- static std::map<std::string, X86Intrinsic> intrinsicMap2[] = {
- // 256 wide 512 wide
- { // AVX
- {"meta.intrinsic.VRCPPS", {{Intrinsic::x86_avx_rcp_ps_256, DOUBLE}, NO_EMU}},
- {"meta.intrinsic.VPERMPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VPERM_EMU}},
- {"meta.intrinsic.VPERMD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VPERM_EMU}},
- {"meta.intrinsic.VGATHERPD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VGATHERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VGATHERDD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::x86_avx_cvt_pd2_ps_256, Intrinsic::not_intrinsic}, NO_EMU}},
- {"meta.intrinsic.VCVTPH2PS", {{Intrinsic::x86_vcvtph2ps_256, Intrinsic::not_intrinsic}, NO_EMU}},
- {"meta.intrinsic.VROUND", {{Intrinsic::x86_avx_round_ps_256, DOUBLE}, NO_EMU}},
- {"meta.intrinsic.VHSUBPS", {{Intrinsic::x86_avx_hsub_ps_256, DOUBLE}, NO_EMU}},
- },
- { // AVX2
- {"meta.intrinsic.VRCPPS", {{Intrinsic::x86_avx_rcp_ps_256, DOUBLE}, NO_EMU}},
- {"meta.intrinsic.VPERMPS", {{Intrinsic::x86_avx2_permps, Intrinsic::not_intrinsic}, VPERM_EMU}},
- {"meta.intrinsic.VPERMD", {{Intrinsic::x86_avx2_permd, Intrinsic::not_intrinsic}, VPERM_EMU}},
- {"meta.intrinsic.VGATHERPD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VGATHERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VGATHERDD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::x86_avx_cvt_pd2_ps_256, Intrinsic::not_intrinsic}, NO_EMU}},
- {"meta.intrinsic.VCVTPH2PS", {{Intrinsic::x86_vcvtph2ps_256, Intrinsic::not_intrinsic}, NO_EMU}},
- {"meta.intrinsic.VROUND", {{Intrinsic::x86_avx_round_ps_256, DOUBLE}, NO_EMU}},
- {"meta.intrinsic.VHSUBPS", {{Intrinsic::x86_avx_hsub_ps_256, DOUBLE}, NO_EMU}},
- },
- { // AVX512
- {"meta.intrinsic.VRCPPS", {{Intrinsic::x86_avx512_rcp14_ps_256, Intrinsic::x86_avx512_rcp14_ps_512}, NO_EMU}},
- {"meta.intrinsic.VPERMPS", {{Intrinsic::x86_avx512_mask_permvar_sf_256, Intrinsic::x86_avx512_mask_permvar_sf_512}, NO_EMU}},
- {"meta.intrinsic.VPERMD", {{Intrinsic::x86_avx512_mask_permvar_si_256, Intrinsic::x86_avx512_mask_permvar_si_512}, NO_EMU}},
- {"meta.intrinsic.VGATHERPD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VGATHERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VGATHERDD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
- {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::x86_avx512_mask_cvtpd2ps_256, Intrinsic::x86_avx512_mask_cvtpd2ps_512 }, NO_EMU}},
- {"meta.intrinsic.VCVTPH2PS", {{Intrinsic::x86_avx512_mask_vcvtph2ps_256, Intrinsic::x86_avx512_mask_vcvtph2ps_512 }, NO_EMU}},
- {"meta.intrinsic.VROUND", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VROUND_EMU}},
- {"meta.intrinsic.VHSUBPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VHSUB_EMU}},
+ using intrinsicMapAdvanced_t = std::vector<std::map<std::string, X86Intrinsic>>;
+
+ static intrinsicMapAdvanced_t& getIntrinsicMapAdvanced()
+ {
+ // clang-format off
+ static intrinsicMapAdvanced_t intrinsicMapAdvanced = {
+ // 256 wide 512 wide
+ {
+ // AVX
+ {"meta.intrinsic.VRCPPS", {{Intrinsic::x86_avx_rcp_ps_256, DOUBLE}, NO_EMU}},
+ {"meta.intrinsic.VPERMPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VPERM_EMU}},
