Date:   Mon, 28 May 2018 13:20:10 +0200
From:   Andrea Parri <andrea.parri@...rulasolutions.com>
To:     "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
Cc:     linux-kernel@...r.kernel.org, linux-arch@...r.kernel.org,
        stern@...land.harvard.edu, will.deacon@....com,
        peterz@...radead.org, boqun.feng@...il.com, npiggin@...il.com,
        dhowells@...hat.com, j.alglave@....ac.uk, luc.maranget@...ia.fr,
        akiyks@...il.com, mingo@...nel.org
Subject: Re: [PATCH RFC tools/memory-model] Add litmus-test naming scheme

On Fri, May 25, 2018 at 12:10:20PM -0700, Paul E. McKenney wrote:
> This commit documents the scheme used to generate the names for the
> litmus tests.
> 
> Signed-off-by: Paul E. McKenney <paulmck@...ux.vnet.ibm.com>
> ---
>  README |  136 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 135 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/memory-model/litmus-tests/README b/tools/memory-model/litmus-tests/README
> index 00140aaf58b7..b81f51054cd3 100644
> --- a/tools/memory-model/litmus-tests/README
> +++ b/tools/memory-model/litmus-tests/README
> @@ -1,4 +1,6 @@
> -This directory contains the following litmus tests:
> +============
> +LITMUS TESTS
> +============
>  
>  CoRR+poonceonce+Once.litmus
>  	Test of read-read coherence, that is, whether or not two
> @@ -151,3 +153,135 @@ Z6.0+pooncerelease+poacquirerelease+mbonceonce.litmus
>  A great many more litmus tests are available here:
>  
>  	https://github.com/paulmckrcu/litmus
> +
> +==================
> +LITMUS TEST NAMING
> +==================
> +
> +Litmus tests are usually named based on their contents, which means that
> +looking at the name tells you what the litmus test does.  The naming
> +scheme covers litmus tests having a single cycle that passes through
> +each process exactly once, so litmus tests not fitting this description
> +are named on an ad-hoc basis.
> +
> +The structure of a litmus-test name is the litmus-test class, a plus
> +sign ("+"), and one string for each process, separated by plus signs.
> +The end of the name is ".litmus".

We used to distinguigh between the test name and the test filename; we
currently have only one test whose name ends with .litmus:

  ISA2+pooncelock+pooncelock+pombonce.litmus

(that I missed until now...).


> +
> +The litmus-test classes may be found in the infamous test6.pdf:
> +https://www.cl.cam.ac.uk/~pes20/ppc-supplemental/test6.pdf
> +Each class defines the pattern of accesses and of the variables accessed.
> +For example, if the one process writes to a pair of variables, and
> +the other process reads from these same variables, the corresponding
> +litmus-test class is "MP" (message passing), which may be found on the
> +left-hand end of the second row of tests on page one of test6.pdf.
> +
> +The strings used to identify the actions carried out by each process are
> +complex due to a desire to have finite-length names.

I'm not sure what you mean here: can you elaborate/rephrase?


> Thus, there is a
> +tool to generate these strings from a given litmus test's actions.  For
> +example, consider the processes from SB+rfionceonce-poonceonces.litmus:
> +
> +	P0(int *x, int *y)
> +	{
> +		int r1;
> +		int r2;
> +
> +		WRITE_ONCE(*x, 1);
> +		r1 = READ_ONCE(*x);
> +		r2 = READ_ONCE(*y);
> +	}
> +
> +	P1(int *x, int *y)
> +	{
> +		int r3;
> +		int r4;
> +
> +		WRITE_ONCE(*y, 1);
> +		r3 = READ_ONCE(*y);
> +		r4 = READ_ONCE(*x);
> +	}
> +
> +The next step is to construct a space-separated list of descriptors,
> +interleaving descriptions of the relation between a pair of consecutive
> +accesses with descriptions of the second access in the pair.
> +
> +P0()'s WRITE_ONCE() is read by its first READ_ONCE(), which is a
> +reads-from link (rf) and internal to the P0() process.  This is
> +"rfi", which is an abbreviation for "reads-from internal".  Because
> +some of the tools string these abbreviations together with space
> +characters separating processes, the first character is capitalized,
> +resulting in "Rfi".
> +
> +P0()'s second access is a READ_ONCE(), as opposed to (for example)
> +smp_load_acquire(), so next is "Once".  Thus far, we have "Rfi Once".
> +
> +P0()'s third access is also a READ_ONCE(), but to y rather than x.
> +This is related to P0()'s second access by program order ("po"),
> +to a different variable ("d"), and both accesses are reads ("RR").
> +The resulting descriptor is "PodRR".  Because P0()'s third access is
> +READ_ONCE(), we add another "Once" descriptor.
> +
> +A from-read ("fre") relation links P0()'s third to P1()'s first
> +access, and the resulting descriptor is "Fre".  P1()'s first access is
> +WRITE_ONCE(), which as before gives the descriptor "Once".  The string
> +thus far is thus "Rfi Once PodRR Once Fre Once".
> +
> +The remainder of P1() is similar to P0(), which means we add
> +"Rfi Once PodRR Once".  Another fre links P1()'s last access to
> +P0()'s first access, which is WRITE_ONCE(), so we add "Fre Once".
> +The full string is thus:
> +
> +	Rfi Once PodRR Once Fre Once Rfi Once PodRR Once Fre Once
> +
> +This string can be given to the "norm7" and "classify7" tools to
> +produce the name:
> +
> +$ norm7 -bell linux-kernel.bell Rfi Once PodRR Once Fre Once Rfi Once PodRR Once Fre Once |  classify7 -bell linux-kernel.bell -diyone | sed -e 's/:.*//g'
> +SB+rfionceonce-poonceonces

We should check for the required version of herdtools7; a quick search
here pointed out:

  ad5681da10fafb ("gen: add the tool 'norm' that normalise and name one cycle given as command line arguments.")

