root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 10:54:36 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

ggml, ci : Windows ARM runner and build fixes (#5979)

Browse Source 
* windows arm ci

* fix `error C2078: too many initializers` with ggml_vld1q_u32 macro for MSVC ARM64

* fix `warning C4146: unary minus operator applied to unsigned type, result still unsigned`

* fix `error C2065: '__fp16': undeclared identifier`
This commit is contained in:
Michael Podvitskiy 2024-03-11 10:28:51 +01:00 committed by GitHub
parent 332bdfd798
commit 3202361c5b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
5 changed files with 21 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
.github/workflows/build.yml vendored
View File

@ -425,6 +425,8 @@ jobs:
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_KOMPUTE=ON -DKOMPUTE_OPT_DISABLE_VULKAN_VERSION_CHECK=ON -DBUILD_SHARED_LIBS=ON' defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_KOMPUTE=ON -DKOMPUTE_OPT_DISABLE_VULKAN_VERSION_CHECK=ON -DBUILD_SHARED_LIBS=ON'
- build: 'vulkan' - build: 'vulkan'
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_VULKAN=ON -DBUILD_SHARED_LIBS=ON' defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_VULKAN=ON -DBUILD_SHARED_LIBS=ON'
- build: 'arm64'
defines: '-A ARM64 -DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DBUILD_SHARED_LIBS=ON'
steps: steps:
- name: Clone - name: Clone
@ -520,7 +522,7 @@ jobs:
- name: Test - name: Test
id: cmake_test id: cmake_test
# not all machines have native AVX-512 # not all machines have native AVX-512
if: ${{ matrix.build != 'clblast' && matrix.build != 'kompute' && matrix.build != 'vulkan' && (matrix.build != 'avx512' || env.HAS_AVX512F == '1') }} if: ${{ matrix.build != 'arm64' && matrix.build != 'clblast' && matrix.build != 'kompute' && matrix.build != 'vulkan' && (matrix.build != 'avx512' || env.HAS_AVX512F == '1') }}
run: | run: |
cd build cd build
ctest -L main -C Release --verbose --timeout 900 ctest -L main -C Release --verbose --timeout 900

8
ggml-impl.h
View File

@ -53,26 +53,30 @@ extern "C" {
// //
#include <arm_neon.h> #include <arm_neon.h>
typedef __fp16 ggml_fp16_internal_t;
#define GGML_COMPUTE_FP16_TO_FP32(x) ggml_compute_fp16_to_fp32(x) #define GGML_COMPUTE_FP16_TO_FP32(x) ggml_compute_fp16_to_fp32(x)
#define GGML_COMPUTE_FP32_TO_FP16(x) ggml_compute_fp32_to_fp16(x) #define GGML_COMPUTE_FP32_TO_FP16(x) ggml_compute_fp32_to_fp16(x)
#define GGML_FP16_TO_FP32(x) ggml_compute_fp16_to_fp32(x) #define GGML_FP16_TO_FP32(x) ggml_compute_fp16_to_fp32(x)
static inline float ggml_compute_fp16_to_fp32(ggml_fp16_t h) { static inline float ggml_compute_fp16_to_fp32(ggml_fp16_t h) {
__fp16 tmp; ggml_fp16_internal_t tmp;
memcpy(&tmp, &h, sizeof(ggml_fp16_t)); memcpy(&tmp, &h, sizeof(ggml_fp16_t));
return (float)tmp; return (float)tmp;
} }
static inline ggml_fp16_t ggml_compute_fp32_to_fp16(float f) { static inline ggml_fp16_t ggml_compute_fp32_to_fp16(float f) {
ggml_fp16_t res; ggml_fp16_t res;
__fp16 tmp = f; ggml_fp16_internal_t tmp = f;
memcpy(&res, &tmp, sizeof(ggml_fp16_t)); memcpy(&res, &tmp, sizeof(ggml_fp16_t));
return res; return res;
} }
#else #else
typedef uint16_t ggml_fp16_internal_t;
#ifdef __wasm_simd128__ #ifdef __wasm_simd128__
#include <wasm_simd128.h> #include <wasm_simd128.h>
#else #else

