Article: AMD's Jaguar Microarchitecture
By: Wilco (Wilco.Dijkstra.delete@this.ntlworld.com), April 6, 2014 4:29 am
Room: Moderated Discussions
computational_scientist (brian.bj.parker99.delete@this.gmail.com) on April 5, 2014 6:50 pm wrote:
> Linus Torvalds (torvalds.delete@this.linux-foundation.org) on April 4, 2014 4:56 pm wrote:
> > TREZA (no.delete@this.ema.il) on April 4, 2014 3:18 pm wrote:
> > >
> > > Denormals is what made IEEE FP superior to VAX. It is the correct way of doing FP math!
> >
> > Don't get me wrong - I actually personally like denormals, and think they are
> > definitely required for serious FP math, but the thing is, 99% of FP use isn't
> > really serious, and most people don't really understand FP anyway.
> >
> > The kind of people who really understand FP and do serious math using it (to the point of
> > caring about the order of operations, never mind denormals), those kinds of people really
> > do know what they are doing, and sometimes (although not always) really do want denormals.
> >
> > But there's the other side of FP math, which really just wants good enough values quickly. There are a lot
> > of people who are ok with single-precision and no denormals. And yes, quite often 24 bits of precision is
> > not enough, and they decide they actually need double precision in order to avoid odd visual artifacts.
> >
> > Yeah, I'm talking about things like games.
> >
> > And the thing is, the defaults tend to be the wrong way around. People who don't know what they
> > are doing with floating point basically *never* need denormals. You will generally hit other issues
> > long before you hit the "oops, I lost precision because I didn't have denormals". But exactly
> > *because* they don't know about floating point, they also don't know to disable them.
> >
> > So I think that from a hardware designer standpoint, it actually
> > would make more sense if the default was "denormals
> > are zero" (or "flush to zero"), because the people who do want denormals also know about them.
> > So while
> > I like denormals, the onus on disabling denormals when not needed is kind of the wrong way around.
>
> This is back to front reasoning: physicists, mathematicians and most computational scientists don't
> know anything about the details of floating point and use the defaults like everyone else.
> Even simple assumptions like x==y iff x-y==0 don't always hold in FTZ mode; default denormal mode
> is the simplest floating point model mathematically. By contrast, people running gaming benchmarks
> who know about these things can easily enable FTZ to speed them up for marketing purposes.
You can't rely on x==y iff x-y==0 even with denormals. IEEE floating point is extremely complex with lots of odd non-intuitive corner cases. Very few mathematical identities work on IEEE, with or without denormals. Denormals are a major pain in algorithms, for example you need special cases to avoid the catastrophic loss of precision (and the infinities/NaNs caused by calculations with a denormal) as well as the huge slowdown on many CPUs.
> http://www.cs.berkeley.edu/~wkahan/19July10.pdf
> has some interesting history and rationale for denormals.
> (There are several readable summaries of the rationale for other IEEE754
> features on Kahan's web site that are well worth reading).
>
> >
> > But it's too late, and big CPU's largely do already handle them right
> > and don't care, so it's mostly a problem for the embedded space.
>
> Using the denormal benchmark at http://charm.cs.uiuc.edu/subnormal/ , I see an (acceptable) 8x slowdown
> of denormals on my 2.3 GHz Intel Core i7 macbook pro in SSE mode and an (unacceptable) 53x slowdown
> in x87 mode (which is particularly egregious as x87 mode is needed for precise numerical work).
> FMAC instructions make fast denormal processing easy to implement,
> which is probably why Jaguar's denormal handling is slow.
8x slowdown is still unacceptable in my book - 10% is acceptable. POWER and ARM are at that level.
> It is a shame that as the importance of computational methods to society and the need for accurate
> and reliable floating point has increased in *absolute* terms over the last decades, the *relative*
> decreased usage compared with multimedia applications has lead to floating point hardware capabilities
> being degraded to meet gamers' needs. Processor speed, memory size, even screen resolution, have
> all increased monotonically over the last decades- floating point precision, range and reliability
> is the only feature that has actually decreased... very unfortunate.
The fact is the contrary is actually true. The non-IEEE compliant 80-bit x87 is finally dead - a huge step forward for floating point. Programming languages and compilers adopted IEEE and by default provide IEEE compliant optimizations as well user selectable adventurous FP optimizations. Libraries have improved hugely as well, providing far more accurate math functions (the norm is now < 1ULP error). 128-bit IEEE is becoming available in hardware. So we've made huge progress in the last decades.
