Ideas on reorganizing terms in perfbook's Index pages
Hi Paul,
With regard to "memory consistency" and "memory model", currently, Index of Terms pages have the following:
- Consistency
- memory
- process
- sequential
- weak
- Memory consistency -> Consistency, memory
, and
- Linux kernel memory consistency model (LKMM)
, which comes from markings for acronym of "LKMM".
The redirection to "Consistency, memory" looks strange to me, as
- Process consistency
- Sequential consistency
- Weak consistency
are about "memory consistency".
So, I'm thinking of reorganizing them into something like, including additional markers for "memory model":
- Memory consistency
- process
- sequential
- weak
- Memory model
- Linux kernel
- Armv7
- Armv8
- etc.
- Memory consistency model -> Memory model
- Linux kernel memory consistency model (LKMM) -> Memory model, Linux kernel
How does the above sound to you?
Thanks, Akira
Quoted 41 lines · Akira Yokosawa
Hi Paul,
With regard to "memory consistency" and "memory model", currently, Index of Terms pages have the following:
- Consistency
- memory
- process
- sequential
- weak- Memory consistency -> Consistency, memory
, and
- Linux kernel memory consistency model (LKMM)
, which comes from markings for acronym of "LKMM".
The redirection to "Consistency, memory" looks strange to me, as
- Process consistency
- Sequential consistency
- Weak consistencyare about "memory consistency".
So, I'm thinking of reorganizing them into something like, including additional markers for "memory model":
- Memory consistency
- process
- sequential
- weak- Memory model
- Linux kernel
- Armv7
- Armv8
- etc.
Quoted 5 lines
- Memory consistency model -> Memory model
- Linux kernel memory consistency model (LKMM) -> Memory model, Linux kernel
How does the above sound to you?
That works for me, and thank you for going through this!
Thanx, Paul
Quoted 1 line · Paul E. McKenney
On Thu, Apr 17, 2025 at 11:02:02AM +0900, Akira Yokosawa wrote:
[...]
Quoted 14 lines
Quoted 14 lines
So, I'm thinking of reorganizing them into something like, including additional markers for "memory model":
- Memory consistency
- process
- sequential
- weak- Memory model
- Linux kernel
- Armv7
- Armv8
- etc.
Quoted 7 lines
Quoted 5 lines
- Memory consistency model -> Memory model
- Linux kernel memory consistency model (LKMM) -> Memory model, Linux kernel
How does the above sound to you?
That works for me, and thank you for going through this!
Hi Paul,
This patch set is my attempt taking care of above.
Thanks, Akira
--
Akira Yokosawa (5):
Refactor index markers of "memory consistency"-related terms
index: List LKMM under "Memory model"
indexsee: Add a couple of redirections to "Cache coherence" together/seqlock: Remove wrong index marker for "sequential consistency"
indexsee: Add redirections for arch specific memory models.
formal/axiomatic.tex | 7 ++++--- formal/ppcmem.tex | 4 ++-- formal/spinhint.tex | 4 +++- future/formalregress.tex | 11 ++++++----- glossary.tex | 6 +++--- indexsee.tex | 27 +++++++++++++++++++++------ memorder/memorder.tex | 39 ++++++++++++++++++++++++--------------- together/seqlock.tex | 2 +- 8 files changed, 64 insertions(+), 36 deletions(-)
base-commit: d4014ed9b2f2ec4074f86b2bdf0cf79953408cb0
Signature · 1 line
2.43.0
PATCH 1/5 Refactor index markers of "memory consistency"-related terms
Commit f24ccfbd345e ("index: Reform indexing tags related to 'memory consistency'") preferred "Sequential consistency", "Process consistency", and "Weak consistency" to "Sequential memory consistency", "Process memory consistency", and "Weak memory consistency" in single level Index pages. As a result, two-level Index pages have the following:
- Consistency
- memory
- sequential
- process
- weak
However, the latter three terms talk about varieties of "memory consistency" models. Having "memory" and the other three at the same level is confusing.
