tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-28 20:25:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

k-quants : fix build on armv7 (android only) (#2920)

Browse Source 
* k-quants : fix build on armv7

* ggml : cleanup unused arm32 specific impl

* k-quants : avoid some unused vzero / mzero define

* ggml-alloc : use 4g for MEASURE_MAX_SIZE in 32-bit arm
This commit is contained in:
Jhen-Jie Hong
2023-09-02 20:23:45 +08:00
committed by GitHub
parent 571083f508
commit 21f3d1be86
3 changed files with 42 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
ggml-alloc.c
View File

@ -284,7 +284,14 @@ struct ggml_allocr * ggml_allocr_new(void * data, size_t size, size_t alignment)
// address and size of the buffer when measuring
// it needs to be large enough to fit all the tensors, but it cannot overlap with other existing buffers
static void * const MEASURE_BASE_ADDR = (void *) 0x1000;
#if defined(__ARM_NEON) && !defined(__aarch64__)
// 32-bit
// TODO: Use for 32-bit x86 as well
static const size_t MEASURE_MAX_SIZE = (1ULL<<32) - 1; // 4 GB
#else
// 64-bit
static const size_t MEASURE_MAX_SIZE = 1ULL<<40; // 1 TB
#endif
struct ggml_allocr * ggml_allocr_new_measure(size_t alignment) {
struct ggml_allocr * alloc = (struct ggml_allocr *)malloc(sizeof(struct ggml_allocr) /* + n_free_blocks * sizeof(struct free_block) */);