diff options
Diffstat (limited to 'libc/sysdeps/x86_64')
-rw-r--r-- | libc/sysdeps/x86_64/configure | 124 | ||||
-rw-r--r-- | libc/sysdeps/x86_64/dl-tlsdesc.h | 6 | ||||
-rw-r--r-- | libc/sysdeps/x86_64/fpu/libm-test-ulps | 174 | ||||
-rw-r--r-- | libc/sysdeps/x86_64/fpu/s_sincosf.S | 558 | ||||
-rw-r--r-- | libc/sysdeps/x86_64/multiarch/init-arch.c | 3 | ||||
-rw-r--r-- | libc/sysdeps/x86_64/multiarch/init-arch.h | 9 | ||||
-rw-r--r-- | libc/sysdeps/x86_64/multiarch/strncmp-ssse3.S | 10 |
7 files changed, 726 insertions, 158 deletions
diff --git a/libc/sysdeps/x86_64/configure b/libc/sysdeps/x86_64/configure index adbaeee32..0fa0e1ecf 100644 --- a/libc/sysdeps/x86_64/configure +++ b/libc/sysdeps/x86_64/configure @@ -1,128 +1,4 @@ -# as_fn_set_status STATUS -# ----------------------- -# Set $? to STATUS, without forking. -as_fn_set_status () -{ - return $1 -} # as_fn_set_status - -# as_fn_exit STATUS -# ----------------- -# Exit the shell with STATUS, even in a "trap 0" or "set -e" context. -as_fn_exit () -{ - set +e - as_fn_set_status $1 - exit $1 -} # as_fn_exit -if expr a : '\(a\)' >/dev/null 2>&1 && - test "X`expr 00001 : '.*\(...\)'`" = X001; then - as_expr=expr -else - as_expr=false -fi - -if (basename -- /) >/dev/null 2>&1 && test "X`basename -- / 2>&1`" = "X/"; then - as_basename=basename -else - as_basename=false -fi - -as_me=`$as_basename -- "$0" || -$as_expr X/"$0" : '.*/\([^/][^/]*\)/*$' \| \ - X"$0" : 'X\(//\)$' \| \ - X"$0" : 'X\(/\)' \| . 2>/dev/null || -$as_echo X/"$0" | - sed '/^.*\/\([^/][^/]*\)\/*$/{ - s//\1/ - q - } - /^X\/\(\/\/\)$/{ - s//\1/ - q - } - /^X\/\(\/\).*/{ - s//\1/ - q - } - s/.*/./; q'` - - - as_lineno_1=$LINENO as_lineno_1a=$LINENO - as_lineno_2=$LINENO as_lineno_2a=$LINENO - eval 'test "x$as_lineno_1'$as_run'" != "x$as_lineno_2'$as_run'" && - test "x`expr $as_lineno_1'$as_run' + 1`" = "x$as_lineno_2'$as_run'"' || { - # Blame Lee E. McMahon (1931-1989) for sed's syntax. :-) - sed -n ' - p - /[$]LINENO/= - ' <$as_myself | - sed ' - s/[$]LINENO.*/&-/ - t lineno - b - :lineno - N - :loop - s/[$]LINENO\([^'$as_cr_alnum'_].*\n\)\(.*\)/\2\1\2/ - t loop - s/-\n.*// - ' >$as_me.lineno && - chmod +x "$as_me.lineno" || - { $as_echo "$as_me: error: cannot create $as_me.lineno; rerun with a POSIX shell" >&2; as_fn_exit 1; } - - # If we had to re-execute with $CONFIG_SHELL, we're ensured to have - # already done that, so ensure we don't try to do so again and fall - # in an infinite loop. This has already happened in practice. - _as_can_reexec=no; export _as_can_reexec - # Don't try to exec as it changes $[0], causing all sort of problems - # (the dirname of $[0] is not the place where we might find the - # original and so on. Autoconf is especially sensitive to this). - . "./$as_me.lineno" - # Exit status is that of the last command. - exit -} - - -# ac_fn_c_try_compile LINENO -# -------------------------- -# Try to compile conftest.$ac_ext, and return whether this succeeded. -ac_fn_c_try_compile () -{ - as_lineno=${as_lineno-"$1"} as_lineno_stack=as_lineno_stack=$as_lineno_stack - rm -f conftest.$ac_objext - if { { ac_try="$ac_compile" -case "(($ac_try" in - *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;; - *) ac_try_echo=$ac_try;; -esac -eval ac_try_echo="\"\$as_me:${as_lineno-$LINENO}: $ac_try_echo\"" -$as_echo "$ac_try_echo"; } >&5 - (eval "$ac_compile") 2>conftest.err - ac_status=$? - if test -s conftest.err; then - grep -v '^ *+' conftest.err >conftest.er1 - cat conftest.er1 >&5 - mv -f conftest.er1 conftest.err - fi - $as_echo "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5 - test $ac_status = 0; } && { - test -z "$ac_c_werror_flag" || - test ! -s conftest.err - } && test -s conftest.$ac_objext; then : - ac_retval=0 -else - $as_echo "$as_me: failed program was:" >&5 -sed 's/^/| /' conftest.$ac_ext >&5 - - ac_retval=1 -fi - eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno - as_fn_set_status $ac_retval - -} # ac_fn_c_try_compile - # ac_fn_c_check_header_compile LINENO HEADER VAR INCLUDES # ------------------------------------------------------- # Tests whether HEADER exists and can be compiled using the include files in diff --git a/libc/sysdeps/x86_64/dl-tlsdesc.h b/libc/sysdeps/x86_64/dl-tlsdesc.h index 06ede0214..ec750904f 100644 --- a/libc/sysdeps/x86_64/dl-tlsdesc.h +++ b/libc/sysdeps/x86_64/dl-tlsdesc.h @@ -20,12 +20,6 @@ #ifndef _X86_64_DL_TLSDESC_H # define _X86_64_DL_TLSDESC_H 1 -/* Use this to access DT_TLSDESC_PLT and DT_TLSDESC_GOT. */ -#ifndef ADDRIDX -# define ADDRIDX(tag) (DT_NUM + DT_THISPROCNUM + DT_VERSIONTAGNUM \ - + DT_EXTRANUM + DT_VALNUM + DT_ADDRTAGIDX (tag)) -#endif - /* Type