int32_t
rte_hash_lookup_with_hash(const struct rte_hash *h,
const void *key, hash_sig_t sig)
{
RETURN_IF_TRUE(((h == NULL) || (key == NULL)), -EINVAL);
return __rte_hash_lookup_with_hash(h, key, sig);
}
void ITimelineModel::Precompute(float fps)
{
RETURN_IF_FALSE(SupportPrecompute());
RETURN_IF_TRUE(mIsPrecomputed);
RETURN_IF(fps < 0.f);
RETURN_IF_EMPTY(mFrames);
mIsPrecomputed = true;
mFPS = fps;
int preCalculatedframeCount = (int)Math::Ceil(mDuration * fps);
RETURN_IF(preCalculatedframeCount <= 0);
mFrameCount = (uint)preCalculatedframeCount;
OnPrecomputeBegin();
float firstTime = mFrames.First().Time;
AddPrecomputedItem(Math::IsZero(firstTime), 0U, 0U, 0.f); //add zero frame
const float frameInterval = 1.f / mFPS;
float time = 0.f;
uint outPrevFrameIndex;
uint outNextFrameIndex;
float outPercent;
uint prevIndex = 0;
FOR_EACH_SIZE(i, mFrameCount)
{
time += frameInterval;
bool isFound = TryGetFrameHelper(time, outPrevFrameIndex, outNextFrameIndex, outPercent, prevIndex);
if (isFound)
{
prevIndex = outPrevFrameIndex;
}
AddPrecomputedItem(isFound, outPrevFrameIndex, outNextFrameIndex, outPercent);
}
void IRenderBatch::CustomDraw(IRenderQueue& renderQueue, RenderingFlags renderingFlags /*= RenderingFlags::None*/)
{
RETURN_IF_TRUE(IsEmpty());
EffectTechnique* technique = mEffect->CurrentTechniqueGroup()->CurrentTechnique();
EffectTechnique::PassCollectionType& renderPasses = technique->RenderPasses();
RenderingStatics::Instance().CountRenderPass(renderPasses.Count());
RenderingStatics::Instance().CountMaterial(mMaterial);
RenderingStatics::Instance().CountMaterialTextures(mMaterial);
FOR_EACH_COLLECTION(i, renderPasses)
{
IRenderPass* renderPass = *i;
RenderingContext::Instance().ApplyRenderPass(renderPass);
RenderingContext::Instance().ApplyMaterial(mMaterial);
RenderingContext::Instance().ApplyState(mStateTreeNode);
RenderingContext::Instance().ApplyBatch(this);
RenderingContext::Instance().ValidateBeforeDraw();
Draw(renderQueue, renderingFlags);
RenderingContext::Instance().RestoreBatch();
RenderingContext::Instance().RestoreState();
RenderingContext::Instance().RestoreMaterial();
RenderingContext::Instance().RestoreRenderPass();
}
void DateTime::ConvertToUTC()
{
RETURN_IF_TRUE(mIsUTC);
mIsUTC = true;
UpdateTimePointToUTCDate();
}
void ShaderUniform::SetArray1(MemoryIntData data)
{
MEDUSA_ASSERT(mDataType == GraphicsUniformDataType::Int || mDataType == GraphicsUniformDataType::Bool, "ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&mIntData.IsContentEqual(data));
#endif
Render::Instance().SetUniformArray1(mLocation, (uint)data.Size(), data.Data());
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mIntData = data;
#endif
}
void ShaderConstant::Set( float x )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::Float,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&Math::IsEqual(mFloatArray[0],x));
#endif
Render::Instance().SetUniform(mLocation,x);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mFloatArray[0]=x;
#endif
}
void ShaderConstant::Set( const Point2F& pos )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::FloatVec2,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&Math::IsEqual(mFloatArray[0],pos.X)&&Math::IsEqual(mFloatArray[1],pos.Y));
#endif
Render::Instance().SetUniform(mLocation,pos.X,pos.Y);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mFloatArray[0]=pos.X;
mFloatArray[1]=pos.Y;
#endif
}
void ShaderConstant::SetArray4( uint count,const float* data )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::FloatVec4,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&mFloatDataCount==count&&mFloatData==data);
#endif
Render::Instance().SetUniformArray4(mLocation,count,data);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mFloatDataCount=count;
mFloatData=data;
#endif
}
void ShaderUniform::SetMatrix2(MemoryFloatData data, bool isTransposed/*=false*/)
{
