����JFIFXX�����    $.' ",#(7),01444'9=82<.342  2!!22222222222222222222222222222222222222222222222222����"��4�� ���,�PG"Z_�4�˷����kjز�Z�,F+��_z�,�© �����zh6�٨�ic�fu���#ډb���_�N�?��wQ���5-�~�I���8����TK<5o�Iv-�����k�_U_�����~b�M��d����Ӝ�U�Hh��?]��E�w��Q���k�{��_}qFW7HTՑ��Y��F�?_�'ϔ��_�Ջt��=||I ��6�έ"�����D���/[�k�9���Y�8ds|\���Ҿp6�Ҵ���]��.����6�z<�v��@]�i%��$j��~�g��J>��no����pM[me�i$[����s�o�ᘨ�˸ nɜG-�ĨU�ycP�3.DB�li�;��hj���x7Z^�N�h������N3u{�:j�x�힞��#M&��jL P@_���� P��&��o8������9�����@Sz6�t7#O�ߋ �s}Yf�T���lmr����Z)'N��k�۞p����w\�Tȯ?�8`�O��i{wﭹW�[�r�� ��Q4F�׊���3m&L�=��h3����z~��#�\�l :�F,j@�� ʱ�wQT����8�"kJO���6�֚l����}���R�>ډK���]��y����&����p�}b��;N�1�m�r$�|��7�>e�@B�TM*-iH��g�D�)� E�m�|�ؘbҗ�a��Ҿ����t4���o���G��*oCN�rP���Q��@z,|?W[0�����:�n,jWiE��W��$~/�hp\��?��{(�0���+�Y8rΟ�+����>S-S����VN;�}�s?.����� w�9��˟<���Mq4�Wv'��{)0�1mB��V����W[�����8�/<� �%���wT^�5���b��)iM� pg�N�&ݝ��VO~�q���u���9� ����!��J27����$O-���! �:�%H��� ـ����y�ΠM=t{!S�� oK8������t<����è:a������[�����ա�H���~��w��Qz`�po�^ ����Q��n� �,uu�C�$ ^���,������8�#��:�6��e�|~���!�3�3.�\0��q��o�4`.|� ����y�Q�`~;�d�ׯ,��O�Zw�������`73�v�܋�<���Ȏ�� ـ4k��5�K�a�u�=9Yd��$>x�A�&�� j0� ���vF��� Y�|�y��� ~�6�@c��1vOp�Ig����4��l�OD���L����� R���c���j�_�uX6��3?nk��Wy�f;^*B� ��@�~a�`��Eu������+���6�L��.ü>��}y���}_�O�6�͐�:�YrG�X��kG�����l^w���~㒶sy��Iu�!� W ��X��N�7BV��O��!X�2����wvG�R�f�T#�����t�/?���%8�^�W�aT��G�cL�M���I��(J����1~�8�?aT ���]����AS�E��(��*E}� 2��#I/�׍qz��^t�̔���b�Yz4x���t�){ OH��+(E��A&�N�������XT��o��"�XC��'���)}�J�z�p� ��~5�}�^����+�6����w��c��Q�|Lp�d�H��}�(�.|����k��c4^�"�����Z?ȕ ��a<�L�!039C� �Eu�C�F�Ew�ç ;�n?�*o���B�8�bʝ���'#Rqf���M}7����]����s2tcS{�\icTx;�\��7K���P���ʇ Z O-��~��c>"��?�������P��E��O�8��@�8��G��Q�g�a�Վ���󁶠�䧘��_%#r�>�1�z�a��eb��qcPѵ��n���#L��� =��׀t� L�7�`��V���A{�C:�g���e@�w1 Xp3�c3�ġ����p��M"'-�@n4���fG��B3�DJ�8[Jo�ߐ���gK)ƛ��$���� ���8�3�����+���� �����6�ʻ���� ���S�kI�*KZlT _`���?��K����QK�d����B`�s}�>���`��*�>��,*@J�d�oF*����弝��O}�k��s��]��y�ߘ��c1G�V���<=�7��7����6�q�PT��tXԀ�!9*4�4Tހ3XΛex�46���Y��D ����� �BdemDa����\�_l,��G�/���֌7���Y�](�xTt^%�GE�����4�}bT���ڹ�����;Y)���B�Q��u��>J/J �⮶.�XԄ��j�ݳ�+E��d ��r�5�_D�1 ��o�� �B�x�΢�#���<��W�����8���R6�@g�M�.��� dr�D��>(otU��@x=��~v���2� ӣ�d�oBd��3�eO�6�㣷�����ݜ6��6Y��Qz`��S��{���\P�~z m5{J/L��1������<�e�ͅPu�b�]�ϔ���'������f�b� Zpw��c`"��i���BD@:)ִ�:�]��hv�E�w���T�l��P���"Ju�}��وV J��G6��. J/�Qgl߭�e�����@�z�Zev2u�)]կ�����7x���s�M�-<ɯ�c��r�v�����@��$�ޮ}lk���a���'����>x��O\�ZFu>�����ck#��&:��`�$�ai�>2Δ����l���oF[h��lE�ܺ�Πk:)���`�� $[6�����9�����kOw�\|���8}������ބ:��񶐕��I�A1/�=�2[�,�!��.}gN#�u����b��� ~��݊��}34q����d�E��Lc��$��"�[q�U�硬g^��%B �z���r�pJ�ru%v\h1Y�ne`ǥ:g���pQM~�^�Xi� ��`S�:V29.�P���V�?B�k�� AEvw%�_�9C�Q����wKekPؠ�\�;Io d�{ ߞo�c1eP����\� `����E=���@K<�Y���eڼ�J���w����{av�F�'�M�@/J��+9p���|]�����Iw &`��8���&M�hg��[�{��Xj��%��Ӓ�$��(����ʹN���<>�I���RY���K2�NPlL�ɀ)��&e����B+ь����( � �JTx���_?EZ� }@ 6�U���뙢ط�z��dWI�n` D����噥�[��uV��"�G&Ú����2g�}&m��?ċ�"����Om#��������� ��{�ON��"S�X��Ne��ysQ���@Fn��Vg���dX�~nj�]J�<�K]:��FW��b�������62�=��5f����JKw��bf�X�55��~J �%^����:�-�QIE��P��v�nZum� z � ~ə ���� ���ة����;�f��\v���g�8�1��f24;�V���ǔ�)����9���1\��c��v�/'Ƞ�w�������$�4�R-��t���� e�6�/�ġ �̕Ecy�J���u�B���<�W�ַ~�w[B1L۲�-JS΂�{���΃������A��20�c#��@ 0!1@AP"#2Q`$3V�%45a6�FRUq��� ����^7ׅ,$n�������+��F�`��2X'��0vM��p�L=������5��8������u�p~���.�`r�����\���O��,ư�0oS ��_�M�����l���4�kv\JSd���x���SW�<��Ae�IX����������$I���w�:S���y���›R��9�Q[���,�5�;�@]�%���u�@ *ro�lbI �� ��+���%m:�͇ZV�����u�̉����θau<�fc�.����{�4Ա� �Q����*�Sm��8\ujqs]{kN���)qO�y�_*dJ�b�7���yQqI&9�ԌK!�M}�R�;������S�T���1���i[U�ɵz�]��U)V�S6���3$K{�ߊ<�(� E]Զ[ǼENg�����'�\?#)Dkf��J���o��v���'�%ƞ�&K�u�!��b�35LX�Ϸ��63$K�a�;�9>,R��W��3�3� d�JeTYE.Mϧ��-�o�j3+y��y^�c�������VO�9NV\nd�1 ��!͕_)a�v;����թ�M�lWR1��)El��P;��yوÏ�u 3�k�5Pr6<�⒲l�!˞*��u־�n�!�l:����UNW ��%��Chx8vL'��X�@��*��)���̮��ˍ��� ���D-M�+J�U�kvK����+�x8��cY������?�Ԡ��~3mo��|�u@[XeY�C�\Kp�x8�oC�C�&����N�~3-H���� ��MX�s�u<`���~"WL��$8ξ��3���a�)|:@�m�\���^�`�@ҷ)�5p+��6���p�%i)P M���ngc�����#0Aruz���RL+xSS?���ʮ}()#�t��mˇ!��0}}y����<�e� �-ή�Ԩ��X������ MF���ԙ~l L.3���}�V뽺�v�����멬��Nl�)�2����^�Iq��a��M��qG��T�����c3#������3U�Ǎ���}��לS�|qa��ڃ�+���-��2�f����/��bz��ڐ�� �ݼ[2�ç����k�X�2�* �Z�d���J�G����M*9W���s{��w���T��x��y,�in�O�v��]���n����P�$�JB@=4�OTI�n��e�22a\����q�d���%�$��(���:���: /*�K[PR�fr\nڙdN���F�n�$�4�[�� U�zƶ����� �mʋ���,�ao�u 3�z� �x��Kn����\[��VFmbE;�_U��&V�Gg�]L�۪&#n%�$ɯ�dG���D�TI=�%+AB�Ru#��b4�1�»x�cs�YzڙJG��f��Il��d�eF'T� iA��T���uC�$����Y��H?����[!G`}���ͪ� �纤Hv\������j�Ex�K���!���OiƸ�Yj�+u-<���'q����uN�*�r\��+�]���<�wOZ.fp�ێ��,-*)V?j-kÊ#�`�r��dV����(�ݽBk�����G�ƛk�QmUڗe��Z���f}|����8�8��a���i��3'J�����~G_�^���d�8w������ R�`(�~�.��u���l�s+g�bv���W���lGc}��u���afE~1�Ue������Z�0�8�=e�� f@/�jqEKQQ�J��oN��J���W5~M>$6�Lt�;$ʳ{���^��6�{����v6���ķܰg�V�cnn �~z�x�«�,2�u�?cE+Ș�H؎�%�Za�)���X>uW�Tz�Nyo����s���FQƤ��$��*�&�LLXL)�1�" L��eO��ɟ�9=���:t��Z���c��Ž���Y?�ӭV�wv�~,Y��r�ۗ�|�y��GaF�����C�����.�+� ���v1���fήJ�����]�S��T��B��n5sW}y�$��~z�'�c ��8 ��� ,! �p��VN�S��N�N�q��y8z˱�A��4��*��'������2n<�s���^ǧ˭P�Jޮɏ�U�G�L�J�*#��<�V��t7�8����TĜ>��i}K%,���)[��z�21z ?�N�i�n1?T�I�R#��m-�����������������1����lA�`��fT5+��ܐ�c�q՝��ʐ��,���3�f2U�եmab��#ŠdQ�y>\��)�SLY����w#��.���ʑ�f��� ,"+�w�~�N�'�c�O�3F�������N<���)j��&��,-� �љ���֊�_�zS���TǦ����w�>��?�������n��U仆�V���e�����0���$�C�d���rP �m�׈e�Xm�Vu� �L��.�bֹ��� �[Դaզ���*��\y�8�Է:�Ez\�0�Kq�C b��̘��cө���Q��=0Y��s�N��S.���3.���O�o:���#���v7�[#߫ ��5�܎�L���Er4���9n��COWlG�^��0k�%<���ZB���aB_���������'=��{i�v�l�$�uC���mƎҝ{�c㱼�y]���W�i ��ߧc��m�H� m�"�"�����;Y�ߝ�Z�Ǔ�����:S#��|}�y�,/k�Ld� TA�(�AI$+I3��;Y*���Z��}|��ӧO��d�v��..#:n��f>�>���ȶI�TX��� 8��y����"d�R�|�)0���=���n4��6ⲑ�+��r<�O�܂~zh�z����7ܓ�HH�Ga롏���nCo�>������a ���~]���R���̲c?�6(�q�;5%� |�uj�~z8R=X��I�V=�|{v�Gj\gc��q����z�؋%M�ߍ����1y��#��@f^���^�>N�����#x#۹��6�Y~�?�dfPO��{��P�4��V��u1E1J �*|���%���JN��`eWu�zk M6���q t[�� ��g�G���v��WIG��u_ft����5�j�"�Y�:T��ɐ���*�;� e5���4����q$C��2d�}���� _S�L#m�Yp��O�.�C�;��c����Hi#֩%+) �Ӎ��ƲV���SYź��g |���tj��3�8���r|���V��1#;.SQ�A[���S������#���`n�+���$��$I �P\[�@�s��(�ED�z���P��])8�G#��0B��[ى��X�II�q<��9�~[Z멜�Z�⊔IWU&A>�P~�#��dp<�?����7���c��'~���5 ��+$���lx@�M�dm��n<=e�dyX��?{�|Aef ,|n3�<~z�ƃ�uۧ�����P��Y,�ӥQ�*g�#먙R�\���;T��i,��[9Qi歉����c>]9�� ��"�c��P�� �Md?٥��If�ت�u��k��/����F��9�c*9��Ǎ:�ØF���z�n*�@|I�ށ9����N3{'��[�'ͬ�Ҳ4��#}��!�V� Fu��,�,mTIk���v C�7v���B�6k�T9��1�*l� '~��ƞF��lU��'�M ����][ΩũJ_�{�i�I�n��$���L�� j��O�dx�����kza۪��#�E��Cl����x˘�o�����V���ɞ�ljr��)�/,�߬h�L��#��^��L�ф�,íMƁe�̩�NB�L�����iL����q�}��(��q��6IçJ$�W�E$��:������=#����(�K�B����zђ <��K(�N�۫K�w��^O{!����)�H���>x�������lx�?>Պ�+�>�W���,Ly!_�D���Ō�l���Q�!�[ �S����J��1��Ɛ�Y}��b,+�Lo�x�ɓ)����=�y�oh�@�꥟/��I��ѭ=��P�y9��� �ۍYӘ�e+�p�Jnϱ?V\SO%�(�t� ���=?MR�[Ș�����d�/ ��n�l��B�7j� ��!�;ӥ�/�[-���A�>�dN�sLj ��,ɪv��=1c�.SQ�O3�U���ƀ�ܽ�E����������̻��9G�ϷD�7(�}��Ävӌ\�y�_0[w ���<΍>����a_��[0+�L��F.�޺��f�>oN�T����q;���y\��bՃ��y�jH�<|q-eɏ�_?