used to represent a TLS descriptor in the GOT. */ struct tlsdesc { diff --git a/libc/sysdeps/x86_64/fpu/libm-test-ulps b/libc/sysdeps/x86_64/fpu/libm-test-ulps index b02b9e6ba..9c5157ff7 100644 --- a/libc/sysdeps/x86_64/fpu/libm-test-ulps +++ b/libc/sysdeps/x86_64/fpu/libm-test-ulps @@ -812,6 +812,12 @@ float: 1 ifloat: 1 ildouble: 1 ldouble: 1 +Test "Real part of: clog (0x0.ffffffffffffffffp0 + 0x0.ffffffffffffffffp-15000 i) == -5.421010862427522170184200798202494495630e-20 + 3.548665303440282824232502561095699343814e-4516 i": +ildouble: 1 +ldouble: 1 +Test "Real part of: clog (0x0.ffffffp0 + 0x0.ffffffp-100 i) == -5.960464655174753498633255797994360530379e-8 + 7.888609052210118054117285652827862296732e-31 i": +float: 1 +ifloat: 1 Test "Real part of: clog (0x1.000566p0 + 0x1.234p-10 i) == 8.298731898331237038231468223024422855654e-5 + 1.110938609507128729312743251313024793990e-3 i": float: 1 ifloat: 1 @@ -841,6 +847,9 @@ ldouble: 1 Test "Real part of: clog (0x1.fp+16383 - 0x1p-16445 i) == 11356.49165759582936919077408168801636572 - 0 i": ildouble: 1 ldouble: 1 +Test "Imaginary part of: clog (0x1367a310575591p-54 + 0x3cfcc0a0541f60p-54 i) == 5.0844550531823026520677817684239496041087e-32 + 1.2627468605458094918919206628466016525397 i": +ildouble: 1 +ldouble: 1 Test "Real part of: clog (0x1p-1074 + 0x1p-1074 i) == -744.0934983311012896593986823853525458290 + pi/4 i": double: 1 idouble: 1 @@ -868,6 +877,12 @@ ldouble: 1 Test "Real part of: clog (0x1p-16445 - 0x1.fp+16383 i) == 11356.49165759582936919077408168801636572 - pi/2 i": ildouble: 1 ldouble: 1 +Test "Imaginary part of: clog (0x2818p-15 + 0x798fp-15 i) == 1.5366822245016167178749091974664853785194e-08 + 1.2522014929038946066987318471922169174157 i": +float: 1 +ifloat: 1 +Test "Real part of: clog (0x2dd46725bp-35 + 0x7783a1284p-35 i) == 4.4469229730850767799109418892826021157328e-20 + 1.2046235979300843056806465045930070146351 i": +ildouble: 1 +ldouble: 1 Test "Real part of: clog (1.0 + 0x1.234566p-10 i) == 6.172834701221959432440126967147726538097e-7 + 1.111110564353742042376451655136933182201e-3 i": float: 1 ifloat: 1 @@ -1009,6 +1024,21 @@ idouble: 1 ifloat: 1 ildouble: 1 ldouble: 1 +Test "Real part of: clog10 (0x0.fffffffffffff8p0 + 0x0.fffffffffffff8p-1000 i) == -4.821637332766435821255375046554377090472e-17 + 4.053112396770095089737411317782466262176e-302 i": +double: 1 +idouble: 1 +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x0.ffffffffffffffffp0 + 0x0.ffffffffffffffffp-15000 i) == -2.354315103889861110220423157644627849164e-20 + 1.541165759405643564697852372112893034397e-4516 i": +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x0.ffffffp0 + 0x0.ffffffp-100 i) == -2.588596909321764128428416045209904492216e-8 + 3.425979381266895667295625489912064603415e-31 i": +double: 1 +float: 2 +idouble: 1 +ifloat: 2 +ildouble: 1 +ldouble: 1 Test "Real part of: clog10 (0x1.000566p0 + 0x1.234p-10 i) == 3.604093470239754109961125085078190708674e-5 + 4.824745078422174667425851670822596859720e-4 i": float: 1 ifloat: 1 @@ -1068,6 +1098,29 @@ ldouble: 1 Test "Real part of: clog10 (0x1.fp+16383 - 0x1p-16445 i) == 4932.061660674182269085496060792589701158 - 0 i": ildouble: 1 ldouble: 1 +Test "Real part of: clog10 (0x10673dd0f2481p-51 + 0x7ef1d17cefbd2p-51 i) == 1.3918041236396763648388478552321724382899e-29 + 0.6263795733790237053262025311642907438291 i": +double: 1 +idouble: 1 +Test "Imaginary part of: clog10 (0x11682p-23 + 0x7ffed1p-23 i) == 5.0916490233953865181284669870035717560498e-13 + 0.6784968969384861816694467029319146542069 i": +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x1367a310575591p-54 + 0x3cfcc0a0541f60p-54 i) == 2.2081507730821788480616336165447731164865e-32 + 0.5484039935757001196548030312819898864760 i": +double: 1 +idouble: 1 +Test "Imaginary part of: clog10 (0x1367a310575591p-54 + 0x3cfcc0a0541f60p-54 i) == 2.2081507730821788480616336165447731164865e-32 + 0.5484039935757001196548030312819898864760 i": +double: 1 +idouble: 1 +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x164c74eea876p-45 + 0x16f393482f77p-45 i) == -1.3155760824064879362415202279780039150764e-26 + 0.3473590599762514228227328130640352044313 i": +double: 1 +idouble: 1 +Test "Real part of: clog10 (0x1a6p-10 + 0x3a5p-10 i) == -6.2126412844802358329771948751248003038444e-07 + 0.4977135139537443711784513409096950995985 i": +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x1a6p-10 + 0x3a5p-10 i) == -6.2126412844802358329771948751248003038444e-07 + 0.4977135139537443711784513409096950995985 i": +double: 1 +idouble: 1 Test "Imaginary part of: clog10 (0x1p-1073 + 0x1p-1073 