vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml_tensor : use 1 bit per flag

Browse source
This commit is contained in:
slaren 2023-07-23 18:04:33 +02:00
parent 261fdaae80
commit e371b716ca
2 changed files with 3 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml.c
View file

@ -4592,9 +4592,9 @@ struct ggml_tensor * ggml_new_tensor_impl(
/*.op =*/ GGML_OP_NONE, /*.op =*/ GGML_OP_NONE,
/*.op_params =*/ {0}, /*.op_params =*/ {0},
/*.is_param =*/ false, /*.is_param =*/ false,
/*.visited =*/ false,
/*.grad =*/ NULL, /*.grad =*/ NULL,
/*.src =*/ { NULL }, /*.src =*/ { NULL },
/*.visited =*/ false,
/*.perf_runs =*/ 0, /*.perf_runs =*/ 0,
/*.perf_cycles =*/ 0, /*.perf_cycles =*/ 0,
/*.perf_time_us =*/ 0, /*.perf_time_us =*/ 0,

5
ggml.h
View file

@ -422,13 +422,12 @@ extern "C" {
// op params - allocated as int32_t for alignment // op params - allocated as int32_t for alignment
int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(uint32_t)]; int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(uint32_t)];
bool is_param; uint32_t is_param:1;
uint32_t visited:1; // used to build graphs
struct ggml_tensor * grad; struct ggml_tensor * grad;
struct ggml_tensor * src[GGML_MAX_SRC]; struct ggml_tensor * src[GGML_MAX_SRC];
bool visited; // used to build graphs
// performance // performance
int perf_runs; int perf_runs;
int64_t perf_cycles; int64_t perf_cycles;