So revert the change in the commit and add a couple of redirections to "Memory consistency" in both single- and two-level Index pages.
Also add index markers for "Memory model".
--- formal/axiomatic.tex | 2 +- formal/ppcmem.tex | 4 ++-- formal/spinhint.tex | 4 +++- future/formalregress.tex | 11 ++++++----- glossary.tex | 6 +++--- indexsee.tex | 16 ++++++++++------ memorder/memorder.tex | 27 ++++++++++++++++----------- together/seqlock.tex | 4 +++- 8 files changed, 44 insertions(+), 30 deletions(-)
formal/axiomatic.tex +1 -1
@@ -67,7 +67,7 @@ suggests that an approach that treated similar traces as one might improve performace. One such approach is the axiomatic approach of \pplsur{Jade}{Alglave} et al.~\cite{Alglave:2014:HCM:2594291.2594347}, -which creates a set of axioms to represent the memory model and then +which creates a set of axioms to represent the \IX{memory model} and then converts litmus tests to theorems that might be proven or disproven over this set of axioms. The resulting tool, called \qco{herd}, conveniently takes as input the
formal/ppcmem.tex +2 -2
@@ -9,7 +9,7 @@ Although Promela and Spin allow you to verify pretty much any (smallish) algorithm, their very generality can sometimes be a curse. -For example, Promela does not understand memory models or any sort +For example, Promela does not understand \IX{memory model}s or any sort of reordering semantics. This section therefore describes some state-space search tools that understand memory models used by production systems, greatly simplifying the
@@ -256,7 +256,7 @@ The next section describes how to run this litmus test. As noted earlier, litmus tests may be run interactively via \url{https://www.cl.cam.ac.uk/~pes20/ppcmem/}, which can help build an -understanding of the memory model. +understanding of the \IX{memory model}. However, this approach requires that the user manually carry out the full state-space search. Because it is very difficult to be sure that you have checked every
formal/spinhint.tex +3 -1
@@ -521,7 +521,9 @@ full ordering. In any given Promela state, all processes agree on both the current state and the order of state changes that caused us to arrive at the current state. -This is analogous to the ``sequentially consistent'' memory model +This is analogous to the +``\IXalth{sequentially consistent}{sequential}{memory consistency}'' +\IX{memory model} used by a few computer systems (such as 1990s MIPS and PA-RISC\@). As noted earlier, and as will be seen in a later example, weak memory ordering must be explicitly coded.
future/formalregress.tex +6 -5
@@ -154,9 +154,9 @@ testing, which is in fact the topic of this section. It is critically important that formal-verification tools correctly model their environment. -One all-too-common omission is the memory model, where a great +One all-too-common omission is the \IX{memory model}, where a great many formal-verification tools, including Promela/\co{spin}, are -restricted to \IXh{sequential}{consistency}. +restricted to \IXalth{sequential consistency}{sequential}{memory consistency}. The QRCU experience related in \cref{sec:formal:Is QRCU Really Correct?} is an important cautionary tale.
@@ -624,8 +624,9 @@ shows a rough-and-ready scorecard for the formal-verification tools covered in this chapter. Shorter wavelengths are better than longer wavelengths. -Promela requires hand translation and supports only sequential -consistency, so its first two cells are red. +Promela requires hand translation and supports only +\IXalth{sequential consistency}{sequential}{memory consistency}, +so its first two cells are red. It has reasonable overhead (for formal verification, anyway) and provides a traceback, so its next two cells are yellow. Despite requiring hand translation, Promela handles assertions
@@ -633,7 +634,7 @@ in a natural way, so its fifth cell is green. PPCMEM usually requires hand translation due to the small size of litmus tests that it supports, so its first cell is orange. -It handles several memory models, so its second cell is green. +It handles several \IX{memory model}s, so its second cell is green. Its overhead is quite high, so its third cell is red. It provides a graphical display of relations among operations, which is not as helpful as a traceback, but is still quite useful, so its
glossary.tex +3 -3
@@ -445,7 +445,7 @@ instructions that executed before the memory barrier to precede those that will execute following that memory barrier. (See also ``read memory barrier'' and ``write memory barrier''.) -\item[\IXGh{Memory}{Consistency}:] +\item[\IXG{Memory Consistency}:] A set of properties that impose constraints on the order in which accesses to groups of variables appear to occur. Memory consistency models range from sequential consistency,
@@ -576,7 +576,7 @@ A source-code memory access that simply mentions the name of the object being accessed. (See ``Marked access''.) -\item[\IXGh{Process}{Consistency}:] +\item[\IXGalth{Process Consistency}{process}{memory consistency}:] A memory-consistency model in which each CPU's stores appear to occur in program order, but in which different CPUs might see accesses from more than one CPU as occurring in different orders.
