From bc33e617040166e971ec1e6822ac1cc417eb6c9c Mon Sep 17 00:00:00 2001 From: Rich Felker Date: Mon, 19 Mar 2012 04:56:07 -0400 Subject: asm for inverse trig functions unlike trig functions, these are easy to do in asm because they do not involve (arbitrary-precision) argument reduction. fpatan automatically takes care of domain issues, and in asin and acos, fsqrt takes care of them for us. --- src/math/i386/acos.s | 24 ++++++++++++++++++++++++ src/math/i386/acosf.s | 1 + src/math/i386/acosl.s | 1 + src/math/i386/asin.s | 23 +++++++++++++++++++++++ src/math/i386/asinf.s | 1 + src/math/i386/asinl.s | 1 + src/math/i386/atan.s | 7 +++++++ src/math/i386/atan2.s | 7 +++++++ src/math/i386/atan2f.s | 7 +++++++ src/math/i386/atan2l.s | 7 +++++++ src/math/i386/atanf.s | 7 +++++++ src/math/i386/atanl.s | 7 +++++++ 12 files changed, 93 insertions(+) create mode 100644 src/math/i386/acos.s create mode 100644 src/math/i386/acosf.s create mode 100644 src/math/i386/acosl.s create mode 100644 src/math/i386/asin.s create mode 100644 src/math/i386/asinf.s create mode 100644 src/math/i386/asinl.s create mode 100644 src/math/i386/atan.s create mode 100644 src/math/i386/atan2.s create mode 100644 src/math/i386/atan2f.s create mode 100644 src/math/i386/atan2l.s create mode 100644 src/math/i386/atanf.s create mode 100644 src/math/i386/atanl.s (limited to 'src') diff --git a/src/math/i386/acos.s b/src/math/i386/acos.s new file mode 100644 index 00000000..6f9ef7f0 --- /dev/null +++ b/src/math/i386/acos.s @@ -0,0 +1,24 @@ +.global acosf +.type acosf,@function +acosf: + flds 4(%esp) + jmp 1f + +.global acosl +.type acosl,@function +acosl: + fldt 4(%esp) + jmp 1f + +.global acos +.type acos,@function +acos: + fldl 4(%esp) +1: fld %st(0) + fmul %st(0) + fld1 + fsubp %st(1) + fsqrt + fxch %st(1) + fpatan + ret diff --git a/src/math/i386/acosf.s b/src/math/i386/acosf.s new file mode 100644 index 00000000..6c95509f --- /dev/null +++ b/src/math/i386/acosf.s @@ -0,0 +1 @@ +# see acos.s diff --git a/src/math/i386/acosl.s b/src/math/i386/acosl.s new file mode 100644 index 00000000..6c95509f --- /dev/null +++ b/src/math/i386/acosl.s @@ -0,0 +1 @@ +# see acos.s diff --git a/src/math/i386/asin.s b/src/math/i386/asin.s new file mode 100644 index 00000000..cab7bfc8 --- /dev/null +++ b/src/math/i386/asin.s @@ -0,0 +1,23 @@ +.global asinf +.type asinf,@function +asinf: + flds 4(%esp) + jmp 1f + +.global asinl +.type asinl,@function +asinl: + fldt 4(%esp) + jmp 1f + +.global asin +.type asin,@function +asin: + fldl 4(%esp) +1: fld %st(0) + fmul %st(0) + fld1 + fsubp %st(1) + fsqrt + fpatan + ret diff --git a/src/math/i386/asinf.s b/src/math/i386/asinf.s new file mode 100644 index 00000000..e07bf599 --- /dev/null +++ b/src/math/i386/asinf.s @@ -0,0 +1 @@ +# see asin.s diff --git a/src/math/i386/asinl.s b/src/math/i386/asinl.s new file mode 100644 index 00000000..e07bf599 --- /dev/null +++ b/src/math/i386/asinl.s @@ -0,0 +1 @@ +# see asin.s diff --git a/src/math/i386/atan.s b/src/math/i386/atan.s new file mode 100644 index 00000000..7e28b395 --- /dev/null +++ b/src/math/i386/atan.s @@ -0,0 +1,7 @@ +.global atan +.type atan,@function +atan: + fldl 4(%esp) + fld1 + fpatan + ret diff --git a/src/math/i386/atan2.s b/src/math/i386/atan2.s new file mode 100644 index 00000000..5c795aff --- /dev/null +++ b/src/math/i386/atan2.s @@ -0,0 +1,7 @@ +.global atan2 +.type atan2,@function +atan2: + fldl 4(%esp) + fldl 12(%esp) + fpatan + ret diff --git a/src/math/i386/atan2f.s b/src/math/i386/atan2f.s new file mode 100644 index 00000000..08627e63 --- /dev/null +++ b/src/math/i386/atan2f.s @@ -0,0 +1,7 @@ +.global atan2f +.type atan2f,@function +atan2f: + flds 4(%esp) + flds 8(%esp) + fpatan + ret diff --git a/src/math/i386/atan2l.s b/src/math/i386/atan2l.s new file mode 100644 index 00000000..adf6e10a --- /dev/null +++ b/src/math/i386/atan2l.s @@ -0,0 +1,7 @@ +.global atan2l +.type atan2l,@function +atan2l: + fldt 4(%esp) + fldt 16(%esp) + fpatan + ret diff --git a/src/math/i386/atanf.s