mir/cpufetch
2
0
Fork 0
mirror of https://github.com/Dr-Noob/cpufetch synced 2025-08-31 06:15:08 +00:00
Code Issues Releases Wiki Activity

Get number of cores (Intel only) with cpuid. This removes sysconf dependency. This implementation breaks cpufetch in NUMA environments

Browse Source
This commit is contained in:
Dr-Noob
2020-06-20 20:54:38 +02:00
parent 6ad5f65c34
commit 525cf1a76f
3 changed files with 95 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/main.c
View File

@@ -24,7 +24,7 @@ Peak FLOPS: 512 GFLOP/s(in simple precision)
***/
static const char* VERSION = "0.42";
static const char* VERSION = "0.43";
void print_help(int argc, char *argv[]) {
printf("Usage: %s [--version] [--help] [--style STYLE]\n\
@@ -69,13 +69,17 @@ int main(int argc, char* argv[]) {
if(freq == NULL)
return EXIT_FAILURE;
struct topology* topo = get_topology_info(cpu);
if(topo == NULL)
return EXIT_FAILURE;
struct ascii* art = set_ascii(get_cpu_vendor(cpu),getStyle());
if(art == NULL)
return EXIT_FAILURE;
char* cpuName = get_str_cpu_name();
char* maxFrequency = get_str_freq(freq);
char* nCores = get_str_ncores(cpu);
char* nCores = get_str_topology(topo);
char* avx = get_str_avx(cpu);
char* sse = get_str_sse(cpu);
char* fma = get_str_fma(cpu);
@@ -84,7 +88,7 @@ int main(int argc, char* argv[]) {
char* l1 = get_str_l1(cach);
char* l2 = get_str_l2(cach);
char* l3 = get_str_l3(cach);
char* pp = get_str_peak_performance(cpu,get_freq(freq));
char* pp = get_str_peak_performance(cpu,topo,get_freq(freq));
setAttribute(art,ATTRIBUTE_NAME,cpuName);
setAttribute(art,ATTRIBUTE_FREQUENCY,maxFrequency);

