danielkrause/colibri-strix
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

reserve 2.5 GB for the page cache in the RAM budget: starving it drops buffered reads 800->180 MB/s (measured)

Browse Source 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
JustVugg
2026-07-06 18:00:35 +02:00
parent e982d2930e
commit 3278ea78d0
1 changed files with 6 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
c/glm.c
+6 -2
View File
@@ -1264,7 +1264,7 @@ static double mem_available_gb(void){
static double expert_avail(Model *m, double ram_gb, int ebits, int max_ctx){ static double expert_avail(Model *m, double ram_gb, int ebits, int max_ctx){
Cfg *c=&m->c; int64_t eb=expert_bytes_probe(m,ebits); Cfg *c=&m->c; int64_t eb=expert_bytes_probe(m,ebits);
if(ram_gb<=0){ ram_gb=g_mem_avail_boot*0.88; if(ram_gb<4) ram_gb=8; } if(ram_gb<=0){ ram_gb=g_mem_avail_boot*0.88; if(ram_gb<4) ram_gb=8; }
double slack = 1.2e9 + 64.0*(double)eb double slack = 1.2e9 + 2.5e9 + 64.0*(double)eb
+ (double)(c->n_layers+1)*max_ctx*(c->kv_lora+c->qk_rope)*4.0 + (double)(c->n_layers+1)*max_ctx*(c->kv_lora+c->qk_rope)*4.0
+ (double)max_ctx*c->n_heads*(c->qk_nope+c->v_head)*4.0; + (double)max_ctx*c->n_heads*(c->qk_nope+c->v_head)*4.0;
return ram_gb*1e9 - (double)m->resident_bytes - slack; return ram_gb*1e9 - (double)m->resident_bytes - slack;
@@ -1288,7 +1288,11 @@ static void cap_for_ram(Model *m, double ram_gb, int ebits, int max_ctx){
double ws_b = 64.0*(double)eb; double ws_b = 64.0*(double)eb;
double kv_b = (double)(c->n_layers+1)*max_ctx*(c->kv_lora+c->qk_rope)*4.0; double kv_b = (double)(c->n_layers+1)*max_ctx*(c->kv_lora+c->qk_rope)*4.0;
double kvb_b = (double)max_ctx*c->n_heads*(c->qk_nope+c->v_head)*4.0; double kvb_b = (double)max_ctx*c->n_heads*(c->qk_nope+c->v_head)*4.0;
double slack = 1.2e9 + ws_b + kv_b + kvb_b; /* RISERVA PAGE-CACHE (misurato 2026-07-06): strangolarla fa crollare le pread
* buffered da ~800 a ~180 MB/s — gli ultimi GB di LRU rendono MENO di quanto
* costino in banda disco persa. 2.5 GB restano SEMPRE al kernel. */
double pc_b = 2.5e9;
double slack = 1.2e9 + pc_b + ws_b + kv_b + kvb_b;
double avail = ram_gb*1e9 - (double)m->resident_bytes - slack; double avail = ram_gb*1e9 - (double)m->resident_bytes - slack;
int capmax = (avail>0 && nsp>0) ? (int)(avail/((double)nsp*eb)) : 0; int capmax = (avail>0 && nsp>0) ? (int)(avail/((double)nsp*eb)) : 0;
if(capmax<1) capmax=1; if(capmax<1) capmax=1;