Skip to content

Commit cb7d9cb

Browse files
RbiessyRbiessy
committed
Add more prints
1 parent 986e89a commit cb7d9cb

File tree

3 files changed

+48
-11
lines changed

3 files changed

+48
-11
lines changed

‎ggml/src/ggml-sycl/common.hpp‎

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -506,7 +506,7 @@ void debug_print_array(const std::string& prefix, const T array[N]) {
506506
GGML_SYCL_DEBUG("%s", ss.str().c_str());
507507
}
508508

509-
inline void debug_print_tensor(const std::string& prefix, const ggml_tensor* tensor) {
509+
inline void debug_print_tensor(const std::string& prefix, const ggml_tensor* tensor, const std::string& suffix = "") {
510510
GGML_SYCL_DEBUG("%s=", prefix.c_str());
511511
if (tensor) {
512512
GGML_SYCL_DEBUG("'%s':type=%s", tensor->name, ggml_type_name(tensor->type));
@@ -521,14 +521,15 @@ inline void debug_print_tensor(const std::string& prefix, const ggml_tensor* ten
521521
} else {
522522
GGML_SYCL_DEBUG("nullptr");
523523
}
524+
GGML_SYCL_DEBUG("%s", suffix.c_str());
524525
}
525526

526527
struct scope_op_debug_print {
527528
scope_op_debug_print(const std::string& func, const ggml_tensor* dst, std::size_t num_src, const std::string& suffix = "") : func(func) {
528529
if (!g_ggml_sycl_debug) {
529530
return;
530531
}
531-
GGML_SYCL_DEBUG("call %s:", func.c_str());
532+
GGML_SYCL_DEBUG("[SYCL][OP] call %s:", func.c_str());
532533
debug_print_tensor(" dst", dst);
533534
if (dst) {
534535
for (std::size_t i = 0; i < num_src; ++i) {
@@ -539,7 +540,7 @@ struct scope_op_debug_print {
539540
}
540541

541542
~scope_op_debug_print() {
542-
GGML_SYCL_DEBUG("call %s done\n", func.c_str());
543+
GGML_SYCL_DEBUG("[SYCL][OP] call %s done\n", func.c_str());
543544
}
544545

545546
private:

‎ggml/src/ggml-sycl/dmmv.cpp‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1092,7 +1092,8 @@ void ggml_sycl_op_dequantize_mul_mat_vec(
10921092
src0->type == GGML_TYPE_Q8_0 || src0->type == GGML_TYPE_F16;
10931093

10941094
if (src1_convert_f16) {
1095-
scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/2, " : converting src1 to fp16");
1095+
scope_op_debug_print scope_dbg_print(std::string(__func__) + "to_fp16_sycl", dst, /*num_src=*/2,
1096+
" : converting src1 to fp16");
10961097
src1_dfloat = src1_dfloat_a.alloc(ne00);
10971098
const to_fp16_sycl_t to_fp16_sycl = ggml_get_to_fp16_sycl(src1->type, dst);
10981099
GGML_ASSERT(to_fp16_sycl != nullptr);

‎ggml/src/ggml-sycl/ggml-sycl.cpp‎

Lines changed: 42 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -346,6 +346,8 @@ static void * ggml_backend_sycl_buffer_get_base(ggml_backend_buffer_t buffer) {
346346
static enum ggml_status
347347
ggml_backend_sycl_buffer_init_tensor(ggml_backend_buffer_t buffer,
348348
ggml_tensor *tensor) try {
349+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
350+
debug_print_tensor(": tensor=", tensor, "\n");
349351
ggml_backend_sycl_buffer_context * ctx = (ggml_backend_sycl_buffer_context *)buffer->context;
350352

351353
if (tensor->view_src != NULL) {
@@ -381,7 +383,9 @@ static void ggml_backend_sycl_buffer_set_tensor(ggml_backend_buffer_t buffer,
381383
ggml_tensor *tensor,
382384
const void *data, size_t offset,
383385
size_t size) try {
384-
386+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
387+
debug_print_tensor(": tensor=", tensor);
388+
GGML_SYCL_DEBUG(" size=%zu offset=%zu\n", size, offset);
385389
ggml_backend_sycl_buffer_context * ctx = ( ggml_backend_sycl_buffer_context *)buffer->context;
386390
ggml_sycl_set_device(ctx->device);
387391
auto stream = &(dpct::dev_mgr::instance().get_device(ctx->device).default_queue());
@@ -406,7 +410,9 @@ static void ggml_backend_sycl_buffer_get_tensor(ggml_backend_buffer_t buffer,
406410
const ggml_tensor *tensor,
407411
void *data, size_t offset,
408412
size_t size) try {
409-
413+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
414+
debug_print_tensor(": tensor=", tensor);
415+
GGML_SYCL_DEBUG(" size=%zu offset=%zu\n", size, offset);
410416
ggml_backend_sycl_buffer_context * ctx = ( ggml_backend_sycl_buffer_context *)buffer->context;
411417

412418
ggml_sycl_set_device(ctx->device);
@@ -434,7 +440,12 @@ static bool
434440
ggml_backend_sycl_buffer_cpy_tensor(ggml_backend_buffer_t buffer,
435441
const ggml_tensor *src,
436442
ggml_tensor *dst) try {
437-
if (ggml_backend_buffer_is_sycl(src->buffer)) {
443+
bool is_cpy_supported = ggml_backend_buffer_is_sycl(src->buffer);
444+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
445+
debug_print_tensor(": dst=", dst);
446+
debug_print_tensor(" src=", src);
447+
GGML_SYCL_DEBUG(" is_cpy_supported=%d\n", is_cpy_supported);
448+
if (is_cpy_supported) {
438449
ggml_backend_sycl_buffer_context * src_ctx = (ggml_backend_sycl_buffer_context *)src->buffer->context;
439450
ggml_backend_sycl_buffer_context * dst_ctx = (ggml_backend_sycl_buffer_context *)dst->buffer->context;
440451

@@ -491,7 +502,8 @@ ggml_backend_sycl_buffer_cpy_tensor(ggml_backend_buffer_t buffer,
491502

492503
static void ggml_backend_sycl_buffer_clear(ggml_backend_buffer_t buffer,
493504
uint8_t value) try {
494-
ggml_backend_sycl_buffer_context * ctx = ( ggml_backend_sycl_buffer_context *)buffer->context;
505+
GGML_SYCL_DEBUG("[SYCL] call %s: size=%zu\n", __func__, buffer->size);
506+
ggml_backend_sycl_buffer_context * ctx = ( ggml_backend_sycl_buffer_context *)buffer->context;
495507

496508
ggml_sycl_set_device(ctx->device);
497509
queue_ptr stream = ctx->stream;
@@ -510,7 +522,9 @@ catch (sycl::exception const &exc) {
510522

511523
static void ggml_backend_sycl_buffer_memset_tensor(ggml_backend_buffer_t buffer, ggml_tensor * tensor, uint8_t value,
512524
size_t offset, size_t size) {
513-
GGML_SYCL_DEBUG(" [SYCL] call %s\n", __func__);
525+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
526+
debug_print_tensor(": tensor=", tensor);
527+
GGML_SYCL_DEBUG(" size=%zu offset=%zu value=%u\n", size, offset, value);
514528
ggml_backend_sycl_buffer_context * ctx = (ggml_backend_sycl_buffer_context *) buffer->context;
515529
SYCL_CHECK(ggml_sycl_set_device(ctx->device));
516530
auto stream = &(dpct::dev_mgr::instance().get_device(ctx->device).default_queue());
@@ -788,6 +802,8 @@ static void * ggml_backend_sycl_split_buffer_get_base(ggml_backend_buffer_t buff
788802
static enum ggml_status
789803
ggml_backend_sycl_split_buffer_init_tensor(ggml_backend_buffer_t buffer,
790804
ggml_tensor *tensor) try {
805+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
806+
debug_print_tensor(": tensor=", tensor, "\n");
791807
GGML_ASSERT(tensor->view_src == nullptr); // views of split tensors are not supported
792808

793809
ggml_backend_sycl_split_buffer_context * ctx = (ggml_backend_sycl_split_buffer_context *)buffer->context;
@@ -872,6 +888,9 @@ static void
872888
ggml_backend_sycl_split_buffer_set_tensor(ggml_backend_buffer_t buffer,
873889
ggml_tensor *tensor, const void *data,
874890
size_t offset, size_t size) try {
891+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
892+
debug_print_tensor(": tensor=", tensor);
893+
GGML_SYCL_DEBUG(" size=%zu offset=%zu\n", size, offset);
875894
// split tensors must always be set in their entirety at once
876895
GGML_ASSERT(offset == 0);
877896
GGML_ASSERT(size == ggml_nbytes(tensor));
@@ -925,6 +944,9 @@ static void
925944
ggml_backend_sycl_split_buffer_get_tensor(ggml_backend_buffer_t buffer,
926945
const ggml_tensor *tensor, void *data,
927946
size_t offset, size_t size) try {
947+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
948+
debug_print_tensor(": tensor=", tensor);
949+
GGML_SYCL_DEBUG(" size=%zu offset=%zu\n", size, offset);
928950
// split tensors must always be set in their entirety at once
929951
GGML_ASSERT(offset == 0);
930952
GGML_ASSERT(size == ggml_nbytes(tensor));
@@ -3723,6 +3745,9 @@ static void ggml_backend_sycl_set_tensor_async(ggml_backend_t backend,
37233745
ggml_tensor *tensor,
37243746
const void *data, size_t offset,
37253747
size_t size) try {
3748+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
3749+
debug_print_tensor(": tensor=", tensor);
3750+
GGML_SYCL_DEBUG(" size=%zu offset=%zu\n", size, offset);
37263751
ggml_backend_sycl_context * sycl_ctx = (ggml_backend_sycl_context *)backend->context;
37273752
ggml_backend_buffer_t buf = tensor->view_src ? tensor->view_src->buffer : tensor->buffer;
37283753

