Fixed CFU implementation

diff --git a/mach/i386/libem/fp8087.s b/mach/i386/libem/fp8087.s
index 39ee20e..42a20df 100644 (file)
--- a/mach/i386/libem/fp8087.s
+++ b/mach/i386/libem/fp8087.s
@@ -329,7 +329,8 @@ bigmin:
        fstcw   4(bx)
        wait
        mov     dx,4(bx)
-       or      4(bx),0xc00     ! truncating mode
+       and     4(bx),0xf3ff
+       or      4(bx),0x400     ! to -infinity
        wait
        fldcw   4(bx)
        cmp     8(bx),4
@@ -337,8 +338,12 @@ bigmin:
                                ! loc 4 loc ? cfu
        flds    12(bx)
        fabs                    ! ???
+       fiaddl  (bigmin)
        fistpl  12(bx)
        wait
+       mov     ax,12(bx)
+       sub     ax,(bigmin)
+       mov     12(bx),ax
 1:
        mov     4(bx),dx
        wait
@@ -348,7 +353,12 @@ bigmin:
                                ! loc 8 loc ? cfu
        fldd    12(bx)
        fabs                    ! ???
+       fiaddl  (bigmin)
        fistpl  16(bx)
+       wait
+       mov     ax,16(bx)
+       sub     ax,(bigmin)
+       mov     16(bx),ax
        jmp     1b
 
 .cff4:

diff --git a/mach/i86/libem/fp8087.s b/mach/i86/libem/fp8087.s
index ef433c2..c6c71e7 100644 (file)
--- a/mach/i86/libem/fp8087.s
+++ b/mach/i86/libem/fp8087.s
@@ -487,7 +487,8 @@ bigmin:
        fstcw   -2(bx)
        wait
        mov     dx,-2(bx)
-       or      -2(bx),0xc00    ! truncating mode
+       and     -2(bx),0xf3ff
+       or      -2(bx),0x400    ! to -infinity
        wait
        fldcw   -2(bx)
        wait
@@ -499,8 +500,12 @@ bigmin:
        wait
        fabs                    ! ???
        wait
+       fiaddl  (bigmin)
        fistpl  6(bx)
        wait
+       mov     ax,8(bx)
+       sub     ax,(bigmin+2)
+       mov     8(bx),ax
        cmp     2(bx),2
        jne     1f
        mov     ax,6(bx)
@@ -517,8 +522,12 @@ bigmin:
        wait
        fabs                    ! ???
        wait
+       fiaddl  (bigmin)
        fistpl  10(bx)
        wait
+       mov     ax,12(bx)
+       sub     ax,(bigmin+2)
+       mov     12(bx),ax
        cmp     2(bx),2
        jne     1b
        mov     ax,10(bx)