@@ -711,7 +711,7 @@ \item[\IXGh{Sequence}{Lock}:] A reader-writer synchronization mechanism in which readers retry their operations if a writer was present. -\item[\IXGh{Sequential}{Consistency}:] +\item[\IXGalth{Sequential Consistency}{sequential}{memory consistency}:] A memory-consistency model where all memory references appear to occur in an order consistent with a single global order, and where each CPU's memory references
indexsee.tex +10 -6
@@ -1,5 +1,6 @@ -% effecive in both single- and two- level index +% effective in both single- and two- level index \index{Full(y) multi-copy atomic|see{Multi-copy atomic}} +\index{Memory consistency model|see{Memory model}} \IfIndexHier{ % effective only in two-level index \index{Address dependency|see{Dependency, address}} \index{Anti-heisenbug|see{Heisenbug, anti-}}
@@ -20,23 +21,26 @@ \index{Full memory barrier|see{Memory barrier, full}} \index{Fully associative cache|see{Cache, fully associative}} \index{Grace-period latency|see{Latency, grace-period}} -\index{Memory consistency|see{Consistency, memory}} \index{Memory latency|see{Latency, memory}} \index{Memory-barrier latency|see{Latency, memory-barrier}} \index{Memory-barrier overhead|see{Overhead, memory-barrier}} \index{Message latency|see{Latency, message}} \index{Non-multi-copy atomic|see{Multi-copy atomic, non-}} \index{Other-multi-copy atomic|see{Multi-copy atomic, other-}} -\index{Process consistency|see{Consistency, process}} +\index{Process consistency|see{Memory consistency, process}} \index{RCU read-side critical section|see{Critical section, RCU read-side}} \index{Read memory barrier|see{Memory barrier, read}} \index{Read-side critical section|see{Critical section, read-side}} \index{Reader-writer lock|see{Lock, reader-writer}} \index{Scheduling latency|see{Latency, scheduling}} \index{Sequence lock|see{Lock, sequence}} -\index{Sequential consistency|see{Consistency, sequential}} -\index{Weak consistency|see{Consistency, weak}} +\index{Sequential consistency|see{Memory consistency, sequential}} +\index{Weak consistency|see{Memory consistency, weak}} \index{Write memory barrier|see{Memory barrier, write}} \index{Write miss|see{Cache miss, write}} \index{Write-side critical section|see{Critical section, write-side}} -}{} +}{ % effective for single-level index +\index{Process consistency|see{Process memory consistency}} +\index{Sequential consistency|see{Sequential memory consistency}} +\index{Weak consistency|see{Weak memory consistency}} +}
memorder/memorder.tex +16 -11
@@ -425,7 +425,7 @@ set of rules of thumb. \label{sec:memorder:Rules of Thumb} The transitive intuitions presented in the previous section are -very appealing, at least as memory models go. +very appealing, at least as \IX{memory model}s go. Unfortunately, hardware is under no obligation to provide temporal cause-and-effect illusions when one thread's store overwrites a value either loaded or stored by some other thread.