_9+P���Hp$�����[ux�K w�Mw��N�ی'$Y2�=��q���KB��P��~������Yul:�[<����F1�2�O���5=d����]Y�sw:���Ϯ���E��j,_Q��X��z`H1,#II ��d�wr��P˂@�ZJV����y$�\y�{}��^~���[:N����ߌ�U�������O��d�����ؾe��${p>G��3c���Ė�lʌ�� ת��[��`ϱ�-W����dg�I��ig2��� ��}s ��ؤ(%#sS@���~���3�X�nRG�~\jc3�v��ӍL��M[JB�T��s3}��j�Nʖ��W����;7��ç?=X�F=-�=����q�ߚ���#���='�c��7���ڑW�I(O+=:uxq�������������e2�zi+�kuG�R��������0�&e�n���iT^J����~\jy���p'dtG��s����O��3����9* �b#Ɋ�� p������[Bws�T�>d4�ۧs���nv�n���U���_�~,�v����ƜJ1��s�� �QIz��)�(lv8M���U=�;����56��G���s#�K���MP�=��LvyGd��}�VwWBF�'�à �?MH�U�g2�� ����!�p�7Q��j��ڴ����=��j�u��� Jn�A s���uM������e��Ɔ�Ҕ�!)'��8Ϣ�ٔ��ޝ(��Vp���צ֖d=�IC�J�Ǡ{q������kԭ�߸���i��@K����u�|�p=..�*+����x�����z[Aqġ#s2a�Ɗ���RR�)*HRsi�~�a &f��M��P����-K�L@��Z��Xy�'x�{}��Zm+���:�)�) IJ�-i�u���� ���ܒH��'�L(7�y�GӜq���� j��� 6ߌg1�g�o���,kر���tY�?W,���p���e���f�OQS��!K�۟cҒA�|ս�j�>��=⬒��˧L[�� �߿2JaB~R��u�:��Q�] �0H~���]�7��Ƽ�I���(}��cq '�ήET���q�?f�ab���ӥvr� �)o��-Q��_'����ᴎo��K������;��V���o��%���~OK ����*��b�f:���-ťIR��`B�5!RB@���ï�� �u �̯e\�_U�_������� g�ES��3�������QT��a����x����U<~�c?�*�#]�MW,[8O�a�x��]�1bC|踤�P��lw5V%�)�{t�<��d��5���0i�XSU��m:��Z�┵�i�"��1�^B�-��P�hJ��&)O��*�D��c�W��vM��)����}���P��ܗ-q����\mmζZ-l@�}��a��E�6��F�@��&Sg@���ݚ�M����� ȹ 4����#p�\H����dYDo�H���"��\��..R�B�H�z_�/5˘����6��KhJR��P�mƶi�m���3�,#c�co��q�a)*Pt����R�m�k�7x�D�E�\Y�閣_X�<���~�)���c[[�BP����6�Yq���S��0����%_����;��Àv�~�| VS؇ ��'O0��F0��\���U�-�d@�����7�SJ*z��3n��y��P����O���������m�~�P�3|Y��ʉr#�C�<�G~�.,! ���bqx���h~0=��!ǫ�jy����l�O,�[B��~��|9��ٱ����Xly�#�i�B��g%�S��������tˋ���e���ې��\[d�t)��.+u�|1 ������#�~Oj����hS�%��i.�~X���I�H�m��0n���c�1uE�q��cF�RF�o���7� �O�ꮧ� ���ۛ{��ʛi5�rw?׌#Qn�TW��~?y$��m\�\o����%W� ?=>S�N@�� �Ʈ���R����N�)�r"C�:��:����� �����#��qb��Y�. �6[��2K����2u�Ǧ�HYR��Q�MV��� �G�$��Q+.>�����nNH��q�^��� ����q��mM��V��D�+�-�#*�U�̒ ���p욳��u:�������IB���m���PV@O���r[b= �� ��1U�E��_Nm�yKbN�O���U�}�the�`�|6֮P>�\2�P�V���I�D�i�P�O;�9�r�mAHG�W�S]��J*�_�G��+kP�2����Ka�Z���H�'K�x�W�MZ%�O�YD�Rc+o��?�q��Ghm��d�S�oh�\�D�|:W������UA�Qc yT�q������~^�H��/��#p�CZ���T�I�1�ӏT����4��"�ČZ�����}��`w�#�*,ʹ�� ��0�i��課�Om�*�da��^gJ݅{���l�e9uF#T�ֲ��̲�ٞC"�q���ߍ ոޑ�o#�XZTp����@ o�8��(jd��xw�]�,f���`~�|,s��^����f�1���t��|��m�򸄭/ctr��5s��7�9Q�4�H1꠲BB@l9@���C�����+�wp�xu�£Yc�9��?`@#�o�mH�s2��)�=��2�.�l����jg�9$�Y�S�%*L������R�Y������7Z���,*=�䷘$�������arm�o�ϰ���UW.|�r�uf����IGw�t����Zwo��~5 ��YյhO+=8fF�)�W�7�L9lM�̘·Y���֘YLf�큹�pRF���99.A �"wz��=E\Z���'a� 2��Ǚ�#;�'}�G���*��l��^"q��+2FQ� hj��kŦ��${���ޮ-�T�٭cf�|�3#~�RJ����t��$b�(R��(����r���dx� >U b�&9,>���%E\� Ά�e�$��'�q't��*�א���ެ�b��-|d���SB�O�O��$�R+�H�)�܎�K��1m`;�J�2�Y~9��O�g8=vqD`K[�F)k�[���1m޼c��n���]s�k�z$@��)!I �x՝"v��9=�ZA=`Ɠi �:�E��)`7��vI��}d�YI�_ �o�:ob���o ���3Q��&D&�2=�� �Ά��;>�h����y.*ⅥS������Ӭ�+q&����j|UƧ����}���J0��WW< ۋS�)jQR�j���Ư��rN)�Gű�4Ѷ(�S)Ǣ�8��i��W52���No˓� ۍ%�5brOn�L�;�n��\G����=�^U�dI���8$�&���h��'���+�(������cȁ߫k�l��S^���cƗjԌE�ꭔ��gF���Ȓ��@���}O���*;e�v�WV���YJ\�]X'5��ղ�k�F��b 6R�o՜m��i N�i����>J����?��lPm�U��}>_Z&�KK��q�r��I�D�Չ~�q�3fL�:S�e>���E���-G���{L�6p�e,8��������QI��h��a�Xa��U�A'���ʂ���s�+טIjP�-��y�8ۈZ?J$��W�P� ��R�s�]��|�l(�ԓ��sƊi��o(��S0��Y� 8�T97.�����WiL��c�~�dxc�E|�2!�X�K�Ƙਫ਼�$((�6�~|d9u+�qd�^3�89��Y�6L�.I�����?���iI�q���9�)O/뚅����O���X��X�V��ZF[�یgQ�L��K1���RҖr@v�#��X�l��F���Нy�S�8�7�kF!A��sM���^rkp�jP�DyS$N���q��nxҍ!U�f�!eh�i�2�m���`�Y�I�9r�6� �TF���C}/�y�^���Η���5d�'��9A-��J��>{�_l+�`��A���[�'��յ�ϛ#w:݅�%��X�}�&�PSt�Q�"�-��\縵�/����$Ɨh�Xb�*�y��BS����;W�ջ_mc�����vt?2}1�;qS�d�d~u:2k5�2�R�~�z+|HE!)�Ǟl��7`��0�<�,�2*���Hl-��x�^����'_TV�gZA�'j� ^�2Ϊ��N7t�����?w�� �x1��f��Iz�C-Ȗ��K�^q�;���-W�DvT�7��8�Z�������� hK�(P:��Q- �8�n�Z���܃e貾�<�1�YT<�,�����"�6{/ �?�͟��|1�:�#g��W�>$����d��J��d�B��=��jf[��%rE^��il:��B���x���Sּ�1հ��,�=��*�7 fcG��#q� �eh?��2�7�����,�!7x��6�n�LC�4x��},Geǝ�tC.��vS �F�43��zz\��;QYC,6����~;RYS/6���|2���5���v��T��i����������mlv��������&� �nRh^ejR�LG�f���? �ۉҬܦƩ��|��Ȱ����>3����!v��i�ʯ�>�v��オ�X3e���_1z�Kȗ\<������!�8���V��]��?b�k41�Re��T�q��mz��TiOʦ�Z��Xq���L������q"+���2ۨ��8}�&N7XU7Ap�d�X��~�׿��&4e�o�F��� �H����O���č�c�� 懴�6���͉��+)��v;j��ݷ�� �UV�� i��� j���Y9GdÒJ1��詞�����V?h��l����l�cGs�ځ�������y�Ac�����\V3�? �� ܙg�>qH�S,�E�W�[�㺨�uch�⍸�O�}���a��>�q�6�n6����N6�q������N ! 1AQaq�0@����"2BRb�#Pr���3C`��Scst���$4D���%Td�� ?���N����a��3��m���C���w��������xA�m�q�m���m������$����4n淿t'��C"w��zU=D�\R+w�p+Y�T�&�պ@��ƃ��3ޯ?�Aﶂ��aŘ���@-�����Q�=���9D��ռ�ѻ@��M�V��P��܅�G5�f�Y<�u=,EC)�<�Fy'�"�&�չ�X~f��l�KԆV��?�� �W�N����=(� �;���{�r����ٌ�Y���h{�١������jW����P���Tc�����X�K�r��}���w�R��%��?���E��m�� �Y�q|����\lEE4���r���}�lsI�Y������f�$�=�d�yO����p�����yBj8jU�o�/�S��?�U��*������ˍ�0������u�q�m [�?f����a�� )Q�>����6#������� ?����0UQ����,IX���(6ڵ[�DI�MNލ�c&���υ�j\��X�R|,4��� j������T�hA�e��^���d���b<����n�� �즇�=!���3�^�`j�h�ȓr��jẕ�c�,ٞX����-����a�ﶔ���#�$��]w�O��Ӫ�1y%��L�Y<�wg#�ǝ�̗`�x�xa�t�w��»1���o7o5��>�m뭛C���Uƃߜ}�C���y1Xνm�F8�jI���]����H���ۺиE@I�i;r�8ӭ����V�F�Շ| ��&?�3|x�B�MuS�Ge�=Ӕ�#BE5G�����Y!z��_e��q�р/W>|-�Ci߇�t�1ޯќd�R3�u��g�=0 5��[?�#͏��q�cf���H��{ ?u�=?�?ǯ���}Z��z���hmΔ�BFTW�����<�q�(v� ��!��z���iW]*�J�V�z��gX֧A�q�&��/w���u�gYӘa���; �i=����g:��?2�dž6�ى�k�4�>�Pxs����}������G�9��3 ���)gG�R<>r h�$��'nc�h�P��Bj��J�ҧH� -��N1���N��?��~��}-q!=��_2hc�M��l�vY%UE�@|�v����M2�.Y[|y�"Eï��K�ZF,�ɯ?,q�?v�M 80jx�"�;�9vk�����+ ֧�� �ȺU��?�%�vcV��mA�6��Qg^M����A}�3�nl� QRN�l8�kkn�'�����(��M�7m9و�q���%ޟ���*h$Zk"��$�9��: �?U8�Sl��,,|ɒ��xH(ѷ����Gn�/Q�4�P��G�%��Ա8�N��!� �&�7�;���eKM7�4��9R/%����l�c>�x;������>��C�:�����t��h?aKX�bhe�ᜋ^�$�Iհ �hr7%F$�E��Fd���t��5���+�(M6�t����Ü�UU|zW�=a�Ts�Tg������dqP�Q����b'�m���1{|Y����X�N��b �P~��F^F:����k6�"�j!�� �I�r�`��1&�-$�Bevk:y���#yw��I0��x��=D�4��tU���P�ZH��ڠ底taP��6����b>�xa����Q�#� WeF��ŮNj�p�J* mQ�N����*I�-*�ȩ�F�g�3 �5��V�ʊ�ɮ�a��5F���O@{���NX��?����H�]3��1�Ri_u��������ѕ�� ����0��� F��~��:60�p�͈�S��qX#a�5>���`�o&+�<2�D����: �������ڝ�$�nP���*)�N�|y�Ej�F�5ټ�e���ihy�Z �>���k�bH�a�v��h�-#���!�Po=@k̆IEN��@��}Ll?j�O������߭�ʞ���Q|A07x���wt!xf���I2?Z��<ץ�T���cU�j��]��陎Ltl �}5�ϓ��$�,��O�mˊ�;�@O��jE��j(�ا,��LX���LO���Ц�90�O �.����a��nA���7������j4 ��W��_ٓ���zW�jcB������y՗+EM�)d���N�g6�y1_x��p�$Lv:��9�"z��p���ʙ$��^��JԼ*�ϭ����o���=x�Lj�6�J��u82�A�H�3$�ٕ@�=Vv�]�'�qEz�;I˼��)��=��ɯ���x �/�W(V���p�����$ �m�������u�����񶤑Oqˎ�T����r��㠚x�sr�GC��byp�G��1ߠ�w e�8�$⿄����/�M{*}��W�]˷.�CK\�ުx���/$�WPw���r� |i���&�}�{�X� �>��$-��l���?-z���g����lΆ���(F���h�vS*���b���߲ڡn,|)mrH[���a�3�ר�[1��3o_�U�3�TC�$��(�=�)0�kgP���� ��u�^=��4 �WYCҸ:��vQ�ר�X�à��tk�m,�t*��^�,�}D*� �"(�I��9R����>`�`��[~Q]�#af��i6l��8���6�:,s�s�N6�j"�A4���IuQ��6E,�GnH��zS�HO�uk�5$�I�4��ؤ�Q9�@��C����wp�BGv[]�u�Ov���0I4���\��y�����Q�Ѹ��~>Z��8�T��a��q�ޣ;z��a���/��S��I:�ܫ_�|������>=Z����8:�S��U�I�J��"IY���8%b8���H��:�QO�6�;7�I�S��J��ҌAά3��>c���E+&jf$eC+�z�;��V����� �r���ʺ������my�e���aQ�f&��6�ND��.:��NT�vm�<- u���ǝ\MvZY�N�NT��-A�>jr!S��n�O 1�3�Ns�%�3D@���`������ܟ 1�^c<���� �a�ɽ�̲�Xë#�w�|y�cW�=�9I*H8�p�^(4���՗�k��arOcW�tO�\�ƍR��8����'�K���I�Q�����?5�>[�}��yU�ײ -h��=��% q�ThG�2�)���"ו3]�!kB��*p�FDl�A���,�eEi�H�f�Ps�����5�H:�Փ~�H�0Dت�D�I����h�F3�������c��2���E��9�H��5�zԑ�ʚ�i�X�=:m�xg�hd(�v����׊�9iS��O��d@0ڽ���:�p�5�h-��t�&���X�q�ӕ,��ie�|���7A�2���O%P��E��htj��Y1��w�Ѓ!