From a51d7b31eb3402cceed85a3ff9044218b01d0dce Mon Sep 17 00:00:00 2001
From: thierry <thierry@home>
Date: Tue, 26 Sep 2023 14:53:54 +0200
Subject: [PATCH] Port to Apple Silicon + uptodate LLVM. Add capture of OpenMP
schedule event
---
includes/kaapi_atomic.h | 44 +++++------
tool/CMakeLists.txt | 6 ++
tool/ompt_callbacks.cpp | 130 ++++++++++++++++++++++++-------
tool/ompt_callbacks.def | 4 +-
tool/tool.cpp | 36 ---------
ukilli/kaapi_trace_reader.cpp | 7 +-
ukilli/kaapi_trace_reader.h | 2 +-
ukilli/kaapi_trace_simulator.cpp | 14 ++--
ukilli/ukilli.cpp | 65 ++++++++--------
9 files changed, 178 insertions(+), 130 deletions(-)
diff --git a/includes/kaapi_atomic.h b/includes/kaapi_atomic.h
index 2fdd0e6..a283f0c 100644
--- a/includes/kaapi_atomic.h
+++ b/includes/kaapi_atomic.h
@@ -62,7 +62,11 @@
#include "kaapi_error.h"
#include <stddef.h>
#include <stdint.h>
-
+#if __cplusplus
+# include <atomic>
+#else
+# include <stdatomic.h>
+#endif
/* ========================= Atomic type ============================= */
/** Atomic type
@@ -246,40 +250,32 @@ static inline void kaapi_slowdown_cpu()
#if defined(__APPLE__)
-# include <libkern/OSAtomic.h>
-static inline void kaapi_writemem_barrier()
+static inline void kaapi_writemem_barrier()
{
-# if defined(__x86_64) || defined(__i386__)
- /* not need sfence on X86 archi: write are ordered __asm__ __volatile__ ("sfence":::"memory"); */
- __asm__ __volatile__ ("":::"memory");
-# else
- OSMemoryBarrier();
-# endif
+#if __cplusplus
+ atomic_thread_fence(std::memory_order_release);
+#else
+ atomic_thread_fence(memory_order_release);
+#endif
}
static inline void kaapi_readmem_barrier()
{
-# if defined(__x86_64) || defined(__i386__)
- __asm__ __volatile__ ("":::"memory");
-// __asm__ __volatile__ ("lfence":::"memory");
-# else
- OSMemoryBarrier();
-# endif
+#if __cplusplus
+ atomic_thread_fence(std::memory_order_acquire);
+#else
+ atomic_thread_fence(memory_order_acquire);
+#endif
}
/* should be both read & write barrier */
static inline void kaapi_mem_barrier()
{
-# if defined(__x86_64) || defined(__i386__)
- /** Mac OS 10.6.8 with gcc 4.2.1 has a buggy __sync_synchronize();
- gcc-4.4.4 pass the test with sync_synchronize
- */
-#if !defined(__MIC__)
- __asm__ __volatile__ ("mfence":::"memory");
+#if __cplusplus
+ atomic_thread_fence(std::memory_order_seq_cst);
+#else
+ atomic_thread_fence(memory_order_seq_cst);
#endif
-# else
- OSMemoryBarrier();
-# endif
}
#elif defined(__linux__)
diff --git a/tool/CMakeLists.txt b/tool/CMakeLists.txt
index 2db4edd..b19e46a 100644
--- a/tool/CMakeLists.txt
+++ b/tool/CMakeLists.txt
@@ -1,3 +1,9 @@
+find_package(OpenMP REQUIRED)
+
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
+
+
unset(OMPT_HEADER CACHE)
find_file(OMPT_HEADER NAMES omp-tools.h HINTS ENV CPLUS_INCLUDE_PATH ENV C_INCLUDE_PATH)
if (${OMPT_HEADER} STREQUAL "OMPT_HEADER-NOTFOUND")
diff --git a/tool/ompt_callbacks.cpp b/tool/ompt_callbacks.cpp
index 82c7018..1caec3a 100644
--- a/tool/ompt_callbacks.cpp
+++ b/tool/ompt_callbacks.cpp
@@ -70,7 +70,7 @@
#include "kaapi_trace_lib.h"
#include "kaapi_error.h"
-#define LOG 0
+#define LOG 1
std::atomic<uint64_t> unique_thread_id(1);
@@ -473,7 +473,7 @@ ompt_callback_task_create_action(
printf("%" PRIu64 ": ompt_task_create: parent_id=<null>, task_id=%" PRIu64\
