Skip to content

Commit 28b3b76

Browse files
slarenslaren
committed
build fixes
1 parent bf79cb3 commit 28b3b76

File tree

7 files changed

+159
-221
lines changed

7 files changed

+159
-221
lines changed

‎ggml/src/CMakeLists.txt

Lines changed: 4 additions & 43 deletions
Original file line numberDiff line numberDiff line change
@@ -29,37 +29,6 @@ endif()
2929
unset(GGML_EXTRA_LIBS_PRIVATE)
3030
unset(GGML_EXTRA_LIBS_PUBLIC)
3131

32-
# musa, hip: add directory with a CMakeLists.txt file, but no source files (use refer to ggml-cuda files as ../ggml-cuda)
33-
if (GGML_MUSA)
34-
list(APPEND CMAKE_MODULE_PATH "/usr/local/musa/cmake/")
35-
find_package(MUSAToolkit)
36-
set(CUDAToolkit_FOUND ${MUSAToolkit_FOUND})
37-
else()
38-
find_package(CUDAToolkit)
39-
endif()
40-
41-
# if (GGML_MUSA)
42-
# set(CMAKE_CUDA_COMPILER ${MUSAToolkit_MCC_EXECUTABLE})
43-
# else()
44-
# if (GGML_MUSA)
45-
# set_source_files_properties(${GGML_SOURCES_CUDA} PROPERTIES LANGUAGE CXX)
46-
# foreach(SOURCE ${GGML_SOURCES_CUDA})
47-
# set_property(SOURCE ${SOURCE} PROPERTY COMPILE_FLAGS "-x musa -mtgpu --cuda-gpu-arch=mp_21 --cuda-gpu-arch=mp_22")
48-
# endforeach()
49-
# endif()
50-
51-
52-
if (GGML_MUSA)
53-
set(CMAKE_C_COMPILER clang)
54-
set(CMAKE_C_EXTENSIONS OFF)
55-
set(CMAKE_CXX_COMPILER clang++)
56-
set(CMAKE_CXX_EXTENSIONS OFF)
57-
58-
set(GGML_CUDA ON)
59-
60-
list(APPEND GGML_CDEF_PUBLIC GGML_USE_MUSA)
61-
endif()
62-
6332
if (GGML_AMX)
6433
if (CMAKE_COMPILER_IS_GNUCC AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 11.0)
6534
else()
@@ -597,12 +566,6 @@ function(get_flags CCID CCVER)
597566
elseif (CCID STREQUAL "GNU")
598567
set(C_FLAGS -Wdouble-promotion)
599568
set(CXX_FLAGS -Wno-array-bounds)
600-
601-
if (NOT GGML_MUSA)
602-
if (CCVER VERSION_GREATER_EQUAL 7.1.0)
603-
list(APPEND CXX_FLAGS -Wno-format-truncation)
604-
endif()
605-
endif()
606569
if (CCVER VERSION_GREATER_EQUAL 8.1.0)
607570
list(APPEND CXX_FLAGS -Wextra-semi)
608571
endif()
@@ -779,18 +742,16 @@ add_library(ggml-base STATIC
779742
ggml-backend.cpp
780743
ggml-threading.cpp
781744
ggml-threading.h
782-
ggml-quants.c # for quantize functions TODO: move dot fns to a separate file
745+
ggml-quants.c
783746
ggml-quants.h
784747
ggml-aarch64.c
785748
ggml-aarch64.h
786-
787-
)
788-
789-
add_subdirectory(ggml-cpu)
749+
)
790750

791751
add_library(ggml
792752
ggml-backend-reg.cpp
793-
)
753+
)
754+
add_subdirectory(ggml-cpu)
794755

795756
target_link_libraries(ggml PUBLIC ggml-base ggml-cpu)
796757

‎ggml/src/ggml-backend-reg.cpp

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,7 @@
1-
#include "ggml-backend.h"
21
#include "ggml-backend-impl.h"
2+
#include "ggml-backend.h"
3+
#include "ggml-cpu.h"
4+
#include "ggml-impl.h"
35
#include <cstring>
46
#include <vector>
57

@@ -45,8 +47,6 @@
4547
#include "ggml-kompute.h"
4648
#endif
4749

48-
#include "ggml-cpu.h"
49-
5050
struct ggml_backend_registry {
5151
std::vector<ggml_backend_reg_t> backends;
5252
std::vector<ggml_backend_dev_t> devices;

