qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] hw/loongarch: Refine default numa id calculation


From: gaosong
Subject: Re: [PATCH] hw/loongarch: Refine default numa id calculation
Date: Tue, 30 Apr 2024 11:59:07 +0800
User-agent: Mozilla/5.0 (X11; Linux loongarch64; rv:68.0) Gecko/20100101 Thunderbird/68.7.0

在 2024/3/19 上午10:26, Bibo Mao 写道:
With numa_test test case, there is subcase named test_def_cpu_split(),
there are 8 sockets and 2 numa nodes. Here is command line:
"-machine smp.cpus=8,smp.sockets=8 -numa node,memdev=ram -numa node"

The required result is:
   node 0 cpus: 0 2 4 6
   node 1 cpus: 1 3 5 7
Test case numa_test fails on LoongArch, since the actual result is:
   node 0 cpus: 0 1 2 3
   node 1 cpus: 4 5 6 7

It will be better if all the cpus in one socket share the same numa
node. Here socket id is used to calculate numa id in function
virt_get_default_cpu_node_id().

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
---
  hw/loongarch/virt.c | 11 +++++------
  1 file changed, 5 insertions(+), 6 deletions(-)
Reviewed-by: Song Gao <gaosong@loongson.cn>

Thanks.
Song Gao
diff --git a/hw/loongarch/virt.c b/hw/loongarch/virt.c
index deb3750d81..29885f6777 100644
--- a/hw/loongarch/virt.c
+++ b/hw/loongarch/virt.c
@@ -1219,15 +1219,14 @@ virt_cpu_index_to_props(MachineState *ms, unsigned 
cpu_index)
static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
  {
-    int64_t nidx = 0;
+    int64_t socket_id;
if (ms->numa_state->num_nodes) {
-        nidx = idx / (ms->smp.cpus / ms->numa_state->num_nodes);
-        if (ms->numa_state->num_nodes <= nidx) {
-            nidx = ms->numa_state->num_nodes - 1;
-        }
+        socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
+        return socket_id % ms->numa_state->num_nodes;
+    } else {
+        return 0;
      }
-    return nidx;
  }
static void loongarch_class_init(ObjectClass *oc, void *data)




reply via email to

[Prev in Thread] Current Thread [Next in Thread]