Add another Rome model (#249)
* Add another Rome model * Add gfx908 4P3L models and support * Revert "Use cached value for detecting GDR support only once" This reverts commit0108a1219d. * Skip using ibverb for GPU direct RDMA detection * Fine tune one Rome model [ROCm/rccl commit:a51e4071e3]
This commit is contained in:
@@ -21,7 +21,7 @@
|
||||
|
||||
DIR="$(cd -P "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
||||
|
||||
for i in {0..25}
|
||||
for i in {0..29}
|
||||
do
|
||||
$DIR/../topo_expl/topo_expl -m $i > "topo_m$i.log"
|
||||
$DIR/../TopoVisual/topo_visual.sh -i "topo_m$i.log"
|
||||
|
||||
@@ -0,0 +1,87 @@
|
||||
<system version="2">
|
||||
<cpu numaid="0" affinity="00000000,00000000,00000000,ffffffff" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:41:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:43:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="0" sm="98" gcn="908" arch="38911" rank="0" gdr="0">
|
||||
<xgmi target="0000:23:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:26:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:03:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
</cpu>
|
||||
<cpu numaid="1" affinity="00000000,00000000,ffffffff,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:21:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:23:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="1" sm="98" gcn="908" arch="38911" rank="1" gdr="0">
|
||||
<xgmi target="0000:43:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:26:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:03:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:24:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:26:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="2" sm="98" gcn="908" arch="38911" rank="2" gdr="0">
|
||||
<xgmi target="0000:43:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:23:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:03:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:01:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:03:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="3" sm="98" gcn="908" arch="38911" rank="3" gdr="0">
|
||||
<xgmi target="0000:43:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:23:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:26:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
</cpu>
|
||||
<cpu numaid="2" affinity="00000000,ffffffff,00000000,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:c1:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:c3:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="4" sm="98" gcn="908" arch="38911" rank="4" gdr="0">
|
||||
<xgmi target="0000:c6:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:a3:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:83:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:c4:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:c6:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="5" sm="98" gcn="908" arch="38911" rank="5" gdr="0">
|
||||
<xgmi target="0000:c3:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:a3:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:83:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:e1:00.0" class="0x020700" link_speed="16 GT/s" link_width="16">
|
||||
<nic>
|
||||
<net name="mlx5_0" dev="0" speed="200000" port="1" guid="0x20cd600003da341c" maxconn="262144" gdr="1"/>
|
||||
</nic>
|
||||
</pci>
|
||||
</cpu>
|
||||
<cpu numaid="3" affinity="ffffffff,00000000,00000000,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:a1:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:a3:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="6" sm="98" gcn="908" arch="38911" rank="6" gdr="0">
|
||||
<xgmi target="0000:c3:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:c6:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:83:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:81:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:83:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="7" sm="98" gcn="908" arch="38911" rank="7" gdr="0">
|
||||
<xgmi target="0000:c3:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:c6:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:a3:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
</cpu>
|
||||
</system>
|
||||
@@ -0,0 +1,84 @@
|
||||
<system version="2">
|
||||