i) == -322.8546703496198318667349645920187712089 + pi/4*log10(e) i": double: 1 idouble: 1 @@ -1151,6 +1204,75 @@ ifloat: 1 Test "Real part of: clog10 (0x1p-8190 + 1.0 i) == 2.920285685286322365786846845062520925172e-4932 + 6.821881769209206737428918127156778851051e-1 i": ildouble: 1 ldouble: 1 +Test "Real part of: clog10 (0x2818p-15 + 0x798fp-15 i) == 6.6737261053986614395049481326819059203910e-09 + 0.5438241985991753781478398141908629586460 i": +double: 1 +float: 1 +idouble: 1 +ifloat: 1 +Test "Imaginary part of: clog10 (0x2818p-15 + 0x798fp-15 i) == 6.6737261053986614395049481326819059203910e-09 + 0.5438241985991753781478398141908629586460 i": +float: 1 +ifloat: 1 +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x2dd46725bp-35 + 0x7783a1284p-35 i) == 1.9312741086596516918394613098872836703188e-20 + 0.5231613813514771042838490538484014771862 i": +double: 1 +idouble: 1 +Test "Imaginary part of: clog10 (0x2dd46725bp-35 + 0x7783a1284p-35 i) == 1.9312741086596516918394613098872836703188e-20 + 0.5231613813514771042838490538484014771862 i": +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x2ede88p-23 + 0x771c3fp-23 i) == -1.9440841725722970687903291200493082253766e-13 + 0.5193774116724956222518530053006822210323 i": +float: 1 +ifloat: 1 +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x2ede88p-23 + 0x771c3fp-23 i) == -1.9440841725722970687903291200493082253766e-13 + 0.5193774116724956222518530053006822210323 i": +double: 1 +idouble: 1 +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x4447d7175p-35 + 0x6c445e00ap-35 i) == -6.4375803621988389731799033530075237868110e-21 + 0.4378257977686804492768642780897650927167 i": +double: 1 +idouble: 1 +Test "Imaginary part of: clog10 (0x4447d7175p-35 + 0x6c445e00ap-35 i) == -6.4375803621988389731799033530075237868110e-21 + 0.4378257977686804492768642780897650927167 i": +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x4d4ep-15 + 0x6605p-15 i) == -7.0781945783414996953799915941870192015212e-09 + 0.4005747524909781155537088181659175147564 i": +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x5b06b680ea2ccp-52 + 0xef452b965da9fp-52 i) == 3.6079845358966994996207055940336690133424e-30 + 0.5243112258263349992771652393178033846555 i": +double: 1 +idouble: 1 +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x659b70ab7971bp-53 + 0x1f5d111e08abecp-53 i) == -1.0893543813872082317104059174982092534059e-30 + 0.5954257879188711495921161433751775633232 i": +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x659feap-24 + 0xeaf6f9p-24 i) == 1.6200701438094619117335617123525612051457e-14 + 0.5049027913635038013499728086604870749732 i": +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x6b10b4f3520217b6p-64 + 0xe8893cbb449253a1p-64 i) == 1.0529283395205396881397407610630442563938e-37 + 0.4947949395762683446121140513971996916447 i": +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x81b7efa81fc35ad1p-65 + 0x1ef4b835f1c79d812p-65 i) == -4.3074341162203896332989394770760901408798e-39 + 0.5709443672155660428417571212549720987784 i": +ildouble: 1 +ldouble: 1 +Test "Real part of: clog10 (0x8ecbf810c4ae6p-52 + 0xd479468b09a37p-52 i) == -4.2289432987513243393180377141513840878196e-30 + 0.4252020027092323591068799049905597805296 i": +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0x9b57bp-20 + 0xcb7b4p-20 i) == -1.7182001068739620267773842120965071561416e-11 + 0.3990121149225253562859800593935899629087 i": +double: 1 +idouble: 1 +Test "Real part of: clog10 (0xf2p-10 + 0x3e3p-10 i) == 2.6921240173351112953324592659528481616879e-06 + 0.5785726025799636431142862788413361783862 i": +double: 1 +idouble: 1 +ildouble: 1 +ldouble: 1 +Test "Imaginary part of: clog10 (0xf2p-10 + 0x3e3p-10 i) == 2.6921240173351112953324592659528481616879e-06 + 0.5785726025799636431142862788413361783862 i": +double: 1 +idouble: 1 +Test "Imaginary part of: clog10 (0xfe961079616p-45 + 0x1bc37e09e6d1p-45 i) == 2.3329549194675052736016290082882121135546e-26 + 0.4561756099441139182878993697611751382976 i": +double: 1 +idouble: 1 Test "Imaginary part of: clog10 (1.0 + 0x1.234566p-10 i) == 2.680828048441605163181684680300513080769e-7 + 4.825491868832381486767558728169977751564e-4 i": double: 1 idouble: 1 @@ -1611,9 +1733,14 @@ ldouble: 1 Test "Imaginary part of: ctan (0x1p1023 + 1 i) == -0.2254627924997545057926782581695274244229 + 0.8786063118883068695462540226219865087189 i": ildouble: 1 ldouble: 1 +Test "Real part of: ctan (0x1p127 + 1 i) == 0.2446359391192790896381501310437708987204 + 0.9101334047676183761532873794426475906201 i": +float: 1 +ifloat: 1 Test "Imaginary part of: ctan (0x1p127 + 1 i) == 0.2446359391192790896381501310437708987204 + 0.9101334047676183761532873794426475906201 i": double: 1 +float: 2 idouble: 1 +ifloat: 2 Test "Real part of: ctan (0x3.243f6cp-1 + 0 i) == -2.287733242885645987394874673945769518150e7 + 0.0 i": float: 1 ifloat: 1 @@ -1758,7 +1885,12 @@ ildouble: 1 ldouble: 1 Test "Real part of: ctanh (1 + 0x1p127 i) == 0.9101334047676183761532873794426475906201 + 0.2446359391192790896381501310437708987204 i": double: 1 +float: 2 idouble: 1 +ifloat: 2 +Test "Imaginary part of: ctanh (1 + 0x1p127 i) == 0.9101334047676183761532873794426475906201 + 0.2446359391192790896381501310437708987204 i": +float: 1 +ifloat: 1 Test "Imaginary part of: ctanh (45 + 1 i) == 1.000000000000000000000000000000000000001 + 1.490158918874345552942703234806348520895e-39 i": ildouble: 1 ldouble: 1 @@ -2399,8 +2531,8 @@ ifloat: 2 ildouble: 1 ldouble: 1 Test "sin_upward (3) == 0.1411200080598672221007448028081102798469": -float: 1 -ifloat: 1 +float: 2 +ifloat: 2 ildouble: 1 ldouble: 1 Test "sin_upward (4) == -0.7568024953079282513726390945118290941359": @@ -2430,12 +2562,6 @@ ifloat: 1 Test "sincos (0.80190127184058835, &sin_res, &cos_res) puts 0.69534156199418473 in cos_res": double: 1 idouble: 1 -Test "sincos (0x1p+120, &sin_res, &cos_res) puts -9.25879022854837867303861764107414946730833e-01 in cos_res": -float: 1 -ifloat: 1 -Test "sincos (0x1p+127, &sin_res, &cos_res) puts 7.81914638714960072263910298466369236613162e-01 in cos_res": -float: 1 -ifloat: 1 Test "sincos (M_PI_6l*2.0, &sin_res, &cos_res) puts 0.5 in cos_res": double: 1 float: 1 @@ -2706,7 +2832,9 @@ ildouble: 2 ldouble: 2 Test "y1 (0x1.27e204p+99) == -8.881610148467797208469612080785210013461e-16": double: 1 +float: 1 idouble: 1 +ifloat: 1 ildouble: 1 ldouble: 1 Test "y1 (1.5) == -0.412308626973911295952829820633445323": @@ -3041,20 +3169,22 @@ ldouble: 1 Function: Imaginary part of "clog": float: 3 ifloat: 3 +ildouble: 1 +ldouble: 1 Function: Real part of "clog10": double: 2 -float: 1 +float: 2 idouble: 2 -ifloat: 1 +ifloat: 2 ildouble: 1 ldouble: 1 Function: Imaginary part of "clog10": double: 1 -float: 5 +float: 1 idouble: 1 -ifloat: 5 +ifloat: 1 ildouble: 1 ldouble: 1 @@ -3085,8 +3215,8 @@ ildouble: 1 ldouble: 1 Function: "cos_upward": -float: 1 -ifloat: 1 +float: 2 +ifloat: 2 ildouble: 1 ldouble: 1 @@ -3174,17 +3304,17 @@ ldouble: 1 Function: Real part of "ctan": double: 1 -float: 1 +float: 2 idouble: 1 -ifloat: 1 +ifloat: 2 ildouble: 2 ldouble: 2 Function: Imaginary part of "ctan": double: 1 -float: 1 +float: 2 idouble: 1 -ifloat: 1 +ifloat: 2 ildouble: 1 ldouble: 1 @@ -3242,9 +3372,9 @@ ldouble: 4 Function: Real part of "ctanh": double: 1 -float: 1 +float: 2 idouble: 1 -ifloat: 1 +ifloat: 2 ildouble: 1 ldouble: 1 @@ -3445,8 +3575,8 @@ ildouble: 1 ldouble: 1 Function: "sin_upward": -float: 1 -ifloat: 1 +float: 2 +ifloat: 2 ildouble: 1 ldouble: 1 diff --git a/libc/sysdeps/x86_64/fpu/s_sincosf.S b/libc/sysdeps/x86_64/fpu/s_sincosf.S new file mode 100644 index 000000000..3334a3a78 --- /dev/null +++ b/libc/sysdeps/x86_64/fpu/s_sincosf.S @@ -0,0 +1,558 @@ +/* Optimized sincosf function. + Copyright (C) 2012 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#include <sysdep.h> +#define __need_Emath +#include <bits/errno.h> + +/* Short algorithm description: + * + * 1) if |x|==0: sin(x)=x, + * cos(x)=1. + * 2) if |x|<2^-27: sin(x)=x-x*DP_SMALL, raising underflow only when needed, + * cos(x)=1-|x|. + * 3) if |x|<2^-5 : sin(x)=x+x*x^2*DP_SIN2_0+x^5*DP_SIN2_1, + * cos(x)=1+1*x^2*DP_COS2_0+x^5*DP_COS2_1 + * 4) if |x|< Pi/4: sin(x)=x+x*x^2*(S0+x^2*(S1+x^2*(S2+x^2*(S3+x^2*S4)))), + * cos(x)=1+1*x^2*(C0+x^2*(C1+x^2*(C2+x^2*(C3+x^2*C4)))). + * 5) if |x| < 9*Pi/4: + * 5.1) Range reduction: + * k=trunc(|x|/(Pi/4)), j=(k+1)&0x0e, n=k+1, t=|x|-j*Pi/4. + * 5.2) Reconstruction: + * sign_sin = sign(x) * (-1.0)^(( n >>2)&1) + * sign_cos = (-1.0)^(((n+2)>>2)&1) + * poly_sin = ((((S4*t^2 + S3)*t^2 + S2)*t^2 + S1)*t^2 + S0)*t^2*t+t + * poly_cos = ((((C4*t^2 + C3)*t^2 + C2)*t^2 + C1)*t^2 + C0)*t^2*s+s + * if(n&2 != 0) { + * using cos(t) and sin(t) polynomials for |t|<Pi/4, results are + * cos(x) = poly_sin * sign_cos + * sin(x) = poly_cos * sign_sin + * } else { + * sin(x) = poly_sin * sign_sin + * cos(x) = poly_cos * sign_cos + * } + * 6) if |x| < 2^23, large args: + * 6.1) Range reduction: + * k=trunc(|x|/(Pi/4)), j=(k+1)&0xfffffffe, n=k+1, t=|x|-j*Pi/4 + * 6.2) Reconstruction same as (5.2). + * 7) if |x| >= 2^23, very large args: + * 7.1) Range reduction: + * k=trunc(|x|/(Pi/4)), j=(k+1)&0xfffffffe, n=k+1, t=|x|-j*Pi/4. + * 7.2) Reconstruction same as (5.2). + * 8) if x is Inf, return x-x, and set errno=EDOM. + * 9) if x is NaN, return