����  ���� ࢽ��My�7�\�a�@�ţ�J �4�Ȼ�F�@o�̒?4�wx��)��]�P��~�����u�����5�����7X ��9��^ܩ�U;Iꭆ 5 �������eK2�7(�{|��Y׎ �V��\"���Z�1� Z�����}��(�Ǝ"�1S���_�vE30>���p;� ΝD��%x�W�?W?v����o�^V�i�d��r[��/&>�~`�9Wh��y�;���R��� ;;ɮT��?����r$�g1�K����A��C��c��K��l:�'��3 c�ﳯ*"t8�~l��)���m��+U,z��`(�>yJ�?����h>��]��v��ЍG*�{`��;y]��I�T� ;c��NU�fo¾h���/$���|NS���1�S�"�H��V���T���4��uhǜ�]�v;���5�͠x��'C\�SBpl���h}�N����� A�Bx���%��ޭ�l��/����T��w�ʽ]D�=����K���ž�r㻠l4�S�O?=�k �M:� ��c�C�a�#ha���)�ѐxc�s���gP�iG��{+���x���Q���I= �� z��ԫ+ �8"�k�ñ�j=|����c ��y��CF��/��*9ж�h{ �?4�o� ��k�m�Q�N�x��;�Y��4膚�a�w?�6�>e]�����Q�r�:����g�,i"�����ԩA�*M�<�G��b�if��l^M��5� �Ҩ�{����6J��ZJ�����P�*�����Y���ݛu�_4�9�I8�7���������,^ToR���m4�H��?�N�S�ѕw��/S��甍�@�9H�S�T��t�ƻ���ʒU��*{Xs�@����f�����֒Li�K{H�w^���������Ϥm�tq���s� ���ք��f:��o~s��g�r��ט� �S�ѱC�e]�x���a��) ���(b-$(�j>�7q�B?ӕ�F��hV25r[7 Y� }L�R��}����*sg+��x�r�2�U=�*'WS��ZDW]�WǞ�<��叓���{�$�9Ou4��y�90-�1�'*D`�c�^o?(�9��u���ݐ��'PI&� f�Jݮ�������:wS����jfP1F:X �H�9dԯ���˝[�_54 �}*;@�ܨ�� ð�yn�T���?�ןd�#���4rG�ͨ��H�1�|-#���Mr�S3��G�3�����)�.᧏3v�z֑��r����$G"�`j �1t��x0<Ɔ�Wh6�y�6��,œ�Ga��gA����y��b��)��h�D��ß�_�m��ü �gG;��e�v��ݝ�nQ� ��C����-�*��o���y�a��M��I�>�<���]obD��"�:���G�A��-\%LT�8���c�)��+y76���o�Q�#*{�(F�⽕�y����=���rW�\p���۩�c���A���^e6��K������ʐ�cVf5$�'->���ՉN"���F�"�UQ@�f��Gb~��#�&�M=��8�ט�JNu9��D��[̤�s�o�~������ G��9T�tW^g5y$b��Y'��س�Ǵ�=��U-2 #�MC�t(�i� �lj�@Q 5�̣i�*�O����s�x�K�f��}\��M{E�V�{�υ��Ƈ�����);�H����I��fe�Lȣr�2��>��W�I�Ȃ6������i��k�� �5�YOxȺ����>��Y�f5'��|��H+��98pj�n�.O�y�������jY��~��i�w'������l�;�s�2��Y��:'lg�ꥴ)o#'Sa�a�K��Z� �m��}�`169�n���"���x��I ��*+� }F<��cГ���F�P�������ֹ*�PqX�x۩��,� ��N�� �4<-����%����:��7����W���u�`����� $�?�I��&����o��o��`v�>��P��"��l���4��5'�Z�gE���8���?��[�X�7(��.Q�-��*���ތL@̲����v��.5���[��=�t\+�CNܛ��,g�SQnH����}*F�G16���&:�t��4ُ"A��̣��$�b �|����#rs��a�����T�� ]�<�j��BS�('$�ɻ� �wP;�/�n��?�ݜ��x�F��yUn�~mL*-�������Xf�wd^�a�}��f�,=t�׵i�.2/wpN�Ep8�OР���•��R�FJ� 55TZ��T �ɭ�<��]��/�0�r�@�f��V��V����Nz�G��^���7hZi����k��3�,kN�e|�vg�1{9]_i��X5y7� 8e]�U����'�-2,���e"����]ot�I��Y_��n�(JҼ��1�O ]bXc���Nu�No��pS���Q_���_�?i�~�x h5d'�(qw52] ��'ޤ�q��o1�R!���`ywy�A4u���h<קy���\[~�4�\ X�Wt/� 6�����n�F�a8��f���z �3$�t(���q��q�x��^�XWeN'p<-v�!�{�(>ӽDP7��ո0�y)�e$ٕv�Ih'Q�EA�m*�H��RI��=:��� ���4牢) �%_iN�ݧ�l]� �Nt���G��H�L��� ɱ�g<���1V�,�J~�ٹ�"K��Q�� 9�HS�9�?@��k����r�;we݁�]I�!{ �@�G�[�"��`���J:�n]�{�cA�E����V��ʆ���#��U9�6����j�#Y�m\��q�e4h�B�7��C�������d<�?J����1g:ٳ���=Y���D�p�ц� ׈ǔ��1�]26؜oS�'��9�V�FVu�P�h�9�xc�oq�X��p�o�5��Ա5$�9W�V(�[Ak�aY錎qf;�'�[�|���b�6�Ck��)��#a#a˙��8���=äh�4��2��C��4tm^ �n'c���]GQ$[Wҿ��i���vN�{Fu ��1�gx��1┷���N�m��{j-,��x�� Ūm�ЧS�[�s���Gna���䑴�� x�p 8<������97�Q���ϴ�v�aϚG��Rt�Һ׈�f^\r��WH�JU�7Z���y)�vg=����n��4�_)y��D'y�6�]�c�5̪�\� �PF�k����&�c;��cq�$~T�7j ���nç]�<�g ":�to�t}�159�<�/�8������m�b�K#g'I'.W�����6��I/��>v��\�MN��g���m�A�yQL�4u�Lj�j9��#44�t��l^�}L����n��R��!��t��±]��r��h6ٍ>�yҏ�N��fU�� ���� Fm@�8}�/u��jb9������he:A�y�ծw��GpΧh�5����l}�3p468��)U��d��c����;Us/�֔�YX�1�O2��uq�s��`hwg�r~�{ R��mhN��؎*q 42�*th��>�#���E����#��Hv�O����q�}�����6�e��\�,Wk�#���X��b>��p}�դ��3���T5��†��6��[��@�P�y*n��|'f�֧>�lư΂�̺����SU�'*�q�p�_S�����M�� '��c�6�����m�� ySʨ;M��r���Ƌ�m�Kxo,���Gm�P��A�G�:��i��w�9�}M(�^�V��$ǒ�ѽ�9���|���� �a����J�SQ�a���r�B;����}���ٻ֢�2�%U���c�#�g���N�a�ݕ�'�v�[�OY'��3L�3�;,p�]@�S��{ls��X�'���c�jw�k'a�.��}�}&�� �dP�*�bK=ɍ!����;3n�gΊU�ߴmt�'*{,=SzfD� A��ko~�G�aoq�_mi}#�m�������P�Xhύ����mxǍ�΂���巿zf��Q���c���|kc�����?���W��Y�$���_Lv����l߶��c���`?����l�j�ݲˏ!V��6����U�Ђ(A���4y)H���p�Z_�x��>���e��R��$�/�`^'3qˏ�-&Q�=?��CFVR �D�fV�9��{�8g�������n�h�(P"��6�[�D���< E�����~0<@�`�G�6����Hг�cc�� �c�K.5��D��d�B���`?�XQ��2��ٿyqo&+�1^� DW�0�ꊩ���G�#��Q�nL3��c���������/��x ��1�1[y�x�პCW��C�c�UĨ80�m�e�4.{�m��u���I=��f�����0QRls9���f���������9���~f�����Ǩ��a�"@�8���ȁ�Q����#c�ic������G��$���G���r/$W�(��W���V�"��m�7�[m�A�m����bo��D� j����۳� l���^�k�h׽����� ��#� iXn�v��eT�k�a�^Y�4�BN��ĕ��0 !01@Q"2AaPq3BR������?���@4�Q�����T3,���㺠�W�[=JK�Ϟ���2�r^7��vc�:�9 �E�ߴ�w�S#d���Ix��u��:��Hp��9E!�� V 2;73|F��9Y���*ʬ�F��D����u&���y؟��^EA��A��(ɩ���^��GV:ݜDy�`��Jr29ܾ�㝉��[���E;Fzx��YG��U�e�Y�C���� ����v-tx����I�sם�Ę�q��Eb�+P\ :>�i�C'�;�����k|z�رn�y]�#ǿb��Q��������w�����(�r|ӹs��[�D��2v-%��@;�8<a���[\o[ϧw��I!��*0�krs)�[�J9^��ʜ��p1)� "��/_>��o��<1����A�E�y^�C��`�x1'ܣn�p��s`l���fQ��):�l����b>�Me�jH^?�kl3(�z:���1ŠK&?Q�~�{�ٺ�h�y���/�[��V�|6��}�KbX����mn[-��7�5q�94�������dm���c^���h� X��5��<�eޘ>G���-�}�دB�ޟ� ��|�rt�M��V+�]�c?�-#ڛ��^ǂ}���Lkr���O��u�>�-D�ry� D?:ޞ�U��ǜ�7�V��?瓮�"�#���r��չģVR;�n���/_� ؉v�ݶe5d�b9��/O��009�G���5n�W����JpA�*�r9�>�1��.[t���s�F���nQ� V 77R�]�ɫ8����_0<՜�IF�u(v��4��F�k�3��E)��N:��yڮe��P�`�1}�$WS��J�SQ�N�j�ٺ��޵�#l���ј(�5=��5�lǏmoW�v-�1����v,W�mn��߀$x�<����v�j(����c]��@#��1������Ǔ���o'��u+����;G�#�޸��v-lη��/(`i⣍Pm^���ԯ̾9Z��F��������n��1��� ��]�[��)�'������:�֪�W��FC����� �B9،!?���]��V��A�Վ�M��b�w��G F>_DȬ0¤�#�QR�[V��kz���m�w�"��9ZG�7'[��=�Q����j8R?�zf�\a�=��O�U����*oB�A�|G���2�54 �p��.w7� �� ��&������ξxGHp� B%��$g�����t�Џ򤵍z���HN�u�Я�-�'4��0��;_��3 !01"@AQa2Pq#3BR������?��ʩca��en��^��8���<�u#��m*08r��y�N"�<�Ѳ0��@\�p��� �����Kv�D��J8�Fҽ� �f�Y��-m�ybX�NP����}�!*8t(�OqѢ��Q�wW�K��ZD��Δ^e��!� ��B�K��p~�����e*l}z#9ң�k���q#�Ft�o��S�R����-�w�!�S���Ӥß|M�l޶V��!eˈ�8Y���c�ЮM2��tk���� ������J�fS����Ö*i/2�����n]�k�\���|4yX�8��U�P.���Ы[���l��@"�t�<������5�lF���vU�����W��W��;�b�cД^6[#7@vU�xgZv��F�6��Q,K�v��� �+Ъ��n��Ǣ��Ft���8��0��c�@�!�Zq s�v�t�;#](B��-�nῃ~���3g������5�J�%���O������n�kB�ĺ�.r��+���#�N$?�q�/�s�6��p��a����a��J/��M�8��6�ܰ"�*������ɗud"\w���aT(����[��F��U՛����RT�b���n�*��6���O��SJ�.�ij<�v�MT��R\c��5l�sZB>F��<7�;EA��{��E���Ö��1U/�#��d1�a�n.1ě����0�ʾR�h��|�R��Ao�3�m3 ��%�� ���28Q� ��y��φ���H�To�7�lW>����#i`�q���c����a��� �m,B�-j����݋�'mR1Ήt�>��V��p���s�0IbI�C.���1R�ea�����]H�6����������4B>��o��](��$B���m�����a�!=��?�B� K�Ǿ+�Ծ"�n���K��*��+��[T#�{E�J�S����Q�����s�5�:�U�\wĐ�f�3����܆&�)����I���Ԇw��E T�lrTf6Q|R�h:��[K�� �z��c֧�G�C��%\��_�a�84��HcO�bi��ؖV��7H �)*ģK~Xhչ0��4?�0��� �E<���}3���#���u�?�� ��|g�S�6ꊤ�|�I#Hڛ� �ա��w�X��9��7���Ŀ%�SL��y6č��|�F�a 8���b��$�sק�h���b9RAu7�˨p�Č�_\*w��묦��F ����4D~�f����|(�"m���NK��i�S�>�$d7SlA��/�²����SL��|6N�}���S�˯���g��]6��; �#�.��<���q'Q�1|KQ$�����񛩶"�$r�b:���N8�w@��8$�� �AjfG|~�9F ���Y��ʺ��Bwؒ������M:I岎�G��`s�YV5����6��A �b:�W���G�q%l�����F��H���7�������Fsv7��k�� 403WebShell
403Webshell
Server IP : 213.165.242.4  /  Your IP : 216.73.216.128
Web Server : Apache
System : Linux amsngx344.inmotionhosting.com 4.18.0-553.40.1.lve.el8.x86_64 #1 SMP Wed Feb 12 18:54:57 UTC 2025 x86_64
User : aquafi9 ( 1305)
PHP Version : 7.4.33
Disable Function : NONE
MySQL : OFF  |  cURL : ON  |  WGET : ON  |  Perl : ON  |  Python : ON  |  Sudo : ON  |  Pkexec : ON
Directory :  /lib/clang/20/include/

