,  X86ISD::CMPM, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_d_128,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_d_256,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_d_512,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_pd_128, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_pd_256, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_pd_512, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_ps_128, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_ps_256, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_ps_512, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_q_128,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_q_256,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+  X86_INTRINSIC_DATA(avx512_mask_compress_q_512,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::COMPRESS, 0),
+
+  X86_INTRINSIC_DATA(avx512_mask_div_pd_512, INTR_TYPE_2OP_MASK, ISD::FDIV,
+                     X86ISD::FDIV_RND),
+  X86_INTRINSIC_DATA(avx512_mask_div_ps_512, INTR_TYPE_2OP_MASK, ISD::FDIV,
+                     X86ISD::FDIV_RND),
+  X86_INTRINSIC_DATA(avx512_mask_expand_d_128,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_d_256,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_d_512,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_pd_128, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_pd_256, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_pd_512, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_ps_128, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_ps_256, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_ps_512, COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_q_128,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_q_256,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_expand_q_512,  COMPRESS_EXPAND_IN_REG,
+                     X86ISD::EXPAND, 0),
+  
+  X86_INTRINSIC_DATA(avx512_mask_mul_pd_512, INTR_TYPE_2OP_MASK, ISD::FMUL,
+                     X86ISD::FMUL_RND),
+  X86_INTRINSIC_DATA(avx512_mask_mul_ps_512, INTR_TYPE_2OP_MASK, ISD::FMUL,
+                     X86ISD::FMUL_RND),
   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),
@@ -195,12 +360,32 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   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_psll_d,        INTR_TYPE_2OP_MASK, X86ISD::VSHL, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psll_q,        INTR_TYPE_2OP_MASK, X86ISD::VSHL, 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_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_q,        INTR_TYPE_2OP_MASK, X86ISD::VSRA, 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_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_psrl_d,        INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
+  X86_INTRINSIC_DATA(avx512_mask_psrl_q,        INTR_TYPE_2OP_MASK, X86ISD::VSRL, 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_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_rndscale_sd,   INTR_TYPE_SCALAR_MASK_RM,
+                     X86ISD::RNDSCALE, 0),
+  X86_INTRINSIC_DATA(avx512_mask_rndscale_ss,   INTR_TYPE_SCALAR_MASK_RM,
+                     X86ISD::RNDSCALE, 0),
+  X86_INTRINSIC_DATA(avx512_mask_sub_pd_512, INTR_TYPE_2OP_MASK, ISD::FSUB,
+                     X86ISD::FSUB_RND),
+  X86_INTRINSIC_DATA(avx512_mask_sub_ps_512, INTR_TYPE_2OP_MASK, ISD::FSUB,
+                     X86ISD::FSUB_RND),
   X86_INTRINSIC_DATA(avx512_mask_ucmp_b_128,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
   X86_INTRINSIC_DATA(avx512_mask_ucmp_b_256,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
   X86_INTRINSIC_DATA(avx512_mask_ucmp_b_512,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
@@ -215,27 +400,118 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_ucmp_w_512,    CMP_MASK_CC,  X86ISD::CMPMU, 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),
+  X86_INTRINSIC_DATA(avx512_rcp28_ss,   INTR_TYPE_SCALAR_MASK_RM, X86ISD::RCP28, 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(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_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(fma_mask_vfmadd_pd_128,    FMA_OP_MASK, X86ISD::FMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmadd_pd_256,    FMA_OP_MASK, X86ISD::FMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmadd_pd_512,    FMA_OP_MASK, X86ISD::FMADD,
+                     X86ISD::FMADD_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfmadd_ps_128,    FMA_OP_MASK, X86ISD::FMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmadd_ps_256,    FMA_OP_MASK, X86ISD::FMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmadd_ps_512,    FMA_OP_MASK, X86ISD::FMADD,
+                     X86ISD::FMADD_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfmaddsub_pd_128, FMA_OP_MASK, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmaddsub_pd_256, FMA_OP_MASK, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmaddsub_pd_512, FMA_OP_MASK, X86ISD::FMADDSUB,
+                     X86ISD::FMADDSUB_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfmaddsub_ps_128, FMA_OP_MASK, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmaddsub_ps_256, FMA_OP_MASK, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmaddsub_ps_512, FMA_OP_MASK, X86ISD::FMADDSUB,
+                     X86ISD::FMADDSUB_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfmsub_pd_128,    FMA_OP_MASK, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsub_pd_256,    FMA_OP_MASK, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsub_pd_512,    FMA_OP_MASK, X86ISD::FMSUB,