x-x. + * + * Special cases: + * sin/cos(+-0) = +-0/1 not raising inexact/underflow, + * sin/cos(subnormal) raises inexact/underflow, + * sin/cos(min_normalized) raises inexact/underflow, + * sin/cos(normalized) raises inexact, + * sin/cos(Inf) = NaN, raises invalid, sets errno to EDOM, + * sin/cos(NaN) = NaN. + */ + +# define ARG_SIN_PTR %rdi +# define ARG_COS_PTR %rsi + + .text +ENTRY(__sincosf) + /* Input: %xmm0 contains single precision argument x */ + /* %rdi points to sin result */ + /* %rsi points to cos result */ + + movd %xmm0, %eax /* Bits of x */ + movaps %xmm0, %xmm7 /* Copy of x */ + cvtss2sd %xmm0, %xmm0 /* DP x */ + movss L(SP_ABS_MASK)(%rip), %xmm3 + movl %eax, %r8d /* Copy of x bits */ + andl $0x7fffffff, %eax /* |x| */ + + cmpl $0x3f490fdb, %eax /* |x|<Pi/4 ? */ + jb L(arg_less_pio4) + + /* Here if |x|>=Pi/4 */ + andps %xmm7, %xmm3 /* SP |x| */ + andpd L(DP_ABS_MASK)(%rip),%xmm0 /* DP |x| */ + movss L(SP_INVPIO4)(%rip), %xmm2 /* SP 1/(Pi/4) */ + + cmpl $0x40e231d6, %eax /* |x|<9*Pi/4 ? */ + jae L(large_args) + + /* Here if Pi/4<=|x|<9*Pi/4 */ + mulss %xmm3, %xmm2 /* SP |x|/(Pi/4) */ + movl %r8d, %ecx /* Load x */ + cvttss2si %xmm2, %eax /* k, number of Pi/4 in x */ + lea L(PIO4J)(%rip), %r9 + shrl $29, %ecx /* (sign of x) << 2 */ + addl $1, %eax /* k+1 */ + movl $0x0e, %edx + andl %eax, %edx /* j = (k+1)&0x0e */ + subsd (%r9,%rdx,8), %xmm0 /* t = |x| - j * Pi/4 */ + +L(reconstruction): + /* Input: %eax=n, %xmm0=t, %ecx=sign(x) */ + + movaps %xmm0, %xmm4 /* t */ + movhpd L(DP_ONES)(%rip), %xmm4 /* 1|t */ + mulsd %xmm0, %xmm0 /* y=t^2 */ + movl $2, %edx + unpcklpd %xmm0, %xmm0 /* y|y */ + addl %eax, %edx /* k+2 */ + movaps %xmm0, %xmm1 /* y|y */ + mulpd %xmm0, %xmm0 /* z=t^4|z=t^4 */ + + movaps L(DP_SC4)(%rip), %xmm2 /* S4 */ + mulpd %xmm0, %xmm2 /* z*S4 */ + movaps L(DP_SC3)(%rip), %xmm3 /* S3 */ + mulpd %xmm0, %xmm3 /* z*S3 */ + xorl %eax, %ecx /* (sign_x ^ (k>>2))<<2 */ + addpd L(DP_SC2)(%rip), %xmm2 /* S2+z*S4 */ + mulpd %xmm0, %xmm2 /* z*(S2+z*S4) */ + shrl $2, %edx /* (k+2)>>2 */ + addpd L(DP_SC1)(%rip), %xmm3 /* S1+z*S3 */ + mulpd %xmm0, %xmm3 /* z*(S1+z*S3) */ + shrl $2, %ecx /* sign_x ^ k>>2 */ + addpd L(DP_SC0)(%rip), %xmm2 /* S0+z*(S2+z*S4) */ + andl $1, %edx /* sign_cos = ((k+2)>>2)&1 */ + mulpd %xmm1, %xmm2 /* y*(S0+z*(S2+z*S4)) */ + andl $1, %ecx /* sign_sin = sign_x ^ ((k>>2)&1) */ + addpd %xmm2, %xmm3 /* y*(S0+y*(S1+y*(S2+y*(S3+y*S4)))) */ + lea L(DP_ONES)(%rip), %r9 + mulpd %xmm4, %xmm3 /*t*y*(S0+y*(S1+y*(S2+y*(S3+y*S4))))*/ + testl $2, %eax /* n&2 != 0 ? */ + addpd %xmm4, %xmm3 /*t+t*y*(S0+y*(S1+y*(S2+y*(S3+y*S4))*/ + jnz L(sin_result_sin_poly) + +/*L(sin_result_cos_poly):*/ + /* + * Here if + * cos(x) = poly_sin * sign_cos + * sin(x) = poly_cos * sign_sin + */ + movsd (%r9,%rcx,8), %xmm4 /* 0|sign_sin */ + movhpd (%r9,%rdx,8), %xmm4 /* sign_cos|sign_sin */ + mulpd %xmm4, %xmm3 /* result_cos|result_sin */ + cvtpd2ps %xmm3, %xmm0 /* SP results */ + movss %xmm0, (ARG_SIN_PTR) /* store sin(x) from xmm0[0] */ + shufps $1, %xmm0, %xmm0 /* move cos(x) to xmm0[0] */ + movss %xmm0, (ARG_COS_PTR) /* store cos(x) */ + ret + + .p2align 4 +L(sin_result_sin_poly): + /* + * Here if + * sin(x) = poly_sin * sign_sin + * cos(x) = poly_cos * sign_cos + */ + movsd (%r9,%rdx,8), %xmm4 /* 0|sign_cos */ + movhpd (%r9,%rcx,8), %xmm4 /* sign_sin|sign_cos */ + mulpd %xmm4, %xmm3 /* result_sin|result_cos */ + cvtpd2ps %xmm3, %xmm0 /* SP results */ + movss %xmm0, (ARG_COS_PTR) /* store cos(x) from xmm0[0] */ + shufps $1, %xmm0, %xmm0 /* move sin(x) to xmm0[0] */ + movss %xmm0, (ARG_SIN_PTR) /* store sin(x) */ + ret + + .p2align 4 +L(large_args): + /* Here if |x|>=9*Pi/4 */ + cmpl $0x7f800000, %eax /* x is Inf or NaN ? */ + jae L(arg_inf_or_nan) + + /* Here if finite |x|>=9*Pi/4 */ + cmpl $0x4b000000, %eax /* |x|<2^23 ? */ + jae L(very_large_args) + + /* Here if 9*Pi/4<=|x|<2^23 */ + movsd L(DP_INVPIO4)(%rip), %xmm1 /* 1/(Pi/4) */ + mulsd %xmm0, %xmm1 /* |x|/(Pi/4) */ + cvttsd2si %xmm1, %eax /* k=trunc(|x|/(Pi/4)) */ + addl $1, %eax /* k+1 */ + movl %eax, %edx + andl $0xfffffffe, %edx /* j=(k+1)&0xfffffffe */ + cvtsi2sdl %edx, %xmm4 /* DP j */ + movl %r8d, %ecx /* Load x */ + movsd L(DP_PIO4HI)(%rip), %xmm2 /* -PIO4HI = high part of -Pi/4 */ + shrl $29, %ecx /* (sign of x) << 2 */ + mulsd %xmm4, %xmm2 /* -j*PIO4HI */ + movsd L(DP_PIO4LO)(%rip), %xmm3 /* -PIO4LO = low part of -Pi/4 */ + addsd %xmm2, %xmm0 /* |x| - j*PIO4HI */ + mulsd %xmm3, %xmm4 /* j*PIO4LO */ + addsd %xmm4, %xmm0 /* t = |x| - j*PIO4HI - j*PIO4LO */ + jmp L(reconstruction) + + .p2align 4 +L(very_large_args): + /* Here if finite |x|>=2^23 */ + + /* bitpos = (ix>>23) - BIAS_32 + 