======================================
Unrolled Inline Assembler Micro Kernel                                    [TOC]
======================================

We unrole the assembler micro kernel manually.  The pattern for unroling the
loop is that

---- CODE(type=txt) ------------------------------------------------------------
for l = 0 to kc-1
    [code for l=1]
--------------------------------------------------------------------------------

get replaced by

---- CODE(type=txt) ------------------------------------------------------------
for i = 0 to kc/4-1
    [code for l=4i]
    [code for l=4i+1]
    [code for l=4i+2]
    [code for l=4i+3]

for i = kc/4 to kc
    [code for l=i]
--------------------------------------------------------------------------------

The `[code for ..]` gets basically just copy and pasted.


Select the demo-sse-asm-unrolled Branch
=======================================
Check out the `demo-sse-asm-unrolled` branch:

    *--[SHELL(path=ulmBLAS)]--------------------------------------------*
    |                                                                   |
    |  git branch -a                                                    |
    |  git checkout -B demo-sse-asm-unrolled                         +++|
    |                  remotes/origin/demo-sse-asm-unrolled             |
    |                                                                   |
    *-------------------------------------------------------------------*

Then we compile the project

    *--[SHELL(path=ulmBLAS,height=15)]----------------------------------*
    |                                                                   |
    |  make                                                             |
    |                                                                   |
    *-------------------------------------------------------------------*


The dgemm_nn Code
=================
:import: ulmBLAS/src/level3/dgemm_nn.c [linenumbers]


Benchmark Results
=================
We run the benchmarks

    *--[SHELL(path=ulmBLAS)]--------------------------------------------*
    |                                                                   |
    |  cd bench                                                         |
    |  ./xdl3blastst > report                                           |
    |  cat report                                                       |
    |                                                                   |
    *-------------------------------------------------------------------*

and filter out the results for the `demo-sse-asm-unrolled`
branch:

    *--[SHELL(path=ulmBLAS/bench)]--------------------------------------*
    |                                                                   |
    |  grep PASS report > demo-sse-asm-unrolled                         |
    |                                                                   |
    *-------------------------------------------------------------------*

With the gnuplot script

:import: ulmBLAS/bench/bench10.gps

we feed gnuplot

    *--[SHELL(path=ulmBLAS/bench)]--------------------------------------*
    |                                                                   |
    |  gnuplot bench10.gps                                              |
    |                                                                   |
    *-------------------------------------------------------------------*

and get

    ---- IMAGE --------------
    ulmBLAS/bench/bench10.svg
    -------------------------

:navigate: __up__    -> doc:index
           __back__  -> doc:page08/index
           __next__  -> doc:page10/index