Wilco
> Linus Torvalds (torvalds.delete@this.linux-foundation.org) on April 4, 2014 4:56 pm wrote:
> > TREZA (no.delete@this.ema.il) on April 4, 2014 3:18 pm wrote:
> > >
> > > Denormals is what made IEEE FP superior to VAX. It is the correct way of doing FP math!
> >
> > Don't get me wrong - I actually personally like denormals, and think they are
> > definitely required for serious FP math, but the thing is, 99% of FP use isn't
> > really serious, and most people don't really understand FP anyway.
> >
> > The kind of people who really understand FP and do serious math using it (to the point of
> > caring about the order of operations, never mind denormals), those kinds of people really
> > do know what they are doing, and sometimes (although not always) really do want denormals.
> >
> > But there's the other side of FP math, which really just wants good enough values quickly. There are a lot
> > of people who are ok with single-precision and no denormals. And yes, quite often 24 bits of precision is
> > not enough, and they decide they actually need double precision in order to avoid odd visual artifacts.
> >
> > Yeah, I'm talking about things like games.
> >
> > And the thing is, the defaults tend to be the wrong way around. People who don't know what they
> > are doing with floating point basically *never* need denormals. You will generally hit other issues
> > long before you hit the "oops, I lost precision because I didn't have denormals". But exactly
> > *because* they don't know about floating point, they also don't know to disable them.
> >
> > So I think that from a hardware designer standpoint, it actually
> > would make more sense if the default was "denormals
> > are zero" (or "flush to zero"), because the people who do want denormals also know about them.
> > So while
> > I like denormals, the onus on disabling denormals when not needed is kind of the wrong way around.
>
> This is back to front reasoning: physicists, mathematicians and most computational scientists don't
> know anything about the details of floating point and use the defaults like everyone else.
> Even simple assumptions like x==y iff x-y==0 don't always hold in FTZ mode; default denormal mode
> is the simplest floating point model mathematically. By contrast, people running gaming benchmarks
> who know about these things can easily enable FTZ to speed them up for marketing purposes.
You can't rely on x==y iff x-y==0 even with denormals. IEEE floating point is extremely complex with lots of odd non-intuitive corner cases. Very few mathematical identities work on IEEE, with or without denormals. Denormals are a major pain in algorithms, for example you need special cases to avoid the catastrophic loss of precision (and the infinities/NaNs caused by calculations with a denormal) as well as the huge slowdown on many CPUs.
> http://www.cs.berkeley.edu/~wkahan/19July10.pdf
> has some interesting history and rationale for denormals.
> (There are several readable summaries of the rationale for other IEEE754
> features on Kahan's web site that are well worth reading).
>
> >
> > But it's too late, and big CPU's largely do already handle them right
> > and don't care, so it's mostly a problem for the embedded space.
>
> Using the denormal benchmark at http://charm.cs.uiuc.edu/subnormal/ , I see an (acceptable) 8x slowdown
> of denormals on my 2.3 GHz Intel Core i7 macbook pro in SSE mode and an (unacceptable) 53x slowdown
> in x87 mode (which is particularly egregious as x87 mode is needed for precise numerical work).
> FMAC instructions make fast denormal processing easy to implement,
> which is probably why Jaguar's denormal handling is slow.
8x slowdown is still unacceptable in my book - 10% is acceptable. POWER and ARM are at that level.
> It is a shame that as the importance of computational methods to society and the need for accurate
> and reliable floating point has increased in *absolute* terms over the last decades, the *relative*
> decreased usage compared with multimedia applications has lead to floating point hardware capabilities
> being degraded to meet gamers' needs. Processor speed, memory size, even screen resolution, have
> all increased monotonically over the last decades- floating point precision, range and reliability
> is the only feature that has actually decreased... very unfortunate.
The fact is the contrary is actually true. The non-IEEE compliant 80-bit x87 is finally dead - a huge step forward for floating point. Programming languages and compilers adopted IEEE and by default provide IEEE compliant optimizations as well user selectable adventurous FP optimizations. Libraries have improved hugely as well, providing far more accurate math functions (the norm is now < 1ULP error). 128-bit IEEE is becoming available in hardware. So we've made huge progress in the last decades.
Wilco