Skip to contentSkip to conversation

whymb: Prevent line breaks of memory-barrier instruction

PATCH Applied Akira Yokosawa 2 participants 6 messages Apr 23, 2017 – Apr 24, 2017 Kbuild

lore
Share & tools
lore
#0 lore raw

cover whymb: Prevent line breaks of memory-barrier instruction

From 343ea80f45c224bf2ecb4bcb84bdc924e8b9cb80 Mon Sep 17 00:00:00 2001
From: Akira Yokosawa <[email protected]>
Date: Mon, 24 Apr 2017 07:23:27 +0900
Subject: [PATCH 0/4] whymb: Prevent line breaks of memory-barrier instruction

Hi Paul,

I wanted to prevent line breaks of memory-barrier instructions of SPARC and zSeries.
However, simply using \nbco{} for them ended up in build error. It turned out \detokenize{} command used in \nbco{} can't hold "#" character.
Patch #1 modifies the definition of \nbco{}.
Also, those instructions are easier to recognize with quotation marks. Instead of writing quotation marks around each instruction, patch #1 also adds a macro named \qco{}.
Patches 2 and 3 use \qco{} for memory-barrier instructions of SPARC and zSeries, respectively.
Patch 4 is to prevent hyphenation of "GPGPU".

Note that \tco{} macro still can't hold "#" character.

Thanks, Akira
--
Akira Yokosawa (4):
perfbook.tex: Modify definition of \nbco{}
whymb: Prevent line breaks between 'membar' and '#Sync', etc.
whymb: Prevent line breaks between "bcr" and "15,0" Add 'GPGPU' to non-hyphenation list

 Makefile                             |  1 +
 appendix/whymb/whymemorybarriers.tex | 26 +++++++++++++-------------
 perfbook.tex                         |  3 ++-
 pfhyphex.tex                         |  2 ++
 4 files changed, 18 insertions(+), 14 deletions(-)
Signature · 1 line
2.7.4
#1 lore raw

PATCH 1/4 perfbook.tex: Modify definition of \nbco{}

From c4f162ce5436bf0105d81cd012f11c4a30691683 Mon Sep 17 00:00:00 2001
From: Akira Yokosawa <[email protected]>
Date: Sat, 22 Apr 2017 00:10:24 +0900
Subject: [PATCH 1/4] perfbook.tex: Modify definition of \nbco{}

It turns out that \nbco{} can't hold "#" character. Modify its definition to use the combination of a \co{} variant and \hbox{}.
Also add a macro named \qco{} which has quotation marks around \nbco{}.

---
 perfbook.tex | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
perfbook.tex +2 -1
index 58b0cd2..2de83f0 100644
@@ -111,7 +111,8 @@
 \makeatother
 % ---  End of workaround for Fedora 23 Texlive bug
 \newcommand{\co}[1]{\lstinline[breaklines=true,breakatwhitespace=true]{#1}}
-\newcommand{\nbco}[1]{\hbox{\texttt{\detokenize{#1}}}} % no break lines for short snippet
+\newcommand{\nbco}[1]{\hbox{\lstinline[breaklines=false,breakatwhitespace=false]{#1}}} % no break lines for short snippet
+\newcommand{\qco}[1]{``\nbco{#1}''} % \nbco with quotation marks
 \newcommand{\tco}[1]{\texttt{\detokenize{#1}}} % for code in tabular environment
 % \tco{} will break at spaces but not at underscores
 \newcommand{\nf}[1]{\textnormal{#1}} % to return to normal font
Signature · 1 line
2.7.4
#2 lore raw

PATCH 2/4 whymb: Prevent line breaks between 'membar' and '#Sync', etc.

In reply to Akira Yokosawa

From 536d7df2df39c0d43bd9e7367b3e4b33c46238a6 Mon Sep 17 00:00:00 2001
From: Akira Yokosawa <[email protected]>
Date: Sat, 22 Apr 2017 00:14:40 +0900
Subject: [PATCH 2/4] whymb: Prevent line breaks between 'membar' and '#Sync', etc.

These instruction variants are easier to recognize with quotation marks. Use \qco{} for them.

---
 appendix/whymb/whymemorybarriers.tex | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)
appendix/whymb/whymemorybarriers.tex +11 -11
index 6c6b28a..502e37f 100644
@@ -2244,28 +2244,28 @@ as in
 \co{membar #LoadLoad | #LoadStore | #StoreStore | #StoreLoad},
 thus fully ordering memory operations.
 
-So, why is \co{membar #MemIssue} needed?
-Because a \co{membar #StoreLoad} could permit a subsequent
+So, why is \qco{membar #MemIssue} needed?
+Because a \qco{membar #StoreLoad} could permit a subsequent
 load to get its value from a store buffer, which would be
 disastrous if the write was to an MMIO register that induced side effects
 on the value to be read.
-In contrast, \co{membar #MemIssue} would wait until the store buffers
+In contrast, \qco{membar #MemIssue} would wait until the store buffers
 were flushed before permitting the loads to execute,
 thereby ensuring that the load actually gets its value from the MMIO register.
-Drivers could instead use \co{membar #Sync}, but the lighter-weight
-\co{membar #MemIssue} is preferred in cases where the additional function
-of the more-expensive \co{membar #Sync} are not required.
+Drivers could instead use \qco{membar #Sync}, but the lighter-weight
+\qco{membar #MemIssue} is preferred in cases where the additional function
+of the more-expensive \qco{membar #Sync} are not required.
 
