1
0
mirror of https://github.com/Interlisp/maiko.git synced 2026-01-14 15:36:34 +00:00
Bruce Mitchener d311b0f4c8
Remove unintentional coercion from float to double. (#256)
By adding `0.0` and not `0.0f` and by calling `fmod()` rather than
`fmodf()`, we were unintentionally coercing the value from a
`float` to a `double`.

This resulted in x86_64 assembler like this:

```
    cvtss2sd %xmm0,%xmm0
    cvtss2sd %xmm1,%xmm1
    callq  403340 <fmod@plt>
    cvtsd2ss %xmm0,%xmm0
```

which is now:

```
    callq  403360 <fmodf@plt>
```

And for the `N_OP_equal` change:

```
    cvtss2sd %xmm0,%xmm0
    xorpd  %xmm2,%xmm2
    addsd  %xmm2,%xmm0
    cvtss2sd %xmm1,%xmm1
    addsd  %xmm2,%xmm1
    xor    %ecx,%ecx
    ucomisd %xmm1,%xmm0
```

is now:

```
    xorps  %xmm2,%xmm2
    addss  %xmm2,%xmm0
    addss  %xmm2,%xmm1
    xor    %ecx,%ecx
    ucomiss %xmm1,%xmm0
```

(Note `ss` rather than `sd`, along with the missing `cvtss2sd` calls.)
2021-01-20 09:36:13 -08:00

80 lines
2.1 KiB
C

/* $Id: ubf2.c,v 1.3 1999/05/31 23:35:44 sybalsky Exp $ (C) Copyright Venue, All Rights Reserved */
/* ubf2.c
*/
/************************************************************************/
/* */
/* (C) Copyright 1989-95 Venue. All Rights Reserved. */
/* Manufactured in the United States of America. */
/* */
/************************************************************************/
#include "version.h"
#include <stdio.h>
#include <math.h>
#include "lispemul.h"
#include "medleyfp.h"
#include "ubf2defs.h"
/************************************************************
OP_ub2 -- op 354 == UBFLOAT2
354/0 ADD
354/1 SUB
354/2 ISUB
354/3 MULT
354/4 DIV
354/5 GREAT
354/6 MAX
354/7 MIN
354/8 REM
354/9 AREF
***********************************************************/
LispPTR N_OP_ubfloat2(int a2, int a1, int alpha) {
float arg1, arg2;
float ans;
int ret;
/* Unboxed floating point args (a1, a2) look like float, but */
/* can't be declared as such because all float args get passed */
/* as double. (That can be avoided in ansi C.) Instead, they */
/* are declared int and accessed as float through cast pointers. */
/* The return value is handled similarly. */
arg1 = *(float *)&a1;
arg2 = *(float *)&a2;
FPCLEAR;
switch (alpha) {
case 0: /* add */ ans = arg1 + arg2; break;
case 1: /* sub */ ans = arg2 - arg1; break;
case 2: /* isub */ ans = arg1 - arg2; break;
case 3: /* mul */ ans = arg1 * arg2; break;
case 4: /* div */ ans = arg2 / arg1; break;
case 5: /* gt */
if (arg2 > arg1)
return (ATOM_T);
else
return (NIL_PTR);
case 6: /* max */
if (arg2 > arg1)
return (a2);
else
return (a1);
case 7: /* min */
if (arg2 > arg1)
return (a1);
else
return (a2);
case 8: /* rem */ ans = fmodf(arg2, arg1); break;
default: ERROR_EXIT(a1);
} /* end switch */
ret = *(int *)&ans; /* Convert it back to int for 32 bit storage */
if (FPTEST(ans)) ERROR_EXIT(a1);
return (ret);
} /* end N_OP_ub2() */
/* end module */