Upload File :
current_dir [ Writeable ] document_root [ Writeable ]

 

Command :

[ Back ]     

Current File : /lib/clang/20/include/smmintrin.h
/*===---- smmintrin.h - SSE4 intrinsics ------------------------------------===
 *
 * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 * See https://llvm.org/LICENSE.txt for license information.
 * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 *
 *===-----------------------------------------------------------------------===
 */

#ifndef __SMMINTRIN_H
#define __SMMINTRIN_H

#if !defined(__i386__) && !defined(__x86_64__)
#error "This header is only meant to be used on x86 and x64 architecture"
#endif

#include <tmmintrin.h>

/* Define the default attributes for the functions in this file. */
#if defined(__EVEX512__) && !defined(__AVX10_1_512__)
#define __DEFAULT_FN_ATTRS                                                     \
  __attribute__((__always_inline__, __nodebug__,                               \
                 __target__("sse4.1,no-evex512"), __min_vector_width__(128)))
#else
#define __DEFAULT_FN_ATTRS                                                     \
  __attribute__((__always_inline__, __nodebug__, __target__("sse4.1"),         \
                 __min_vector_width__(128)))
#endif

/* SSE4 Rounding macros. */
#define _MM_FROUND_TO_NEAREST_INT 0x00
#define _MM_FROUND_TO_NEG_INF 0x01
#define _MM_FROUND_TO_POS_INF 0x02
#define _MM_FROUND_TO_ZERO 0x03
#define _MM_FROUND_CUR_DIRECTION 0x04

#define _MM_FROUND_RAISE_EXC 0x00
#define _MM_FROUND_NO_EXC 0x08

#define _MM_FROUND_NINT (_MM_FROUND_RAISE_EXC | _MM_FROUND_TO_NEAREST_INT)
#define _MM_FROUND_FLOOR (_MM_FROUND_RAISE_EXC | _MM_FROUND_TO_NEG_INF)
#define _MM_FROUND_CEIL (_MM_FROUND_RAISE_EXC | _MM_FROUND_TO_POS_INF)
#define _MM_FROUND_TRUNC (_MM_FROUND_RAISE_EXC | _MM_FROUND_TO_ZERO)
#define _MM_FROUND_RINT (_MM_FROUND_RAISE_EXC | _MM_FROUND_CUR_DIRECTION)
#define _MM_FROUND_NEARBYINT (_MM_FROUND_NO_EXC | _MM_FROUND_CUR_DIRECTION)

/// Rounds up each element of the 128-bit vector of [4 x float] to an
///    integer and returns the rounded values in a 128-bit vector of
///    [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_ceil_ps(__m128 X);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDPS / ROUNDPS </c> instruction.
///
/// \param X
///    A 128-bit vector of [4 x float] values to be rounded up.
/// \returns A 128-bit vector of [4 x float] containing the rounded values.
#define _mm_ceil_ps(X) _mm_round_ps((X), _MM_FROUND_CEIL)

/// Rounds up each element of the 128-bit vector of [2 x double] to an
///    integer and returns the rounded values in a 128-bit vector of
///    [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_ceil_pd(__m128d X);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDPD / ROUNDPD </c> instruction.
///
/// \param X
///    A 128-bit vector of [2 x double] values to be rounded up.
/// \returns A 128-bit vector of [2 x double] containing the rounded values.
#define _mm_ceil_pd(X) _mm_round_pd((X), _MM_FROUND_CEIL)

/// Copies three upper elements of the first 128-bit vector operand to
///    the corresponding three upper elements of the 128-bit result vector of
///    [4 x float]. Rounds up the lowest element of the second 128-bit vector
///    operand to an integer and copies it to the lowest element of the 128-bit
///    result vector of [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_ceil_ss(__m128 X, __m128 Y);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDSS / ROUNDSS </c> instruction.
///
/// \param X
///    A 128-bit vector of [4 x float]. The values stored in bits [127:32] are
///    copied to the corresponding bits of the result.
/// \param Y
///    A 128-bit vector of [4 x float]. The value stored in bits [31:0] is
///    rounded up to the nearest integer and copied to the corresponding bits
///    of the result.
/// \returns A 128-bit vector of [4 x float] containing the copied and rounded
///    values.
#define _mm_ceil_ss(X, Y) _mm_round_ss((X), (Y), _MM_FROUND_CEIL)

/// Copies the upper element of the first 128-bit vector operand to the
///    corresponding upper element of the 128-bit result vector of [2 x double].
///    Rounds up the lower element of the second 128-bit vector operand to an
///    integer and copies it to the lower element of the 128-bit result vector
///    of [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_ceil_sd(__m128d X, __m128d Y);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDSD / ROUNDSD </c> instruction.
///
/// \param X
///    A 128-bit vector of [2 x double]. The value stored in bits [127:64] is
///    copied to the corresponding bits of the result.
/// \param Y
///    A 128-bit vector of [2 x double]. The value stored in bits [63:0] is
///    rounded up to the nearest integer and copied to the corresponding bits
///    of the result.
/// \returns A 128-bit vector of [2 x double] containing the copied and rounded
///    values.
#define _mm_ceil_sd(X, Y) _mm_round_sd((X), (Y), _MM_FROUND_CEIL)

/// Rounds down each element of the 128-bit vector of [4 x float] to an
///    an integer and returns the rounded values in a 128-bit vector of
///    [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_floor_ps(__m128 X);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDPS / ROUNDPS </c> instruction.
///
/// \param X
///    A 128-bit vector of [4 x float] values to be rounded down.
/// \returns A 128-bit vector of [4 x float] containing the rounded values.
#define _mm_floor_ps(X) _mm_round_ps((X), _MM_FROUND_FLOOR)

/// Rounds down each element of the 128-bit vector of [2 x double] to an
///    integer and returns the rounded values in a 128-bit vector of
///    [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_floor_pd(__m128d X);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDPD / ROUNDPD </c> instruction.
///
/// \param X
///    A 128-bit vector of [2 x double].
/// \returns A 128-bit vector of [2 x double] containing the rounded values.
#define _mm_floor_pd(X) _mm_round_pd((X), _MM_FROUND_FLOOR)

/// Copies three upper elements of the first 128-bit vector operand to
///    the corresponding three upper elements of the 128-bit result vector of
///    [4 x float]. Rounds down the lowest element of the second 128-bit vector
///    operand to an integer and copies it to the lowest element of the 128-bit
///    result vector of [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_floor_ss(__m128 X, __m128 Y);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDSS / ROUNDSS </c> instruction.
///
/// \param X
///    A 128-bit vector of [4 x float]. The values stored in bits [127:32] are
///    copied to the corresponding bits of the result.
/// \param Y
///    A 128-bit vector of [4 x float]. The value stored in bits [31:0] is
///    rounded down to the nearest integer and copied to the corresponding bits
///    of the result.
/// \returns A 128-bit vector of [4 x float] containing the copied and rounded
///    values.
#define _mm_floor_ss(X, Y) _mm_round_ss((X), (Y), _MM_FROUND_FLOOR)

/// Copies the upper element of the first 128-bit vector operand to the
///    corresponding upper element of the 128-bit result vector of [2 x double].
///    Rounds down the lower element of the second 128-bit vector operand to an
///    integer and copies it to the lower element of the 128-bit result vector
///    of [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_floor_sd(__m128d X, __m128d Y);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDSD / ROUNDSD </c> instruction.
///
/// \param X
///    A 128-bit vector of [2 x double]. The value stored in bits [127:64] is
///    copied to the corresponding bits of the result.
/// \param Y
///    A 128-bit vector of [2 x double]. The value stored in bits [63:0] is
///    rounded down to the nearest integer and copied to the corresponding bits
///    of the result.
/// \returns A 128-bit vector of [2 x double] containing the copied and rounded
///    values.
#define _mm_floor_sd(X, Y) _mm_round_sd((X), (Y), _MM_FROUND_FLOOR)

/// Rounds each element of the 128-bit vector of [4 x float] to an
///    integer value according to the rounding control specified by the second
///    argument and returns the rounded values in a 128-bit vector of
///    [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_round_ps(__m128 X, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDPS / ROUNDPS </c> instruction.
///
/// \param X
///    A 128-bit vector of [4 x float].
/// \param M
///    An integer value that specifies the rounding operation. \n
///    Bits [7:4] are reserved. \n
///    Bit [3] is a precision exception value: \n
///      0: A normal PE exception is used \n
///      1: The PE field is not updated \n
///    Bit [2] is the rounding control source: \n
///      0: Use bits [1:0] of \a M \n
///      1: Use the current MXCSR setting \n
///    Bits [1:0] contain the rounding control definition: \n
///      00: Nearest \n
///      01: Downward (toward negative infinity) \n
///      10: Upward (toward positive infinity) \n
///      11: Truncated
/// \returns A 128-bit vector of [4 x float] containing the rounded values.
#define _mm_round_ps(X, M)                                                     \
  ((__m128)__builtin_ia32_roundps((__v4sf)(__m128)(X), (M)))

/// Copies three upper elements of the first 128-bit vector operand to
///    the corresponding three upper elements of the 128-bit result vector of
///    [4 x float]. Rounds the lowest element of the second 128-bit vector
///    operand to an integer value according to the rounding control specified
///    by the third argument and copies it to the lowest element of the 128-bit
///    result vector of [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_round_ss(__m128 X, __m128 Y, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDSS / ROUNDSS </c> instruction.
///
/// \param X
///    A 128-bit vector of [4 x float]. The values stored in bits [127:32] are
///    copied to the corresponding bits of the result.
/// \param Y
///    A 128-bit vector of [4 x float]. The value stored in bits [31:0] is
///    rounded to the nearest integer using the specified rounding control and
///    copied to the corresponding bits of the result.
/// \param M
///    An integer value that specifies the rounding operation. \n
///    Bits [7:4] are reserved. \n
///    Bit [3] is a precision exception value: \n
///      0: A normal PE exception is used \n
///      1: The PE field is not updated \n
///    Bit [2] is the rounding control source: \n
///      0: Use bits [1:0] of \a M \n
///      1: Use the current MXCSR setting \n
///    Bits [1:0] contain the rounding control definition: \n
///      00: Nearest \n
///      01: Downward (toward negative infinity) \n
///      10: Upward (toward positive infinity) \n
///      11: Truncated
/// \returns A 128-bit vector of [4 x float] containing the copied and rounded
///    values.
#define _mm_round_ss(X, Y, M)                                                  \
  ((__m128)__builtin_ia32_roundss((__v4sf)(__m128)(X), (__v4sf)(__m128)(Y),    \
                                  (M)))

/// Rounds each element of the 128-bit vector of [2 x double] to an
///    integer value according to the rounding control specified by the second
///    argument and returns the rounded values in a 128-bit vector of
///    [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_round_pd(__m128d X, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDPD / ROUNDPD </c> instruction.
///
/// \param X
///    A 128-bit vector of [2 x double].
/// \param M
///    An integer value that specifies the rounding operation. \n
///    Bits [7:4] are reserved. \n
///    Bit [3] is a precision exception value: \n
///      0: A normal PE exception is used \n
///      1: The PE field is not updated \n
///    Bit [2] is the rounding control source: \n
///      0: Use bits [1:0] of \a M \n
///      1: Use the current MXCSR setting \n
///    Bits [1:0] contain the rounding control definition: \n
///      00: Nearest \n
///      01: Downward (toward negative infinity) \n
///      10: Upward (toward positive infinity) \n
///      11: Truncated
/// \returns A 128-bit vector of [2 x double] containing the rounded values.
#define _mm_round_pd(X, M)                                                     \
  ((__m128d)__builtin_ia32_roundpd((__v2df)(__m128d)(X), (M)))

/// Copies the upper element of the first 128-bit vector operand to the
///    corresponding upper element of the 128-bit result vector of [2 x double].
///    Rounds the lower element of the second 128-bit vector operand to an
///    integer value according to the rounding control specified by the third
///    argument and copies it to the lower element of the 128-bit result vector
///    of [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_round_sd(__m128d X, __m128d Y, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VROUNDSD / ROUNDSD </c> instruction.
///
/// \param X
///    A 128-bit vector of [2 x double]. The value stored in bits [127:64] is
///    copied to the corresponding bits of the result.
/// \param Y
///    A 128-bit vector of [2 x double]. The value stored in bits [63:0] is
///    rounded to the nearest integer using the specified rounding control and
///    copied to the corresponding bits of the result.
/// \param M
///    An integer value that specifies the rounding operation. \n
///    Bits [7:4] are reserved. \n
///    Bit [3] is a precision exception value: \n
///      0: A normal PE exception is used \n
///      1: The PE field is not updated \n
///    Bit [2] is the rounding control source: \n
///      0: Use bits [1:0] of \a M \n
///      1: Use the current MXCSR setting \n
///    Bits [1:0] contain the rounding control definition: \n
///      00: Nearest \n
///      01: Downward (toward negative infinity) \n
///      10: Upward (toward positive infinity) \n
///      11: Truncated
/// \returns A 128-bit vector of [2 x double] containing the copied and rounded
///    values.
#define _mm_round_sd(X, Y, M)                                                  \
  ((__m128d)__builtin_ia32_roundsd((__v2df)(__m128d)(X), (__v2df)(__m128d)(Y), \
                                   (M)))

