thelilfrog/cpufetch
1
0
Fork 0
mirror of https://github.com/Dr-Noob/cpufetch.git synced 2026-03-25 16:00:39 +01:00
Code Issues Packages Projects Releases Wiki Activity

[v1.04][ARM] Experimental new backend for computing peak performance, considering the number of VPUs and width (previously this was highly simplified)

Browse Source
This commit is contained in:
Dr-Noob
2023-05-06 15:36:00 +02:00
parent b3719dc216
commit 53e5dd19f7
3 changed files with 108 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
src/arm/midr.c
View File

@@ -80,26 +80,21 @@ int64_t get_peak_performance(struct cpuInfo* cpu) {
}
}
int64_t flops = 0;
int64_t total_flops = 0;
ptr = cpu;
if(cpu->soc->soc_vendor == SOC_VENDOR_APPLE) {
// Special case for M1/M2
// First we find the E cores, then the P
// M1 have 2 (E cores) or 4 (P cores) FMA units
// Source: https://dougallj.github.io/applecpu/firestorm-simd.html
flops += ptr->topo->total_cores * (get_freq(ptr->freq) * 1000000) * 2 * 4 * 2;
ptr = ptr->next_cpu;
flops += ptr->topo->total_cores * (get_freq(ptr->freq) * 1000000) * 2 * 4 * 4;
}
else {
for(int i=0; i < cpu->num_cpus; ptr = ptr->next_cpu, i++) {
flops += ptr->topo->total_cores * (get_freq(ptr->freq) * 1000000);
}
if(cpu->feat->NEON) flops = flops * 4;
for(int i=0; i < cpu->num_cpus; ptr = ptr->next_cpu, i++) {
int vpus = get_number_of_vpus(ptr);
int vpus_width = get_vpus_width(ptr);
bool has_fma = has_fma_support(ptr);
int64_t flops = ptr->topo->total_cores * get_freq(ptr->freq) * 1000000 * vpus * (vpus_width/32);
if(has_fma) flops = flops * 2;
total_flops += flops;
}
return flops;
return total_flops;
}
uint32_t fill_ids_from_midr(uint32_t* midr_array, int32_t* freq_array, uint32_t* ids_array, int len) {