+                     X86ISD::FMSUB_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfmsub_ps_128,    FMA_OP_MASK, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsub_ps_256,    FMA_OP_MASK, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsub_ps_512,    FMA_OP_MASK, X86ISD::FMSUB,
+                     X86ISD::FMSUB_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfmsubadd_pd_128, FMA_OP_MASK, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsubadd_pd_256, FMA_OP_MASK, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsubadd_pd_512, FMA_OP_MASK, X86ISD::FMSUBADD,
+                     X86ISD::FMSUBADD_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfmsubadd_ps_128, FMA_OP_MASK, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsubadd_ps_256, FMA_OP_MASK, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfmsubadd_ps_512, FMA_OP_MASK, X86ISD::FMSUBADD,
+                     X86ISD::FMSUBADD_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfnmadd_pd_128,   FMA_OP_MASK, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmadd_pd_256,   FMA_OP_MASK, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmadd_pd_512,   FMA_OP_MASK, X86ISD::FNMADD,
+                     X86ISD::FNMADD_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfnmadd_ps_128,   FMA_OP_MASK, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmadd_ps_256,   FMA_OP_MASK, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmadd_ps_512,   FMA_OP_MASK, X86ISD::FNMADD,
+                     X86ISD::FNMADD_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfnmsub_pd_128,   FMA_OP_MASK, X86ISD::FNMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmsub_pd_256,   FMA_OP_MASK, X86ISD::FNMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmsub_pd_512,   FMA_OP_MASK, X86ISD::FNMSUB,
+                     X86ISD::FNMSUB_RND),
+  X86_INTRINSIC_DATA(fma_mask_vfnmsub_ps_128,   FMA_OP_MASK, X86ISD::FNMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmsub_ps_256,   FMA_OP_MASK, X86ISD::FNMSUB, 0),
+  X86_INTRINSIC_DATA(fma_mask_vfnmsub_ps_512,   FMA_OP_MASK, X86ISD::FNMSUB,
+                     X86ISD::FNMSUB_RND),
+  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),
+  X86_INTRINSIC_DATA(fma_vfmadd_ps_256,    INTR_TYPE_3OP, X86ISD::FMADD, 0),
+  X86_INTRINSIC_DATA(fma_vfmaddsub_pd,     INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmaddsub_pd_256, INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmaddsub_ps,     INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmaddsub_ps_256, INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmsub_pd,        INTR_TYPE_3OP, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmsub_pd_256,    INTR_TYPE_3OP, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmsub_ps,        INTR_TYPE_3OP, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmsub_ps_256,    INTR_TYPE_3OP, X86ISD::FMSUB, 0),
+  X86_INTRINSIC_DATA(fma_vfmsubadd_pd,     INTR_TYPE_3OP, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_vfmsubadd_pd_256, INTR_TYPE_3OP, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_vfmsubadd_ps,     INTR_TYPE_3OP, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_vfmsubadd_ps_256, INTR_TYPE_3OP, X86ISD::FMSUBADD, 0),
+  X86_INTRINSIC_DATA(fma_vfnmadd_pd,       INTR_TYPE_3OP, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_vfnmadd_pd_256,   INTR_TYPE_3OP, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_vfnmadd_ps,       INTR_TYPE_3OP, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_vfnmadd_ps_256,   INTR_TYPE_3OP, X86ISD::FNMADD, 0),
+  X86_INTRINSIC_DATA(fma_vfnmsub_pd,       INTR_TYPE_3OP, X86ISD::FNMSUB, 0),
+  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(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),
   X86_INTRINSIC_DATA(sse2_comile_sd,    COMI, X86ISD::COMI, ISD::SETLE),
   X86_INTRINSIC_DATA(sse2_comilt_sd,    COMI, X86ISD::COMI, ISD::SETLT),
   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_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_pmaxs_w,      INTR_TYPE_2OP, X86ISD::SMAX, 0),
   X86_INTRINSIC_DATA(sse2_pmaxu_b,      INTR_TYPE_2OP, X86ISD::UMAX, 0),
   X86_INTRINSIC_DATA(sse2_pmins_w,      INTR_TYPE_2OP, X86ISD::SMIN, 0),
   X86_INTRINSIC_DATA(sse2_pminu_b,      INTR_TYPE_2OP, X86ISD::UMIN, 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_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),
@@ -266,6 +542,7 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(sse3_hsub_pd,      INTR_TYPE_2OP, X86ISD::FHSUB, 0),
   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, X86ISD::SMAX, 0),
   X86_INTRINSIC_DATA(sse41_pmaxsd,      INTR_TYPE_2OP, X86ISD::SMAX, 0),
   X86_INTRINSIC_DATA(sse41_pmaxud,      INTR_TYPE_2OP, X86ISD::UMAX, 0),
@@ -274,12 +551,27 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(sse41_pminsd,      INTR_TYPE_2OP, X86ISD::SMIN, 0),
   X86_INTRINSIC_DATA(sse41_pminud,      INTR_TYPE_2OP, X86ISD::UMIN, 0),
   X86_INTRINSIC_DATA(sse41_pminuw,      INTR_TYPE_2OP, X86ISD::UMIN, 0),
+  X86_INTRINSIC_DATA(sse41_pmovsxbd,    INTR_TYPE_1OP, X86ISD::VSEXT, 0),
+  X86_INTRINSIC_DATA(sse41_pmovsxbq,    INTR_TYPE_1OP, X86ISD::VSEXT, 0),
+  X86_INTRINSIC_DATA(sse41_pmovsxbw,    INTR_TYPE_1OP, X86ISD::VSEXT, 0),
+  X86_INTRINSIC_DATA(sse41_pmovsxdq,    INTR_TYPE_1OP, X86ISD::VSEXT, 0),
+  X86_INTRINSIC_DATA(sse41_pmovsxwd,    INTR_TYPE_1OP, X86ISD::VSEXT, 0),
+  X86_INTRINSIC_DATA(sse41_pmovsxwq,    INTR_TYPE_1OP, X86ISD::VSEXT, 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(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_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),
@@ -290,7 +582,11 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   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_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)
 };
 
 /*