+ {"meta.intrinsic.VPERMD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VPERM_EMU}},
+ {"meta.intrinsic.VGATHERPD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VGATHERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VGATHERDD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VSCATTERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VSCATTER_EMU}},
+ {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::x86_avx_cvt_pd2_ps_256, Intrinsic::not_intrinsic}, NO_EMU}},
+ {"meta.intrinsic.VROUND", {{Intrinsic::x86_avx_round_ps_256, DOUBLE}, NO_EMU}},
+ {"meta.intrinsic.VHSUBPS", {{Intrinsic::x86_avx_hsub_ps_256, DOUBLE}, NO_EMU}},
+ },
+ {
+ // AVX2
+ {"meta.intrinsic.VRCPPS", {{Intrinsic::x86_avx_rcp_ps_256, DOUBLE}, NO_EMU}},
+ {"meta.intrinsic.VPERMPS", {{Intrinsic::x86_avx2_permps, Intrinsic::not_intrinsic}, VPERM_EMU}},
+ {"meta.intrinsic.VPERMD", {{Intrinsic::x86_avx2_permd, Intrinsic::not_intrinsic}, VPERM_EMU}},
+ {"meta.intrinsic.VGATHERPD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VGATHERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VGATHERDD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VSCATTERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VSCATTER_EMU}},
+ {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::x86_avx_cvt_pd2_ps_256, DOUBLE}, NO_EMU}},
+ {"meta.intrinsic.VROUND", {{Intrinsic::x86_avx_round_ps_256, DOUBLE}, NO_EMU}},
+ {"meta.intrinsic.VHSUBPS", {{Intrinsic::x86_avx_hsub_ps_256, DOUBLE}, NO_EMU}},
+ },
+ {
+ // AVX512
+ {"meta.intrinsic.VRCPPS", {{Intrinsic::x86_avx512_rcp14_ps_256, Intrinsic::x86_avx512_rcp14_ps_512}, NO_EMU}},
+ #if LLVM_VERSION_MAJOR < 7
+ {"meta.intrinsic.VPERMPS", {{Intrinsic::x86_avx512_mask_permvar_sf_256, Intrinsic::x86_avx512_mask_permvar_sf_512}, NO_EMU}},
+ {"meta.intrinsic.VPERMD", {{Intrinsic::x86_avx512_mask_permvar_si_256, Intrinsic::x86_avx512_mask_permvar_si_512}, NO_EMU}},
+ #else
+ {"meta.intrinsic.VPERMPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VPERM_EMU}},
+ {"meta.intrinsic.VPERMD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VPERM_EMU}},
+ #endif
+ {"meta.intrinsic.VGATHERPD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VGATHERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VGATHERDD", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VGATHER_EMU}},
+ {"meta.intrinsic.VSCATTERPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VSCATTER_EMU}},
+ #if LLVM_VERSION_MAJOR < 7
+ {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::x86_avx512_mask_cvtpd2ps_256, Intrinsic::x86_avx512_mask_cvtpd2ps_512}, NO_EMU}},
+ #else
+ {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VCONVERT_EMU}},
+ #endif
+ {"meta.intrinsic.VROUND", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VROUND_EMU}},
+ {"meta.intrinsic.VHSUBPS", {{Intrinsic::not_intrinsic, Intrinsic::not_intrinsic}, VHSUB_EMU}}
+ }};
+ // clang-format on
+ return intrinsicMapAdvanced;
+ }
+
+ static uint32_t getBitWidth(VectorType *pVTy)
+ {
+#if LLVM_VERSION_MAJOR >= 11
+ return pVTy->getNumElements() * pVTy->getElementType()->getPrimitiveSizeInBits();
+#else
+ return pVTy->getBitWidth();
+#endif