@@ -509,7 +509,8 @@ This rule underlies the release-acquire intuitions presented in \cref{sec:memorder:Release-Acquire Intuitions}. You can replace a given acquire with a dependency in environments permitting -this, keeping in mind that the C11 standard's memory model does \emph{not} +this, keeping in mind that the +\IXalthmr{C11 standard's memory model}{C11}{memory model} does \emph{not} fully respect dependencies. Therefore, a dependency leading to a load must be headed by a \co{READ_ONCE()} or an \co{rcu_dereference()}:
@@ -599,7 +600,8 @@ will be discussed in \cref{sec:memorder:Hardware Specifics}, but these guarantees may only be relied upon in code that runs only for that architecture. -In addition, more accurate memory models~\cite{Alglave:2018:FSC:3173162.3177156} +In addition, more accurate +\IX{memory model}s~\cite{Alglave:2018:FSC:3173162.3177156} may give stronger guarantees with lower-overhead operations than do these rules of thumb, albeit at the expense of greater complexity. In these more formal memory-ordering papers, a store-to-load link is an
@@ -5467,14 +5469,15 @@ semaphores, RCU, \ldots) include any needed ordering primitives. So if you are working with code that uses these primitives properly, you need not worry about Linux's memory-ordering primitives. -That said, deep knowledge of each CPU's memory-consistency model +That said, deep knowledge of each CPU's +\IXalt{memory-consistency model}{memory model} can be very helpful when debugging, to say nothing of when writing architecture-specific code or synchronization primitives. Besides, they say that a little knowledge is a very dangerous thing. Just imagine the damage you could do with a lot of knowledge! For those who wish to understand more about individual CPUs' -\IXh{memory}{consistency} models, the next sections describe those of a few +memory consistency models, the next sections describe those of a few popular and prominent CPUs. Although there is no substitute for actually reading a given CPU's documentation, these sections do give a good overview.
@@ -5707,7 +5710,7 @@ For more on Alpha, see its reference manual~\cite{ALPHA2002}. The \ARM\ family of CPUs is popular in deep embedded applications, particularly for power-constrained microcontrollers. -Its memory model is similar to that of \Power{} +Its \IXalthmr{memory model}{Armv7}{memory model} is similar to that of \Power{} (see \cref{sec:memorder:POWER / PowerPC}), but \ARM\ uses a different set of memory-barrier instructions~\cite{ARMv7A:2010}:
@@ -5795,7 +5798,8 @@ Note that inserting an additional \co{ISB} instruction somewhere between includes 64-bit capabilities, in contrast to their 32-bit-only CPU described in \cref{sec:memorder:ARMv7-A/R}. -\ARMv8's memory model closely resembles its \ARMv7 counterpart, +\IXalthmr{\ARMv8's memory model}{Armv8}{memory model} +closely resembles its \ARMv7 counterpart, but adds load-acquire (\co{LDLARB}, \co{LDLARH}, and \co{LDLAR}) and store-release (\co{STLLRB}, \co{STLLRH}, and \co{STLLR}) instructions.