59; */ + shrl $23, %eax /* eb = biased exponent of x */ + subl $68, %eax /* bitpos=eb-0x7f+59, where 0x7f */ + /*is exponent bias */ + movl $28, %ecx /* %cl=28 */ + movl %eax, %edx /* bitpos copy */ + + /* j = bitpos/28; */ + div %cl /* j in register %al=%ax/%cl */ + movapd %xmm0, %xmm3 /* |x| */ + andl $0xff, %eax /* clear unneeded remainder from %ah*/ + + imull $28, %eax, %ecx /* j*28 */ + lea L(_FPI)(%rip), %r9 + movsd L(DP_HI_MASK)(%rip), %xmm4 /* DP_HI_MASK */ + movapd %xmm0, %xmm5 /* |x| */ + mulsd -16(%r9,%rax,8), %xmm3 /* tmp3 = FPI[j-2]*|x| */ + movapd %xmm0, %xmm1 /* |x| */ + mulsd -8(%r9,%rax,8), %xmm5 /* tmp2 = FPI[j-1]*|x| */ + mulsd (%r9,%rax,8), %xmm0 /* tmp0 = FPI[j]*|x| */ + addl $19, %ecx /* j*28+19 */ + mulsd 8(%r9,%rax,8), %xmm1 /* tmp1 = FPI[j+1]*|x| */ + cmpl %ecx, %edx /* bitpos>=j*28+19 ? */ + jl L(very_large_skip1) + + /* Here if bitpos>=j*28+19 */ + andpd %xmm3, %xmm4 /* HI(tmp3) */ + subsd %xmm4, %xmm3 /* tmp3 = tmp3 - HI(tmp3) */ +L(very_large_skip1): + + movsd L(DP_2POW52)(%rip), %xmm6 + movapd %xmm5, %xmm2 /* tmp2 copy */ + addsd %xmm3, %xmm5 /* tmp5 = tmp3 + tmp2 */ + movl $1, %edx + addsd %xmm5, %xmm6 /* tmp6 = tmp5 + 2^52 */ + movsd 8+L(DP_2POW52)(%rip), %xmm4 + movd %xmm6, %eax /* k = I64_LO(tmp6); */ + addsd %xmm6, %xmm4 /* tmp4 = tmp6 - 2^52 */ + movl %r8d, %ecx /* Load x */ + comisd %xmm5, %xmm4 /* tmp4 > tmp5 ? */ + jbe L(very_large_skip2) + + /* Here if tmp4 > tmp5 */ + subl $1, %eax /* k-- */ + addsd 8+L(DP_ONES)(%rip), %xmm4 /* tmp4 -= 1.0 */ +L(very_large_skip2): + + andl %eax, %edx /* k&1 */ + lea L(DP_ZERONE)(%rip), %r9 + subsd %xmm4, %xmm3 /* tmp3 -= tmp4 */ + addsd (%r9,%rdx,8), %xmm3 /* t = DP_ZERONE[k&1] + tmp3 */ + addsd %xmm2, %xmm3 /* t += tmp2 */ + shrl $29, %ecx /* (sign of x) << 2 */ + addsd %xmm3, %xmm0 /* t += tmp0 */ + addl $1, %eax /* n=k+1 */ + addsd %xmm1, %xmm0 /* t += tmp1 */ + mulsd L(DP_PIO4)(%rip), %xmm0 /* t *= PI04 */ + + jmp L(reconstruction) /* end of very_large_args peth */ + + .p2align 4 +L(arg_less_pio4): + /* Here if |x|<Pi/4 */ + cmpl $0x3d000000, %eax /* |x|<2^-5 ? */ + jl L(arg_less_2pn5) + + /* Here if 2^-5<=|x|<Pi/4 */ + movaps %xmm0, %xmm3 /* DP x */ + movhpd L(DP_ONES)(%rip), %xmm3 /* DP 1|x */ + mulsd %xmm0, %xmm0 /* DP y=x^2 */ + unpcklpd %xmm0, %xmm0 /* DP y|y */ + movaps %xmm0, %xmm1 /* y|y */ + mulpd %xmm0, %xmm0 /* z=x^4|z=x^4 */ + + movapd L(DP_SC4)(%rip), %xmm4 /* S4 */ + mulpd %xmm0, %xmm4 /* z*S4 */ + movapd L(DP_SC3)(%rip), %xmm5 /* S3 */ + mulpd %xmm0, %xmm5 /* z*S3 */ + addpd L(DP_SC2)(%rip), %xmm4 /* S2+z*S4 */ + mulpd %xmm0, %xmm4 /* z*(S2+z*S4) */ + addpd L(DP_SC1)(%rip), %xmm5 /* S1+z*S3 */ + mulpd %xmm0, %xmm5 /* z*(S1+z*S3) */ + addpd L(DP_SC0)(%rip), %xmm4 /* S0+z*(S2+z*S4) */ + mulpd %xmm1, %xmm4 /* y*(S0+z*(S2+z*S4)) */ + mulpd %xmm3, %xmm5 /* x*z*(S1+z*S3) */ + mulpd %xmm3, %xmm4 /* x*y*(S0+z*(S2+z*S4)) */ + addpd %xmm5, %xmm4 /*x*y*(S0+y*(S1+y*(S2+y*(S3+y*S4))))*/ + addpd %xmm4, %xmm3 /*x+x*y*(S0+y*(S1+y*(S2+y*(S3+y*S4))*/ + cvtpd2ps %xmm3, %xmm0 /* SP results */ + movss %xmm0, (ARG_SIN_PTR) /* store sin(x) from xmm0[0] */ + shufps $1, %xmm0, %xmm0 /* move cos(x) to xmm0[0] */ + movss %xmm0, (ARG_COS_PTR) /* store cos(x) */ + ret + + .p2align 4 +L(arg_less_2pn5): + /* Here if |x|<2^-5 */ + cmpl $0x32000000, %eax /* |x|<2^-27 ? */ + jl L(arg_less_2pn27) + + /* Here if 2^-27<=|x|<2^-5 */ + movaps %xmm0, %xmm1 /* DP x */ + movhpd L(DP_ONES)(%rip), %xmm1 /* DP 1|x */ + mulsd %xmm0, %xmm0 /* DP x^2 */ + unpcklpd %xmm0, %xmm0 /* DP x^2|x^2 */ + + movaps L(DP_SINCOS2_1)(%rip), %xmm3 /* DP DP_SIN2_1 */ + mulpd %xmm0, %xmm3 /* DP x^2*DP_SIN2_1 */ + addpd L(DP_SINCOS2_0)(%rip), %xmm3 /* DP DP_SIN2_0+x^2*DP_SIN2_1 */ + mulpd %xmm0, %xmm3 /* DP x^2*DP_SIN2_0+x^4*DP_SIN2_1 */ + mulpd %xmm1, %xmm3 /* DP x^3*DP_SIN2_0+x^5*DP_SIN2_1 */ + addpd %xmm1, %xmm3 /* DP x+x^3*DP_SIN2_0+x^5*DP_SIN2_1 */ + cvtpd2ps %xmm3, %xmm0 /* SP results */ + movss %xmm0, (ARG_SIN_PTR) /* store sin(x) from xmm0[0] */ + shufps $1, %xmm0, %xmm0 /* move cos(x) to xmm0[0] */ + movss %xmm0, (ARG_COS_PTR) /* store cos(x) */ + ret + + .p2align 4 +L(arg_less_2pn27): + cmpl $0, %eax /* x=0 ? */ + je L(arg_zero) /* in case x=0 return sin(+-0)==+-0 */ + /* Here if |x|<2^-27 */ + /* + * Special cases here: + * sin(subnormal) raises inexact/underflow + * sin(min_normalized) raises inexact/underflow + * sin(normalized) raises inexact + * cos(here)=1-|x| (raising inexact) + */ + movaps %xmm0, %xmm3 /* DP x */ + mulsd L(DP_SMALL)(%rip), %xmm0/* DP x*DP_SMALL */ + subsd %xmm0, %xmm3 /* DP sin result is x-x*DP_SMALL */ + andps L(SP_ABS_MASK)(%rip), %xmm7/* SP |x| */ + cvtsd2ss %xmm3, %xmm0 /* sin(x) */ + movss L(SP_ONE)(%rip), %xmm1 /* SP 1.0 */ + movss %xmm0, (ARG_SIN_PTR) /* sin(x) store */ + subss %xmm7, %xmm1 /* cos(x) */ + movss %xmm1, (ARG_COS_PTR) /* cos(x) store */ + ret + + .p2align 4 +L(arg_zero): + movss L(SP_ONE)(%rip), %xmm0 /* 1.0 */ + movss %xmm7, (ARG_SIN_PTR) /* sin(+-0)==x */ + movss %xmm0, (ARG_COS_PTR) /* cos(+-0)==1 */ + ret + + .p2align 4 +L(arg_inf_or_nan): + /* Here if |x| is Inf or NAN */ + jne L(skip_errno_setting) /* in case of x is NaN */ + + /* Here if x is Inf. Set errno to EDOM. */ + call JUMPTARGET(__errno_location) + movl $EDOM, (%rax) + + .p2align 4 +L(skip_errno_setting): + /* Here if |x| is Inf or NAN. Continued. */ + subss %xmm7, %xmm7 /* x-x, result is NaN */ + movss %xmm7, (ARG_SIN_PTR) + movss %xmm7, (ARG_COS_PTR) + ret +END(__sincosf) + + .section .rodata, "a" + .p2align 3 +L(PIO4J): /* Table of j*Pi/4, for j=0,1,..,10 */ + .long 0x00000000,0x00000000 + .long 0x54442d18,0x3fe921fb + .long 0x54442d18,0x3ff921fb + .long 0x7f3321d2,0x4002d97c + .long 0x54442d18,0x400921fb + .long 0x2955385e,0x400f6a7a + .long 0x7f3321d2,0x4012d97c + .long 0xe9bba775,0x4015fdbb + .long 0x54442d18,0x401921fb + .long 0xbeccb2bb,0x401c463a + .long 0x2955385e,0x401f6a7a + .type L(PIO4J), @object + ASM_SIZE_DIRECTIVE(L(PIO4J)) + + .p2align 3 +L(_FPI): /* 4/Pi broken into sum of positive DP values */ + .long 0x00000000,0x00000000 + .long 0x6c000000,0x3ff45f30 + .long 0x2a000000,0x3e3c9c88 + .long 0xa8000000,0x3c54fe13 + .long 0xd0000000,0x3aaf47d4 + .long 0x6c000000,0x38fbb81b + .long 0xe0000000,0x3714acc9 + .long 0x7c000000,0x3560e410 + .long 0x56000000,0x33bca2c7 + .long 0xac000000,0x31fbd778 + .long 0xe0000000,0x300b7246 + .long 0xe8000000,0x2e5d2126 + .long 0x48000000,0x2c970032 + .long 0xe8000000,0x2ad77504 + .long 0xe0000000,0x290921cf + .long 0xb0000000,0x274deb1c + .long 0xe0000000,0x25829a73 + .long 0xbe000000,0x23fd1046 + .long 0x10000000,0x2224baed + .long 0x8e000000,0x20709d33 + .long 0x80000000,0x1e535a2f + .long 0x64000000,0x1cef904e + .long 0x30000000,0x1b0d6398 + .long 0x24000000,0x1964ce7d + .long 0x16000000,0x17b908bf + .type L(_FPI), @object + ASM_SIZE_DIRECTIVE(L(_FPI)) + +/* Coefficients of polynomials for */ +/* sin(x)~=x+x*x^2*(DP_SIN2_0+x^2*DP_SIN2_1) in low DP part, */ +/* cos(x)~=1+1*x^2*(DP_COS2_0+x^2*DP_COS2_1) in high DP part, */ +/* for |x|<2^-5. */ + .p2align 4 +L(DP_SINCOS2_0): + .long 0x5543d49d,0xbfc55555 + .long 0xff5cc6fd,0xbfdfffff + .type L(DP_SINCOS2_0), @object + ASM_SIZE_DIRECTIVE(L(DP_SINCOS2_0)) + + .p2align 4 +L(DP_SINCOS2_1): + .long 0x75cec8c5,0x3f8110f4 + .long 0xb178dac5,0x3fa55514 + .type L(DP_SINCOS2_1), @object + ASM_SIZE_DIRECTIVE(L(DP_SINCOS2_1)) + + + .p2align 3 +L(DP_ZERONE): + .long 0x00000000,0x00000000 /* 0.0 */ + .long 0x00000000,0xbff00000 /* 1.0 */ + .type L(DP_ZERONE), @object + ASM_SIZE_DIRECTIVE(L(DP_ZERONE)) + + .p2align 3 +L(DP_ONES): + .long 0x00000000,0x3ff00000 /* +1.0 */ + .long 0x00000000,0xbff00000 /* -1.0 */ + .type L(DP_ONES), @object + ASM_SIZE_DIRECTIVE(L(DP_ONES)) + +/* Coefficients of polynomials for */ +/* sin(t)~=t+t*t^2*(S0+t^2*(S1+t^2*(S2+t^2*(S3+t^2*S4)))) in low DP part, */ +/* cos(t)~=1+1*t^2*(C0+t^2*(C1+t^2*(C2+t^2*(C3+t^2*C4)))) in high DP part, */ +/* for |t|<Pi/4. */ + .p2align 4 +L(DP_SC4): + .long 0x1674b58a,0xbe5a947e + .long 0xdd8844d7,0xbe923c97 + .type L(DP_SC4), @object + ASM_SIZE_DIRECTIVE(L(DP_SC4)) + + .p2align 4 +L(DP_SC3): + .long 0x64e6b5b4,0x3ec71d72 + .long 0x9ac43cc0,0x3efa00eb + .type L(DP_SC3), @object + ASM_SIZE_DIRECTIVE(L(DP_SC3)) + + .p2align 4 +L(DP_SC2): + .long 0x8b4bd1f9,0xbf2a019f + .long 0x348b6874,0xbf56c16b + .type L(DP_SC2), @object + ASM_SIZE_DIRECTIVE(L(DP_SC2)) + + .p2align 4 +L(DP_SC1): + .long 0x10c2688b,0x3f811111 + .long 0x545c50c7,0x3fa55555 + .type L(DP_SC1), @object + ASM_SIZE_DIRECTIVE(L(DP_SC1)) + + .p2align 4 +L(DP_SC0): + .long 0x55551cd9,0xbfc55555 + .long 0xfffe98ae,0xbfdfffff + .type L(DP_SC0), @object + ASM_SIZE_DIRECTIVE(L(DP_SC0)) + + .p2align 3 +L(DP_SMALL): + .long 0x00000000,0x3cd00000 /* 2^(-50) */ + .type L(DP_SMALL), @object + ASM_SIZE_DIRECTIVE(L(DP_SMALL)) + + .p2align 3 +L(DP_PIO4): + .long 0x54442d18,0x3fe921fb /* Pi/4 */ + .type L(DP_PIO4), @object + ASM_SIZE_DIRECTIVE(L(DP_PIO4)) + + .p2align 3 +L(DP_2POW52): + .long 0x00000000,0x43300000 /* +2^52 */ + .long 0x00000000,0xc3300000 /* -2^52 */ + .type L(DP_2POW52), @object + ASM_SIZE_DIRECTIVE(L(DP_2POW52)) + + .p2align 3 +L(DP_INVPIO4): + .long 0x6dc9c883,0x3ff45f30 /* 4/Pi */ + .type L(DP_INVPIO4), @object + ASM_SIZE_DIRECTIVE(L(DP_INVPIO4)) + + .p2align 3 +L(DP_PIO4HI): + .long 0x54000000,0xbfe921fb /* High part of Pi/4 */ + .type