<cpu numaid="0" affinity="00000000,00000000,00000000,ffffffff" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:41:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:43:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="0" sm="96" gcn="906" arch="38911" rank="0" gdr="1">
|
||||
<xgmi target="0000:03:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:c6:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:61:00.0" class="0x020700" link_speed="16 GT/s" link_width="16">
|
||||
<nic>
|
||||
<net name="mlx5_0" dev="0" speed="200000" port="1" guid="0xa8134300039f59b8" maxconn="262144" gdr="1"/>
|
||||
</nic>
|
||||
</pci>
|
||||
</cpu>
|
||||
<cpu numaid="1" affinity="00000000,00000000,ffffffff,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:21:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:23:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="1" sm="96" gcn="906" arch="38911" rank="1" gdr="1">
|
||||
<xgmi target="0000:e3:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:83:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:24:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:26:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="2" sm="96" gcn="906" arch="38911" rank="2" gdr="1">
|
||||
<xgmi target="0000:03:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:c6:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:01:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:03:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="3" sm="96" gcn="906" arch="38911" rank="3" gdr="1">
|
||||
<xgmi target="0000:43:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:26:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
</cpu>
|
||||
<cpu numaid="2" affinity="00000000,ffffffff,00000000,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:e1:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:e3:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="4" sm="96" gcn="906" arch="38911" rank="4" gdr="1">
|
||||
<xgmi target="0000:23:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:c3:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:c1:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:c3:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="5" sm="96" gcn="906" arch="38911" rank="5" gdr="1">
|
||||
<xgmi target="0000:e3:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:83:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:c4:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:c6:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="6" sm="96" gcn="906" arch="38911" rank="6" gdr="1">
|
||||
<xgmi target="0000:43:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:26:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
</cpu>
|
||||
<cpu numaid="3" affinity="ffffffff,00000000,00000000,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
|
||||
<pci busid="0000:81:00.0" class="0x060400" link_speed="16 GT/s" link_width="16">
|
||||
<pci busid="0000:83:00.0" class="0x038000" link_speed="16 GT/s" link_width="16">
|
||||
<gpu dev="7" sm="96" gcn="906" arch="38911" rank="7" gdr="1">
|
||||
<xgmi target="0000:23:00.0" count="1" tclass="0x038000"/>
|
||||
<xgmi target="0000:c3:00.0" count="1" tclass="0x038000"/>
|
||||
</gpu>
|
||||
</pci>
|
||||
</pci>
|
||||
<pci busid="0000:a1:00.0" class="0x020700" link_speed="16 GT/s" link_width="16">
|
||||
<nic>
|
||||
<net name="mlx5_2" dev="1" speed="200000" port="1" guid="0x38815600039f59b8" maxconn="262144" gdr="1"/>
|
||||
</nic>
|
||||
</pci>
|
||||
</cpu>
|
||||
</system>
|
||||
@@ -90,11 +90,15 @@ NodeModelDesc model_descs[] = {
|
||||
{4, "topo_8p6l_4nic.xml", "4 nodes gfx908 8P6L 4 NICs"},
|
||||
{4, "topo_8p6l_5nic.xml", "4 nodes gfx908 8P6L 5 NICs"},
|
||||
{4, "topo_8p6l_6nic.xml", "4 nodes gfx908 8P6L 6 NICs"},
|
||||
{1, "topo_8p_rome_n2.xml", "singel node 8 VEGA20 Rome NPS=2"},
|
||||
{1, "topo_8p_rome_n2.xml", "single node 8 VEGA20 Rome NPS=2"},
|
||||
{4, "topo_8p_rome_n2.xml", "4 nodes 8 VEGA20 Rome NPS=2"},
|
||||
{4, "topo_8p_rome_n2_1.xml", "4 nodes 8 VEGA20 Rome NPS=2 Alt. Model"},
|
||||
{1, "topo_8p_ts1.xml", "Single node 6 VEGA20 TS1"},
|
||||
{1, "topo_8p_ts1.xml", "single node 8 VEGA20 TS1"},
|
||||
{4, "topo_8p_ts1.xml", "4 nodes 8 VEGA20 TS1"},
|
||||
{1, "topo_8p_ts1_1.xml", "single node 8 VEGA20 TS1 Alt. Model"},
|
||||
{4, "topo_8p_ts1_1.xml", "4 nodes 8 VEGA20 TS1 Alt. Model"},
|
||||
{1, "topo_4p3l_2h.xml", "single node 8 gfx908 Rome"},
|
||||
{4, "topo_4p3l_2h.xml", "4 nodes 8 gfx908 Rome"},
|
||||
};
|
||||
|
||||
int main(int argc,char* argv[])
|
||||
|
||||
Reference in New Issue
Block a user