You can download some testing queries from here: [human] [mouse] [bacteria].

You can download the human build 36 database from here, and the mouse build 36 database from
here
.

Ackowledgement:
The human and mouse query sets are from NCBI ftp server. .
The bacteria query set is from http://www.ncbi.nlm.nih.gov/sra/SRX338063 .

1) We assume the following directory hierarchy under /home/blastn: 
.
├── bin
│   ├── blastn          # the orignal blastn
│   └── gblastn         # the G-BLASTN (you need to copy the compiled binary "blastn" into this direcotry, and rename to "gblastn" or others; 
├── blast
│   └── src
│       └── gpu
│           ├── c++
│           ├── install
│           ├── log
│           ├── ncbi-blast-2.2.28+-src
│           └── 2.28.zip
├── data                # NCBI nt database
│   └── nt2m
│       ├── nohup.out
│       ├── nt.2m.00.nhr
│       ├── ...
│       ├── nt.2m.counts
│       └── nt.2m.nal
├── database           # human and mouse database
│   └── blastdb
│       ├── human
│       │   ├── human.1-8.wm.body
│       │   ├── ...
│       │   ├── human.9-Y.wm.nsq
│       │   └── human.wm.nal
│       └── mouse
│           ├── mouse.1-10.wm.nhr
│           ├── ...
│           ├── mouse.11-Y.wm.nsq
│           ├── mouse.nal
│           └── mouse.wm.nal
├── output              # output directory
│   ├── query.101.fa.out
│   └── ...
├── query               # query directory
│   ├── queries
│   │   ├── human
│   │   │   ├── qlarge
│   │   │   │   ├── query.11.fa
│   │   │   │   └── ...
│   │   │   ├── qmedium
│   │   │   │   ├── query.100.fa
│   │   │   │   └── ...
│   │   │   └── qsmall
│   │   │       ├── query.101.fa
│   │   │       └── ...
│   │   └── mouse
│   │       ├── qlarge
│   │       │   ├── query.201.fa
│   │       │   └── ...
│   │       ├── qmedium
│   │       │   ├── query.101.fa
│   │       │   └── ...
│   │       └── qsmall
│   │           ├── query.100.fa
│   │           └── ...
│   └── SRR955707
│       ├── 1.fa
│       └── ...
├── script             # shell directory
│   └── new_query_human_mouse
│       ├── g.m.sh     
│       └── m.ls
└── wm_counts          # mask database for human and mouse
├── human.stat
└── mouse.stat

2) Prepare the database

The following steps show how to prepare the masked database for G-BLASTN, using human build 36 as an example:

>makeblastdb -in human.fa -dbtype nucl -title human_database -parse_seqids -out human.wm -logfile human.log

>windowmasker -in human.wm -infmt blastdb -mk_counts  -parse_seqids -out human.stat


Remember to put the human.stat into the wm_counts directory.

3) Prepare the query list file:

Create a text file (e.g., h.ls) which contains the path names of all your queries, e.g.:

##############################################
/home/blastn/query/queries/human/qmedium/query.10.fa
/home/blastn/query/queries/human/qmedium/query.12.fa
##############################################

4) Create a shell script (e.g., g.h.sh) to execute G-BLASTN 

##############################################
#!/bin/sh
echo "***************start*****************"
blastn="/home/blastn/bin/gblastn"
data_dir="/home/blastn/"
database="${data_dir}database/blastdb/human/human.wm"
maskdb="${data_dir}wm_counts/human.stat"
query_list="h.ls"
outpath="/home/blastn/output/log.txt"

$blastn -db $database -window_masker_db $maskdb -query_list $query_list -outfmt "7" -out $outpath -use_gpu true

echo "***************complete*****************"
##############################################

>chmod +x g.h.sh

5) Run G-BLASTN;

>./g.h.sh

The results can be found in the output directory.