", type=%i, has_dep=%i, ptr=%" PRIu64 "\n", thread_id,
task->id, type,
- has_dependences, task->task_ptr);
+ has_dependences, (uint64_t)task->task_ptr);
}
#endif
}
@@ -486,11 +486,17 @@ ompt_callback_task_schedule_action(
ompt_data_t *next_task_data)
{
uint64_t thread_id = ompt_get_thread_data()->value;
+ tikki_task_id_t *prior_task =0;
+ if (prior_task_data->ptr) prior_task = (tikki_task_id_t *)prior_task_data->ptr;
+ tikki_task_id_t *next_task = 0;
+ if (next_task_data->ptr) next_task = (tikki_task_id_t *)next_task_data->ptr;
+#if LOG
+ printf("%" PRIu64 ": ompt_callback_task_schedule: prior_task=%" PRIu64 ", status=%i, next_task: %" PRIu64 "\n", thread_id, prior_task->id, prior_task_status, next_task->id);
+#endif
if (prior_task_data->ptr) {
- tikki_task_id_t *prior_task = (tikki_task_id_t *)prior_task_data->ptr;
// We are ending a task
#if LOG
- printf("%" PRIu64 ": ompt_event_task_end: task_id=%" PRIu64 "\n", thread_id, prior_task->id );
+ printf("%" PRIu64 ": ompt_callback_task_schedule-> task_end: task_id=%" PRIu64 "\n", thread_id, prior_task->id );
#endif
/* thread 0 ends the parallel region
But because the end of implicit task is not related to
@@ -514,11 +520,10 @@ ompt_callback_task_schedule_action(
//free(prior_task);
}
if (next_task_data->ptr)
-{
- tikki_task_id_t *new_task = (tikki_task_id_t *)next_task_data->ptr;
+ {
// We are starting a task
#if LOG
- printf("%" PRIu64 ": ompt_event_task_begin: task_id=%" PRIu64 ", status=%i, deps: %p\n", thread_id, new_task->id, prior_task_status, new_task->deps);
+ printf("%" PRIu64 ": ompt_callback_task_schedule-> task_begin: task_id=%" PRIu64 ", status=%i, deps: %p\n", thread_id, next_task->id, prior_task_status, next_task->deps);
#endif
/* This is code for explicit task begin.
*/
@@ -526,28 +531,28 @@ ompt_callback_task_schedule_action(
kaapi_tracelib_thread_switchstate(koti->kproc);
char buff[30];
const char* taskname = 0;
- if (new_task->name == 0)
+ if (next_task->name == 0)
{
- sprintf(buff, "<undef-%p-%i>\0", new_task->task_ptr, new_task->id);
+ snprintf(buff, 30, "<undef-%p-%" PRIu64 ">", next_task->task_ptr, next_task->id);
taskname = buff;
}
else
- taskname = new_task->name;
+ taskname = next_task->name;
kaapi_descrformat_t* fdescr = kaapi_tracelib_register_fmtdescr(
0,
// TODO: get codeptr_ra there
- (void *)new_task->task_ptr,
+ (void *)next_task->task_ptr,
0, //loc
taskname,
libomp_filter_func
);
int idxtop = koti->pstack.top;
koti->pstack.stack[idxtop].fdescr = fdescr;
- koti->pstack.stack[idxtop].task = (void*)new_task->id;
+ koti->pstack.stack[idxtop].task = (void*)next_task->id;
kaapi_tracelib_task_begin(
koti->kproc,
- (kaapi_task_id_t)new_task->id,
+ (kaapi_task_id_t)next_task->id,
fdescr->fmtid,
1,
0, 0, 0,
@@ -556,29 +561,29 @@ ompt_callback_task_schedule_action(
koti->pstack.top ==0 ? 0 : koti->pstack.stack[idxtop-1].accum
);
++koti->pstack.top;
- if (new_task->ndeps > 0) {
+ if (next_task->ndeps > 0) {
kaapi_tracelib_task_access(
koti->kproc,
- (kaapi_task_id_t)new_task->id,
- new_task->ndeps,
- (void*)new_task->deps,
+ (kaapi_task_id_t)next_task->id,
+ next_task->ndeps,
+ (void*)next_task->deps,
0,
(void*)0,
(void (*)(void*, int, void**, size_t*, int*))ompt_decoder
);
}
- if (new_task->datainfo.count)
+ if (next_task->datainfo.count)
{
kaapi_tracelib_task_data(
koti->kproc,
- (kaapi_task_id_t)new_task->id,
- new_task->datainfo.count,
- new_task->datainfo.data,
- new_task->datainfo.size,
- new_task->datainfo.mode,
+ (kaapi_task_id_t)next_task->id,
+ next_task->datainfo.count,
+ next_task->datainfo.data,
+ next_task->datainfo.size,
+ next_task->datainfo.mode,
(void (*)(int, int*))ompt_mode_decoder
);
- new_task->datainfo.count = 0;
+ next_task->datainfo.count = 0;
}
}
}
@@ -594,7 +599,8 @@ ompt_callback_implicit_task_action (
)
{
uint64_t thread_id = ompt_get_thread_data()->value;
- if (endpoint == ompt_scope_begin) {
+ if (endpoint == ompt_scope_begin)
+ {
tikki_task_id_t *task = (tikki_task_id_t *)malloc(sizeof(tikki_task_id_t));
task->id = ompt_get_unique_id();
task->task_ptr = 0;
@@ -604,7 +610,7 @@ ompt_callback_implicit_task_action (
printf("%" PRIu64 ": ompt_event_implicit_task_action: begin. parallel_id=%" PRIu64 ", task_id=%" PRIu64 "\n", thread_id, parallel_data->value, task->id);
#endif
char buff[30];
- sprintf(buff, "<implicit>") ;
+ snprintf(buff, 30, "<implicit>") ;
kaapi_ompt_thread_info_t* koti = &__kaapi_oth_info[thread_id];
kaapi_descrformat_t* fdescr = kaapi_tracelib_register_fmtdescr(
1,
@@ -651,7 +657,7 @@ ompt_callback_implicit_task_action (
);
//free(task);
} else {
- fprintf(stderr, "%" PRIu64 ": ompt_event_implicit_task_action: unknown endpoint. task_id=%" PRIu64 "\n", thread_id, task_data->ptr);
+ fprintf(stderr, "%" PRIu64 ": ompt_event_implicit_task_action: unknown endpoint. task_id=%" PRIu64 "\n", thread_id, (uint64_t)task_data->ptr);
}
/* This is code for implicit task begin.
*/
@@ -727,3 +733,73 @@ ompt_callback_task_dependence_action (
#endif
}
+void ompt_callback_sync_region_action (
+ ompt_sync_region_t kind,
+ ompt_scope_endpoint_t endpoint,
+ ompt_data_t *parallel_data,
+ ompt_data_t *task_data,
+ const void *codeptr_ra
+)
+{
+#if LOG
+ uint64_t thread_id = ompt_get_thread_data()->value;
+ tikki_task_id_t *task = (tikki_task_id_t *)task_data->ptr;
+ printf("%" PRIu64 ": ompt_callback_sync_region: ", thread_id );
+ switch (endpoint) {
+ case ompt_scope_begin:
+ printf("BEGIN of kind <");
+ break;
+
+ case ompt_scope_end:
+ printf("END of kind <");
+ break;
+
+ case ompt_scope_beginend:
+ printf("BEGINEND of kind <");
+ break;
+ }
+
+ switch (kind) {
+ case ompt_sync_region_barrier:
+ printf("barrier");
+ break;
+
+ case ompt_sync_region_barrier_implicit:
+ printf("barrier implicit");
+ break;
+
+ case ompt_sync_region_barrier_explicit:
+ printf("barrier explicit");
+ break;
+
+ case ompt_sync_region_barrier_implicit_workshare:
+ printf("barrier implicit workshare");
+ break;
+
+ case ompt_sync_region_barrier_implicit_parallel:
+ printf("barrier implicit parallel");
+ break;
+
+ case ompt_sync_region_barrier_teams:
+ printf("barrier teams");
+ break;
+
+ case ompt_sync_region_barrier_implementation:
+ printf("barrier implementation");
+ break;
+
+ case ompt_sync_region_taskwait:
+ printf("taskwait");
+ break;
+
+ case ompt_sync_region_taskgroup:
+ printf("taskgroup");
+ break;
+
+ case ompt_sync_region_reduction:
+ printf("reduction");
+ break;
+ };
+ printf(">, parallel_data:%p, task_data: %p, task_id:%" PRIu64 ", codeptr_ra: %p\n", parallel_data, task_data, task->id, codeptr_ra);
+#endif
+}
diff --git a/tool/ompt_callbacks.def b/tool/ompt_callbacks.def
index e4b2b71..afc18ea 100644
--- a/tool/ompt_callbacks.def
+++ b/tool/ompt_callbacks.def
@@ -14,6 +14,9 @@ CALLBACK(ompt_callback_task_schedule, ompt_data_t *, ompt_task_status_t, ompt_da
CALLBACK(ompt_callback_implicit_task, ompt_scope_endpoint_t, ompt_data_t *, ompt_data_t*, unsigned int, unsigned int, int)
CALLBACK(ompt_callback_dependences, ompt_data_t *, const ompt_dependence_t *, int)
CALLBACK(ompt_callback_task_dependence, ompt_data_t *, ompt_data_t * )
+CALLBACK(ompt_callback_sync_region, ompt_sync_region_t kind, ompt_scope_endpoint_t endpoint, ompt_data_t *parallel_data, ompt_data_t *task_data, const void *codeptr_ra )
+CALLBACK(ompt_callback_sync_region, ompt_sync_region_t, ompt_scope_endpoint_t, ompt_data_t *, ompt_data_t *, const void * )
+
UNUSED(ompt_callback_target)
UNUSED(ompt_callback_target_data_op)
UNUSED(ompt_callback_target_submit)
@@ -27,7 +30,6 @@ UNUSED(ompt_callback_mutex_released)
UNUSED(ompt_callback_work)
UNUSED(ompt_callback_master)
UNUSED(ompt_callback_target_map)
-UNUSED(ompt_callback_sync_region)
UNUSED(ompt_callback_lock_init)
UNUSED(ompt_callback_lock_destroy)
UNUSED(ompt_callback_mutex_acquire)
diff --git a/tool/tool.cpp b/tool/tool.cpp
index 1f68149..41e2c57 100644
--- a/tool/tool.cpp
+++ b/tool/tool.cpp
@@ -107,39 +107,3 @@ extern "C" {
}
}
-
-
-#if 0
-struct F {
- void operator() ( ... )
- {
- Inst0;
- Fork<F1>() ( e1 );
- Inst1;
- Fork<F2>() ( e2 );
- Inst2;
- ...
- Fork<Fn>() ( en );
- Instn;
- }
-};
-
-struct F {
- void operator() ( ... )
- {
- Inst0;
- e1 = eval(e1);
- Inst1;
- e2 = eval(e2);
- Inst2;
- ...
- en= eval(en);
- Instn;
-
- Fork<F1>() ( e1 );
- Fork<F2>() ( e2 );
- ...
- Fork<Fn>() ( en );
- }
-};
-#endif
diff --git a/ukilli/kaapi_trace_reader.cpp b/ukilli/kaapi_trace_reader.cpp
index 77c8cb1..f30bc67 100644
--- a/ukilli/kaapi_trace_reader.cpp
+++ b/ukilli/kaapi_trace_reader.cpp
@@ -62,9 +62,10 @@
#include "kaapi_trace_reader.h"
/* Reader for one file */
+#define FILE_EVENT_NAME_SIZE 128
struct file_event {
int fd;
- char name[128]; /* container name */
+ char name[FILE_EVENT_NAME_SIZE]; /* container name */
kaapi_eventfile_header_t* header; /* pointer to the header of the file, at least format v1, may be new format */
kaapi_event_t* base; /* base for event */
size_t rpos; /* next position to read event */
@@ -267,7 +268,7 @@ std::cout << "In vector = " << fdset->perfcounter_name[kk] << std::endl;
int ReadFiles(
FileSet* fdset,
void* arg,
- void (*callback)( void* arg, char* name, const kaapi_event_t* event)
+ void (*callback)( void* arg, char* name, int size, const kaapi_event_t* event)
)
{
if (callback ==0) return EINVAL;
@@ -281,7 +282,7 @@ int ReadFiles(
/* The container name is passed in/out: first event can initialize them */
- callback( arg, fe->name, &fe->base[fe->rpos++] );
+ callback( arg, fe->name, FILE_EVENT_NAME_SIZE, &fe->base[fe->rpos++] );
if (fe->rpos < fe->end)
fdset->eventqueue.push( next_event_t(fe->base[fe->rpos].date, ne.fds) );
diff --git a/ukilli/kaapi_trace_reader.h b/ukilli/kaapi_trace_reader.h
index 218b7e1..59dadcb 100644
--- a/ukilli/kaapi_trace_reader.h
+++ b/ukilli/kaapi_trace_reader.h
@@ -93,7 +93,7 @@ extern const char** GetPerfCounterName(struct FileSet* fdset, int* size );
event.