MEDUSA_ASSERT(mDataType == GraphicsUniformDataType::FloatMat2, "ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&mFloatData.IsContentEqual(data)&&mIsTransposed==isTransposed);
#endif
Render::Instance().SetUniformMatrix2(mLocation, (uint)data.Size() / sizeof(Matrix2), data.Data(), isTransposed);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mFloatData = data;
mIsTransposed = isTransposed;
#endif
}
void ShaderConstant::Set( const Point3I& pos )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::IntVec3||mDataType==GraphicsUniformDataType::BoolVec3,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&mIntArray[0]==pos.X&&mIntArray[1]==pos.Y&&mIntArray[2]==pos.Z);
#endif
Render::Instance().SetUniform(mLocation,pos.X,pos.Y,pos.Z);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mIntArray[0]=pos.X;
mIntArray[1]=pos.Y;
mIntArray[2]=pos.Z;
#endif
}
void ShaderConstant::SetMatrix4( uint count,const float* data,bool isTransposed/*=false*/ )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::FloatMat4,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&mFloatDataCount==count&&mFloatData==data&&mIsTransposed==isTransposed);
#endif
Render::Instance().SetUniformMatrix4(mLocation,count,data,isTransposed);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mFloatDataCount=count;
mFloatData=data;
mIsTransposed=isTransposed;
#endif
}
void ShaderConstant::Set( const Color3F& color )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::FloatVec3,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&Math::IsEqual(mFloatArray[0],color.R)&&Math::IsEqual(mFloatArray[1],color.G)&&Math::IsEqual(mFloatArray[2],color.B));
#endif
Render::Instance().SetUniform(mLocation,color.R,color.G,color.B);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mFloatArray[0]=color.R;
mFloatArray[1]=color.G;
mFloatArray[2]=color.B;
#endif
}
void ShaderConstant::Set( const Color4B& color )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::IntVec4||mDataType==GraphicsUniformDataType::BoolVec4,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&mIntArray[0]==(int)color.R&&mIntArray[1]==(int)color.G&&mIntArray[2]==(int)color.B&&mIntArray[3]==color.A);
#endif
Render::Instance().SetUniform(mLocation,color.R,color.G,color.B,color.A);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mIntArray[0]=color.R;
mIntArray[1]=color.G;
mIntArray[2]=color.B;
mIntArray[3]=color.A;
#endif
}
void ShaderConstant::Set( int x )
{
MEDUSA_ASSERT(mDataType==GraphicsUniformDataType::Bool
||mDataType==GraphicsUniformDataType::Int
||mDataType==GraphicsUniformDataType::Sampler2D
||mDataType==GraphicsUniformDataType::SamplerCube,"ErrorDataType");
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
RETURN_IF_TRUE(mIsAssigned&&mIntArray[0]==x);
#endif
Render::Instance().SetUniform(mLocation,x);
MarkAssigned();
#ifdef MEDUSA_RENDER_STATE_CACHE_ENABLED
mIntArray[0]=x;
#endif
}
int
rte_hash_lookup_bulk(const struct rte_hash *h, const void **keys,
uint32_t num_keys, int32_t *positions)
{
uint32_t i, j, bucket_index;
hash_sig_t sigs[RTE_HASH_LOOKUP_BULK_MAX];
RETURN_IF_TRUE(((h == NULL) || (keys == NULL) || (num_keys == 0) ||
(num_keys > RTE_HASH_LOOKUP_BULK_MAX) ||
(positions == NULL)), -EINVAL);
/* Get the hash signature and bucket index */
for (i = 0; i < num_keys; i++) {
sigs[i] = h->hash_func(keys[i], h->key_len,
h->hash_func_init_val) | h->sig_msb;
bucket_index = sigs[i] & h->bucket_bitmask;
/* Pre-fetch relevant buckets */
rte_prefetch1((void *) get_sig_tbl_bucket(h, bucket_index));
rte_prefetch1((void *) get_key_tbl_bucket(h, bucket_index));
}
/* Check if key is already present in the hash */
for (i = 0; i < num_keys; i++) {
bucket_index = sigs[i] & h->bucket_bitmask;
hash_sig_t *sig_bucket = get_sig_tbl_bucket(h, bucket_index);
uint8_t *key_bucket = get_key_tbl_bucket(h, bucket_index);
positions[i] = -ENOENT;
for (j = 0; j < h->bucket_entries; j++) {
if ((sigs[i] == sig_bucket[j]) &&
likely(memcmp(keys[i],
get_key_from_bucket(h, key_bucket, j),
h->key_len) == 0)) {
positions[i] = bucket_index *