@@ -3741,6 +3766,9 @@ static void ggml_backend_sycl_get_tensor_async(ggml_backend_t backend,
37413766
const ggml_tensor *tensor,
37423767
void *data, size_t offset,
37433768
size_t size) try {
3769+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
3770+
debug_print_tensor(": tensor=", tensor);
3771+
GGML_SYCL_DEBUG(" size=%zu offset=%zu\n", size, offset);
37443772
ggml_backend_sycl_context * sycl_ctx = (ggml_backend_sycl_context *)backend->context;
37453773
ggml_backend_buffer_t buf = tensor->view_src ? tensor->view_src->buffer : tensor->buffer;
37463774

@@ -3759,7 +3787,12 @@ static bool ggml_backend_sycl_cpy_tensor_async(ggml_backend_t backend,
37593787
const ggml_tensor *src,
37603788
ggml_tensor *dst) try {
37613789
ggml_backend_sycl_context * sycl_ctx = (ggml_backend_sycl_context *)backend->context;
3762-
if (dst->buffer->buft == ggml_backend_sycl_buffer_type(sycl_ctx->device) && ggml_backend_buffer_is_sycl(src->buffer)) {
3790+
bool is_cpy_supported = dst->buffer->buft == ggml_backend_sycl_buffer_type(sycl_ctx->device) && ggml_backend_buffer_is_sycl(src->buffer);
3791+
GGML_SYCL_DEBUG("[SYCL] call %s", __func__);
3792+
debug_print_tensor(": dst=", dst);
3793+
debug_print_tensor(" src=", src);
3794+
GGML_SYCL_DEBUG(" is_cpy_supported=%d\n", is_cpy_supported);
3795+
if (is_cpy_supported) {
37633796
/*
37643797
DPCT1009:215: SYCL uses exceptions to report errors and does not use the
37653798
error codes. The original code was commented out and a warning string
@@ -3780,6 +3813,7 @@ catch (sycl::exception const &exc) {
37803813
}
37813814

37823815
static void ggml_backend_sycl_synchronize(ggml_backend_t backend) try {
3816+
GGML_SYCL_DEBUG("[SYCL] call %s\n", __func__);
37833817
ggml_backend_sycl_context * sycl_ctx = (ggml_backend_sycl_context *)backend->context;
37843818
const queue_ptr stream = sycl_ctx->stream(sycl_ctx->device, 0);
37853819
SYCL_CHECK(CHECK_TRY_ERROR((stream)->wait()));
@@ -3881,7 +3915,7 @@ catch (sycl::exception const &exc)
38813915
}
38823916

38833917
static void ggml_backend_sycl_event_wait(ggml_backend_t backend, ggml_backend_event_t event) try {
3884-
3918+
GGML_SYCL_DEBUG("[SYCL] call %s\n", __func__);
38853919
sycl::event* sycl_event = static_cast<sycl::event*>(event->context);
38863920

38873921
if (ggml_backend_is_sycl(backend)) {
@@ -4276,6 +4310,7 @@ static void ggml_backend_sycl_device_event_free(ggml_backend_dev_t dev, ggml_bac
42764310

42774311
static void ggml_backend_sycl_device_event_synchronize(ggml_backend_dev_t dev, ggml_backend_event_t event) try {
42784312
GGML_UNUSED(dev);
4313+
GGML_SYCL_DEBUG("[SYCL] call %s\n", __func__);
42794314

42804315
sycl::event *sycl_event = static_cast<sycl::event *>(event->context);
42814316
SYCL_CHECK(CHECK_TRY_ERROR(sycl_event->wait()));

0 commit comments

Comments
 (0)