Return 0 in case of success.
*/
-extern int ReadFiles(struct FileSet* fdset, void* arg, void (*callback)( void* arg, char* name, const kaapi_event_t* event) );
+extern int ReadFiles(struct FileSet* fdset, void* arg, void (*callback)( void* arg, char* name, int sz, const kaapi_event_t* event) );
/* Return true iff no more event
*/
diff --git a/ukilli/kaapi_trace_simulator.cpp b/ukilli/kaapi_trace_simulator.cpp
index 8773a58..828d2c1 100644
--- a/ukilli/kaapi_trace_simulator.cpp
+++ b/ukilli/kaapi_trace_simulator.cpp
@@ -251,8 +251,8 @@ Simulator* OpenSimulator( FileSet* fds,
char* buffer = comment;
/* write format for the file */
char unitdate[15];
- sprintf(unitdate,"date(%s)",header.event_date_unit);
- int c = sprintf(buffer, "%5s %12s %12s %8s", "kid", unitdate, "duration", "task_id");
+ snprintf(unitdate, 15, "date(%s)",header.event_date_unit);
+ int c = snprintf(buffer, size, "%5s %12s %12s %8s", "kid", unitdate, "duration", "task_id");
buffer += c;
int i, count_perfcounter = header.perfcounter_count & 0xFF;
@@ -265,7 +265,7 @@ Simulator* OpenSimulator( FileSet* fds,
kaapi_assert( comment - buffer < size );
}
}
- c = sprintf(buffer, "\n");
+ c = snprintf(buffer, comment+size-buffer, "\n");
buffer += c;
kaapi_assert( comment - buffer < size );
write(sim->fd_taskdump, comment, strlen(comment));
@@ -670,7 +670,7 @@ static void processor_simulate_event(
{
static char tmp[2048];
char* buffer = tmp;
- int c = sprintf(buffer, "%5lu %12llu %12lld %8u",
+ int c = snprintf(buffer, 2048, "%5lu %12llu %12lld %8u",
(unsigned long)event->kid, /* kid */
(unsigned long long) proc->stacktask[proc->sp].date, /* start date */
(long long) (event->date - proc->stacktask[proc->sp].date), /* duration */
@@ -687,14 +687,14 @@ static void processor_simulate_event(
kaapi_perf_idset_clear(&proc->lasttask_perfctr_mask, idx );
//printf("PerfIdx: %i ->%s\n", idx, kaapi_perfctr_info[idx].name );
if (kaapi_perfctr_info[idx].ns2s)
- c = sprintf(buffer, " %16e", 1e-9*(double)proc->lasttask_perfctr[idx] );
+ c = snprintf(buffer, tmp+2048-buffer, " %16e", 1e-9*(double)proc->lasttask_perfctr[idx] );
else
- c = sprintf(buffer, " %16" PRIi64, proc->lasttask_perfctr[idx]);
+ c = snprintf(buffer, tmp+2048-buffer, " %16" PRIi64, proc->lasttask_perfctr[idx]);
buffer += c;
kaapi_assert( buffer <= tmp + 256);
}
}
- c = sprintf(buffer, "\n");
+ c = snprintf(buffer, tmp+2048-buffer, "\n");
write(proc->sim->fd_taskdump,tmp, strlen(tmp));
}
kaapi_perf_idset_zero(&proc->lasttask_perfctr_mask);
diff --git a/ukilli/ukilli.cpp b/ukilli/ukilli.cpp
index b57d3cc..6909646 100644
--- a/ukilli/ukilli.cpp
+++ b/ukilli/ukilli.cpp
@@ -238,6 +238,7 @@ static inline const char* int2affinitykind(uint8_t v)
static void callback_print_event(
void* context,
char* container_name,
+ int sz_name,
const kaapi_event_t* event
)
{
@@ -727,7 +728,7 @@ static void callback_display_paje_event_gpuevent(
case KAAPI_EVT_OFFLOAD_DTOH_BEG:
if(katracereader_options.gputransfer){
kid = event->kid;
- sprintf(tmp,"d2h-%i",kid);
+ snprintf(tmp,128,"d2h-%i",kid);
kaapi_trace_poti_PushState (d0, tmp, "STATE", "kd2h");