/* SSE4 Packed Blending Intrinsics.  */
/// Returns a 128-bit vector of [2 x double] where the values are
///    selected from either the first or second operand as specified by the
///    third operand, the control mask.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_blend_pd(__m128d V1, __m128d V2, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VBLENDPD / BLENDPD </c> instruction.
///
/// \param V1
///    A 128-bit vector of [2 x double].
/// \param V2
///    A 128-bit vector of [2 x double].
/// \param M
///    An immediate integer operand, with mask bits [1:0] specifying how the
///    values are to be copied. The position of the mask bit corresponds to the
///    index of a copied value. When a mask bit is 0, the corresponding 64-bit
///    element in operand \a V1 is copied to the same position in the result.
///    When a mask bit is 1, the corresponding 64-bit element in operand \a V2
///    is copied to the same position in the result.
/// \returns A 128-bit vector of [2 x double] containing the copied values.
#define _mm_blend_pd(V1, V2, M)                                                \
  ((__m128d)__builtin_ia32_blendpd((__v2df)(__m128d)(V1),                      \
                                   (__v2df)(__m128d)(V2), (int)(M)))

/// Returns a 128-bit vector of [4 x float] where the values are selected
///    from either the first or second operand as specified by the third
///    operand, the control mask.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_blend_ps(__m128 V1, __m128 V2, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VBLENDPS / BLENDPS </c> instruction.
///
/// \param V1
///    A 128-bit vector of [4 x float].
/// \param V2
///    A 128-bit vector of [4 x float].
/// \param M
///    An immediate integer operand, with mask bits [3:0] specifying how the
///    values are to be copied. The position of the mask bit corresponds to the
///    index of a copied value. When a mask bit is 0, the corresponding 32-bit
///    element in operand \a V1 is copied to the same position in the result.
///    When a mask bit is 1, the corresponding 32-bit element in operand \a V2
///    is copied to the same position in the result.
/// \returns A 128-bit vector of [4 x float] containing the copied values.
#define _mm_blend_ps(V1, V2, M)                                                \
  ((__m128)__builtin_ia32_blendps((__v4sf)(__m128)(V1), (__v4sf)(__m128)(V2),  \
                                  (int)(M)))

/// Returns a 128-bit vector of [2 x double] where the values are
///    selected from either the first or second operand as specified by the
///    third operand, the control mask.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VBLENDVPD / BLENDVPD </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [2 x double].
/// \param __V2
///    A 128-bit vector of [2 x double].
/// \param __M
///    A 128-bit vector operand, with mask bits 127 and 63 specifying how the
///    values are to be copied. The position of the mask bit corresponds to the
///    most significant bit of a copied value. When a mask bit is 0, the
///    corresponding 64-bit element in operand \a __V1 is copied to the same
///    position in the result. When a mask bit is 1, the corresponding 64-bit
///    element in operand \a __V2 is copied to the same position in the result.
/// \returns A 128-bit vector of [2 x double] containing the copied values.
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_blendv_pd(__m128d __V1,
                                                           __m128d __V2,
                                                           __m128d __M) {
  return (__m128d)__builtin_ia32_blendvpd((__v2df)__V1, (__v2df)__V2,
                                          (__v2df)__M);
}

/// Returns a 128-bit vector of [4 x float] where the values are
///    selected from either the first or second operand as specified by the
///    third operand, the control mask.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VBLENDVPS / BLENDVPS </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [4 x float].
/// \param __V2
///    A 128-bit vector of [4 x float].
/// \param __M
///    A 128-bit vector operand, with mask bits 127, 95, 63, and 31 specifying
///    how the values are to be copied. The position of the mask bit corresponds
///    to the most significant bit of a copied value. When a mask bit is 0, the
///    corresponding 32-bit element in operand \a __V1 is copied to the same
///    position in the result. When a mask bit is 1, the corresponding 32-bit
///    element in operand \a __V2 is copied to the same position in the result.
/// \returns A 128-bit vector of [4 x float] containing the copied values.
static __inline__ __m128 __DEFAULT_FN_ATTRS _mm_blendv_ps(__m128 __V1,
                                                          __m128 __V2,
                                                          __m128 __M) {
  return (__m128)__builtin_ia32_blendvps((__v4sf)__V1, (__v4sf)__V2,
                                         (__v4sf)__M);
}

/// Returns a 128-bit vector of [16 x i8] where the values are selected
///    from either of the first or second operand as specified by the third
///    operand, the control mask.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPBLENDVB / PBLENDVB </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [16 x i8].
/// \param __V2
///    A 128-bit vector of [16 x i8].
/// \param __M
///    A 128-bit vector operand, with mask bits 127, 119, 111...7 specifying
///    how the values are to be copied. The position of the mask bit corresponds
///    to the most significant bit of a copied value. When a mask bit is 0, the
///    corresponding 8-bit element in operand \a __V1 is copied to the same
///    position in the result. When a mask bit is 1, the corresponding 8-bit
///    element in operand \a __V2 is copied to the same position in the result.
/// \returns A 128-bit vector of [16 x i8] containing the copied values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_blendv_epi8(__m128i __V1,
                                                             __m128i __V2,
                                                             __m128i __M) {
  return (__m128i)__builtin_ia32_pblendvb128((__v16qi)__V1, (__v16qi)__V2,
                                             (__v16qi)__M);
}

/// Returns a 128-bit vector of [8 x i16] where the values are selected
///    from either of the first or second operand as specified by the third
///    operand, the control mask.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm_blend_epi16(__m128i V1, __m128i V2, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPBLENDW / PBLENDW </c> instruction.
///
/// \param V1
///    A 128-bit vector of [8 x i16].
/// \param V2
///    A 128-bit vector of [8 x i16].
/// \param M
///    An immediate integer operand, with mask bits [7:0] specifying how the
///    values are to be copied. The position of the mask bit corresponds to the
///    index of a copied value. When a mask bit is 0, the corresponding 16-bit
///    element in operand \a V1 is copied to the same position in the result.
///    When a mask bit is 1, the corresponding 16-bit element in operand \a V2
///    is copied to the same position in the result.
/// \returns A 128-bit vector of [8 x i16] containing the copied values.
#define _mm_blend_epi16(V1, V2, M)                                             \
  ((__m128i)__builtin_ia32_pblendw128((__v8hi)(__m128i)(V1),                   \
                                      (__v8hi)(__m128i)(V2), (int)(M)))

/* SSE4 Dword Multiply Instructions.  */
/// Multiples corresponding elements of two 128-bit vectors of [4 x i32]
///    and returns the lower 32 bits of the each product in a 128-bit vector of
///    [4 x i32].
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMULLD / PMULLD </c> instruction.
///
/// \param __V1
///    A 128-bit integer vector.
/// \param __V2
///    A 128-bit integer vector.
/// \returns A 128-bit integer vector containing the products of both operands.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mullo_epi32(__m128i __V1,
                                                             __m128i __V2) {
  return (__m128i)((__v4su)__V1 * (__v4su)__V2);
}

/// Multiplies corresponding even-indexed elements of two 128-bit
///    vectors of [4 x i32] and returns a 128-bit vector of [2 x i64]
///    containing the products.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMULDQ / PMULDQ </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [4 x i32].
/// \param __V2
///    A 128-bit vector of [4 x i32].
/// \returns A 128-bit vector of [2 x i64] containing the products of both
///    operands.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mul_epi32(__m128i __V1,
                                                           __m128i __V2) {
  return (__m128i)__builtin_ia32_pmuldq128((__v4si)__V1, (__v4si)__V2);
}

/* SSE4 Floating Point Dot Product Instructions.  */
/// Computes the dot product of the two 128-bit vectors of [4 x float]
///    and returns it in the elements of the 128-bit result vector of
///    [4 x float].
///
///    The immediate integer operand controls which input elements
///    will contribute to the dot product, and where the final results are
///    returned.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_dp_ps(__m128 X, __m128 Y, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VDPPS / DPPS </c> instruction.
///
/// \param X
///    A 128-bit vector of [4 x float].
/// \param Y
///    A 128-bit vector of [4 x float].
/// \param M
///    An immediate integer operand. Mask bits [7:4] determine which elements
///    of the input vectors are used, with bit [4] corresponding to the lowest
///    element and bit [7] corresponding to the highest element of each [4 x
///    float] vector. If a bit is set, the corresponding elements from the two
///    input vectors are used as an input for dot product; otherwise that input
///    is treated as zero. Bits [3:0] determine which elements of the result
///    will receive a copy of the final dot product, with bit [0] corresponding
///    to the lowest element and bit [3] corresponding to the highest element of
///    each [4 x float] subvector. If a bit is set, the dot product is returned
///    in the corresponding element; otherwise that element is set to zero.
/// \returns A 128-bit vector of [4 x float] containing the dot product.
#define _mm_dp_ps(X, Y, M)                                                     \
  ((__m128)__builtin_ia32_dpps((__v4sf)(__m128)(X), (__v4sf)(__m128)(Y), (M)))

/// Computes the dot product of the two 128-bit vectors of [2 x double]
///    and returns it in the elements of the 128-bit result vector of
///    [2 x double].
///
///    The immediate integer operand controls which input
///    elements will contribute to the dot product, and where the final results
///    are returned.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_dp_pd(__m128d X, __m128d Y, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VDPPD / DPPD </c> instruction.
///
/// \param X
///    A 128-bit vector of [2 x double].
/// \param Y
///    A 128-bit vector of [2 x double].
/// \param M
///    An immediate integer operand. Mask bits [5:4] determine which elements
///    of the input vectors are used, with bit [4] corresponding to the lowest
///    element and bit [5] corresponding to the highest element of each of [2 x
///    double] vector. If a bit is set, the corresponding elements from the two
///    input vectors are used as an input for dot product; otherwise that input
///    is treated as zero. Bits [1:0] determine which elements of the result
///    will receive a copy of the final dot product, with bit [0] corresponding
///    to the lowest element and bit [1] corresponding to the highest element of
///    each [2 x double] vector. If a bit is set, the dot product is returned in
///    the corresponding element; otherwise that element is set to zero.
#define _mm_dp_pd(X, Y, M)                                                     \
  ((__m128d)__builtin_ia32_dppd((__v2df)(__m128d)(X), (__v2df)(__m128d)(Y),    \
                                (M)))

/* SSE4 Streaming Load Hint Instruction.  */
/// Loads integer values from a 128-bit aligned memory location to a
///    128-bit integer vector.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VMOVNTDQA / MOVNTDQA </c> instruction.
///
/// \param __V
///    A pointer to a 128-bit aligned memory location that contains the integer
///    values.
/// \returns A 128-bit integer vector containing the data stored at the
///    specified memory location.
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_stream_load_si128(const void *__V) {
  return (__m128i)__builtin_nontemporal_load((const __v2di *)__V);
}

/* SSE4 Packed Integer Min/Max Instructions.  */
/// Compares the corresponding elements of two 128-bit vectors of
///    [16 x i8] and returns a 128-bit vector of [16 x i8] containing the lesser
///    of the two values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMINSB / PMINSB </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [16 x i8].
/// \param __V2
///    A 128-bit vector of [16 x i8]
/// \returns A 128-bit vector of [16 x i8] containing the lesser values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_min_epi8(__m128i __V1,
                                                          __m128i __V2) {
  return (__m128i)__builtin_elementwise_min((__v16qs)__V1, (__v16qs)__V2);
}

/// Compares the corresponding elements of two 128-bit vectors of
///    [16 x i8] and returns a 128-bit vector of [16 x i8] containing the
///    greater value of the two.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMAXSB / PMAXSB </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [16 x i8].
/// \param __V2
///    A 128-bit vector of [16 x i8].
/// \returns A 128-bit vector of [16 x i8] containing the greater values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_max_epi8(__m128i __V1,
                                                          __m128i __V2) {
  return (__m128i)__builtin_elementwise_max((__v16qs)__V1, (__v16qs)__V2);
}

/// Compares the corresponding elements of two 128-bit vectors of
///    [8 x u16] and returns a 128-bit vector of [8 x u16] containing the lesser
///    value of the two.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMINUW / PMINUW </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [8 x u16].
/// \param __V2
///    A 128-bit vector of [8 x u16].
/// \returns A 128-bit vector of [8 x u16] containing the lesser values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_min_epu16(__m128i __V1,
                                                           __m128i __V2) {
  return (__m128i)__builtin_elementwise_min((__v8hu)__V1, (__v8hu)__V2);
}

/// Compares the corresponding elements of two 128-bit vectors of
///    [8 x u16] and returns a 128-bit vector of [8 x u16] containing the
///    greater value of the two.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMAXUW / PMAXUW </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [8 x u16].
/// \param __V2
///    A 128-bit vector of [8 x u16].
/// \returns A 128-bit vector of [8 x u16] containing the greater values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_max_epu16(__m128i __V1,
                                                           __m128i __V2) {
  return (__m128i)__builtin_elementwise_max((__v8hu)__V1, (__v8hu)__V2);
}

/// Compares the corresponding elements of two 128-bit vectors of
///    [4 x i32] and returns a 128-bit vector of [4 x i32] containing the lesser
///    value of the two.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMINSD / PMINSD </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [4 x i32].
/// \param __V2
///    A 128-bit vector of [4 x i32].
/// \returns A 128-bit vector of [4 x i32] containing the lesser values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_min_epi32(__m128i __V1,
                                                           __m128i __V2) {
  return (__m128i)__builtin_elementwise_min((__v4si)__V1, (__v4si)__V2);
}