@@ -5832,8 +5836,8 @@ defined its memory ordering with an executable formal model~\cite{ARMv8A:2017}. \subsection{Itanium} \label{sec:memorder:Itanium} -Itanium offers a \IXh{weak}{consistency} -model, so that in absence of explicit +Itanium offers a \IXalth{weak consistency}{weak}{memory consistency} +\IXalthmr{model}{Itanium}{memory model}, so that in absence of explicit memory-barrier instructions or dependencies, Itanium is within its rights to arbitrarily reorder memory references~\cite{IntelItanium02v2}. Itanium has a memory-fence instruction named \co{mf}, but also has
@@ -5997,7 +6001,7 @@ CPU, including those to the same variable. \subsection{MIPS} -The MIPS memory model~\cite[page~479]{MIPSvII-A-2016} +The \IXhmr{MIPS}{memory model}~\cite[page~479]{MIPSvII-A-2016} appears to resemble that of \ARM, Itanium, and \Power{}, being weakly ordered by default, but respecting dependencies. MIPS has a wide variety of memory-barrier instructions, but ties them
@@ -6227,7 +6231,8 @@ the \co{smp_mb()} and \co{smp_rmb()} primitives expanded to \co{lock;addl}. This atomic instruction acts as a barrier to both loads and stores. But those were ancient times. -More recently, Intel has published a memory model for +More recently, Intel has published a +\IXalthmr{memory model}{x86}{memory model} for x86~\cite{Intelx86MemoryOrdering2007}. It turns out that Intel's modern CPUs enforce tighter ordering than was claimed in the previous specifications, so this model simply mandates
together/seqlock.tex +3 -1
@@ -132,7 +132,9 @@ notably in the dcache subsystem~\cite{NeilBrown2015RCUwalk}. Note that it is likely that similar schemes also work with hazard pointers. -This approach provides \IXh{sequential}{consistency} to successful readers, +This approach provides +\IXalth{sequential consistency}{sequential}{memory consistency} +to successful readers, each of which will either see the effects of a given update or not, with any partial updates resulting in a read-side retry. Sequential consistency is an extremely strong guarantee, incurring equally
Signature · 1 line
2.43.0
PATCH 2/5 index: List LKMM under "Memory model"
In reply to Akira YokosawaGiven that the Linux kernel memory model (LKMM) is a class of "memory model", it is better be listed under "Memory model" in Index.
As it is currently impossible to make acronym-related terms go to second level, remove existing index markers of acronym (meaning \IXacr --> \acr) and use additional index markers of \IXalthmr.
Given it goes under "Memory model", add a redirection of "Linux kernel consistency model (LKMM)" to "Memory model, Linux kernel" as well.
--- formal/axiomatic.tex | 5 +++-- indexsee.tex | 2 ++ memorder/memorder.tex | 12 ++++++++---- 3 files changed, 13 insertions(+), 6 deletions(-)
formal/axiomatic.tex +3 -2
@@ -142,8 +142,9 @@ Axiomatic approaches may also be applied to higher-level languages and also to higher-level synchronization primitives, as exemplified by the lock-based litmus test shown in \cref{lst:formal:Locking Example} (\path{C-Lock1.litmus}). -This litmus test can be modeled by -the \IXacrf{lkmm}~\cite{Alglave:2018:FSC:3173162.3177156,LucMaranget2018lock.cat}. +This litmus test can be modeled by the +\IXalthmr{\acrf{lkmm}}{Linux kernel}{memory model}~% +\cite{Alglave:2018:FSC:3173162.3177156,LucMaranget2018lock.cat}. As expected, the \co{herd} tool's output features the string \co{Never}, correctly indicating that \co{P1()} cannot see \co{x} having a value of one.\footnote{
indexsee.tex +2 -0
@@ -21,6 +21,7 @@ \index{Full memory barrier|see{Memory barrier, full}} \index{Fully associative cache|see{Cache, fully associative}} \index{Grace-period latency|see{Latency, grace-period}} +\index{Linux kernel memory consistency model (LKMM)|see{Memory model, Linux kernel}} \index{Memory latency|see{Latency, memory}} \index{Memory-barrier latency|see{Latency, memory-barrier}} \index{Memory-barrier overhead|see{Overhead, memory-barrier}}
@@ -40,6 +41,7 @@ \index{Write miss|see{Cache miss, write}} \index{Write-side critical section|see{Critical section, write-side}} }{ % effective for single-level index +\index{Linux kernel memory consistency model (LKMM)|see{Linux kernel memory model}} \index{Process consistency|see{Process memory consistency}} \index{Sequential consistency|see{Sequential memory consistency}} \index{Weak consistency|see{Weak memory consistency}}
memorder/memorder.tex +8 -4
@@ -69,7 +69,8 @@ follows up with more detail on a few representative hardware platforms. The \IXacrf{kcsan}~\cite{MarcoElver2020FearDataRaceDetector1,MarcoElver2020FearDataRaceDetector2}, based in part on RacerD~\cite{SamBlackshear2018RacerD} - and implementing \IXacr{lkmm}, has also been added to the Linux kernel + and implementing \IXalthmr{\acr{lkmm}}{Linux kernel}{memory model}, + has also been added to the Linux kernel and is now heavily used. Finally, there are now better ways of describing the Linux-kernel
@@ -86,7 +87,8 @@ follows up with more detail on a few representative hardware platforms. {F. W. Nichol} This section is for people who would like to avoid learning all the -details of the Linux-kernel memory model (\IXacr{lkmm}), the better +details of the \IXalthmr{Linux-kernel memory model (\acr{lkmm})} +{Linux kernel}{memory model}, the better to get on with their concurrency lives. The good news is that learning a very small fraction of LKMM can get you most of its benefits.