h->bucket_entries + j;
break;
}
}
}
return 0;
}
static inline void
tkz_file_tk_io_block_process(s_io_block_t *io_block, s_tk_t *tk_head,
s_tkz_lang_t *tkz_lang)
{
assert_exit(tk_structure_legal_p(tk_head));
assert_exit(tkz_io_block_structure_legal_p(io_block));
RETURN_IF_TRUE(tkz_io_block_empty_p(io_block));
switch (tkz_lang->type) {
case TKZ_LANG_C:
tkz_io_block_lang_c_match(tkz_lang, tk_head, io_block);
break;
default:
assert_exit(false);
break;
}
tkz_io_block_cleanup(io_block);
}
void print_i7z_socket_single(struct cpu_socket_info socket_0, int printw_offset, int PLATFORM_INFO_MSR, int PLATFORM_INFO_MSR_high, int PLATFORM_INFO_MSR_low,
int* online_cpus, double cpu_freq_cpuinfo, struct timespec one_second_sleep, char TURBO_MODE,
char* HT_ON_str, int* kk_1, U_L_L_I * old_val_CORE, U_L_L_I * old_val_REF, U_L_L_I * old_val_C3, U_L_L_I * old_val_C6, U_L_L_I * old_val_C7,
U_L_L_I * old_TSC, int estimated_mhz, U_L_L_I * new_val_CORE, U_L_L_I * new_val_REF, U_L_L_I * new_val_C3,
U_L_L_I * new_val_C6, U_L_L_I * new_val_C7, U_L_L_I * new_TSC, double* _FREQ, double* _MULT, long double * C0_time, long double * C1_time,
long double * C3_time, long double * C6_time, long double * C7_time, struct timeval* tvstart, struct timeval* tvstop, int *max_observed_cpu)
{
int numPhysicalCores, numLogicalCores;
double TRUE_CPU_FREQ;
//Print a slew of information on the ncurses window
mvprintw (0, 0, "Cpu speed from cpuinfo %0.2fMhz\n", cpu_freq_cpuinfo);
mvprintw (1, 0,
"cpuinfo might be wrong if cpufreq is enabled. To guess correctly try estimating via tsc\n");
mvprintw (2, 0, "Linux's inbuilt cpu_khz code emulated now\n\n");
//estimate the freq using the estimate_MHz() code that is almost mhz accurate
cpu_freq_cpuinfo = estimate_MHz ();
mvprintw (3, 0, "True Frequency (without accounting Turbo) %0.0f MHz\n",
cpu_freq_cpuinfo);
int i, ii;
//int k;
int CPU_NUM;
int* core_list;
unsigned long int IA32_MPERF, IA32_APERF;
int CPU_Multiplier, error_indx;
unsigned long long int CPU_CLK_UNHALTED_CORE, CPU_CLK_UNHALTED_REF, CPU_CLK_C3, CPU_CLK_C6, CPU_CLK_C1, CPU_CLK_C7;
//current blck value
float BLCK;
char print_core[32];
long double c1_time;
//use this variable to monitor the max number of cores ever online
*max_observed_cpu = (socket_0.max_cpu > *max_observed_cpu)? socket_0.max_cpu: *max_observed_cpu;
int core_list_size_phy, core_list_size_log;
if (socket_0.max_cpu > 0) {
//set the variable print_core to 0, use it to check if a core is online and doesnt
//have any garbage values
memset(print_core, 0, 6*sizeof(char));
//We just need one CPU (we use Core-1) to figure out the multiplier and the bus clock freq.
//multiplier doesnt automatically include turbo
//note turbo is not guaranteed, only promised
//So this msr will only reflect the actual multiplier, rest has to be figured out
//Now get all the information about the socket from the structure
CPU_NUM = socket_0.processor_num[0];
core_list = socket_0.processor_num;
core_list_size_phy = socket_0.num_physical_cores;
core_list_size_log = socket_0.num_logical_cores;
/*if (CPU_NUM == -1)
{
sleep (1); //sleep for a bit hoping that the offline socket becomes online
continue;
}*/
//number of CPUs is as told via cpuinfo
int numCPUs = socket_0.num_physical_cores;
CPU_Multiplier = get_msr_value (CPU_NUM, PLATFORM_INFO_MSR, PLATFORM_INFO_MSR_high, PLATFORM_INFO_MSR_low, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
//Blck is basically the true speed divided by the multiplier
BLCK = cpu_freq_cpuinfo / CPU_Multiplier;
//Use Core-1 as the one to check for the turbo limit
//Core number shouldnt matter
//bits from 0-63 in this store the various maximum turbo limits
int MSR_TURBO_RATIO_LIMIT = 429;
// 3B defines till Max 4 Core and the rest bit values from 32:63 were reserved.
int MAX_TURBO_1C=0, MAX_TURBO_2C=0, MAX_TURBO_3C=0,
MAX_TURBO_4C=0, MAX_TURBO_5C=0, MAX_TURBO_6C=0;
if ( E7_mp_present){
//e7 mp dont have 429 register so dont read the register.
