systemsoftware
/
Programming-Guides
25
1
Fork 0
Code Issues 17 Pull Requests Projects Releases Wiki Activity
Browse Source

Add pcv descriptions for vec_reve & vec_extract_fp32_from_short[hl]

pull/69/head
Bill Schmidt 3 years ago
parent
7c3dacd77b
commit
b47b40b21a
1 changed files with 104 additions and 9 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 113
      Intrinsics_Reference/ch_vec_reference.xml

113
Intrinsics_Reference/ch_vec_reference.xml
Unescape View File

@ -13210,9 +13210,16 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref"> @@ -13210,9 +13210,16 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref">
single-precision IEEE numbers.</para>
<para><emphasis role="bold">Endian considerations:</emphasis>
The element numbering within a register is left-to-right for big-endian
targets, and right-to-left for little-endian targets. Thus the
permute control vector at address <emphasis role="bold">pcv</emphasis>
in the example implementation will differ for big- and little-endian.
targets, and right-to-left for little-endian targets.
</para>
<para><emphasis role="bold">Notes:</emphasis> The example
implementation assumes that the vperm instruction is used for
big-endian, and the vpermr instruction is used for
little-endian. The permute control vector for the vperm or
vpermr instruction is in a memory location identified by pcv.
The value located at pcv is identical in natural element order
for big- and little-endian: { 15, 14, 0, 0, 13, 12, 0, 0, 11,
10, 0, 0, 9, 8, 0, 0 }.
</para>
<indexterm>
@ -13223,6 +13230,10 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref"> @@ -13223,6 +13230,10 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref">
<primary>vperm</primary>
<secondary>vec_extract_fp32_from_shorth</secondary>
</indexterm>
<indexterm>
<primary>vpermr</primary>
<secondary>vec_extract_fp32_from_shorth</secondary>
</indexterm>
<indexterm>
<primary>xvcvhpsp</primary>
<secondary>vec_extract_fp32_from_shorth</secondary>
@ -13266,7 +13277,7 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref"> @@ -13266,7 +13277,7 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref">
<entry>
<programlisting>
lxv t,0(pcv)
vperm u,a,a,t
vperm[r] u,a,a,t
xvcvhpsp r,u
</programlisting>
</entry>
@ -13300,9 +13311,16 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref"> @@ -13300,9 +13311,16 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref">
single-precision IEEE numbers.</para>
<para><emphasis role="bold">Endian considerations:</emphasis>
The element numbering within a register is left-to-right for big-endian
targets, and right-to-left for little-endian targets. Thus the
permute control vector at address <emphasis role="bold">pcv</emphasis>
in the example implementation will differ for big- and little-endian.
targets, and right-to-left for little-endian targets.
</para>
<para><emphasis role="bold">Notes:</emphasis> The example
implementation assumes that the vperm instruction is used for
big-endian, and the vpermr instruction is used for
little-endian. The permute control vector for the vperm or
vpermr instruction is in a memory location identified by pcv.
The value located at pcv is identical in natural element order
for big- and little-endian: { 7, 6, 0, 0, 5, 4, 0, 0, 3, 2, 0,
0, 1, 0, 0, 0 }.
</para>
<indexterm>
@ -13313,6 +13331,10 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref"> @@ -13313,6 +13331,10 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref">
<primary>vperm</primary>
<secondary>vec_extract_fp32_from_shortl</secondary>
</indexterm>
<indexterm>
<primary>vpermr</primary>
<secondary>vec_extract_fp32_from_shortl</secondary>
</indexterm>
<indexterm>
<primary>xvcvhpsp</primary>
<secondary>vec_extract_fp32_from_shortl</secondary>
@ -13356,7 +13378,7 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref"> @@ -13356,7 +13378,7 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref">
<entry>
<programlisting>
lxv t,0(pcv)
vperm u,a,a,t
vperm[r] u,a,a,t
xvcvhpsp r,u
</programlisting>
</entry>
@ -25636,8 +25658,81 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref"> @@ -25636,8 +25658,81 @@ xmlns:xlink="http://www.w3.org/1999/xlink" xml:id="VIPR.vec-ref">
<para><emphasis role="bold">Notes:</emphasis> The example
implementations assume that the permute control vector for the
vperm or vpermr instruction is in a register identified by pcv.
The value of pcv differs based on the element size.
The value of pcv differs based on the element size, and is the
same (in natural element order) for big- and little-endian,
assuming the use of vperm for big-endian and vpermr for
little-endian.
</para>
<informaltable>
<tgroup cols="2">
<colspec colname="c0" colwidth="10*" />
<colspec colname="c1" colwidth="10*" />
<thead>
<row>
<entry align="center">
<emphasis role="bold">Vector types</emphasis>
</entry>
<entry align="center">
<emphasis role="bold">Permute control vector</emphasis>
</entry>
</row>
</thead>
<tbody>
<row>
<entry align="center">
<para>
vector char
</para>
</entry>
<entry align="center">
<para>
{ 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1,
0 }
</para>
</entry>
</row>
<row>
<entry align="center">
<para>
vector short
</para>
</entry>
<entry align="center">
<para>
{ 14, 15, 12, 13, 10, 11, 8, 9, 6, 7, 4, 5, 2, 3, 0,
1 }
</para>
</entry>
</row>
<row>
<entry align="center">
<para>
vector int, vector float
</para>
</entry>
<entry align="center">
<para>
{ 12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2,
3 }
</para>
</entry>
</row>
<row>
<entry align="center">
<para>
vector long long, vector double
</para>
</entry>
<entry align="center">
<para>
{ 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6,
7 }
</para>
</entry>
</row>
</tbody>
</tgroup>
</informaltable>
<indexterm>
<primary>vperm</primary>

Write Preview
Loading…
Cancel
Save