‎ggml/src/ggml-cpu/ggml-cpu-quants.c

Lines changed: 0 additions & 23 deletions
Original file line numberDiff line numberDiff line change
@@ -10783,22 +10783,6 @@ void ggml_vec_dot_iq4_xs_q8_K(int n, float * restrict s, size_t bs, const void *
1078310783
#endif
1078410784
}
1078510785

10786-
//
10787-
// ============================================= 3-bit using D4 lattice
10788-
//
10789-
10790-
void quantize_row_iq3_xxs(const float * restrict x, void * restrict vy, int64_t k) {
10791-
assert(k % QK_K == 0);
10792-
block_iq3_xxs * restrict y = vy;
10793-
quantize_row_iq3_xxs_ref(x, y, k);
10794-
}
10795-
10796-
void quantize_row_iq3_s(const float * restrict x, void * restrict vy, int64_t k) {
10797-
assert(k % QK_K == 0);
10798-
block_iq3_s * restrict y = vy;
10799-
quantize_row_iq3_s_ref(x, y, k);
10800-
}
10801-
1080210786
// ============================ 4-bit non-linear quants
1080310787

1080410788
void quantize_row_iq4_nl(const float * restrict x, void * restrict y, int64_t k) {
@@ -10810,10 +10794,3 @@ void quantize_row_iq4_xs(const float * restrict x, void * restrict y, int64_t k)
1081010794
assert(k % QK_K == 0);
1081110795
quantize_iq4_xs(x, y, 1, k, NULL);
1081210796
}
10813-
10814-
// =============================== 2.5625 bpw
10815-
10816-
void quantize_row_iq2_s(const float * restrict x, void * restrict y, int64_t k) {
10817-
assert(k % QK_K == 0);
10818-
quantize_iq2_s(x, y, 1, k, NULL);
10819-
}

‎ggml/src/ggml-cpu/ggml-cpu-quants.h

Lines changed: 0 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -29,11 +29,8 @@ void quantize_row_q8_K(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, in
2929
void quantize_row_tq1_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
3030
void quantize_row_tq2_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
3131

32-
void quantize_row_iq3_xxs(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
3332
void quantize_row_iq4_nl (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
3433
void quantize_row_iq4_xs (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
35-
void quantize_row_iq3_s (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
36-
void quantize_row_iq2_s (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
3734

3835
// Dot product
3936
void ggml_vec_dot_q4_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);

‎ggml/src/ggml-cpu/ggml-cpu.c

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -356,19 +356,20 @@ static const struct ggml_type_traits_cpu type_traits_cpu[GGML_TYPE_COUNT] = {
356356
.nrows = 1,
357357
},
358358
[GGML_TYPE_IQ3_XXS] = {
359-
.from_float = quantize_row_iq3_xxs,
359+
// NOTE: from_float for iq3 and iq2_s was removed because these quants require initialization in ggml_quantize_init
360+
//.from_float = quantize_row_iq3_xxs,
360361
.vec_dot = ggml_vec_dot_iq3_xxs_q8_K,
361362
.vec_dot_type = GGML_TYPE_Q8_K,
362363
.nrows = 1,
363364
},
364365
[GGML_TYPE_IQ3_S] = {
365-
.from_float = quantize_row_iq3_s,
366+
//.from_float = quantize_row_iq3_s,
366367
.vec_dot = ggml_vec_dot_iq3_s_q8_K,
367368
.vec_dot_type = GGML_TYPE_Q8_K,
368369
.nrows = 1,
369370
},
370371
[GGML_TYPE_IQ2_S] = {
371-
.from_float = quantize_row_iq2_s,
372+
//.from_float = quantize_row_iq2_s,
372373
.vec_dot = ggml_vec_dot_iq2_s_q8_K,
373374
.vec_dot_type = GGML_TYPE_Q8_K,
374375
.nrows = 1,

‎ggml/src/ggml-cpu/ggml-cpu.cpp

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,14 @@
1111
#include <sys/sysctl.h>
1212
#endif
1313

14+
#if defined(_WIN32)
15+
#define WIN32_LEAN_AND_MEAN
16+
#ifndef NOMINMAX
17+
#define NOMINMAX
18+
#endif
19+
#include <windows.h>
20+
#endif
21+
1422
// ggml-backend interface
1523

1624
#ifdef GGML_USE_CPU_HBM

0 commit comments

Comments
 (0)