} else {
//Bits:0-7 - core1
MAX_TURBO_1C = get_msr_value (CPU_NUM, MSR_TURBO_RATIO_LIMIT, 7, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
//Bits:15-8 - core2
MAX_TURBO_2C = get_msr_value (CPU_NUM, MSR_TURBO_RATIO_LIMIT, 15, 8, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
//Bits:23-16 - core3
MAX_TURBO_3C = get_msr_value (CPU_NUM, MSR_TURBO_RATIO_LIMIT, 23, 16, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
//Bits:31-24 - core4
MAX_TURBO_4C = get_msr_value (CPU_NUM, MSR_TURBO_RATIO_LIMIT, 31, 24, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
//gulftown/Hexacore support
//technically these should be the bits to get for core 5,6
//Bits:39-32 - core4
MAX_TURBO_5C = get_msr_value (CPU_NUM, MSR_TURBO_RATIO_LIMIT, 39, 32, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
//Bits:47-40 - core4
MAX_TURBO_6C = get_msr_value (CPU_NUM, MSR_TURBO_RATIO_LIMIT, 47, 40, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
}
//fflush (stdout);
//sleep (1);
char string_ptr1[200], string_ptr2[200];
int IA32_PERF_GLOBAL_CTRL = 911; //38F
int IA32_PERF_GLOBAL_CTRL_Value;
IA32_PERF_GLOBAL_CTRL_Value = get_msr_value (CPU_NUM, IA32_PERF_GLOBAL_CTRL, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
int IA32_FIXED_CTR_CTL = 909; //38D
int IA32_FIXED_CTR_CTL_Value;
IA32_FIXED_CTR_CTL_Value = get_msr_value (CPU_NUM, IA32_FIXED_CTR_CTL, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
IA32_MPERF = get_msr_value (CPU_NUM, 231, 7, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
IA32_APERF = get_msr_value (CPU_NUM, 232, 7, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
CPU_CLK_UNHALTED_CORE = get_msr_value (CPU_NUM, 778, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
CPU_CLK_UNHALTED_REF = get_msr_value (CPU_NUM, 779, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
//SLEEP FOR 1 SECOND (500ms is also alright)
nanosleep (&one_second_sleep, NULL);
IA32_MPERF = get_msr_value (CPU_NUM, 231, 7, 0, &error_indx) - IA32_MPERF;
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
IA32_APERF = get_msr_value (CPU_NUM, 232, 7, 0, &error_indx) - IA32_APERF;
SET_IF_TRUE(error_indx,online_cpus[0],-1);
RETURN_IF_TRUE(online_cpus[0]==-1);
mvprintw (4 + printw_offset, 0," CPU Multiplier %dx || Bus clock frequency (BCLK) %0.2f MHz \n", CPU_Multiplier, BLCK);
if (numCPUs <= 0) {
sprintf (string_ptr1, " Max TURBO Multiplier (if Enabled) with 0 cores is");
sprintf (string_ptr2, " %dx/%dx ", MAX_TURBO_1C, MAX_TURBO_2C);
}
if (numCPUs >= 1 && numCPUs < 4) {
sprintf (string_ptr1, " Max TURBO Multiplier (if Enabled) with 1/2 Cores is");
sprintf (string_ptr2, " ");
}
if (numCPUs >= 2 && numCPUs < 6) {
sprintf (string_ptr1, " Max TURBO Multiplier (if Enabled) with 1/2/3/4 Cores is");
sprintf (string_ptr2, " %dx/%dx/%dx/%dx ", MAX_TURBO_1C, MAX_TURBO_2C, MAX_TURBO_3C, MAX_TURBO_4C);
}
if (numCPUs >= 2 && numCPUs >= 6) { // Gulftown 6-cores, Nehalem-EX
sprintf (string_ptr1, " Max TURBO Multiplier (if Enabled) with 1/2/3/4/5/6 Cores is");
sprintf (string_ptr2, " %dx/%dx/%dx/%dx/%dx/%dx ", MAX_TURBO_1C, MAX_TURBO_2C, MAX_TURBO_3C, MAX_TURBO_4C,
MAX_TURBO_5C, MAX_TURBO_6C);
}
numCPUs = core_list_size_phy;
numPhysicalCores = core_list_size_phy;
numLogicalCores = core_list_size_log;
