fix excess precision in return value of i386 acos[f] and asin[f]

analogous to commit 1c9afd69051a64cf085c6fb3674a444ff9a43857 for
atan[2][f].

6 files changed, 75 insertions, 42 deletions

diff --git a/src/math/i386/acos.s b/src/math/i386/acos.s
index 47f365ef..af423a2f 100644
--- a/src/math/i386/acos.s
+++ b/src/math/i386/acos.s
@@ -1,22 +1,10 @@
 # use acos(x) = atan2(fabs(sqrt((1-x)*(1+x))), x)
 
-.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)
+	fld %st(0)
 	fld1
 	fsub %st(0),%st(1)
 	fadd %st(2)
@@ -25,4 +13,6 @@ acos:
 	fabs         # fix sign of zero (matters in downward rounding mode)
 	fxch %st(1)
 	fpatan
+	fstpl 4(%esp)
+	fldl 4(%esp)
 	ret
diff --git a/src/math/i386/acosf.s b/src/math/i386/acosf.s
index 6c95509f..d2cdfdbf 100644
--- a/src/math/i386/acosf.s
+++ b/src/math/i386/acosf.s
@@ -1 +1,16 @@
-# see acos.s
+.global acosf
+.type acosf,@function
+acosf:
+	flds 4(%esp)
+	fld %st(0)
+	fld1
+	fsub %st(0),%st(1)
+	fadd %st(2)
+	fmulp
+	fsqrt
+	fabs         # fix sign of zero (matters in downward rounding mode)
+	fxch %st(1)
+	fpatan
+	fstps 4(%esp)
+	flds 4(%esp)
+	ret
diff --git a/src/math/i386/acosl.s b/src/math/i386/acosl.s
index 6c95509f..599c8230 100644
--- a/src/math/i386/acosl.s
+++ b/src/math/i386/acosl.s
@@ -1 +1,14 @@
-# see acos.s
+.global acosl
+.type acosl,@function
+acosl:
+	fldt 4(%esp)
+	fld %st(0)
+	fld1
+	fsub %st(0),%st(1)
+	fadd %st(2)
+	fmulp
+	fsqrt
+	fabs         # fix sign of zero (matters in downward rounding mode)
+	fxch %st(1)
+	fpatan
+	ret
diff --git a/src/math/i386/asin.s b/src/math/i386/asin.s
index 920d967a..2bc8356f 100644
--- a/src/math/i386/asin.s
+++ b/src/math/i386/asin.s
@@ -1,23 +1,3 @@
-.global asinf
-.type asinf,@function
-asinf:
-	flds 4(%esp)
-	mov 4(%esp),%eax
-	add %eax,%eax
-	cmp $0x01000000,%eax
-	jae 1f
-		# subnormal x, return x with underflow
-	fld %st(0)
-	fmul %st(1)
-	fstps 4(%esp)
-	ret
-
-.global asinl
-.type asinl,@function
-asinl:
-	fldt 4(%esp)
-	jmp 1f
-
 .global asin
 .type asin,@function
 asin:
@@ -25,15 +5,17 @@ asin:
 	mov 8(%esp),%eax
 	add %eax,%eax
 	cmp $0x00200000,%eax
-	jae 1f
-		# subnormal x, return x with underflow
-	fsts 4(%esp)
-	ret
-1:	fld %st(0)
+	jb 1f
+	fld %st(0)
 	fld1
 	fsub %st(0),%st(1)
 	fadd %st(2)
 	fmulp
 	fsqrt
 	fpatan
+	fstpl 4(%esp)
+	fldl 4(%esp)
+	ret
+		# subnormal x, return x with underflow
+1:	fsts 4(%esp)
 	ret
diff --git a/src/math/i386/asinf.s b/src/math/i386/asinf.s
index e07bf599..05909753 100644
--- a/src/math/i386/asinf.s
+++ b/src/math/i386/asinf.s
@@ -1 +1,23 @@
-# see asin.s
+.global asinf
+.type asinf,@function
+asinf:
+	flds 4(%esp)
+	mov 4(%esp),%eax
+	add %eax,%eax
+	cmp $0x01000000,%eax
+	jb 1f
+	fld %st(0)
+	fld1
+	fsub %st(0),%st(1)
+	fadd %st(2)
+	fmulp
+	fsqrt
+	fpatan
+	fstps 4(%esp)
+	flds 4(%esp)
+	ret
+		# subnormal x, return x with underflow
+1:	fld %st(0)
+	fmul %st(1)
+	fstps 4(%esp)
+	ret
diff --git a/src/math/i386/asinl.s b/src/math/i386/asinl.s
index e07bf599..e973fc85 100644
--- a/src/math/i386/asinl.s
+++ b/src/math/i386/asinl.s
@@ -1 +1,12 @@
-# see asin.s
+.global asinl
+.type asinl,@function
+asinl:
+	fldt 4(%esp)
+	fld %st(0)
+	fld1
+	fsub %st(0),%st(1)
+	fadd %st(2)
+	fmulp
+	fsqrt
+	fpatan
+	ret