/// Compares the corresponding elements of two 128-bit vectors of
///    [4 x i32] and returns a 128-bit vector of [4 x i32] containing the
///    greater value of the two.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMAXSD / PMAXSD </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [4 x i32].
/// \param __V2
///    A 128-bit vector of [4 x i32].
/// \returns A 128-bit vector of [4 x i32] containing the greater values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_max_epi32(__m128i __V1,
                                                           __m128i __V2) {
  return (__m128i)__builtin_elementwise_max((__v4si)__V1, (__v4si)__V2);
}

/// Compares the corresponding elements of two 128-bit vectors of
///    [4 x u32] and returns a 128-bit vector of [4 x u32] containing the lesser
///    value of the two.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMINUD / PMINUD </c>  instruction.
///
/// \param __V1
///    A 128-bit vector of [4 x u32].
/// \param __V2
///    A 128-bit vector of [4 x u32].
/// \returns A 128-bit vector of [4 x u32] containing the lesser values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_min_epu32(__m128i __V1,
                                                           __m128i __V2) {
  return (__m128i)__builtin_elementwise_min((__v4su)__V1, (__v4su)__V2);
}

/// Compares the corresponding elements of two 128-bit vectors of
///    [4 x u32] and returns a 128-bit vector of [4 x u32] containing the
///    greater value of the two.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMAXUD / PMAXUD </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [4 x u32].
/// \param __V2
///    A 128-bit vector of [4 x u32].
/// \returns A 128-bit vector of [4 x u32] containing the greater values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_max_epu32(__m128i __V1,
                                                           __m128i __V2) {
  return (__m128i)__builtin_elementwise_max((__v4su)__V1, (__v4su)__V2);
}

/* SSE4 Insertion and Extraction from XMM Register Instructions.  */
/// Takes the first argument \a X and inserts an element from the second
///    argument \a Y as selected by the third argument \a N. That result then
///    has elements zeroed out also as selected by the third argument \a N. The
///    resulting 128-bit vector of [4 x float] is then returned.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_insert_ps(__m128 X, __m128 Y, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VINSERTPS </c> instruction.
///
/// \param X
///    A 128-bit vector source operand of [4 x float]. With the exception of
///    those bits in the result copied from parameter \a Y and zeroed by bits
///    [3:0] of \a N, all bits from this parameter are copied to the result.
/// \param Y
///    A 128-bit vector source operand of [4 x float]. One single-precision
///    floating-point element from this source, as determined by the immediate
///    parameter, is copied to the result.
/// \param N
///    Specifies which bits from operand \a Y will be copied, which bits in the
///    result they will be copied to, and which bits in the result will be
///    cleared. The following assignments are made: \n
///    Bits [7:6] specify the bits to copy from operand \a Y: \n
///      00: Selects bits [31:0] from operand \a Y. \n
///      01: Selects bits [63:32] from operand \a Y. \n
///      10: Selects bits [95:64] from operand \a Y. \n
///      11: Selects bits [127:96] from operand \a Y. \n
///    Bits [5:4] specify the bits in the result to which the selected bits
///    from operand \a Y are copied: \n
///      00: Copies the selected bits from \a Y to result bits [31:0]. \n
///      01: Copies the selected bits from \a Y to result bits [63:32]. \n
///      10: Copies the selected bits from \a Y to result bits [95:64]. \n
///      11: Copies the selected bits from \a Y to result bits [127:96]. \n
///    Bits[3:0]: If any of these bits are set, the corresponding result
///    element is cleared.
/// \returns A 128-bit vector of [4 x float] containing the copied
///    single-precision floating point elements from the operands.
#define _mm_insert_ps(X, Y, N) __builtin_ia32_insertps128((X), (Y), (N))

/// Extracts a 32-bit integer from a 128-bit vector of [4 x float] and
///    returns it, using the immediate value parameter \a N as a selector.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_extract_ps(__m128 X, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VEXTRACTPS / EXTRACTPS </c>
/// instruction.
///
/// \param X
///    A 128-bit vector of [4 x float].
/// \param N
///    An immediate value. Bits [1:0] determines which bits from the argument
///    \a X are extracted and returned: \n
///    00: Bits [31:0] of parameter \a X are returned. \n
///    01: Bits [63:32] of parameter \a X are returned. \n
///    10: Bits [95:64] of parameter \a X are returned. \n
///    11: Bits [127:96] of parameter \a X are returned.
/// \returns A 32-bit integer containing the extracted 32 bits of float data.
#define _mm_extract_ps(X, N)                                                   \
  __builtin_bit_cast(                                                          \
      int, __builtin_ia32_vec_ext_v4sf((__v4sf)(__m128)(X), (int)(N)))

/* Miscellaneous insert and extract macros.  */
/* Extract a single-precision float from X at index N into D.  */
#define _MM_EXTRACT_FLOAT(D, X, N)                                             \
  do {                                                                         \
    (D) = __builtin_ia32_vec_ext_v4sf((__v4sf)(__m128)(X), (int)(N));          \
  } while (0)

/* Or together 2 sets of indexes (X and Y) with the zeroing bits (Z) to create
   an index suitable for _mm_insert_ps.  */
#define _MM_MK_INSERTPS_NDX(X, Y, Z) (((X) << 6) | ((Y) << 4) | (Z))

/* Extract a float from X at index N into the first index of the return.  */
#define _MM_PICK_OUT_PS(X, N)                                                  \
  _mm_insert_ps(_mm_setzero_ps(), (X), _MM_MK_INSERTPS_NDX((N), 0, 0x0e))

/* Insert int into packed integer array at index.  */
/// Constructs a 128-bit vector of [16 x i8] by first making a copy of
///    the 128-bit integer vector parameter, and then inserting the lower 8 bits
///    of an integer parameter \a I into an offset specified by the immediate
///    value parameter \a N.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm_insert_epi8(__m128i X, int I, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPINSRB / PINSRB </c> instruction.
///
/// \param X
///    A 128-bit integer vector of [16 x i8]. This vector is copied to the
///    result and then one of the sixteen elements in the result vector is
///    replaced by the lower 8 bits of \a I.
/// \param I
///    An integer. The lower 8 bits of this operand are written to the result
///    beginning at the offset specified by \a N.
/// \param N
///    An immediate value. Bits [3:0] specify the bit offset in the result at
///    which the lower 8 bits of \a I are written. \n
///    0000: Bits [7:0] of the result are used for insertion. \n
///    0001: Bits [15:8] of the result are used for insertion. \n
///    0010: Bits [23:16] of the result are used for insertion. \n
///    0011: Bits [31:24] of the result are used for insertion. \n
///    0100: Bits [39:32] of the result are used for insertion. \n
///    0101: Bits [47:40] of the result are used for insertion. \n
///    0110: Bits [55:48] of the result are used for insertion. \n
///    0111: Bits [63:56] of the result are used for insertion. \n
///    1000: Bits [71:64] of the result are used for insertion. \n
///    1001: Bits [79:72] of the result are used for insertion. \n
///    1010: Bits [87:80] of the result are used for insertion. \n
///    1011: Bits [95:88] of the result are used for insertion. \n
///    1100: Bits [103:96] of the result are used for insertion. \n
///    1101: Bits [111:104] of the result are used for insertion. \n
///    1110: Bits [119:112] of the result are used for insertion. \n
///    1111: Bits [127:120] of the result are used for insertion.
/// \returns A 128-bit integer vector containing the constructed values.
#define _mm_insert_epi8(X, I, N)                                               \
  ((__m128i)__builtin_ia32_vec_set_v16qi((__v16qi)(__m128i)(X), (int)(I),      \
                                         (int)(N)))

/// Constructs a 128-bit vector of [4 x i32] by first making a copy of
///    the 128-bit integer vector parameter, and then inserting the 32-bit
///    integer parameter \a I at the offset specified by the immediate value
///    parameter \a N.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm_insert_epi32(__m128i X, int I, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPINSRD / PINSRD </c> instruction.
///
/// \param X
///    A 128-bit integer vector of [4 x i32]. This vector is copied to the
///    result and then one of the four elements in the result vector is
///    replaced by \a I.
/// \param I
///    A 32-bit integer that is written to the result beginning at the offset
///    specified by \a N.
/// \param N
///    An immediate value. Bits [1:0] specify the bit offset in the result at
///    which the integer \a I is written. \n
///    00: Bits [31:0] of the result are used for insertion. \n
///    01: Bits [63:32] of the result are used for insertion. \n
///    10: Bits [95:64] of the result are used for insertion. \n
///    11: Bits [127:96] of the result are used for insertion.
/// \returns A 128-bit integer vector containing the constructed values.
#define _mm_insert_epi32(X, I, N)                                              \
  ((__m128i)__builtin_ia32_vec_set_v4si((__v4si)(__m128i)(X), (int)(I),        \
                                        (int)(N)))

#ifdef __x86_64__
/// Constructs a 128-bit vector of [2 x i64] by first making a copy of
///    the 128-bit integer vector parameter, and then inserting the 64-bit
///    integer parameter \a I, using the immediate value parameter \a N as an
///    insertion location selector.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm_insert_epi64(__m128i X, long long I, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPINSRQ / PINSRQ </c> instruction.
///
/// \param X
///    A 128-bit integer vector of [2 x i64]. This vector is copied to the
///    result and then one of the two elements in the result vector is replaced
///    by \a I.
/// \param I
///    A 64-bit integer that is written to the result beginning at the offset
///    specified by \a N.
/// \param N
///    An immediate value. Bit [0] specifies the bit offset in the result at
///    which the integer \a I is written. \n
///    0: Bits [63:0] of the result are used for insertion. \n
///    1: Bits [127:64] of the result are used for insertion. \n
/// \returns A 128-bit integer vector containing the constructed values.
#define _mm_insert_epi64(X, I, N)                                              \
  ((__m128i)__builtin_ia32_vec_set_v2di((__v2di)(__m128i)(X), (long long)(I),  \
                                        (int)(N)))
#endif /* __x86_64__ */

/* Extract int from packed integer array at index.  This returns the element
 * as a zero extended value, so it is unsigned.
 */
/// Extracts an 8-bit element from the 128-bit integer vector of
///    [16 x i8], using the immediate value parameter \a N as a selector.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_extract_epi8(__m128i X, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPEXTRB / PEXTRB </c> instruction.
///
/// \param X
///    A 128-bit integer vector.
/// \param N
///    An immediate value. Bits [3:0] specify which 8-bit vector element from
///    the argument \a X to extract and copy to the result. \n
///    0000: Bits [7:0] of parameter \a X are extracted. \n
///    0001: Bits [15:8] of the parameter \a X are extracted. \n
///    0010: Bits [23:16] of the parameter \a X are extracted. \n
///    0011: Bits [31:24] of the parameter \a X are extracted. \n
///    0100: Bits [39:32] of the parameter \a X are extracted. \n
///    0101: Bits [47:40] of the parameter \a X are extracted. \n
///    0110: Bits [55:48] of the parameter \a X are extracted. \n
///    0111: Bits [63:56] of the parameter \a X are extracted. \n
///    1000: Bits [71:64] of the parameter \a X are extracted. \n
///    1001: Bits [79:72] of the parameter \a X are extracted. \n
///    1010: Bits [87:80] of the parameter \a X are extracted. \n
///    1011: Bits [95:88] of the parameter \a X are extracted. \n
///    1100: Bits [103:96] of the parameter \a X are extracted. \n
///    1101: Bits [111:104] of the parameter \a X are extracted. \n
///    1110: Bits [119:112] of the parameter \a X are extracted. \n
///    1111: Bits [127:120] of the parameter \a X are extracted.
/// \returns  An unsigned integer, whose lower 8 bits are selected from the
///    128-bit integer vector parameter and the remaining bits are assigned
///    zeros.
#define _mm_extract_epi8(X, N)                                                 \
  ((int)(unsigned char)__builtin_ia32_vec_ext_v16qi((__v16qi)(__m128i)(X),     \
                                                    (int)(N)))

/// Extracts a 32-bit element from the 128-bit integer vector of
///    [4 x i32], using the immediate value parameter \a N as a selector.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_extract_epi32(__m128i X, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPEXTRD / PEXTRD </c> instruction.
///
/// \param X
///    A 128-bit integer vector.
/// \param N
///    An immediate value. Bits [1:0] specify which 32-bit vector element from
///    the argument \a X to extract and copy to the result. \n
///    00: Bits [31:0] of the parameter \a X are extracted. \n
///    01: Bits [63:32] of the parameter \a X are extracted. \n
///    10: Bits [95:64] of the parameter \a X are extracted. \n
///    11: Bits [127:96] of the parameter \a X are exracted.
/// \returns  An integer, whose lower 32 bits are selected from the 128-bit
///    integer vector parameter and the remaining bits are assigned zeros.
#define _mm_extract_epi32(X, N)                                                \
  ((int)__builtin_ia32_vec_ext_v4si((__v4si)(__m128i)(X), (int)(N)))

/// Extracts a 64-bit element from the 128-bit integer vector of
///    [2 x i64], using the immediate value parameter \a N as a selector.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// long long _mm_extract_epi64(__m128i X, const int N);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPEXTRQ / PEXTRQ </c> instruction
/// in 64-bit mode.
///
/// \param X
///    A 128-bit integer vector.
/// \param N
///    An immediate value. Bit [0] specifies which 64-bit vector element from
///    the argument \a X to return. \n
///    0: Bits [63:0] are returned. \n
///    1: Bits [127:64] are returned. \n
/// \returns  A 64-bit integer.
#define _mm_extract_epi64(X, N)                                                \
  ((long long)__builtin_ia32_vec_ext_v2di((__v2di)(__m128i)(X), (int)(N)))