//if (socket_0.socket_num == 0) {
mvprintw (19, 0, "C0 = Processor running without halting ");
mvprintw (20, 0, "C1 = Processor running with halts (States >C0 are power saver modes with cores idling)");
mvprintw (21, 0, "C3 = Cores running with PLL turned off and core cache turned off");
mvprintw (22, 0, "C6, C7 = Everything in C3 + core state saved to last level cache, C7 is deeper than C6");
mvprintw (23, 0, " Above values in table are in percentage over the last 1 sec");
// mvprintw (24, 0, "Total Logical Cores: [%d], Total Physical Cores: [%d] \n", numLogicalCores, numPhysicalCores);
mvprintw (24, 0, "[core-id] refers to core-id number in /proc/cpuinfo");
mvprintw (25, 0, "'Garbage Values' message printed when garbage values are read");
mvprintw (26, 0, " Ctrl+C to exit");
//}
mvprintw (6 + printw_offset, 0, "Socket [%d] - [physical cores=%d, logical cores=%d, max online cores ever=%d] \n", socket_0.socket_num, numPhysicalCores, numLogicalCores,*max_observed_cpu);
mvprintw (9 + printw_offset, 0, "%s %s\n", string_ptr1, string_ptr2);
if (TURBO_MODE == 1) {
mvprintw (7 + printw_offset, 0, " TURBO ENABLED on %d Cores, %s\n", numPhysicalCores, HT_ON_str);
TRUE_CPU_FREQ = BLCK * ((double) CPU_Multiplier + 1);
mvprintw (8 + printw_offset, 0, " Max Frequency without considering Turbo %0.2f MHz (%0.2f x [%d]) \n", TRUE_CPU_FREQ, BLCK, CPU_Multiplier + 1);
} else {
mvprintw (7 + printw_offset, 0, " TURBO DISABLED on %d Cores, %s\n", numPhysicalCores, HT_ON_str);
TRUE_CPU_FREQ = BLCK * ((double) CPU_Multiplier);
mvprintw (8 + printw_offset, 0," Max Frequency without considering Turbo %0.2f MHz (%0.2f x [%d]) \n", TRUE_CPU_FREQ, BLCK, CPU_Multiplier);
}
//Primarily for 32-bit users, found that after sometimes the counters loopback, so inorder
//to prevent loopback, reset the counters back to 0 after 10 iterations roughly 10 secs
if (*kk_1 > 10) {
*kk_1 = 0;
for (i = 0; i < numCPUs; i++) {
//Set up the performance counters and then start reading from them
assert(i < MAX_SK_PROCESSORS);
CPU_NUM = core_list[i];
ii = core_list[i];
assert(i < MAX_PROCESSORS); //online_cpus[i]
assert(ii < numCPUs_max);
IA32_PERF_GLOBAL_CTRL_Value = get_msr_value (CPU_NUM, IA32_PERF_GLOBAL_CTRL, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
set_msr_value (CPU_NUM, IA32_PERF_GLOBAL_CTRL, 0x700000003LLU);
IA32_FIXED_CTR_CTL_Value = get_msr_value (CPU_NUM, IA32_FIXED_CTR_CTL, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
set_msr_value (CPU_NUM, IA32_FIXED_CTR_CTL, 819);
IA32_PERF_GLOBAL_CTRL_Value = get_msr_value (CPU_NUM, IA32_PERF_GLOBAL_CTRL, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
IA32_FIXED_CTR_CTL_Value = get_msr_value (CPU_NUM, IA32_FIXED_CTR_CTL, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
old_val_CORE[ii] = get_msr_value (CPU_NUM, 778, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
old_val_REF[ii] = get_msr_value (CPU_NUM, 779, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
old_val_C3[ii] = get_msr_value (CPU_NUM, 1020, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
old_val_C6[ii] = get_msr_value (CPU_NUM, 1021, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
//table b-20 in 325384 and only for sandy bridge
old_val_C7[ii] = get_msr_value (CPU_NUM, 1022, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