(so after v7.49), but I do remember a 'norm7' tool during my 'Parisian
days', mmh...

I also notice that, with the current version, the above command can be
simplified to:

  $ norm7 -bell linux-kernel.bell Rfi Once PodRR Once Fre Once Rfi Once PodRR Once Fre Once | sed -e 's/:.*//g'

but we might want to list other commands for backward compatibility.


> +
> +Adding the ".litmus" suffix: SB+rfionceonce-poonceonces.litmus
> +
> +
> +=======================
> +LITMUS TEST DESCRIPTORS
> +=======================
> +
> +These descriptors cover connections between consecutive accesses:

Maybe expand to recall that we're referring to a particular cycle (the
cycle referred to in the previous section)? (the term 'consecutive' is
overloaded ;-)


> +
> +Fre: From-read external.  The current process wrote a variable that
> +	the previous process read.  Example: The SB (store buffering) test.
> +Fri: From-read internal.  This process read a variable and then
> +	immediately wrote to it.  Example: ???
> +PodRR: Program-order different variable, read followed by read.
> +	This process read a variable and again read a different variable.
> +	Example: The read-side process in the MP (message-passing) test.
> +PodRW: Program-order different variable, read followed by write.
> +	This process read a variable and then wrote a different variable.
> +	Example: The LB (load buffering) test.
> +PodWR: Program-order different variable, write followed by read.
> +	This process wrote a variable and then read a different variable.
> +	Example: The SB (store buffering) test.
> +PodWW: Program-order different variable, write followed by write.
> +	This process wrote a variable and again wrote a different variable.
> +	Example: The write-side process in the MP (message-passing) test.
> +PosRR: Program-order same variable, read followed by read.
> +	This process read a variable and again read that same variable.
> +	Example: ???
> +PosRW: Program-order same variable, read followed by write.
> +	This process read a variable and then wrote that same variable.
> +	Example: ???
> +PosWR: Program-order same variable, write followed by read.
> +	This process wrote a variable and then read that same variable.
> +	Example: ???
> +PosWW: Program-order same variable, write followed by write.
> +	This process wrote a variable and again rrote that same variable.

s/rrote/wrote


> +	Example: ???
> +Rfe: Read-from external.  The current process read a variable written
> +	by the previous process.  Example: The MP (message passing) test.
> +Rfi: Read-from internal.  The current process wrote a variable and then
> +	immediately read the value back from it.  Example:  ???
> +	Comparison to PosWR???

I'm not sure if it is worth commenting on this, but compare, e.g., the
'exists' clauses of the following two tests (thinking at 'coherence'):

$ diyone7 -arch LISA PosWR PodRR Fre PosWR PodRR Fre
LISA A
"PosWR PodRR Fre PosWR PodRR Fre"
Generator=diyone7 (version 7.49+02(dev))
Prefetch=0:x=F,0:y=T,1:y=F,1:x=T
Com=Fr Fr
Orig=PosWR PodRR Fre PosWR PodRR Fre
{
}
 P0       | P1       ;
 w[] x 1  | w[] y 1  ;
 r[] r0 x | r[] r0 y ;
 r[] r1 y | r[] r1 x ;
exists
(0:r1=0 /\ 1:r1=0)

$ diyone7 -arch LISA Rfi PodRR Fre Rfi PodRR Fre
LISA A
"Rfi PodRR Fre Rfi PodRR Fre"
Generator=diyone7 (version 7.49+02(dev))
Prefetch=0:x=F,0:y=T,1:y=F,1:x=T
Com=Fr Fr
Orig=Rfi PodRR Fre Rfi PodRR Fre
{
}
 P0       | P1       ;
 w[] x 1  | w[] y 1  ;
 r[] r0 x | r[] r0 y ;
 r[] r1 y | r[] r1 x ;
exists
(0:r0=1 /\ 0:r1=0 /\ 1:r0=1 /\ 1:r1=0)


> +Wse: Write same external.  The current process wrote to a variable that
> +	was also written to by the previous process.  Example:  ???
> +Wsi: Write same internal.  The current process wrote to a variable and
> +	then immediately wrote to it again.  Example:  ???

The list of descriptors is incomplete; the command:

  $ diyone7 -bell linux-kernel.bell -show edges

shows other descriptors (including fences and dependencies).  We might
want to list this command; searching the commit history, I found:

   3c24730ef6c662 ("gen: Add a new command line option -show (edges|fences|annotations) that list various categories of candidate relaxations.")

I also notice that our current names for tests with fences (and cycle)
deviate from the corresponding 'norm7' results; e.g.,

  $ norm7 -bell linux-kernel.bell FenceWmbdWW Once Rfe Once FenceRmbdRR Once Fre Once | sed -e 's/:.*//g'
  MP+fencewmbonceonce+fencermbonceonce

while we use 'MP+wmbonceonce+rmbonceonce' (that is, we omit the 'fence'
prefixes).

  Andrea


> 

Powered by blists - more mailing lists