Re: [Qemu-devel] [PATCH] fix fdiv instruction

[David Gibson]

On Sun, Jun 24, 2018 at 11:24:17PM -0400, G 3 wrote:
> 
> On Jun 24, 2018, at 8:46 PM, David Gibson wrote:
> 
> > On Fri, Jun 22, 2018 at 10:22:58PM -0400, John Arbuckle wrote:
> > > When the fdiv instruction divides a finite number by zero,
> > > the result actually depends on the FPSCR[ZE] bit. If this
> > > bit is set, the return value is zero. If it is not set
> > > the result should be either positive or negative infinity.
> > > The sign of this result would depend on the sign of the
> > > two inputs. What currently happens is only infinity is
> > > returned even if the FPSCR[ZE] bit is set. This patch
> > > fixes this problem by actually checking the FPSCR[ZE] bit
> > > when deciding what the answer should be.
> > > 
> > > fdiv is suppose to only set the FPSCR's FPRF bits during a
> > > division by zero situation when the FPSCR[ZE] is not set.
> > > What currently happens is these bits are always set. This
> > > patch fixes this problem by checking the FPSCR[ZE] bit to
> > > decide if the FPRF bits should be set.
> > > 
> > > https://www.pdfdrive.net/powerpc-microprocessor-family-the-programming-environments-for-32-e3087633.html
> > > This document has the information on the fdiv. Page 133 has the
> > > information on what action is executed when a division by zero
> > > situation takes place.
> > 
> > I'm looking at that table and there is definitely no mention of a 0
> > *result* in any situation.  So this patch is definitely wrong on that
> > point.  If there's something else this is fixing, then the commit
> > message needs to describe that.
> 
> I did not find any mention of a zero result *yet*. There could be mention of
> something in another document. I will see what I can find. Right now I do
> have example code that I ran on a PowerPC 970 CPU and the results are as
> follows:
> 
> When dividing by zero:
> 
> if FPSCR[ZE] is enabled
>       then answer = 0

Really?  Or is it just that the result register already contained zero
and is being left unchanged as the document says should happen?

> if FPSCR[ZE] is disabled
>       then answer = 0x7ff0000000000000
> 
> I think this feature may be undocumented.
> 
> Here is the example program I used. When I ran this program, the FPSCR is
> set to zero initially (all bits disabled). When the mtfsb1 instruction is
> executed, the ZE bit is set. I ran this program twice. Once with the mtfsb1
> line uncommented, and another time with the line commented. The result for
> the answer was different in each situation.
> 
> #include <stdio.h>
> #include <stdint.h>
> #include <inttypes.h>
> 
> // Used to convert unsigned integer <--> double
> union Converter
> {
>     double d;
>     uint64_t i;
> };
> typedef union Converter Converter;
> 
> int main (int argc, const char * argv[]) {
>     Converter answer;
>     //asm volatile("mtfsb1 27"); /* Set ZE bit */
>     asm volatile("fdiv %0, %1, %2" : "=f"(answer.d) : "f"(1.0), "f"(0.0));
>     printf("answer = 0x%" PRIx64 "\n", answer.i);
>     return 0;
> }
> 
> 
> 

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature