Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

hexagon: support for op-trace (fine-grain tracing of HVX/HMX/DMA events) (#24592)

Browse Source 
* hex-optrace: add support for optrace and instrument matmul and flash-atten code

* hex-trace: improve trace event and prefetto generator

* hex-trace: add new script dedicated to handling traces, specifically perfetto traces

* hex-trace: add --head/--tail options to profile and trace tools

* hex-trace: fix whitespaces

* hex-trace: fix flake8 warnings

* hex-trace: fix flake8 warnings

* hmx-fa: restore q_tiles clearing

* hex-profile: remove circular dep in includes

* hex-trace: simplify trace sizing check

* hex-profile: sort events in the summary by name
This commit is contained in:
Max Krasnyansky
2026-06-18 08:35:02 -07:00
committed by GitHub
parent 7b6c5a2aed
commit d2c67959b3
17 changed files with 1121 additions and 224 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-hexagon/htp/hex-profile.h
+64
View File
@@ -0,0 +1,64 @@
#ifndef HEX_PROFILE_H
#define HEX_PROFILE_H
#include <stdbool.h>
#include <stdint.h>
#include <qurt.h>
#include "hex-utils.h"
#include "htp-ops.h"
#define HTP_TRACE_EVT_START 0
#define HTP_TRACE_EVT_STOP 1
#ifndef HEX_NUM_PMU_COUNTERS
#define HEX_NUM_PMU_COUNTERS 8
#endif
static inline void hex_get_pmu(uint32_t counters[]) {
#if __HVX_ARCH__ >= 79
asm volatile("%0 = upmucnt0" : "=r"(counters[0]));
asm volatile("%0 = upmucnt1" : "=r"(counters[1]));
asm volatile("%0 = upmucnt2" : "=r"(counters[2]));
asm volatile("%0 = upmucnt3" : "=r"(counters[3]));
asm volatile("%0 = upmucnt4" : "=r"(counters[4]));
asm volatile("%0 = upmucnt5" : "=r"(counters[5]));
asm volatile("%0 = upmucnt6" : "=r"(counters[6]));
asm volatile("%0 = upmucnt7" : "=r"(counters[7]));
#else
counters[0] = qurt_pmu_get(QURT_PMUCNT0);
counters[1] = qurt_pmu_get(QURT_PMUCNT1);
counters[2] = qurt_pmu_get(QURT_PMUCNT2);
counters[3] = qurt_pmu_get(QURT_PMUCNT3);
counters[4] = qurt_pmu_get(QURT_PMUCNT4);
counters[5] = qurt_pmu_get(QURT_PMUCNT5);
counters[6] = qurt_pmu_get(QURT_PMUCNT6);
counters[7] = qurt_pmu_get(QURT_PMUCNT7);
#endif
}
struct htp_thread_trace {
uint32_t count;
uint32_t max_events;
struct htp_trace_desc * events;
};
static inline void htp_trace_event(struct htp_thread_trace * tr, uint16_t id, uint16_t info, uint32_t type) {
if (tr && tr->events && tr->count < tr->max_events) {
uint32_t idx = tr->count;
tr->events[idx].id = id;
tr->events[idx].info = info | (type == HTP_TRACE_EVT_STOP ? 0x8000 : 0);
tr->events[idx].cycles = (uint32_t) hex_get_cycles();
tr->count++;
}
}
static inline void htp_trace_event_start(struct htp_thread_trace * tr, uint16_t id, uint16_t info) {
htp_trace_event(tr, id, info, HTP_TRACE_EVT_START);
}
static inline void htp_trace_event_stop(struct htp_thread_trace * tr, uint16_t id, uint16_t info) {
htp_trace_event(tr, id, info, HTP_TRACE_EVT_STOP);
}
#endif /* HEX_PROFILE_H */