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

ggml : reduce hash table reset cost (#8698)

Browse source 
* ggml : reduce hash table reset cost

* fix unreachable code warnings after GGML_ASSERT(false)

* GGML_ASSERT(false) -> GGML_ABORT("fatal error")

* GGML_ABORT use format string
This commit is contained in:
slaren 2024-07-27 04:41:55 +02:00 • committed by GitHub
parent 01245f5b16
commit 2b1f616b20
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
46 changed files with 851 additions and 754 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml/src/ggml-cuda/mmq.cuh
View file

@ -75,7 +75,7 @@ static mmq_q8_1_ds_layout mmq_get_q8_1_ds_layout(const ggml_type type_x) {
case GGML_TYPE_IQ4_NL:
return MMQ_Q8_1_DS_LAYOUT_D4;
default:
GGML_ASSERT(false);
GGML_ABORT("fatal error");
break;
}
}
@ -2898,7 +2898,7 @@ void mul_mat_q_case(ggml_backend_cuda_context & ctx, const mmq_args & args, cuda
break;
default:
fprintf(stderr, "mmq_x_best=%d\n", mmq_x_best);
GGML_ASSERT(false);
GGML_ABORT("fatal error");
break;
}
}