}
old_TSC[ii] = rdtsc ();
}
}
(*kk_1)++;
nanosleep (&one_second_sleep, NULL);
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
mvprintw (11 + printw_offset, 0, "\tCore [core-id] :Actual Freq (Mult.)\t C0%% Halt(C1)%% C3 %% C6 %% C7 %% Temp VCore\n");
}else{
mvprintw (11 + printw_offset, 0, "\tCore [core-id] :Actual Freq (Mult.)\t C0%% Halt(C1)%% C3 %% C6 %% Temp VCore\n");
}
//estimate the CPU speed
estimated_mhz = estimate_MHz();
for (i = 0; i < numCPUs; i++) {
//read from the performance counters
//things like halted unhalted core cycles
assert(i < MAX_SK_PROCESSORS);
CPU_NUM = core_list[i];
ii = core_list[i];
assert(i < MAX_PROCESSORS); //online_cpus[i]
assert(ii < numCPUs_max);
new_val_CORE[ii] = get_msr_value (CPU_NUM, 778, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
new_val_REF[ii] = get_msr_value (CPU_NUM, 779, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
new_val_C3[ii] = get_msr_value (CPU_NUM, 1020, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
new_val_C6[ii] = get_msr_value (CPU_NUM, 1021, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
new_val_C7[ii] = get_msr_value (CPU_NUM, 1022, 63, 0, &error_indx);
SET_IF_TRUE(error_indx,online_cpus[i],-1);
CONTINUE_IF_TRUE(online_cpus[i]==-1);
}
new_TSC[ii] = rdtsc ();
if (old_val_CORE[ii] > new_val_CORE[ii]) { //handle overflow
CPU_CLK_UNHALTED_CORE = (UINT64_MAX - old_val_CORE[ii]) + new_val_CORE[ii];
} else {
CPU_CLK_UNHALTED_CORE = new_val_CORE[ii] - old_val_CORE[ii];
}
//number of TSC cycles while its in halted state
if ((new_TSC[ii] - old_TSC[ii]) < CPU_CLK_UNHALTED_CORE) {
CPU_CLK_C1 = 0;
} else {
CPU_CLK_C1 = ((new_TSC[ii] - old_TSC[ii]) - CPU_CLK_UNHALTED_CORE);
}
if (old_val_REF[ii] > new_val_REF[ii]) { //handle overflow
CPU_CLK_UNHALTED_REF = (UINT64_MAX - old_val_REF[ii]) + new_val_REF[ii]; //3.40282366921e38
} else {
CPU_CLK_UNHALTED_REF = new_val_REF[ii] - old_val_REF[ii];
}
if (old_val_C3[ii] > new_val_C3[ii]) { //handle overflow
CPU_CLK_C3 = (UINT64_MAX - old_val_C3[ii]) + new_val_C3[ii];
} else {
CPU_CLK_C3 = new_val_C3[ii] - old_val_C3[ii];
}
if (old_val_C6[ii] > new_val_C6[ii]) { //handle overflow
CPU_CLK_C6 = (UINT64_MAX - old_val_C6[ii]) + new_val_C6[ii];
} else {
CPU_CLK_C6 = new_val_C6[ii] - old_val_C6[ii];
}
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
if (old_val_C7[ii] > new_val_C7[ii]) { //handle overflow
CPU_CLK_C7 = (UINT64_MAX - old_val_C7[ii]) + new_val_C7[ii];
} else {
CPU_CLK_C7 = new_val_C7[ii] - old_val_C7[ii];
}
}
_FREQ[ii] =
THRESHOLD_BETWEEN_0_6000(estimated_mhz * ((long double) CPU_CLK_UNHALTED_CORE /
(long double) CPU_CLK_UNHALTED_REF));
_MULT[ii] = _FREQ[ii] / BLCK;
C0_time[ii] = ((long double) CPU_CLK_UNHALTED_REF /
(long double) (new_TSC[ii] - old_TSC[ii]));
C1_time[ii] = ((long double) CPU_CLK_C1 /
(long double) (new_TSC[ii] - old_TSC[ii]));
C3_time[ii] = ((long double) CPU_CLK_C3 /
(long double) (new_TSC[ii] - old_TSC[ii]));
C6_time[ii] = ((long double) CPU_CLK_C6 /
(long double) (new_TSC[ii] - old_TSC[ii]));
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
C7_time[ii] = ((long double) CPU_CLK_C7 /
(long double) (new_TSC[ii] - old_TSC[ii]));
}
if (C0_time[ii] < 1e-2) {
if (C0_time[ii] > 1e-4) {
C0_time[ii] = 0.01;