b/src/math/i386/atanf.s new file mode 100644 index 00000000..3cd40233 --- /dev/null +++ b/src/math/i386/atanf.s @@ -0,0 +1,7 @@ +.global atanf +.type atanf,@function +atanf: + flds 4(%esp) + fld1 + fpatan + ret diff --git a/src/math/i386/atanl.s b/src/math/i386/atanl.s new file mode 100644 index 00000000..c508bc46 --- /dev/null +++ b/src/math/i386/atanl.s @@ -0,0 +1,7 @@ +.global atanl +.type atanl,@function +atanl: + fldt 4(%esp) + fld1 + fpatan + ret -- cgit v1.2.1 From 58ff9e8eaf99f2294232be860daa2ca6f3674faf Mon Sep 17 00:00:00 2001 From: Rich Felker Date: Mon, 19 Mar 2012 05:15:30 -0400 Subject: asm for scalbn family unlike some implementations, these functions perform the equivalent of gcc's -ffloat-store on the result before returning. this is necessary to raise underflow/overflow/inexact exceptions, perform the correct rounding with denormals, etc. --- src/math/i386/ldexp.s | 1 + src/math/i386/ldexpf.s | 1 + src/math/i386/ldexpl.s | 1 + src/math/i386/scalbln.s | 1 + src/math/i386/scalblnf.s | 1 + src/math/i386/scalblnl.s | 1 + src/math/i386/scalbn.s | 20 ++++++++++++++++++++ src/math/i386/scalbnf.s | 20 ++++++++++++++++++++ src/math/i386/scalbnl.s | 18 ++++++++++++++++++ 9 files changed, 64 insertions(+) create mode 100644 src/math/i386/ldexp.s create mode 100644 src/math/i386/ldexpf.s create mode 100644 src/math/i386/ldexpl.s create mode 100644 src/math/i386/scalbln.s create mode 100644 src/math/i386/scalblnf.s create mode 100644 src/math/i386/scalblnl.s create mode 100644 src/math/i386/scalbn.s create mode 100644 src/math/i386/scalbnf.s create mode 100644 src/math/i386/scalbnl.s (limited to 'src') diff --git a/src/math/i386/ldexp.s b/src/math/i386/ldexp.s new file mode 100644 index 00000000..c430f784 --- /dev/null +++ b/src/math/i386/ldexp.s @@ -0,0 +1 @@ +# see scalbn.s diff --git a/src/math/i386/ldexpf.s b/src/math/i386/ldexpf.s new file mode 100644 index 00000000..3f8e4b95 --- /dev/null +++ b/src/math/i386/ldexpf.s @@ -0,0 +1 @@ +# see scalbnf.s diff --git a/src/math/i386/ldexpl.s b/src/math/i386/ldexpl.s new file mode 100644 index 00000000..86fe5621 --- /dev/null +++ b/src/math/i386/ldexpl.s @@ -0,0 +1 @@ +# see scalbnl.s diff --git a/src/math/i386/scalbln.s b/src/math/i386/scalbln.s new file mode 100644 index 00000000..c430f784 --- /dev/null +++ b/src/math/i386/scalbln.s @@ -0,0 +1 @@ +# see scalbn.s diff --git a/src/math/i386/scalblnf.s b/src/math/i386/scalblnf.s new file mode 100644 index 00000000..3f8e4b95 --- /dev/null +++ b/src/math/i386/scalblnf.s @@ -0,0 +1 @@ +# see scalbnf.s diff --git a/src/math/i386/scalblnl.s b/src/math/i386/scalblnl.s new file mode 100644 index 00000000..86fe5621 --- /dev/null +++ b/src/math/i386/scalblnl.s @@ -0,0 +1 @@ +# see scalbnl.s diff --git a/src/math/i386/scalbn.s b/src/math/i386/scalbn.s new file mode 100644 index 00000000..e275d14f --- /dev/null +++ b/src/math/i386/scalbn.s @@ -0,0 +1,20 @@ +.global ldexp +.type ldexp,@function +ldexp: + nop + +.global scalbln +.type scalbln,@function +scalbln: + nop + +.global scalbn +.type scalbn,@function +scalbn: + fildl 12(%esp) + fldl 4(%esp) + fscale + fstp %st(1) + fstpl 4(%esp) + fldl 4(%esp) + ret diff --git a/src/math/i386/scalbnf.s b/src/math/i386/scalbnf.s new file mode 100644 index 00000000..40232b6a --- /dev/null +++ b/src/math/i386/scalbnf.s @@ -0,0 +1,20 @@ +.global ldexpf +.type ldexpf,@function +ldexpf: + nop + +.global scalblnf +.type scalblnf,@function +scalblnf: + nop + +.global scalbnf +.type scalbnf,@function +scalbnf: + fildl 8(%esp) + flds 4(%esp) + fscale + fstp %st(1) + fstps 4(%esp) + flds 4(%esp) + ret diff --git a/src/math/i386/scalbnl.s b/src/math/i386/scalbnl.s new file mode 100644 index 00000000..224b1bef --- /dev/null +++ b/src/math/i386/scalbnl.s @@ -0,0 +1,18 @@ +.global ldexpl +.type ldexpl,@function +ldexpl: + nop + +.global scalblnl +.type scalblnl,@function +scalblnl: + nop + +.global scalbnl +.type scalbnl,@function +scalbnl: + fildl 16(%esp) + fldt 4(%esp) + fscale + fstp %st(1) + ret -- cgit v1.2.1