-The \co{membar #Lookaside} is a lighter-weight version of
-\co{membar #MemIssue}, which is useful when writing to a given MMIO register
+The \qco{membar #Lookaside} is a lighter-weight version of
+\qco{membar #MemIssue}, which is useful when writing to a given MMIO register
 affects the value that will next be read from that register.
-However, the heavier-weight \co{membar #MemIssue} must be used when
+However, the heavier-weight \qco{membar #MemIssue} must be used when
 a write to a given MMIO register affects the value that will next be
 read from {\em some other} MMIO register.
 
 It is not clear why SPARC does not define \co{wmb()} to be
-\co{membar #MemIssue} and \co{smp_wmb()} to be
-\co{membar #StoreStore},
+\qco{membar #MemIssue} and \co{smp_wmb()} to be
+\qco{membar #StoreStore},
 as the current definitions seem vulnerable to bugs in some drivers.
 It is quite possible that all the SPARC CPUs that Linux runs on
 implement a more conservative memory-ordering model than the architecture
Signature · 1 line
2.7.4
#3 lore raw

PATCH 3/4 whymb: Prevent line breaks between "bcr" and "15,0"

In reply to Akira Yokosawa

From 9b6444fcf2c6a57c2edb88ed81f0b35231a03d10 Mon Sep 17 00:00:00 2001
From: Akira Yokosawa <[email protected]>
Date: Sat, 22 Apr 2017 00:19:54 +0900
Subject: [PATCH 3/4] whymb: Prevent line breaks between "bcr" and "15,0"

---
 appendix/whymb/whymemorybarriers.tex | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
appendix/whymb/whymemorybarriers.tex +2 -2
index 502e37f..96daa7b 100644
@@ -2345,7 +2345,7 @@ mainframe family, previously
 known as the 360, 370, and 390~\cite{IBMzSeries04a}.
 Parallelism came late to zSeries, but given that these mainframes first
 shipped in the mid 1960s, this is not saying much.
-The \co{bcr 15,0} instruction is used for the Linux \co{smp_mb()},
+The \qco{bcr 15,0} instruction is used for the Linux \co{smp_mb()},
 \co{smp_rmb()}, and \co{smp_wmb()} primitives.
 It also has comparatively strong memory-ordering semantics, as shown in
 Table~\ref{tab:app:whymb:Summary of Memory Ordering}, which should allow the
@@ -2363,7 +2363,7 @@ That said, many actual zSeries machines do in fact accommodate self-modifying
 code without serializing instructions.
 The zSeries instruction set provides a large set of serializing instructions,
 including compare-and-swap, some types of branches (for example, the
-aforementioned \co{bcr 15,0} instruction), and test-and-set,
+aforementioned \qco{bcr 15,0} instruction), and test-and-set,
 among others.
 
 \section{Are Memory Barriers Forever?}
Signature · 1 line
2.7.4
#4 lore raw

PATCH 4/4 Add 'GPGPU' to non-hyphenation list

In reply to Akira Yokosawa

From 343ea80f45c224bf2ecb4bcb84bdc924e8b9cb80 Mon Sep 17 00:00:00 2001
From: Akira Yokosawa <[email protected]>
Date: Sat, 22 Apr 2017 00:20:41 +0900
Subject: [PATCH 4/4] Add 'GPGPU' to non-hyphenation list

Also add hyphenation list files to LATEXSOURCES in Makefile

---
 Makefile     | 1 +
 pfhyphex.tex | 2 ++
 2 files changed, 3 insertions(+)
Makefile +1 -0
index 88a23e2..8799313 100644
@@ -4,6 +4,7 @@ LATEXSOURCES = \
 	glossary.tex \
 	qqz.sty origpub.sty \
 	book.cls \
+	ushyphex.tex pfhyphex.tex \
 	*/*.tex \
 	*/*/*.tex
 
pfhyphex.tex +2 -0
index 768de86..fbcae67 100644
@@ -1,4 +1,6 @@
 \hyphenation{
   REQACK
   COPART
+  GPGPU
+  GPGPUs
 }
Signature · 1 line
2.7.4
#5 lore raw
In reply to Akira Yokosawa
Quoted 20 lines · Akira Yokosawa
Quoted 1 line

From 343ea80f45c224bf2ecb4bcb84bdc924e8b9cb80 Mon Sep 17 00:00:00 2001

From: Akira Yokosawa <[email protected]>
Date: Mon, 24 Apr 2017 07:23:27 +0900
Subject: [PATCH 0/4] whymb: Prevent line breaks of memory-barrier instruction

Hi Paul,

I wanted to prevent line breaks of memory-barrier instructions of SPARC and zSeries.
However, simply using \nbco{} for them ended up in build error. It turned out \detokenize{} command used in \nbco{} can't hold "#" character.
Patch #1 modifies the definition of \nbco{}.
Also, those instructions are easier to recognize with quotation marks. Instead of writing quotation marks around each instruction, patch #1 also adds a macro named \qco{}.
Patches 2 and 3 use \qco{} for memory-barrier instructions of SPARC and zSeries, respectively.
Patch 4 is to prevent hyphenation of "GPGPU".

Quoted 1 line

Note that \tco{} macro still can't hold "#" character.

Queued and pushed, thank you!

Thanx, Paul

Quoted 17 lines

Thanks, Akira
--
Akira Yokosawa (4):
perfbook.tex: Modify definition of \nbco{}
whymb: Prevent line breaks between 'membar' and '#Sync', etc.
whymb: Prevent line breaks between "bcr" and "15,0" Add 'GPGPU' to non-hyphenation list

 Makefile                             |  1 +
 appendix/whymb/whymemorybarriers.tex | 26 +++++++++++++-------------
 perfbook.tex                         |  3 ++-
 pfhyphex.tex                         |  2 ++
 4 files changed, 18 insertions(+), 14 deletions(-)
Signature · 1 line
2.7.4