119
src/standart.c
View File

@@ -44,9 +44,6 @@ struct cpuInfo {
VENDOR cpu_vendor;
int nThreads;
// Threads per core
int HT;
// Max cpuids levels
unsigned int maxLevels;
// Max cpuids extended levels
@@ -65,6 +62,13 @@ struct frequency {
long max;
};
struct topology {
int physical_cores;
int logical_cores;
int smt;
bool ht;
};
void init_cpu_info(struct cpuInfo* cpu) {
cpu->AVX = false;
cpu->AVX2 = false;
@@ -129,29 +133,12 @@ struct cpuInfo* get_cpu_info() {
}
//Get max extended level
eax = 0x8000000;
eax = 0x80000000;
ebx = 0;
ecx = 0;
edx = 0;
cpuid(&eax, &ebx, &ecx, &edx);
cpu->maxExtendedLevels = eax;
//Fill cores and threads
cpu->nThreads = sysconf(_SC_NPROCESSORS_ONLN);
//Always check we can fetch data
if (cpu->maxLevels >= 0x0000000B) {
eax = 0x0000000B;
ecx = 0x00000000;
cpuid(&eax, &ebx, &ecx, &edx);
cpu->HT = ebx & 0xF;
if(cpu->HT == 0) {
//AMD should not work with this, returning 0
//Suppose we have 1
cpu->HT = 1;
}
}
else {
printWarn("Can't read topology information from cpuid (needed level is 0x%.8X, max is 0x%.8X). Assuming HT is disabled", 0x0000000B, cpu->maxLevels);
cpu->HT = 1;
}
cpu->maxExtendedLevels = eax;
//Fill instructions support
if (cpu->maxLevels >= 0x00000001){
@@ -173,6 +160,7 @@ struct cpuInfo* get_cpu_info() {
else {
printWarn("Can't read features information from cpuid (needed level is 0x%.8X, max is 0x%.8X)", 0x00000001, cpu->maxLevels);
}
if (cpu->maxLevels >= 0x00000007){
eax = 0x00000007;
ecx = 0x00000000;
@@ -191,6 +179,7 @@ struct cpuInfo* get_cpu_info() {
else {
printWarn("Can't read features information from cpuid (needed level is 0x%.8X, max is 0x%.8X)", 0x00000007, cpu->maxLevels);
}
if (cpu->maxExtendedLevels >= 0x80000001){
eax = 0x80000001;
cpuid(&eax, &ebx, &ecx, &edx);
@@ -204,6 +193,65 @@ struct cpuInfo* get_cpu_info() {
return cpu;
}
struct topology* get_topology_info(struct cpuInfo* cpu) {
struct topology* topo = malloc(sizeof(struct cache));
unsigned int eax, ebx, ecx, edx;
int type;
if (cpu->maxLevels >= 0x00000001) {
eax = 0x00000001;
cpuid(&eax, &ebx, &ecx, &edx);
topo->ht = edx & (1 << 28);
}
else {
printWarn("Can't read HT information from cpuid (needed level is 0x%.8X, max is 0x%.8X). Assuming HT is disabled", 0x00000001, cpu->maxLevels);
topo->ht = false;
}
switch(cpu->cpu_vendor) {
case VENDOR_INTEL:
if (cpu->maxLevels >= 0x0000000B) {
//TODO: This idea only works with no NUMA systems
eax = 0x0000000B;
ecx = 0x00000000;
cpuid(&eax, &ebx, &ecx, &edx);
type = (ecx >> 8) & 0xFF;
if (type != 1) {
printBug("Unexpected type in cpuid 0x0000000B (expected 1, got %d)", type);
return NULL;
}
topo->smt = ebx & 0xFFFF;
eax = 0x0000000B;
ecx = 0x00000001;
cpuid(&eax, &ebx, &ecx, &edx);
type = (ecx >> 8) & 0xFF;
if (type < 2) {
printBug("Unexpected type in cpuid 0x0000000B (expected < 2, got %d)", type);
return NULL;
}
topo->logical_cores = ebx & 0xFFFF;
topo->physical_cores = topo->logical_cores / topo->smt;
}
else {
printWarn("Can't read topology information from cpuid (needed level is 0x%.8X, max is 0x%.8X)", 0x0000000B, cpu->maxLevels);
topo->physical_cores = 1;
topo->logical_cores = 1;
topo->smt = 1;
}
break;
case VENDOR_AMD:
printBug("Unimplemented!");
break;
default:
printBug("Cant get topology because VENDOR is empty");
return NULL;
}
return topo;
}
// see https://stackoverflow.com/questions/12594208/c-program-to-determine-levels-size-of-cache
struct cache* get_cache_info(struct cpuInfo* cpu) {
struct cache* cach = malloc(sizeof(struct cache));
@@ -333,7 +381,7 @@ void debug_frequency(struct frequency* freq) {
/*** STRING FUNCTIONS ***/
char* get_str_peak_performance(struct cpuInfo* cpu, long freq) {
char* get_str_peak_performance(struct cpuInfo* cpu, struct topology* topo, long freq) {
/***
PP = PeakPerformance
SP = SinglePrecision
@@ -358,7 +406,7 @@ char* get_str_peak_performance(struct cpuInfo* cpu, long freq) {
return string;
}
float flops = (cpu->nThreads/cpu->HT)*(freq*1000000)*2;
float flops = topo->physical_cores*(freq*1000000)*2;
if(cpu->FMA3 || cpu->FMA4)
flops = flops*2;
@@ -379,20 +427,19 @@ char* get_str_peak_performance(struct cpuInfo* cpu, long freq) {
return string;
}
char* get_str_ncores(struct cpuInfo* cpu) {
if(cpu->HT > 1) {
char* get_str_topology(struct topology* topo) {
char* string;
if(topo->smt > 1) {
//2(N.Cores)7(' cores(')3(N.Threads)9(' threads)')
int size = 2+7+3+9+1;
char* string = malloc(sizeof(char)*size);
snprintf(string,size,"%d cores(%d threads)",cpu->nThreads/cpu->HT,cpu->nThreads);
return string;
string = malloc(sizeof(char)*size);
snprintf(string,size,"%d cores (%d threads)",topo->physical_cores,topo->logical_cores);
}
else {
char* string = malloc(sizeof(char)*2+7+1);
snprintf(string,2+7+1,"%d cores",cpu->nThreads);
return string;
string = malloc(sizeof(char)*2+7+1);
snprintf(string,2+7+1,"%d cores",topo->physical_cores);
}
return string;
}
char* get_str_avx(struct cpuInfo* cpu) {

6
src/standart.h
View File

@@ -9,6 +9,7 @@
struct cpuInfo;
struct frequency;
struct cache;
struct topology;
typedef int VENDOR;
@@ -17,6 +18,7 @@ VENDOR get_cpu_vendor(struct cpuInfo* cpu);
long get_freq(struct frequency* freq);
struct cache* get_cache_info(struct cpuInfo* cpu);
struct frequency* get_frequency_info(struct cpuInfo* cpu);
struct topology* get_topology_info(struct cpuInfo* cpu);
char* get_str_ncores(struct cpuInfo* cpu);
char* get_str_avx(struct cpuInfo* cpu);
@@ -31,7 +33,9 @@ char* get_str_l3(struct cache* cach);
char* get_str_freq(struct frequency* freq);
char* get_str_peak_performance(struct cpuInfo* cpu, long freq);
char* get_str_topology(struct topology* topo);
char* get_str_peak_performance(struct cpuInfo* cpu, struct topology* topo, long freq);
void free_cpuinfo_struct(struct cpuInfo* cpu);
void free_cache_struct(struct cache* cach);