} else {
C0_time[ii] = 0;
}
}
if (C1_time[ii] < 1e-2) {
if (C1_time[ii] > 1e-4) {
C1_time[ii] = 0.01;
} else {
C1_time[ii] = 0;
}
}
if (C3_time[ii] < 1e-2) {
if (C3_time[ii] > 1e-4) {
C3_time[ii] = 0.01;
} else {
C3_time[ii] = 0;
}
}
if (C6_time[ii] < 1e-2) {
if (C6_time[ii] > 1e-4) {
C6_time[ii] = 0.01;
} else {
C6_time[ii] = 0;
}
}
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
if (C7_time[ii] < 1e-2) {
if (C7_time[ii] > 1e-4) {
C7_time[ii] = 0.01;
} else {
C7_time[ii] = 0;
}
}
}
}
//CHECK IF ALL COUNTERS ARE CORRECT AND NO GARBAGE VALUES ARE PRESENT
//If there is any garbage values set print_core[i] to 0
for (ii = 0; ii < numCPUs; ii++) {
assert(ii < MAX_SK_PROCESSORS);
i = core_list[ii];
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
if ( !IS_THIS_BETWEEN_0_100(C0_time[i] * 100) ||
!IS_THIS_BETWEEN_0_100(C1_time[i] * 100 - (C3_time[i] + C6_time[i]) * 100) ||
!IS_THIS_BETWEEN_0_100(C3_time[i] * 100) ||
!IS_THIS_BETWEEN_0_100(C6_time[i] * 100) ||
!IS_THIS_BETWEEN_0_100(C7_time[i] * 100) ||
isinf(_FREQ[i]) )
print_core[ii]=0;
else
print_core[ii]=1;
}else{
if ( !IS_THIS_BETWEEN_0_100(C0_time[i] * 100) ||
!IS_THIS_BETWEEN_0_100(C1_time[i] * 100 - (C3_time[i] + C6_time[i]) * 100) ||
!IS_THIS_BETWEEN_0_100(C3_time[i] * 100) ||
!IS_THIS_BETWEEN_0_100(C6_time[i] * 100) ||
isinf(_FREQ[i]) )
print_core[ii]=0;
else
print_core[ii]=1;
}
}
//Now print the information about the cores. Print garbage values message if there is garbage
for (ii = 0; ii < numCPUs; ii++) {
assert(ii < MAX_SK_PROCESSORS);
i = core_list[ii];
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
//there is a bit of leeway to be had as the total counts might deviate
//if this happens c1_time might be negative so just adjust so that it is thresholded to 0
c1_time = C1_time[i] * 100 - (C3_time[i] + C6_time[i] + C7_time[i]) * 100;
if (!isnan(c1_time) && !isinf(c1_time)) {
if (c1_time <= 0) {
c1_time=0;
}
}
if (print_core[ii])
mvprintw (12 + ii + printw_offset, 0, "\tCore %d [%d]:\t %0.2f (%.2fx)\t%4.3Lg\t%4.3Lg\t%4.3Lg\t%4.3Lg\t%4.3Lg\t%d\t%0.4f\n",
ii + 1, core_list[ii], _FREQ[i], _MULT[i], THRESHOLD_BETWEEN_0_100(C0_time[i] * 100),
THRESHOLD_BETWEEN_0_100(c1_time), THRESHOLD_BETWEEN_0_100(C3_time[i] * 100), THRESHOLD_BETWEEN_0_100(C6_time[i] * 100),THRESHOLD_BETWEEN_0_100(C7_time[i] * 100),
Read_Thermal_Status_CPU(core_list[ii]), //C0_time[i]*100+C1_time[i]*100 around 100
Read_Voltage_CPU(core_list[ii]));
else
mvprintw (12 + ii + printw_offset, 0, "\tCore %d [%d]:\t Garbage Values\n", ii + 1, core_list[ii]);
}else{
//there is a bit of leeway to be had as the total counts might deviate
//if this happens c1_time might be negative so just adjust so that it is thresholded to 0
c1_time = C1_time[i] * 100 - (C3_time[i] + C6_time[i]) * 100;
if (!isnan(c1_time) && !isinf(c1_time)) {
if (c1_time <= 0) {
c1_time=0;
}
}
if (print_core[ii])
mvprintw (12 + ii + printw_offset, 0, "\tCore %d [%d]:\t %0.2f (%.2fx)\t%4.3Lg\t%4.3Lg\t%4.3Lg\t%4.3Lg\t%d\t%0.4f\n",
ii + 1, core_list[ii], _FREQ[i], _MULT[i], THRESHOLD_BETWEEN_0_100(C0_time[i] * 100),
THRESHOLD_BETWEEN_0_100(c1_time), THRESHOLD_BETWEEN_0_100(C3_time[i] * 100), THRESHOLD_BETWEEN_0_100(C6_time[i] * 100),Read_Thermal_Status_CPU(core_list[ii]), //C0_time[i]*100+C1_time[i]*100 around 100