L(DP_PIO4HI), @object + ASM_SIZE_DIRECTIVE(L(DP_PIO4HI)) + + .p2align 3 +L(DP_PIO4LO): + .long 0x11A62633,0xbe010b46 /* Low part of Pi/4 */ + .type L(DP_PIO4LO), @object + ASM_SIZE_DIRECTIVE(L(DP_PIO4LO)) + + .p2align 2 +L(SP_INVPIO4): + .long 0x3fa2f983 /* 4/Pi */ + .type L(SP_INVPIO4), @object + ASM_SIZE_DIRECTIVE(L(SP_INVPIO4)) + + .p2align 4 +L(DP_ABS_MASK): /* Mask for getting DP absolute value */ + .long 0xffffffff,0x7fffffff + .long 0xffffffff,0x7fffffff + .type L(DP_ABS_MASK), @object + ASM_SIZE_DIRECTIVE(L(DP_ABS_MASK)) + + .p2align 3 +L(DP_HI_MASK): /* Mask for getting high 21 bits of DP value */ + .long 0x00000000,0xffffffff + .type L(DP_HI_MASK), @object + ASM_SIZE_DIRECTIVE(L(DP_HI_MASK)) + + .p2align 4 +L(SP_ABS_MASK): /* Mask for getting SP absolute value */ + .long 0x7fffffff,0x7fffffff + .long 0x7fffffff,0x7fffffff + .type L(SP_ABS_MASK), @object + ASM_SIZE_DIRECTIVE(L(SP_ABS_MASK)) + + .p2align 2 +L(SP_ONE): + .long 0x3f800000 /* 1.0 */ + .type L(SP_ONE), @object + ASM_SIZE_DIRECTIVE(L(SP_ONE)) + +weak_alias(__sincosf, sincosf) diff --git a/libc/sysdeps/x86_64/multiarch/init-arch.c b/libc/sysdeps/x86_64/multiarch/init-arch.c index fb44dcfcf..fed5ab898 100644 --- a/libc/sysdeps/x86_64/multiarch/init-arch.c +++ b/libc/sysdeps/x86_64/multiarch/init-arch.c @@ -156,6 +156,9 @@ __init_cpu_features (void) /* Determine if AVX is usable. */ if (CPUID_AVX) __cpu_features.feature[index_AVX_Usable] |= bit_AVX_Usable; + /* Determine if FMA is usable. */ + if (CPUID_FMA) + __cpu_features.feature[index_FMA_Usable] |= bit_FMA_Usable; /* Determine if FMA4 is usable. */ if (CPUID_FMA4) __cpu_features.feature[index_FMA4_Usable] |= bit_FMA4_Usable; diff --git a/libc/sysdeps/x86_64/multiarch/init-arch.h b/libc/sysdeps/x86_64/multiarch/init-arch.h index 45e265167..f33f1c81e 100644 --- a/libc/sysdeps/x86_64/multiarch/init-arch.h +++ b/libc/sysdeps/x86_64/multiarch/init-arch.h @@ -22,7 +22,8 @@ #define bit_Fast_Unaligned_Load (1 << 4) #define bit_Prefer_PMINUB_for_stringop (1 << 5) #define bit_AVX_Usable (1 << 6) -#define bit_FMA4_Usable (1 << 7) +#define bit_FMA_Usable (1 << 7) +#define bit_FMA4_Usable (1 << 8) /* CPUID Feature flags. */ #define bit_SSE2 (1 << 26) @@ -56,6 +57,7 @@ # define index_Fast_Unaligned_Load FEATURE_INDEX_1*FEATURE_SIZE # define index_Prefer_PMINUB_for_stringop FEATURE_INDEX_1*FEATURE_SIZE # define index_AVX_Usable FEATURE_INDEX_1*FEATURE_SIZE +# define index_FMA_Usable FEATURE_INDEX_1*FEATURE_SIZE # define index_FMA4_Usable FEATURE_INDEX_1*FEATURE_SIZE #else /* __ASSEMBLER__ */ @@ -131,6 +133,8 @@ extern const struct cpu_features *__get_cpu_features (void) HAS_CPUID_FLAG (COMMON_CPUID_INDEX_1, ecx, bit_OSXSAVE) # define CPUID_AVX \ HAS_CPUID_FLAG (COMMON_CPUID_INDEX_1, ecx, bit_AVX) +# define CPUID_FMA \ + HAS_CPUID_FLAG (COMMON_CPUID_INDEX_1, ecx, bit_FMA) # define CPUID_FMA4 \ HAS_CPUID_FLAG (COMMON_CPUID_INDEX_80000001, ecx, bit_FMA4) @@ -140,7 +144,6 @@ extern const struct cpu_features *__get_cpu_features (void) # define HAS_SSSE3 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSSE3) # define HAS_SSE4_1 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_1) # define HAS_SSE4_2 HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_2) -# define HAS_FMA HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_FMA) # define index_Fast_Rep_String FEATURE_INDEX_1 # define index_Fast_Copy_Backward FEATURE_INDEX_1 @@ -148,6 +151,7 @@ extern const struct cpu_features *__get_cpu_features (void) # define index_Prefer_SSE_for_memop FEATURE_INDEX_1 # define index_Fast_Unaligned_Load FEATURE_INDEX_1 # define index_AVX_Usable FEATURE_INDEX_1 +# define index_FMA_Usable FEATURE_INDEX_1 # define index_FMA4_Usable FEATURE_INDEX_1 # define HAS_ARCH_FEATURE(name) \ @@ -159,6 +163,7 @@ extern const struct cpu_features *__get_cpu_features (void) # define HAS_PREFER_SSE_FOR_MEMOP HAS_ARCH_FEATURE (Prefer_SSE_for_memop) # define HAS_FAST_UNALIGNED_LOAD HAS_ARCH_FEATURE (Fast_Unaligned_Load) # define HAS_AVX HAS_ARCH_FEATURE (AVX_Usable) +# define HAS_FMA HAS_ARCH_FEATURE (FMA_Usable) # define HAS_FMA4 HAS_ARCH_FEATURE (FMA4_Usable) #endif /* __ASSEMBLER__ */ diff --git a/libc/sysdeps/x86_64/multiarch/strncmp-ssse3.S b/libc/sysdeps/x86_64/multiarch/strncmp-ssse3.S index a320a3e94..96380a46b 100644 --- a/libc/sysdeps/x86_64/multiarch/strncmp-ssse3.S +++ b/libc/sysdeps/x86_64/multiarch/strncmp-ssse3.S @@ -1,4 +1,6 @@ -#define USE_SSSE3 1 -#define STRCMP __strncmp_ssse3 -#define USE_AS_STRNCMP -#include "../strcmp.S" +#ifdef SHARED +# define USE_SSSE3 1 +# define STRCMP __strncmp_ssse3 +# define USE_AS_STRNCMP +# include "../strcmp.S" +#endif |