/* SSE4 128-bit Packed Integer Comparisons.  */
/// Tests whether the specified bits in a 128-bit integer vector are all
///    zeros.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPTEST / PTEST </c> instruction.
///
/// \param __M
///    A 128-bit integer vector containing the bits to be tested.
/// \param __V
///    A 128-bit integer vector selecting which bits to test in operand \a __M.
/// \returns TRUE if the specified bits are all zeros; FALSE otherwise.
static __inline__ int __DEFAULT_FN_ATTRS _mm_testz_si128(__m128i __M,
                                                         __m128i __V) {
  return __builtin_ia32_ptestz128((__v2di)__M, (__v2di)__V);
}

/// Tests whether the specified bits in a 128-bit integer vector are all
///    ones.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPTEST / PTEST </c> instruction.
///
/// \param __M
///    A 128-bit integer vector containing the bits to be tested.
/// \param __V
///    A 128-bit integer vector selecting which bits to test in operand \a __M.
/// \returns TRUE if the specified bits are all ones; FALSE otherwise.
static __inline__ int __DEFAULT_FN_ATTRS _mm_testc_si128(__m128i __M,
                                                         __m128i __V) {
  return __builtin_ia32_ptestc128((__v2di)__M, (__v2di)__V);
}

/// Tests whether the specified bits in a 128-bit integer vector are
///    neither all zeros nor all ones.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPTEST / PTEST </c> instruction.
///
/// \param __M
///    A 128-bit integer vector containing the bits to be tested.
/// \param __V
///    A 128-bit integer vector selecting which bits to test in operand \a __M.
/// \returns TRUE if the specified bits are neither all zeros nor all ones;
///    FALSE otherwise.
static __inline__ int __DEFAULT_FN_ATTRS _mm_testnzc_si128(__m128i __M,
                                                           __m128i __V) {
  return __builtin_ia32_ptestnzc128((__v2di)__M, (__v2di)__V);
}

/// Tests whether the specified bits in a 128-bit integer vector are all
///    ones.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_test_all_ones(__m128i V);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPTEST / PTEST </c> instruction.
///
/// \param V
///    A 128-bit integer vector containing the bits to be tested.
/// \returns TRUE if the bits specified in the operand are all set to 1; FALSE
///    otherwise.
#define _mm_test_all_ones(V) _mm_testc_si128((V), _mm_set1_epi32(-1))

/// Tests whether the specified bits in a 128-bit integer vector are
///    neither all zeros nor all ones.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_test_mix_ones_zeros(__m128i M, __m128i V);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPTEST / PTEST </c> instruction.
///
/// \param M
///    A 128-bit integer vector containing the bits to be tested.
/// \param V
///    A 128-bit integer vector selecting which bits to test in operand \a M.
/// \returns TRUE if the specified bits are neither all zeros nor all ones;
///    FALSE otherwise.
#define _mm_test_mix_ones_zeros(M, V) _mm_testnzc_si128((M), (V))

/// Tests whether the specified bits in a 128-bit integer vector are all
///    zeros.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_test_all_zeros(__m128i M, __m128i V);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPTEST / PTEST </c> instruction.
///
/// \param M
///    A 128-bit integer vector containing the bits to be tested.
/// \param V
///    A 128-bit integer vector selecting which bits to test in operand \a M.
/// \returns TRUE if the specified bits are all zeros; FALSE otherwise.
#define _mm_test_all_zeros(M, V) _mm_testz_si128((M), (V))

/* SSE4 64-bit Packed Integer Comparisons.  */
/// Compares each of the corresponding 64-bit values of the 128-bit
///    integer vectors for equality.
///
///    Each comparison returns 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPCMPEQQ / PCMPEQQ </c> instruction.
///
/// \param __V1
///    A 128-bit integer vector.
/// \param __V2
///    A 128-bit integer vector.
/// \returns A 128-bit integer vector containing the comparison results.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi64(__m128i __V1,
                                                             __m128i __V2) {
  return (__m128i)((__v2di)__V1 == (__v2di)__V2);
}

/* SSE4 Packed Integer Sign-Extension.  */
/// Sign-extends each of the lower eight 8-bit integer elements of a
///    128-bit vector of [16 x i8] to 16-bit values and returns them in a
///    128-bit vector of [8 x i16]. The upper eight elements of the input vector
///    are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVSXBW / PMOVSXBW </c> instruction.
///
/// \param __V
///    A 128-bit vector of [16 x i8]. The lower eight 8-bit elements are
///    sign-extended to 16-bit values.
/// \returns A 128-bit vector of [8 x i16] containing the sign-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi16(__m128i __V) {
  /* This function always performs a signed extension, but __v16qi is a char
     which may be signed or unsigned, so use __v16qs. */
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v16qs)__V, (__v16qs)__V, 0, 1, 2, 3, 4, 5, 6,
                              7),
      __v8hi);
}

/// Sign-extends each of the lower four 8-bit integer elements of a
///    128-bit vector of [16 x i8] to 32-bit values and returns them in a
///    128-bit vector of [4 x i32]. The upper twelve elements of the input
///    vector are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVSXBD / PMOVSXBD </c> instruction.
///
/// \param __V
///    A 128-bit vector of [16 x i8]. The lower four 8-bit elements are
///    sign-extended to 32-bit values.
/// \returns A 128-bit vector of [4 x i32] containing the sign-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi32(__m128i __V) {
  /* This function always performs a signed extension, but __v16qi is a char
     which may be signed or unsigned, so use __v16qs. */
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v16qs)__V, (__v16qs)__V, 0, 1, 2, 3), __v4si);
}

/// Sign-extends each of the lower two 8-bit integer elements of a
///    128-bit integer vector of [16 x i8] to 64-bit values and returns them in
///    a 128-bit vector of [2 x i64]. The upper fourteen elements of the input
///    vector are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVSXBQ / PMOVSXBQ </c> instruction.
///
/// \param __V
///    A 128-bit vector of [16 x i8]. The lower two 8-bit elements are
///    sign-extended to 64-bit values.
/// \returns A 128-bit vector of [2 x i64] containing the sign-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi64(__m128i __V) {
  /* This function always performs a signed extension, but __v16qi is a char
     which may be signed or unsigned, so use __v16qs. */
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v16qs)__V, (__v16qs)__V, 0, 1), __v2di);
}

/// Sign-extends each of the lower four 16-bit integer elements of a
///    128-bit integer vector of [8 x i16] to 32-bit values and returns them in
///    a 128-bit vector of [4 x i32]. The upper four elements of the input
///    vector are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVSXWD / PMOVSXWD </c> instruction.
///
/// \param __V
///    A 128-bit vector of [8 x i16]. The lower four 16-bit elements are
///    sign-extended to 32-bit values.
/// \returns A 128-bit vector of [4 x i32] containing the sign-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi16_epi32(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v8hi)__V, (__v8hi)__V, 0, 1, 2, 3), __v4si);
}

/// Sign-extends each of the lower two 16-bit integer elements of a
///    128-bit integer vector of [8 x i16] to 64-bit values and returns them in
///    a 128-bit vector of [2 x i64]. The upper six elements of the input
///    vector are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVSXWQ / PMOVSXWQ </c> instruction.
///
/// \param __V
///    A 128-bit vector of [8 x i16]. The lower two 16-bit elements are
///     sign-extended to 64-bit values.
/// \returns A 128-bit vector of [2 x i64] containing the sign-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi16_epi64(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v8hi)__V, (__v8hi)__V, 0, 1), __v2di);
}

/// Sign-extends each of the lower two 32-bit integer elements of a
///    128-bit integer vector of [4 x i32] to 64-bit values and returns them in
///    a 128-bit vector of [2 x i64]. The upper two elements of the input vector
///    are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVSXDQ / PMOVSXDQ </c> instruction.
///
/// \param __V
///    A 128-bit vector of [4 x i32]. The lower two 32-bit elements are
///    sign-extended to 64-bit values.
/// \returns A 128-bit vector of [2 x i64] containing the sign-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi32_epi64(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v4si)__V, (__v4si)__V, 0, 1), __v2di);
}

/* SSE4 Packed Integer Zero-Extension.  */
/// Zero-extends each of the lower eight 8-bit integer elements of a
///    128-bit vector of [16 x i8] to 16-bit values and returns them in a
///    128-bit vector of [8 x i16]. The upper eight elements of the input vector
///    are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVZXBW / PMOVZXBW </c> instruction.
///
/// \param __V
///    A 128-bit vector of [16 x i8]. The lower eight 8-bit elements are
///    zero-extended to 16-bit values.
/// \returns A 128-bit vector of [8 x i16] containing the zero-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi16(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1, 2, 3, 4, 5, 6,
                              7),
      __v8hi);
}

/// Zero-extends each of the lower four 8-bit integer elements of a
///    128-bit vector of [16 x i8] to 32-bit values and returns them in a
///    128-bit vector of [4 x i32]. The upper twelve elements of the input
///    vector are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVZXBD / PMOVZXBD </c> instruction.
///
/// \param __V
///    A 128-bit vector of [16 x i8]. The lower four 8-bit elements are
///    zero-extended to 32-bit values.
/// \returns A 128-bit vector of [4 x i32] containing the zero-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi32(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1, 2, 3), __v4si);
}

/// Zero-extends each of the lower two 8-bit integer elements of a
///    128-bit integer vector of [16 x i8] to 64-bit values and returns them in
///    a 128-bit vector of [2 x i64]. The upper fourteen elements of the input
///    vector are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVZXBQ / PMOVZXBQ </c> instruction.
///
/// \param __V
///    A 128-bit vector of [16 x i8]. The lower two 8-bit elements are
///    zero-extended to 64-bit values.
/// \returns A 128-bit vector of [2 x i64] containing the zero-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi64(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1), __v2di);
}

/// Zero-extends each of the lower four 16-bit integer elements of a
///    128-bit integer vector of [8 x i16] to 32-bit values and returns them in
///    a 128-bit vector of [4 x i32]. The upper four elements of the input
///    vector are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVZXWD / PMOVZXWD </c> instruction.
///
/// \param __V
///    A 128-bit vector of [8 x i16]. The lower four 16-bit elements are
///    zero-extended to 32-bit values.
/// \returns A 128-bit vector of [4 x i32] containing the zero-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu16_epi32(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v8hu)__V, (__v8hu)__V, 0, 1, 2, 3), __v4si);
}

/// Zero-extends each of the lower two 16-bit integer elements of a
///    128-bit integer vector of [8 x i16] to 64-bit values and returns them in
///    a 128-bit vector of [2 x i64]. The upper six elements of the input vector
///    are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVZXWQ / PMOVZXWQ </c> instruction.
///
/// \param __V
///    A 128-bit vector of [8 x i16]. The lower two 16-bit elements are
///    zero-extended to 64-bit values.
/// \returns A 128-bit vector of [2 x i64] containing the zero-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu16_epi64(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v8hu)__V, (__v8hu)__V, 0, 1), __v2di);
}

/// Zero-extends each of the lower two 32-bit integer elements of a
///    128-bit integer vector of [4 x i32] to 64-bit values and returns them in
///    a 128-bit vector of [2 x i64]. The upper two elements of the input vector
///    are unused.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPMOVZXDQ / PMOVZXDQ </c> instruction.
///
/// \param __V
///    A 128-bit vector of [4 x i32]. The lower two 32-bit elements are
///    zero-extended to 64-bit values.
/// \returns A 128-bit vector of [2 x i64] containing the zero-extended values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu32_epi64(__m128i __V) {
  return (__m128i) __builtin_convertvector(
      __builtin_shufflevector((__v4su)__V, (__v4su)__V, 0, 1), __v2di);
}

/* SSE4 Pack with Unsigned Saturation.  */
/// Converts, with saturation, 32-bit signed integers from both 128-bit integer
///    vector operands into 16-bit unsigned integers, and returns the packed
///    result.
///
///    Values greater than 0xFFFF are saturated to 0xFFFF. Values less than
///    0x0000 are saturated to 0x0000.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPACKUSDW / PACKUSDW </c> instruction.
///
/// \param __V1
///    A 128-bit vector of [4 x i32]. The converted [4 x i16] values are
///    written to the lower 64 bits of the result.
/// \param __V2
///    A 128-bit vector of [4 x i32]. The converted [4 x i16] values are
///    written to the higher 64 bits of the result.
/// \returns A 128-bit vector of [8 x i16] containing the converted values.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packus_epi32(__m128i __V1,
                                                              __m128i __V2) {
  return (__m128i)__builtin_ia32_packusdw128((__v4si)__V1, (__v4si)__V2);
}

/* SSE4 Multiple Packed Sums of Absolute Difference.  */
/// Subtracts 8-bit unsigned integer values and computes the absolute
///    values of the differences to the corresponding bits in the destination.
///    Then sums of the absolute differences are returned according to the bit
///    fields in the immediate operand.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm_mpsadbw_epu8(__m128i X, __m128i Y, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VMPSADBW / MPSADBW </c> instruction.
///
/// \param X
///    A 128-bit vector of [16 x i8].
/// \param Y
///    A 128-bit vector of [16 x i8].
/// \param M
///    An 8-bit immediate operand specifying how the absolute differences are to
///    be calculated, according to the following algorithm:
///    \code
///    // M2 represents bit 2 of the immediate operand
///    // M10 represents bits [1:0] of the immediate operand
///    i = M2 * 4;
///    j = M10 * 4;
///    for (k = 0; k < 8; k = k + 1) {
///      d0 = abs(X[i + k + 0] - Y[j + 0]);
///      d1 = abs(X[i + k + 1] - Y[j + 1]);
///      d2 = abs(X[i + k + 2] - Y[j + 2]);
///      d3 = abs(X[i + k + 3] - Y[j + 3]);
///      r[k] = d0 + d1 + d2 + d3;
///    }
///    \endcode
/// \returns A 128-bit integer vector containing the sums of the sets of
///    absolute differences between both operands.
#define _mm_mpsadbw_epu8(X, Y, M)                                              \
  ((__m128i)__builtin_ia32_mpsadbw128((__v16qi)(__m128i)(X),                   \
                                      (__v16qi)(__m128i)(Y), (M)))