}
break;
@@ -735,7 +736,7 @@ static void callback_display_paje_event_gpuevent(
case KAAPI_EVT_OFFLOAD_DTOH_END:
if(katracereader_options.gputransfer){
kid = event->kid;
- sprintf(tmp,"d2h-%i",kid);
+ snprintf(tmp,128,"d2h-%i",kid);
kaapi_trace_poti_PopState (d0, tmp, "STATE");
}
break;
@@ -743,7 +744,7 @@ static void callback_display_paje_event_gpuevent(
case KAAPI_EVT_OFFLOAD_HTOD_BEG:
if(katracereader_options.gputransfer){
kid = event->kid;
- sprintf(tmp,"h2d-%i",kid);
+ snprintf(tmp,128,"h2d-%i",kid);
kaapi_trace_poti_PushState (d0, tmp, "STATE", "kh2d");
}
break;
@@ -751,20 +752,20 @@ static void callback_display_paje_event_gpuevent(
case KAAPI_EVT_OFFLOAD_HTOD_END:
if(katracereader_options.gputransfer){
kid = event->kid;
- sprintf(tmp,"h2d-%i",kid);
+ snprintf(tmp,128,"h2d-%i",kid);
kaapi_trace_poti_PopState (d0, tmp, "STATE");
}
break;
case KAAPI_EVT_OFFLOAD_KERNEL_BEG:
kid = event->kid;
- sprintf(tmp,"gpu-%i",kid);
+ snprintf(tmp,128,"gpu-%i",kid);
kaapi_trace_poti_PushState (d0, tmp, "STATE", "kgpu");
break;
case KAAPI_EVT_OFFLOAD_KERNEL_END:
kid = event->kid;
- sprintf(tmp,"gpu-%i",kid);
+ snprintf(tmp,128,"gpu-%i",kid);
kaapi_trace_poti_PopState (d0, tmp, "STATE");
break;
@@ -801,14 +802,14 @@ static void callback_display_paje_event_stealevent(
case KAAPI_EVT_REQUEST_BEG:
kid = (unsigned int)KAAPI_EVENT_DATA(event,0,i);
serial = (int)KAAPI_EVENT_DATA(event,1,i);
- sprintf(tmp,"thief-%i",kid);
+ snprintf(tmp,128,"thief-%i",kid);
kaapi_trace_poti_PushState (d0, tmp, "STEAL", "r");
break;
case KAAPI_EVT_REQUEST_END:
kid = (unsigned int)KAAPI_EVENT_DATA(event,0,i);
serial = (int)KAAPI_EVENT_DATA(event,1,i);
- sprintf(tmp,"thief-%i",kid);
+ snprintf(tmp,128,"thief-%i",kid);
kaapi_trace_poti_PopState (d0, tmp, "STEAL");
break;
@@ -882,6 +883,7 @@ static void callback_display_paje_event_stealevent(
static void callback_display_paje_event(
void* context,
char* name,
+ int sz_name,
const kaapi_event_t* event
)
{
@@ -898,9 +900,9 @@ static void callback_display_paje_event(
case KAAPI_EVT_KPROC_START:
if (gantt_container_created.find( kid ) == gantt_container_created.end())
{
- sprintf(name,"thread-%i",kid);
+ snprintf(name,sz_name,"thread-%i",kid);
char rootname[64];
- sprintf(rootname,"NUMA-%i", (int)KAAPI_EVENT_DATA(event,1,i));
+ snprintf(rootname,64,"NUMA-%i", (int)KAAPI_EVENT_DATA(event,1,i));
//printf("NUMA %i %i %lu\n", kid,(int)KAAPI_EVENT_DATA(event,1,i), (unsigned long)KAAPI_EVENT_DATA(event,1,u));
kaapi_trace_poti_CreateContainer (d0, name, "THREAD", rootname, name);
gantt_container_created.insert( std::make_pair( kid, event->date ) );
@@ -937,7 +939,7 @@ static void callback_display_paje_event(
if (KAAPI_EVENT_DATA(event,1,u) !=0)
{
char stateid[64];
- sprintf(stateid, "t%lu", (unsigned long)KAAPI_EVENT_DATA(event,1,u));
+ snprintf(stateid, 64, "t%lu", (unsigned long)KAAPI_EVENT_DATA(event,1,u));
kaapi_trace_poti_PushState (d0, name, "STATE", stateid);
}
else
@@ -1195,7 +1197,7 @@ static int fnc_paje_gantt_header(kaapi_eventfile_header_t* header)
for (i=0; i<header->taskfmt_count;++i)
{
char tmpident[64];
