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

Windows fixes (#31)

Browse source 
* Apply fixes suggested to build on windows

Issue: https://github.com/ggerganov/llama.cpp/issues/22

* Remove unsupported VLAs

* MSVC: Remove features that are only available on MSVC C++20.

* Fix zero initialization of the other fields.

* Change the use of vector for stack allocations.
This commit is contained in:
Sebastián A 2023-03-12 17:15:00 -03:00 • committed by GitHub
parent 7027a97837
commit eb062bb012
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 30 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

16
utils.cpp
View file

@ -5,6 +5,12 @@
#include <fstream>
#include <regex>
#if defined(_MSC_VER) || defined(__MINGW32__)
#include <malloc.h> // using malloc.h with MSC/MINGW
#elif !defined(__FreeBSD__)
#include <alloca.h>
#endif
bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
for (int i = 1; i < argc; i++) {
std::string arg = argv[i];
@ -472,7 +478,8 @@ size_t ggml_quantize_q4_0(float * src, void * dst, int n, int k, int qk, int64_t
assert(k % qk == 0);
uint8_t pp[qk/2];
const size_t pp_size = qk / 2;
uint8_t *pp = static_cast<uint8_t*>(alloca(pp_size));
char * pdst = (char *) dst;
@ -511,7 +518,7 @@ size_t ggml_quantize_q4_0(float * src, void * dst, int n, int k, int qk, int64_t
pp[l/2] = vi0 | (vi1 << 4);
}
memcpy(pb, pp, sizeof(pp));
memcpy(pb, pp, pp_size);
pb += bs;
}
}
@ -526,7 +533,8 @@ size_t ggml_quantize_q4_1(float * src, void * dst, int n, int k, int qk, int64_t
assert(k % qk == 0);
uint8_t pp[qk/2];
const size_t pp_size = qk / 2;
uint8_t *pp = static_cast<uint8_t*>(alloca(pp_size));
char * pdst = (char *) dst;
@ -570,7 +578,7 @@ size_t ggml_quantize_q4_1(float * src, void * dst, int n, int k, int qk, int64_t
pp[l/2] = vi0 | (vi1 << 4);
}
memcpy(pb + i*qk/2, pp, sizeof(pp));
memcpy(pb + i*qk/2, pp, pp_size);
}
}
}