16
ggml-quants.c
View File

@ -9374,15 +9374,15 @@ void ggml_vec_dot_iq3_s_q8_K (int n, float * restrict s, size_t bs, const void *
const uint8x16_t idx_l = vld1q_u8(qs); qs += 16; const uint8x16_t idx_l = vld1q_u8(qs); qs += 16;
idx.vec_index = vorrq_u16(vmovl_u8(vget_low_u8 (idx_l)), vandq_u16(vshlq_u16(vdupq_n_u16(qh[ib32+0]), hshift), m256)); idx.vec_index = vorrq_u16(vmovl_u8(vget_low_u8 (idx_l)), vandq_u16(vshlq_u16(vdupq_n_u16(qh[ib32+0]), hshift), m256));
const uint32x4_t aux32x4_0 = {iq3s_grid[idx.index[0]], iq3s_grid[idx.index[1]], const uint32x4_t aux32x4_0 = ggml_vld1q_u32(iq3s_grid[idx.index[0]], iq3s_grid[idx.index[1]],
iq3s_grid[idx.index[2]], iq3s_grid[idx.index[3]]}; iq3s_grid[idx.index[2]], iq3s_grid[idx.index[3]]);
const uint32x4_t aux32x4_1 = {iq3s_grid[idx.index[4]], iq3s_grid[idx.index[5]], const uint32x4_t aux32x4_1 = ggml_vld1q_u32(iq3s_grid[idx.index[4]], iq3s_grid[idx.index[5]],
iq3s_grid[idx.index[6]], iq3s_grid[idx.index[7]]}; iq3s_grid[idx.index[6]], iq3s_grid[idx.index[7]]);
idx.vec_index = vorrq_u16(vmovl_u8(vget_high_u8(idx_l)), vandq_u16(vshlq_u16(vdupq_n_u16(qh[ib32+1]), hshift), m256)); idx.vec_index = vorrq_u16(vmovl_u8(vget_high_u8(idx_l)), vandq_u16(vshlq_u16(vdupq_n_u16(qh[ib32+1]), hshift), m256));
const uint32x4_t aux32x4_2 = {iq3s_grid[idx.index[0]], iq3s_grid[idx.index[1]], const uint32x4_t aux32x4_2 = ggml_vld1q_u32(iq3s_grid[idx.index[0]], iq3s_grid[idx.index[1]],
iq3s_grid[idx.index[2]], iq3s_grid[idx.index[3]]}; iq3s_grid[idx.index[2]], iq3s_grid[idx.index[3]]);
const uint32x4_t aux32x4_3 = {iq3s_grid[idx.index[4]], iq3s_grid[idx.index[5]], const uint32x4_t aux32x4_3 = ggml_vld1q_u32(iq3s_grid[idx.index[4]], iq3s_grid[idx.index[5]],
iq3s_grid[idx.index[6]], iq3s_grid[idx.index[7]]}; iq3s_grid[idx.index[6]], iq3s_grid[idx.index[7]]);
vs.val[0] = vreinterpretq_u8_u32(vdupq_n_u32(signs[0] | (signs[1] << 16))); vs.val[0] = vreinterpretq_u8_u32(vdupq_n_u32(signs[0] | (signs[1] << 16)));

4
ggml.c
View File

@ -857,7 +857,7 @@ inline static float vaddvq_f32(float32x4_t v) {
#define GGML_F16x8 float16x8_t #define GGML_F16x8 float16x8_t
#define GGML_F16x8_ZERO vdupq_n_f16(0.0f) #define GGML_F16x8_ZERO vdupq_n_f16(0.0f)
#define GGML_F16x8_SET1(x) vdupq_n_f16(x) #define GGML_F16x8_SET1(x) vdupq_n_f16(x)
#define GGML_F16x8_LOAD(x) vld1q_f16((const __fp16 *)(x)) #define GGML_F16x8_LOAD(x) vld1q_f16((const ggml_fp16_internal_t *)(x))
#define GGML_F16x8_STORE vst1q_f16 #define GGML_F16x8_STORE vst1q_f16
#define GGML_F16x8_FMA(a, b, c) vfmaq_f16(a, b, c) #define GGML_F16x8_FMA(a, b, c) vfmaq_f16(a, b, c)
#define GGML_F16x8_ADD vaddq_f16 #define GGML_F16x8_ADD vaddq_f16
@ -900,7 +900,7 @@ inline static float vaddvq_f32(float32x4_t v) {
#define GGML_F32Cx4 float32x4_t #define GGML_F32Cx4 float32x4_t
#define GGML_F32Cx4_ZERO vdupq_n_f32(0.0f) #define GGML_F32Cx4_ZERO vdupq_n_f32(0.0f)
#define GGML_F32Cx4_SET1(x) vdupq_n_f32(x) #define GGML_F32Cx4_SET1(x) vdupq_n_f32(x)
#define GGML_F32Cx4_LOAD(x) vcvt_f32_f16(vld1_f16((const __fp16 *)(x))) #define GGML_F32Cx4_LOAD(x) vcvt_f32_f16(vld1_f16((const ggml_fp16_internal_t *)(x)))
#define GGML_F32Cx4_STORE(x, y) vst1_f16(x, vcvt_f16_f32(y)) #define GGML_F32Cx4_STORE(x, y) vst1_f16(x, vcvt_f16_f32(y))
#define GGML_F32Cx4_FMA(a, b, c) vfmaq_f32(a, b, c) #define GGML_F32Cx4_FMA(a, b, c) vfmaq_f32(a, b, c)
#define GGML_F32Cx4_ADD vaddq_f32 #define GGML_F32Cx4_ADD vaddq_f32

4
llama.cpp
View File

@ -13980,7 +13980,7 @@ int32_t llama_token_to_piece(const struct llama_model * model, llama_token token
} else if (llama_is_user_defined_token(model->vocab, token)) { } else if (llama_is_user_defined_token(model->vocab, token)) {
std::string result = model->vocab.id_to_token[token].text; std::string result = model->vocab.id_to_token[token].text;
if (length < (int) result.length()) { if (length < (int) result.length()) {
return -result.length(); return -(int) result.length();
} }
memcpy(buf, result.c_str(), result.length()); memcpy(buf, result.c_str(), result.length());
return result.length(); return result.length();
@ -14015,7 +14015,7 @@ int32_t llama_token_to_piece(const struct llama_model * model, llama_token token
} else if (llama_is_user_defined_token(model->vocab, token)) { } else if (llama_is_user_defined_token(model->vocab, token)) {
std::string result = model->vocab.id_to_token[token].text; std::string result = model->vocab.id_to_token[token].text;
if (length < (int) result.length()) { if (length < (int) result.length()) {
return -result.length(); return -(int) result.length();
} }
memcpy(buf, result.c_str(), result.length()); memcpy(buf, result.c_str(), result.length());
return result.length(); return result.length();