- sprintf(tmpident, "t%" PRIu64, header->fmtdefs[i].fmtid );
+ snprintf(tmpident, 64, "t%" PRIu64, header->fmtdefs[i].fmtid );
kaapi_trace_poti_DefineEntityValue(tmpident, "STATE",
convert2paje(header->fmtdefs[i].name),
goodcolor[i%MAX_COLOR]/*header->fmtdefs[i].color*/);
@@ -1259,7 +1261,7 @@ static int fnc_paje_gantt_header(kaapi_eventfile_header_t* header)
for (int i=0; i<header->numacount; ++i)
{
char tmpident[64];
- sprintf(tmpident, "NUMA-%i", i );
+ snprintf(tmpident, 64, "NUMA-%i", i );
kaapi_trace_poti_CreateContainer (0.00, tmpident, "ROOT", "0", tmpident);
}
@@ -1274,7 +1276,7 @@ static int fnc_paje_gantt_close(kaapi_eventfile_header_t* header, uint64_t gantt
while (ibeg != iend)
{
char name[64];
- sprintf(name, "thread-%i", (int)ibeg->first );
+ snprintf(name, 64, "thread-%i", (int)ibeg->first );
kaapi_trace_poti_DestroyContainer (1e-9*(double)(ibeg->second-gantt_tmin), "THREAD", name);
++ibeg;
}
@@ -1648,6 +1650,7 @@ rastello_parallel_region_t* rastello_parallel_region_factory_t::create(uintptr_t
static void callback_display_rastello(
void* context,
char* name,
+ int sz_name,
const kaapi_event_t* event
)
{
@@ -2144,7 +2147,7 @@ printf("pred[id:%i/%p] = {id:%i}/%p\n", succ->id, (void*)succ->addr, task->id, (
*/
int rastello_parallel_region_t::openfile(kaapi_eventfile_header_t* header)
{
- sprintf(filename, "rastello_%i.c", (int)parallel_id);
+ snprintf(filename, 128, "rastello_%i.c", (int)parallel_id);
fout = fopen(filename,"w");
if (fout ==0)
{
@@ -2287,7 +2290,7 @@ parallel_region_t* dot_parallel_region_factory_t::create(uintptr_t prid)
int dot_parallel_region_t::openfile(kaapi_eventfile_header_t* header)
{
FILE* file = 0;
- sprintf(filename, "graph_%i.dot", (int)parallel_id);
+ snprintf(filename, 128, "graph_%i.dot", (int)parallel_id);
fout = file = fopen(filename,"w");
if (file ==0)
{
@@ -2538,7 +2541,7 @@ int csv_parallel_region_t::openfile(kaapi_eventfile_header_t* header)
if (csv_parallel_region_t::fout_pr ==0)
{
char filename[128];
- sprintf(filename, "parallels.csv");
+ snprintf(filename, 128, "parallels.csv");
csv_parallel_region_t::fout_pr = fopen(filename,"w");
if (csv_parallel_region_t::fout_pr ==0)
{
@@ -2572,7 +2575,7 @@ int csv_parallel_region_t::closefile(int cpucount)
if (csv_parallel_region_t::fout ==0)
{
char filename[128];
- sprintf(filename, "tasks.csv");
+ snprintf(filename, 128, "tasks.csv");
csv_parallel_region_t::fout = fopen(filename,"w");
if (csv_parallel_region_t::fout ==0)
{
@@ -2613,7 +2616,7 @@ int csv_parallel_region_t::closefile(int cpucount)
if (csv_parallel_region_t::fout_complexity ==0)
{
char filename[128];
- sprintf(filename, "complexity.csv");
+ snprintf(filename, 128, "complexity.csv");
csv_parallel_region_t::fout_complexity = fopen(filename,"w");
if (csv_parallel_region_t::fout_complexity ==0)
{
@@ -2672,7 +2675,7 @@ int csv_parallel_region_t::closefile(int cpucount)
if (csv_parallel_region_t::fout_threads ==0)
{
char filename[128];
- sprintf(filename, "threads.csv");
+ snprintf(filename, 128, "threads.csv");
csv_parallel_region_t::fout_threads = fopen(filename,"w");
if (csv_parallel_region_t::fout_threads ==0)