@@ -4495,7 +4497,8 @@ For their part, weakly ordered systems might well choose to execute the memory-barrier instructions required to guarantee both orderings, possibly simplifying code making advanced use of combinations of locked and lockless accesses. -However, as noted earlier, \IXacr{lkmm} chooses not to provide these additional +However, as noted earlier, \IXalthmr{\acr{lkmm}}{Linux kernel}{memory model} +chooses not to provide these additional orderings, in part to avoid imposing performance penalties on the simpler and more prevalent locking use cases. Instead, the \co{smp_mb__after_spinlock()} and \co{smp_mb__after_unlock_lock()}
@@ -5746,7 +5749,8 @@ None of these instructions exactly match the semantics of Linux's The \co{DMB} and \co{DSB} instructions have a recursive definition of accesses ordered before and after the barrier, which has an effect similar to that of \Power{}'s cumulativity, both of which are -stronger than \IXacr{lkmm}'s cumulativity described in +stronger than \IXalthmr{\acr{lkmm}}{Linux kernel}{memory model}'s +cumulativity described in \cref{sec:memorder:Cumulativity}. \ARM\ also implements \IXalth{control dependencies}{control}{dependency},
Signature · 1 line
2.43.0
PATCH 3/5 indexsee: Add a couple of redirections to "Cache coherence"
In reply to Akira YokosawaIn perfbook, "single-copy atomic" and "single-variable SC" are mentioned as "cache coherence". Plain "coherence" is used as its synonym as well.
--- indexsee.tex | 3 +++ 1 file changed, 3 insertions(+)
indexsee.tex +3 -0
@@ -1,6 +1,9 @@ % effective in both single- and two- level index +\index{Coherence|see{Cache coherence}} \index{Full(y) multi-copy atomic|see{Multi-copy atomic}} \index{Memory consistency model|see{Memory model}} +\index{Single-copy atomic|see{Cache coherence}} +\index{Single-variable SC|see{Cache coherence}} \IfIndexHier{ % effective only in two-level index \index{Address dependency|see{Dependency, address}} \index{Anti-heisenbug|see{Heisenbug, anti-}}
Signature · 1 line
2.43.0
PATCH 4/5 together/seqlock: Remove wrong index marker for "sequential consistency"
In reply to Akira YokosawaCommit 1efb66c2b391 ("index: Add index and acronym tags, take two") made this "sequential consistency" be listed under "Memory consistency". But it is nothing to do with "memory consistency". So it should not appear in the Index pages.
