tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-09 10:32:38 -04:00
Code Issues Packages Projects Releases Wiki Activity

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
46 changed files with 851 additions and 754 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
ggml/src/ggml-cuda/fattn.cu
View File

@@ -38,7 +38,7 @@ static void ggml_cuda_flash_attn_ext_wmma_f16(ggml_backend_cuda_context & ctx, g
ggml_cuda_flash_attn_ext_wmma_f16_case<256, cols_per_block, float>(ctx, dst);
break;
default:
GGML_ASSERT(false);
GGML_ABORT("fatal error");
break;
}
} else {
@@ -63,7 +63,7 @@ static void ggml_cuda_flash_attn_ext_wmma_f16(ggml_backend_cuda_context & ctx, g
// ggml_cuda_flash_attn_ext_wmma_f16_case<128, cols_per_block, float>(ctx, dst);
// break;
default:
GGML_ASSERT(false);
GGML_ABORT("fatal error");
break;
}
}
@@ -86,7 +86,7 @@ static void ggml_cuda_flash_attn_ext_wmma_f16(ggml_backend_cuda_context & ctx, g
ggml_cuda_flash_attn_ext_wmma_f16_case<256, cols_per_block, half>(ctx, dst);
break;
default:
GGML_ASSERT(false);
GGML_ABORT("fatal error");
break;
}
return;
@@ -114,7 +114,7 @@ static void ggml_cuda_flash_attn_ext_wmma_f16(ggml_backend_cuda_context & ctx, g
ggml_cuda_flash_attn_ext_wmma_f16_case<256, cols_per_block, half>(ctx, dst);
break;
default:
GGML_ASSERT(false);
GGML_ABORT("fatal error");
break;
}
return;
@@ -141,7 +141,7 @@ static void ggml_cuda_flash_attn_ext_wmma_f16(ggml_backend_cuda_context & ctx, g
ggml_cuda_flash_attn_ext_wmma_f16_case<256, cols_per_block, half>(ctx, dst);
break;
default:
GGML_ASSERT(false);
GGML_ABORT("fatal error");
break;
}
}