{
@@ -2717,7 +2720,7 @@ int csv_parallel_region_t::closefile(int cpucount)
if (csv_parallel_region_t::fout_loops ==0)
{
char filename[128];
- sprintf(filename, "loops.csv");
+ snprintf(filename, 128, "loops.csv");
csv_parallel_region_t::fout_loops = fopen(filename,"w");
if (csv_parallel_region_t::fout_loops ==0)
{
@@ -2746,7 +2749,7 @@ int csv_parallel_region_t::closefile(int cpucount)
if (csv_parallel_region_t::fout_memory ==0)
{
char filename[128];
- sprintf(filename, "memory.csv");
+ snprintf(filename, 128, "memory.csv");
csv_parallel_region_t::fout_memory = fopen(filename,"w");
if (csv_parallel_region_t::fout_memory ==0)
{
@@ -2772,7 +2775,7 @@ int csv_parallel_region_t::closefile(int cpucount)
if (csv_parallel_region_t::fout_counter ==0)
{
char filename[128];
- sprintf(filename, "counters.csv");
+ snprintf(filename, 128, "counters.csv");
csv_parallel_region_t::fout_counter = fopen(filename,"w");
if (csv_parallel_region_t::fout_counter ==0)
{
@@ -2798,7 +2801,7 @@ int csv_parallel_region_t::closefile(int cpucount)
fclose(csv_parallel_region_t::fout_counter);
csv_parallel_region_t::fout_counter = 0;
- sprintf(filename, "uncoreperfctr.csv");
+ snprintf(filename, 128, "uncoreperfctr.csv");
csv_parallel_region_t::fout_counter = fopen(filename,"w");
if (csv_parallel_region_t::fout_counter ==0)
{
@@ -2988,7 +2991,7 @@ parallel_region_t* somp_parallel_region_factory_t::create(uintptr_t prid)
int somp_parallel_region_t::openfile(kaapi_eventfile_header_t* header)
{
FILE* file = 0;
- sprintf(filename, "trace_%i.rec", (int)parallel_id);
+ snprintf(filename, 128, "trace_%i.rec", (int)parallel_id);
fout = file = fopen(filename,"w");
if (file ==0)
{
@@ -3055,10 +3058,10 @@ void somp_parallel_region_t::dump_task_info( task_info* ti )
fprintf(stdout, "cannot find fmtid: %" PRIu64 "\n", ti->fmtid);
fprintf(fout,"Name: %s\n"
- "JobId: %li\n"
+ "JobId: %" PRIu64 "\n"
"StartTime: %f\n"
"EndTime: %f\n"
- "MemoryNode: %li\n",
+ "MemoryNode: %i\n",
(fmt == rastello_fmtname.end() ? "<unknown>" : fmt->second ),
ti->id,
(double)ti->start*1e-6 /* ns -> millis */,
@@ -3071,7 +3074,7 @@ void somp_parallel_region_t::dump_task_info( task_info* ti )
for (int i=0; i<ti->param.size(); ++i)
{
uint64_t ptr = ti->param[i].ptr;
- fprintf(fout," %p", ptr);
+ fprintf(fout," %p", (void*)ptr);
}
fprintf(fout,"\n");
@@ -3104,7 +3107,7 @@ void somp_parallel_region_t::dump_task_info( task_info* ti )
for (int i=0; i<ti->data.size(); ++i)
{
uint64_t ptr = ti->data[i].ptr;
- fprintf(fout," %p", ptr);
+ fprintf(fout," %p", (void*)ptr);
}
fprintf(fout,"\n");
@@ -3113,7 +3116,7 @@ void somp_parallel_region_t::dump_task_info( task_info* ti )
for (int i=0; i<ti->data.size(); ++i)
{
uint64_t ptr = ti->data[i].size;
- fprintf(fout," %li", ptr);
+ fprintf(fout," %" PRIu64, ptr);
}
fprintf(fout,"\n");
@@ -3140,7 +3143,7 @@ void somp_parallel_region_t::dump_task_info( task_info* ti )
fprintf(fout,"DependsOn:");
for (int i=0; i<ti->pred.size(); ++i)
{
- fprintf(fout," %li", ti->pred[i]->id);
+ fprintf(fout," %" PRIu64, ti->pred[i]->id);
}
fprintf(fout,"\n\n");
}
--
GitLab