Move page cache via mbind to prevent cross-NUMA access

[vishalc-ibm]

page cache pages are retained in memory of the node after running llama-bench bound to a node on multi-node systems, incuring cross-NUMA memory access penalty for subsequent runs of llama-bench bound to a different node. This commit introduces an mbind call as best effort basis to move the pages to the target node where llama-bench is executed, ensuring optimal NUMA locality. Additionally, necessary NUMA headers are included and the build is updated to link against the NUMA library.

Experiments:

1. Run llama-bench on node 1 (base)
2. Run llama-bench on node 0 (regression observed)
3. Run patched llama-bench on node 0 (throughput same as base)

+ /usr/bin/time -p numactl -N 1 -m 1 $llama-bench -m $models/llama-2-7b-chat.Q8_0.gguf -ngl 0 --prio 0 -b 1 -t 24

model	size	params	backend	threads	n_batch	test	t/s
llama 7B Q8_0	6.67 GiB	6.74 B	CPU	24	1	pp512	5.39 ± 0.01
llama 7B Q8_0	6.67 GiB	6.74 B	CPU	24	1	tg128	5.49 ± 0.03

build: 35782ae (5014)
real 687.60
user 15653.73
sys 42.67

+ /usr/bin/time -p numactl -N 0 -m 0 $llama-bench -m $models/llama-2-7b-chat.Q8_0.gguf -ngl 0 --prio 0 -b 1 -t 24

model	size	params	backend	threads	n_batch	test	t/s
llama 7B Q8_0	6.67 GiB	6.74 B	CPU	24	1	pp512	4.60 ± 0.01
llama 7B Q8_0	6.67 GiB	6.74 B	CPU	24	1	tg128	4.67 ± 0.03

build: 35782ae (5014)
real 805.99
user 18187.26
sys 48.93

+ /usr/bin/time -p numactl -N 0 -m 0 $patched-llama-bench -m $models/llama-2-7b-chat.Q8_0.gguf -ngl 0 --prio 0 -b 1 -t 24

model	size	params	backend	threads	n_batch	test	t/s
llama 7B Q8_0	6.67 GiB	6.74 B	CPU	24	1	pp512	5.35 ± 0.01
llama 7B Q8_0	6.67 GiB	6.74 B	CPU	24	1	tg128	5.46 ± 0.02

build: 35782ae (5014)
real 696.12
user 15735.41
sys 44.08

Suggested-by: Ritesh Harjani (IBM) ritesh.list@gmail.com
Signed-off-by: Vishal Chourasia vishalc@linux.ibm.com

[vishalc-ibm]

Hi @ggerganov can you approve the workflows

[vishalc-ibm]

28:   CPY(type_src=f32,type_dst=q5_1,ne=[256,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0]): [CPY] NMSE = 0.000001863 > 0.000001000 FAIL

Following test is failing.

This test checks a CPY operation involving conversion from f32 to q5_1 with a specific permutation. The Normalized Mean Squared Error (NMSE) exceeds the allowed threshold of 0.000001, resulting in a failure.

The changes introduced in this PR relate to moving stale page cache pages to the NUMA node where the benchmark most recently ran.

At this point, it's unclear how or why this change would affect the precision of this operation. If anyone has insights into how memory locality or page migration might influence quantization accuracy or the copy path, please share your thoughts.