@ -116,8 +116,7 @@ e_ident[EI_DATA] ELFDATA2LSB For all little-endian implementations.</progra
</row>
<row>
<entry>
<para>.plt
<footnote>
<para>.plt<footnote>
<para>The type of the OpenPOWER ABI .plt section is
SHT_NOBITS, not SHT_PROGBITS as on most other processors.</para>
</footnote></para>
@ -184,7 +183,7 @@ e_ident[EI_DATA] ELFDATA2LSB For all little-endian implementations.</progra
is linker generated. The linker must ensure that .got is aligned to an
8-byte boundary. In an executable or shared library, it may contain
part or all of the TOC. For more information, see
<xref revisionflag='deleted' linkend="dbdoclet.50655240___RefHeading___Toc377640591" /> and
<xref linkend="dbdoclet.50655240___RefHeading___Toc377640591" /> and
<xref linkend="dbdoclet.50655242_47739" />.</para>
</listitem>
<listitem>
@ -319,7 +318,7 @@ e_ident[EI_DATA] ELFDATA2LSB For all little-endian implementations.</progra
pointer can use a common entry point for the local and global
entry points.</para>
<note>
<para>Note: If the function is not a leaf function, it must
<para>If the function is not a leaf function, it must
call subroutines using the R_PPC64_REL24_NOTOC relocation
to indicate that the TOC register is not initialized. In
turn, this may lead to more expensive procedure linkage
@ -852,7 +851,7 @@ my_func:
</entry>
</row>
<row>
<entry align="left">
<entry align="right">
<para>0</para>
</entry>
<entry>
@ -2230,7 +2229,7 @@ my_func:
</entry>
<entry>
<para>Denotes the high adjusted value: bits 16 - 63 of the
indicated value, compensating for #lo() being treated as a
indicated value, compensating for #lo( ) being treated as a
signed number. That is:</para>
<para>#ha(x) = (x + 0x8000) >> 16</para>
</entry>
@ -2324,7 +2323,7 @@ my_func:
<para>If n is the offset computed:</para>
<para>GOT[n] = dtpmod</para>
<para>GOT[n + 1] = dtprel</para>
<para>The call to __tls_get_addr () happens as:</para>
<para>The call to __tls_get_addr ( ) happens as:</para>
<para>__tls_get_addr ((tls_index *) &GOT[n])</para>
</entry>
</row>
@ -2339,7 +2338,7 @@ my_func:
<para>If n is the offset computed:</para>
<para>GOT[n] = dtpmod</para>
<para>GOT[n + 1] = 0</para>
<para>The call to __tls_get_addr () happens as:</para>
<para>The call to __tls_get_addr ( ) happens as:</para>
<para>__tls_get_addr ((tls_index *) &GOT[n])</para>
</entry>
</row>
@ -2360,7 +2359,7 @@ my_func:
</tgroup>
</informaltable>
<note>
<para>Note: Relocations flagged with an asterisk(*) will
<para>Relocations flagged with an asterisk(*) will
trigger a relocation failure if the value computed does
not fit in the field specified.</para>
</note>
@ -2426,11 +2425,12 @@ my_func:
<tfoot>
<row>
<entry nameend="c4" namest="c1" align="left">
<para><emphasis role="bold">Note:</emphasis>Relocation values 8, 9, 12, 13, 18, 23, 32,
<note>
<para>Relocation values 8, 9, 12, 13, 18, 23, 32,
and 247 are not used. This is to maintain a
correspondence to the relocation values used by the
32-bit PowerPC ELF ABI.
</para>
</para></note>
</entry>
</row>
</tfoot>
@ -3204,7 +3204,7 @@ my_func:
<para>half16ds*</para>
</entry>
<entry>
<para></para>
<para>(R + A) >> 2</para>
</entry>
</row>
<row>
@ -4119,9 +4119,7 @@ my_func:
resolved through a call to the symbol’
Additionally, it instructs the link editor to build a procedure linkage
table for the executable or shared object if one is not created.</para>
<para>
<anchor xml:id="dbdoclet.50655241_R_PPC_COPY"
xreflabel="" /> R_PPC64_COPY</para>
<para><anchor xml:id="dbdoclet.50655241_R_PPC_COPY" xreflabel="" />R_PPC64_COPY</para>
<para>This relocation type is created by the link editor for dynamic
linking. Its offset member refers to a location in a writable segment.
