Geert Uytterhoeven wrote:
> On Mon, Sep 24, 2012 at 3:56 PM, George Spelvin wrote:
>> SPARCv8 UMUL puts the high half of the 64-bit result into the Y
>> register, and SPARCv7 has a multiply-step instruction (MULScc) which
>> does likewise.
>
> Early SPARCs don't even have a multiply instruction.
On Mon, Sep 24, 2012 at 3:56 PM, George Spelvin wrote:
> Michal Nazarewicz wrote:
>> Didn't some SPARCs have 32x32->32 multiply? I remember reading some
>> rant from a GMP developer about how SPARC is broken that way.
>
> SPARCv9 only has 64x64->64; there's no 128-bit result version.
> That cuts
On Mon, Sep 24, 2012 at 2:35 PM, George Spelvin wrote:
>> Here is the comparison of the x86-32 assembly
>> of the fragment which does "x / 1" thing,
>> before and after the patch:
>
>> -01 c6 add%eax,%esi
>> -b8 59 17 b7 d1 mov$0xd1b71759,%eax
>> -f7 e6
Michal Nazarewicz wrote:
> Didn't some SPARCs have 32x32->32 multiply? I remember reading some
> rant from a GMP developer about how SPARC is broken that way.
SPARCv9 only has 64x64->64; there's no 128-bit result version.
That cuts large-integer math speed by a factor of 4 (very
crude approximat
On Mon, Sep 24 2012, George Spelvin wrote:
>> You are using a 64-bit multiply in a path that is designed for 32-bit
>> processors, which makes me feel that it will be slower.
>
> Slower than the divide it's replacing?
OK, granted, it might be faster after all. ;) Still, I'd love to see
some bench
> Here is the comparison of the x86-32 assembly
> of the fragment which does "x / 1" thing,
> before and after the patch:
> -01 c6 add%eax,%esi
> -b8 59 17 b7 d1 mov$0xd1b71759,%eax
> -f7 e6 mul%esi
> -89 d3 mov%edx,%eb
> You are using a 64-bit multiply in a path that is designed for 32-bit
> processors, which makes me feel that it will be slower.
Slower than the divide it's replacing?
The following 32-bit processors have 32x32->64-bit multiply:
x86
ARM (as of ARMv4 = ARM7TDMI, the lowest version in common use)
On Fri, Aug 3, 2012 at 7:21 AM, George Spelvin wrote:
> The same multiply-by-inverse technique can be used to
> convert division by 1 to a 32x32->64-bit multiply.
>
> Signed-off-by: George Spelvin
> ---
> lib/vsprintf.c | 60
> +++-
> 1
On Fri, Aug 03 2012, George Spelvin wrote:
> The same multiply-by-inverse technique can be used to
> convert division by 1 to a 32x32->64-bit multiply.
>
> Signed-off-by: George Spelvin
You are using a 64-bit multiply in a path that is designed for 32-bit
processors, which makes me feel that
The same multiply-by-inverse technique can be used to
convert division by 1 to a 32x32->64-bit multiply.
Signed-off-by: George Spelvin
---
lib/vsprintf.c | 60 +++-
1 file changed, 33 insertions(+), 27 deletions(-)
This is something of a
10 matches
Mail list logo