Read_Voltage_CPU(core_list[ii]));
else
mvprintw (12 + ii + printw_offset, 0, "\tCore %d [%d]:\t Garbage Values\n", ii + 1, core_list[ii]);
}
}
/*k=0;
for (ii = 00; ii < *max_observed_cpu; ii++)
{
if (in_core_list(ii,core_list)){
continue;
}else{
mvprintw (12 + k + numCPUs + printw_offset, 0, "\tProcessor %d [%d]: OFFLINE\n", k + numCPUs + 1, ii);
}
k++;
}*/
//FOR THE REST OF THE CORES (i.e. the offline cores+non-present cores=6 )
//I have space allocated for 6 cores to be printed per socket so from all the present cores
//till 6 print a blank line
//for(ii=*max_observed_cpu; ii<6; ii++)
for (ii = numCPUs; ii<6; ii++)
mvprintw (12 + ii + printw_offset, 0, "\n");
TRUE_CPU_FREQ = 0;
logOpenFile_single();
//time_t time_to_save;
//logCpuFreq_single_d(time(&time_to_save));
clock_gettime(CLOCK_REALTIME, &global_ts);
logCpuFreq_single_ts( &global_ts);
logCpuCstates_single_ts( &global_ts);
for (ii = 0; ii < numCPUs; ii++) {
assert(ii < MAX_SK_PROCESSORS);
i = core_list[ii];
if ( (_FREQ[i] > TRUE_CPU_FREQ) && (print_core[ii]) && !isinf(_FREQ[i]) ) {
TRUE_CPU_FREQ = _FREQ[i];
}
if ( (print_core[ii]) && !isinf(_FREQ[i]) ) {
logCpuFreq_single(_FREQ[i]);
}
logCpuCstates_single_c(" [");
logCpuCstates_single((float)THRESHOLD_BETWEEN_0_100(C0_time[i] * 100)); logCpuCstates_single_c(",");
c1_time = C1_time[i] * 100 - (C3_time[i] + C6_time[i] + C7_time[i]) * 100;
logCpuCstates_single((float)THRESHOLD_BETWEEN_0_100(c1_time)); logCpuCstates_single_c(",");
logCpuCstates_single((float)THRESHOLD_BETWEEN_0_100(C3_time[i] * 100)); logCpuCstates_single_c(",");
logCpuCstates_single((float)THRESHOLD_BETWEEN_0_100(C6_time[i] * 100));
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
logCpuCstates_single_c(",");
logCpuCstates_single((float)THRESHOLD_BETWEEN_0_100(C7_time[i] * 100));
}
logCpuCstates_single_c("]\t");
}
// logCpuCstates_single_c("\n");
logCloseFile_single();
mvprintw (10 + printw_offset, 0,
" Real Current Frequency %0.2f MHz [%0.2f x %0.2f] (Max of below)\n", TRUE_CPU_FREQ, BLCK, TRUE_CPU_FREQ/BLCK);
refresh ();
//shift the new values to the old counter values
//so that the next time we use those to find the difference
memcpy (old_val_CORE, new_val_CORE,
sizeof (*old_val_CORE) * numCPUs);
memcpy (old_val_REF, new_val_REF, sizeof (*old_val_REF) * numCPUs);
memcpy (old_val_C3, new_val_C3, sizeof (*old_val_C3) * numCPUs);
memcpy (old_val_C6, new_val_C6, sizeof (*old_val_C6) * numCPUs);
if(prog_options.i7_version.sandy_bridge || prog_options.i7_version.ivy_bridge || prog_options.i7_version.haswell){
memcpy (old_val_C7, new_val_C7, sizeof (*old_val_C7) * numCPUs);
}
memcpy (tvstart, tvstop, sizeof (*tvstart) * numCPUs);
memcpy (old_TSC, new_TSC, sizeof (*old_TSC) * numCPUs);
} else {
// If all the cores in the socket go offline, just erase the whole screen
//WELL for single socket machine this code will never be executed. lol
//atleast 1 core will be online so ...
//for (ii = 0 ; ii<14; ii++)
// mvprintw (3 + ii + printw_offset, 0, "Ending up here\n");
//print_socket_information(&socket_0);
}
}
int32_t
rte_hash_del_key(const struct rte_hash *h, const void *key)
{
RETURN_IF_TRUE(((h == NULL) || (key == NULL)), -EINVAL);
return __rte_hash_del_key_with_hash(h, key, rte_hash_hash(h, key));
}