The symbol table index specifies a symbol that should exist both in the
@ -4219,7 +4217,7 @@ ld r3,x@got@l(r3)</programlisting>
<para> </para>
<note>
<para>
<emphasis role="bold">Note:</emphasis> If X is a variable stored in the TOC,
If X is a variable stored in the TOC,
then X@got is the offset within the TOC of a doubleword whose
value is X@toc.</para>
</note>
@ -4274,7 +4272,7 @@ lwz rt, offset(r2)</programlisting>
<para>Compilers and programmers
<emphasis>must</emphasis> ensure that r2 is live at the actual data access
point associated with extended displacement addressing.</para>
</section>
<section xml:id="dbdoclet.50655241_19147">
<title>TOC Pointer Usage</title>
<para>To enable linker-based optimizations when global data is accessed,
@ -4320,6 +4318,8 @@ target:
rewrite address references created using GOT-indirect loads and bl+4
sequences to use TOC-relative address computation.</para>
</section>
</section>
<section>
<title>Fusion</title>
<para>Code generation in compilers, linkers, and by programmers should
@ -4426,14 +4426,14 @@ addi r4, r4, lower</programlisting>
tlsoffset(m + 1) = round(tlsoffset(m) + tlssize(m), align(m + 1))</programlisting>
<itemizedlist>
<listitem>
<para>The function round() returns its first argument rounded up to
<para>The function round( ) returns its first argument rounded up to
the next multiple of its second argument:</para>
</listitem>
</itemizedlist>
<programlisting>round(x, y) = y ×
<itemizedlist>
<listitem>
<para>The function ceiling() returns the smallest integer greater
<para>The function ceiling( ) returns the smallest integer greater
than or equal to its argument, where n is an integer satisfying: n -
1 < x ≤ n:</para>
</listitem>
@ -4441,9 +4441,9 @@ tlsoffset(m + 1) = round(tlsoffset(m) + tlssize(m), align(m + 1))</programlistin
<programlisting>ceiling(x) = n</programlisting>
<para>In the case of dynamic shared objects (DSO), TLS blocks are
allocated on an as-needed basis, with the details of allocation
abstracted away by the __tls_get_addr() function, which is used to
abstracted away by the __tls_get_addr( ) function, which is used to
retrieve the address of any TLS variable.</para>
<para>The prototype for the __tls_get_addr() function, is defined as
<para>The prototype for the __tls_get_addr( ) function, is defined as
follows.</para>
<programlisting>typedef struct
{
@ -4514,7 +4514,7 @@ extern void *__tls_get_addr (tls_index *ti);</programlisting>
code model, which is the default for the ELF V2 ABI.</para>
</note>
<para>Given the following code fragment, to determine the address of a
thread-local variable x, the __tls_get_addr() function is called with one
thread-local variable x, the __tls_get_addr( ) function is called with one
parameter. That parameter is a pointer to a data object of type
tls_index.</para>
<programlisting>extern __thread unsigned int x;
@ -5717,8 +5717,7 @@ static __thread unsigned int x3;
the following code, which makes no reference to GOT entries. The GOT
entries in
<xref linkend="dbdoclet.50655241_16273" /> can be removed from the GOT by
the linker when performing this code transformation.
<footnote xml:id="pgfId-1134055">
the linker when performing this code transformation.<footnote xml:id="pgfId-1134055">
<para>To further optimize the code in
<xref linkend="dbdoclet.50655241_16273" />, a linker may reschedule the
sequence to exploit fusion by generating a sequence that may be fused
@ -6251,7 +6250,7 @@ nop</programlisting>
<para>
<orderedlist>
<listitem xml:id="dbdoclet.50655241_21152">
<para>1. The linker may prefer to schedule the addis and
<para>The linker may prefer to schedule the addis and
addi to be adjacent to take advantage of fusion as a
microarchitecture optimization opportunity.</para>
</listitem>
@ -6703,7 +6702,7 @@ nop</programlisting>
<para>The result of performing a relocation for a TLS symbol is the
module ID and its offset within the TLS block. These are then stored in
the GOT. Later, they are obtained by the dynamic linker at run-time and
passed to __tls_get_addr(), which returns the address for the variable
passed to __tls_get_addr( ), which returns the address for the variable
for the current thread.</para>
<para>For more information, see
<xref linkend="dbdoclet.50655241_18894" />. For TLS relocations, see
Jeff Scheel
8 years ago