--- together/seqlock.tex | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
together/seqlock.tex +1 -3
@@ -132,9 +132,7 @@ notably in the dcache subsystem~\cite{NeilBrown2015RCUwalk}. Note that it is likely that similar schemes also work with hazard pointers. -This approach provides -\IXalth{sequential consistency}{sequential}{memory consistency} -to successful readers, +This approach provides sequential consistency to successful readers, each of which will either see the effects of a given update or not, with any partial updates resulting in a read-side retry. Sequential consistency is an extremely strong guarantee, incurring equally
Signature · 1 line
2.43.0
PATCH 5/5 indexsee: Add redirections for arch specific memory models
In reply to Akira Yokosawa--- indexsee.tex | 6 ++++++ 1 file changed, 6 insertions(+)
indexsee.tex +6 -0
@@ -5,12 +5,15 @@ \index{Single-copy atomic|see{Cache coherence}} \index{Single-variable SC|see{Cache coherence}} \IfIndexHier{ % effective only in two-level index +\index{Armv7 memory model|see{Memory model, Armv7}} +\index{Armv8 memory model|see{Memory model, Armv8}} \index{Address dependency|see{Dependency, address}} \index{Anti-heisenbug|see{Heisenbug, anti-}} \index{Associativity|see{Cache associativity}} \index{Associativity miss|see{Cache miss, associativity}} \index{Bounded wait free|see{Wait free, bounded}} \index{Bounded population-oblivious wait free|see{Wait free, bounded population-oblivious}} +\index{C11 memory model|see{Memory model, C11}} \index{Cache-invalidation latency|see{Latency, cache-invalidation}} \index{Cache-miss latency|see{Latency, cache-miss}} \index{Capacity miss|see{Cache miss, capacity}}
@@ -24,11 +27,13 @@ \index{Full memory barrier|see{Memory barrier, full}} \index{Fully associative cache|see{Cache, fully associative}} \index{Grace-period latency|see{Latency, grace-period}} +\index{Itanium memory model|see{Memory model, Itanium}} \index{Linux kernel memory consistency model (LKMM)|see{Memory model, Linux kernel}} \index{Memory latency|see{Latency, memory}} \index{Memory-barrier latency|see{Latency, memory-barrier}} \index{Memory-barrier overhead|see{Overhead, memory-barrier}} \index{Message latency|see{Latency, message}} +\index{MIPS memory model|see{Memory model, MIPS}} \index{Non-multi-copy atomic|see{Multi-copy atomic, non-}} \index{Other-multi-copy atomic|see{Multi-copy atomic, other-}} \index{Process consistency|see{Memory consistency, process}}
@@ -43,6 +48,7 @@ \index{Write memory barrier|see{Memory barrier, write}} \index{Write miss|see{Cache miss, write}} \index{Write-side critical section|see{Critical section, write-side}} +\index{x86 memory model|see{Memory model, x86}} }{ % effective for single-level index \index{Linux kernel memory consistency model (LKMM)|see{Linux kernel memory model}} \index{Process consistency|see{Process memory consistency}}
Signature · 1 line
2.43.0
Quoted 21 lines · Akira Yokosawa
Quoted 1 line · Paul E. McKenney
On Thu, Apr 17, 2025 at 11:02:02AM +0900, Akira Yokosawa wrote:
[...]
Quoted 18 lines
Quoted 18 lines
So, I'm thinking of reorganizing them into something like, including additional markers for "memory model":
- Memory consistency
- process
- sequential
- weak- Memory model
- Linux kernel
- Armv7
- Armv8
- etc.- Memory consistency model -> Memory model
- Linux kernel memory consistency model (LKMM) -> Memory model, Linux kernel
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
How does the above sound to you?
That works for me, and thank you for going through this!
Hi Paul,
This patch set is my attempt taking care of above.
Looks good to me, queued and pushed, thank you!
Thanx, Paul
Quoted 27 lines
Thanks, Akira
--
Akira Yokosawa (5):
Refactor index markers of "memory consistency"-related terms
index: List LKMM under "Memory model"
indexsee: Add a couple of redirections to "Cache coherence" together/seqlock: Remove wrong index marker for "sequential consistency"
indexsee: Add redirections for arch specific memory models.formal/axiomatic.tex | 7 ++++--- formal/ppcmem.tex | 4 ++-- formal/spinhint.tex | 4 +++- future/formalregress.tex | 11 ++++++----- glossary.tex | 6 +++--- indexsee.tex | 27 +++++++++++++++++++++------ memorder/memorder.tex | 39 ++++++++++++++++++++++++--------------- together/seqlock.tex | 2 +- 8 files changed, 64 insertions(+), 36 deletions(-)base-commit:
d4014ed9b2f2ec4074f86b2bdf0cf79953408cb0Signature · 1 line
2.43.0