/// Finds the minimum unsigned 16-bit element in the input 128-bit
///    vector of [8 x u16] and returns it and along with its index.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPHMINPOSUW / PHMINPOSUW </c>
/// instruction.
///
/// \param __V
///    A 128-bit vector of [8 x u16].
/// \returns A 128-bit value where bits [15:0] contain the minimum value found
///    in parameter \a __V, bits [18:16] contain the index of the minimum value
///    and the remaining bits are set to 0.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_minpos_epu16(__m128i __V) {
  return (__m128i)__builtin_ia32_phminposuw128((__v8hi)__V);
}

/* Handle the sse4.2 definitions here. */

/* These definitions are normally in nmmintrin.h, but gcc puts them in here
   so we'll do the same.  */

#undef __DEFAULT_FN_ATTRS
#define __DEFAULT_FN_ATTRS                                                     \
  __attribute__((__always_inline__, __nodebug__, __target__("sse4.2")))

/* These specify the type of data that we're comparing.  */
#define _SIDD_UBYTE_OPS 0x00
#define _SIDD_UWORD_OPS 0x01
#define _SIDD_SBYTE_OPS 0x02
#define _SIDD_SWORD_OPS 0x03

/* These specify the type of comparison operation.  */
#define _SIDD_CMP_EQUAL_ANY 0x00
#define _SIDD_CMP_RANGES 0x04
#define _SIDD_CMP_EQUAL_EACH 0x08
#define _SIDD_CMP_EQUAL_ORDERED 0x0c

/* These macros specify the polarity of the operation.  */
#define _SIDD_POSITIVE_POLARITY 0x00
#define _SIDD_NEGATIVE_POLARITY 0x10
#define _SIDD_MASKED_POSITIVE_POLARITY 0x20
#define _SIDD_MASKED_NEGATIVE_POLARITY 0x30

/* These macros are used in _mm_cmpXstri() to specify the return.  */
#define _SIDD_LEAST_SIGNIFICANT 0x00
#define _SIDD_MOST_SIGNIFICANT 0x40

/* These macros are used in _mm_cmpXstri() to specify the return.  */
#define _SIDD_BIT_MASK 0x00
#define _SIDD_UNIT_MASK 0x40

/* SSE4.2 Packed Comparison Intrinsics.  */
/// Uses the immediate operand \a M to perform a comparison of string
///    data with implicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns a 128-bit integer vector representing the result
///    mask of the comparison.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm_cmpistrm(__m128i A, __m128i B, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPISTRM / PCMPISTRM </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words, the type of comparison to perform, and the format of the return
///    value. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
///    Bit [6]: Determines whether the result is zero-extended or expanded to 16
///             bytes. \n
///      0: The result is zero-extended to 16 bytes. \n
///      1: The result is expanded to 16 bytes (this expansion is performed by
///         repeating each bit 8 or 16 times).
/// \returns Returns a 128-bit integer vector representing the result mask of
///    the comparison.
#define _mm_cmpistrm(A, B, M)                                                  \
  ((__m128i)__builtin_ia32_pcmpistrm128((__v16qi)(__m128i)(A),                 \
                                        (__v16qi)(__m128i)(B), (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with implicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns an integer representing the result index of the
///    comparison.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpistri(__m128i A, __m128i B, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPISTRI / PCMPISTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words, the type of comparison to perform, and the format of the return
///    value. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
///    Bit [6]: Determines whether the index of the lowest set bit or the
///             highest set bit is returned. \n
///      0: The index of the least significant set bit. \n
///      1: The index of the most significant set bit. \n
/// \returns Returns an integer representing the result index of the comparison.
#define _mm_cmpistri(A, B, M)                                                  \
  ((int)__builtin_ia32_pcmpistri128((__v16qi)(__m128i)(A),                     \
                                    (__v16qi)(__m128i)(B), (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with explicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns a 128-bit integer vector representing the result
///    mask of the comparison.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm_cmpestrm(__m128i A, int LA, __m128i B, int LB, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPESTRM / PCMPESTRM </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LA
///    An integer that specifies the length of the string in \a A.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LB
///    An integer that specifies the length of the string in \a B.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words, the type of comparison to perform, and the format of the return
///    value. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
///    Bit [6]: Determines whether the result is zero-extended or expanded to 16
///             bytes. \n
///      0: The result is zero-extended to 16 bytes. \n
///      1: The result is expanded to 16 bytes (this expansion is performed by
///         repeating each bit 8 or 16 times). \n
/// \returns Returns a 128-bit integer vector representing the result mask of
///    the comparison.
#define _mm_cmpestrm(A, LA, B, LB, M)                                          \
  ((__m128i)__builtin_ia32_pcmpestrm128((__v16qi)(__m128i)(A), (int)(LA),      \
                                        (__v16qi)(__m128i)(B), (int)(LB),      \
                                        (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with explicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns an integer representing the result index of the
///    comparison.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpestri(__m128i A, int LA, __m128i B, int LB, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPESTRI / PCMPESTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LA
///    An integer that specifies the length of the string in \a A.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LB
///    An integer that specifies the length of the string in \a B.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words, the type of comparison to perform, and the format of the return
///    value. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
///    Bit [6]: Determines whether the index of the lowest set bit or the
///             highest set bit is returned. \n
///      0: The index of the least significant set bit. \n
///      1: The index of the most significant set bit. \n
/// \returns Returns an integer representing the result index of the comparison.
#define _mm_cmpestri(A, LA, B, LB, M)                                          \
  ((int)__builtin_ia32_pcmpestri128((__v16qi)(__m128i)(A), (int)(LA),          \
                                    (__v16qi)(__m128i)(B), (int)(LB),          \
                                    (int)(M)))

/* SSE4.2 Packed Comparison Intrinsics and EFlag Reading.  */
/// Uses the immediate operand \a M to perform a comparison of string
///    data with implicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the bit mask is zero and the length of the
///    string in \a B is the maximum, otherwise, returns 0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpistra(__m128i A, __m128i B, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPISTRI / PCMPISTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
/// \returns Returns 1 if the bit mask is zero and the length of the string in
///    \a B is the maximum; otherwise, returns 0.
#define _mm_cmpistra(A, B, M)                                                  \
  ((int)__builtin_ia32_pcmpistria128((__v16qi)(__m128i)(A),                    \
                                     (__v16qi)(__m128i)(B), (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with implicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the bit mask is non-zero, otherwise, returns
///    0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpistrc(__m128i A, __m128i B, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPISTRI / PCMPISTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B.
/// \returns Returns 1 if the bit mask is non-zero, otherwise, returns 0.
#define _mm_cmpistrc(A, B, M)                                                  \
  ((int)__builtin_ia32_pcmpistric128((__v16qi)(__m128i)(A),                    \
                                     (__v16qi)(__m128i)(B), (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with implicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns bit 0 of the resulting bit mask.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpistro(__m128i A, __m128i B, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPISTRI / PCMPISTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
/// \returns Returns bit 0 of the resulting bit mask.
#define _mm_cmpistro(A, B, M)                                                  \
  ((int)__builtin_ia32_pcmpistrio128((__v16qi)(__m128i)(A),                    \
                                     (__v16qi)(__m128i)(B), (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with implicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the length of the string in \a A is less than
///    the maximum, otherwise, returns 0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpistrs(__m128i A, __m128i B, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPISTRI / PCMPISTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
/// \returns Returns 1 if the length of the string in \a A is less than the
///    maximum, otherwise, returns 0.
#define _mm_cmpistrs(A, B, M)                                                  \
  ((int)__builtin_ia32_pcmpistris128((__v16qi)(__m128i)(A),                    \
                                     (__v16qi)(__m128i)(B), (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with implicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the length of the string in \a B is less than
///    the maximum, otherwise, returns 0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpistrz(__m128i A, __m128i B, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPISTRI / PCMPISTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B.
/// \returns Returns 1 if the length of the string in \a B is less than the
///    maximum, otherwise, returns 0.
#define _mm_cmpistrz(A, B, M)                                                  \
  ((int)__builtin_ia32_pcmpistriz128((__v16qi)(__m128i)(A),                    \
                                     (__v16qi)(__m128i)(B), (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with explicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the bit mask is zero and the length of the
///    string in \a B is the maximum, otherwise, returns 0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpestra(__m128i A, int LA, __m128i B, int LB, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPESTRI / PCMPESTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LA
///    An integer that specifies the length of the string in \a A.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LB
///    An integer that specifies the length of the string in \a B.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B.
/// \returns Returns 1 if the bit mask is zero and the length of the string in
///    \a B is the maximum, otherwise, returns 0.
#define _mm_cmpestra(A, LA, B, LB, M)                                          \
  ((int)__builtin_ia32_pcmpestria128((__v16qi)(__m128i)(A), (int)(LA),         \
                                     (__v16qi)(__m128i)(B), (int)(LB),         \
                                     (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with explicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the resulting mask is non-zero, otherwise,
///    returns 0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpestrc(__m128i A, int LA, __m128i B, int LB, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPESTRI / PCMPESTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LA
///    An integer that specifies the length of the string in \a A.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LB
///    An integer that specifies the length of the string in \a B.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
/// \returns Returns 1 if the resulting mask is non-zero, otherwise, returns 0.
#define _mm_cmpestrc(A, LA, B, LB, M)                                          \
  ((int)__builtin_ia32_pcmpestric128((__v16qi)(__m128i)(A), (int)(LA),         \
                                     (__v16qi)(__m128i)(B), (int)(LB),         \
                                     (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with explicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns bit 0 of the resulting bit mask.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpestro(__m128i A, int LA, __m128i B, int LB, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPESTRI / PCMPESTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LA
///    An integer that specifies the length of the string in \a A.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LB
///    An integer that specifies the length of the string in \a B.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B.
/// \returns Returns bit 0 of the resulting bit mask.
#define _mm_cmpestro(A, LA, B, LB, M)                                          \
  ((int)__builtin_ia32_pcmpestrio128((__v16qi)(__m128i)(A), (int)(LA),         \
                                     (__v16qi)(__m128i)(B), (int)(LB),         \
                                     (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with explicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the length of the string in \a A is less than
///    the maximum, otherwise, returns 0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpestrs(__m128i A, int LA, __m128i B, int LB, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPESTRI / PCMPESTRI </c>
/// instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LA
///    An integer that specifies the length of the string in \a A.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LB
///    An integer that specifies the length of the string in \a B.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement in the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B. \n
/// \returns Returns 1 if the length of the string in \a A is less than the
///    maximum, otherwise, returns 0.
#define _mm_cmpestrs(A, LA, B, LB, M)                                          \
  ((int)__builtin_ia32_pcmpestris128((__v16qi)(__m128i)(A), (int)(LA),         \
                                     (__v16qi)(__m128i)(B), (int)(LB),         \
                                     (int)(M)))

/// Uses the immediate operand \a M to perform a comparison of string
///    data with explicitly defined lengths that is contained in source operands
///    \a A and \a B. Returns 1 if the length of the string in \a B is less than
///    the maximum, otherwise, returns 0.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// int _mm_cmpestrz(__m128i A, int LA, __m128i B, int LB, const int M);
/// \endcode
///
/// This intrinsic corresponds to the <c> VPCMPESTRI </c> instruction.
///
/// \param A
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LA
///    An integer that specifies the length of the string in \a A.
/// \param B
///    A 128-bit integer vector containing one of the source operands to be
///    compared.
/// \param LB
///    An integer that specifies the length of the string in \a B.
/// \param M
///    An 8-bit immediate operand specifying whether the characters are bytes or
///    words and the type of comparison to perform. \n
///    Bits [1:0]: Determine source data format. \n
///      00: 16 unsigned bytes  \n
///      01: 8 unsigned words \n
///      10: 16 signed bytes \n
///      11: 8 signed words \n
///    Bits [3:2]: Determine comparison type and aggregation method. \n
///      00: Subset: Each character in \a B is compared for equality with all
///          the characters in \a A. \n
///      01: Ranges: Each character in \a B is compared to \a A. The comparison
///          basis is greater than or equal for even-indexed elements in \a A,
///          and less than or equal for odd-indexed elements in \a A. \n
///      10: Match: Compare each pair of corresponding characters in \a A and
///          \a B for equality. \n
///      11: Substring: Search \a B for substring matches of \a A. \n
///    Bits [5:4]: Determine whether to perform a one's complement on the bit
///                mask of the comparison results. \n
///      00: No effect. \n
///      01: Negate the bit mask. \n
///      10: No effect. \n
///      11: Negate the bit mask only for bits with an index less than or equal
///          to the size of \a A or \a B.
/// \returns Returns 1 if the length of the string in \a B is less than the
///    maximum, otherwise, returns 0.
#define _mm_cmpestrz(A, LA, B, LB, M)                                          \
  ((int)__builtin_ia32_pcmpestriz128((__v16qi)(__m128i)(A), (int)(LA),         \
                                     (__v16qi)(__m128i)(B), (int)(LB),         \
                                     (int)(M)))

/* SSE4.2 Compare Packed Data -- Greater Than.  */
/// Compares each of the corresponding 64-bit values of the 128-bit
///    integer vectors to determine if the values in the first operand are
///    greater than those in the second operand.
///
///    Each comparison returns 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to the <c> VPCMPGTQ / PCMPGTQ </c> instruction.
///
/// \param __V1
///    A 128-bit integer vector.
/// \param __V2
///    A 128-bit integer vector.
/// \returns A 128-bit integer vector containing the comparison results.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi64(__m128i __V1,
                                                             __m128i __V2) {
  return (__m128i)((__v2di)__V1 > (__v2di)__V2);
}

#undef __DEFAULT_FN_ATTRS

#include <popcntintrin.h>

#include <crc32intrin.h>

#endif /* __SMMINTRIN_H */

Youez - 2016 - github.com/yon3zu
LinuXploit