x86_64 math asm, long double functions only
[musl] / src / math / x86_64 / expl.s
diff --git a/src/math/x86_64/expl.s b/src/math/x86_64/expl.s
new file mode 100644 (file)
index 0000000..64c1c78
--- /dev/null
@@ -0,0 +1,85 @@
+.global expm1l
+.type expm1l,@function
+expm1l:
+       fldt 8(%rsp)
+1:     fldl2e
+       fmulp
+       fld1
+       fld %st(1)
+       fabs
+       fucom %st(1)
+       fnstsw %ax
+       fstp %st(0)
+       fstp %st(0)
+       sahf
+       ja 1f
+       f2xm1
+       ret
+1:     push %rax
+       call 1f
+       pop %rax
+       fld1
+       fsubrp
+       ret
+
+.global expl
+.type expl,@function
+expl:
+       fldt 8(%rsp)
+       fldl2e
+       fmulp
+       jmp 1f
+
+.global exp2l
+.type exp2l,@function
+exp2l:
+       fldt 8(%rsp)
+1:     mov $0x467ff000,%eax
+       mov %eax,-16(%rsp)
+       mov $0x80000000,%eax
+       mov %eax,-20(%rsp)
+       xor %eax,%eax
+       mov %eax,-24(%rsp)
+       flds -16(%rsp)
+       fld %st(1)
+       fabs
+       fucom %st(1)
+       fnstsw
+       fstp %st(0)
+       fstp %st(0)
+       sahf
+       ja 2f
+       fld %st(0)
+       fistpl -16(%rsp)
+       fildl -16(%rsp)
+       fxch %st(1)
+       fsub %st(1)
+       mov $0x3fff,%eax
+       add %eax,-16(%rsp)
+       f2xm1
+       fld1
+       faddp
+       fldt -24(%rsp)
+       fmulp
+       fstp %st(1)
+       ret
+
+2:     fld %st(0)
+       fstpt -24(%rsp)
+       mov -15(%rsp),%ah
+       and $0x7f,%ah
+       cmp $0x7f,%ah
+       jne 1f
+       decb -15(%rsp)
+       fstp %st(0)
+       fldt -24(%rsp)
+1:     fld %st(0)
+       frndint
+       fxch %st(1)
+       fsub %st(1)
+       f2xm1
+       fld1
+       faddp
+       fscale
+       fstp %st(1)
+       ret