diff options
| -rw-r--r-- | ggml.c | 4 | ||||
| -rw-r--r-- | llama-util.h | 40 | ||||
| -rw-r--r-- | llama.cpp | 7 |
3 files changed, 26 insertions, 25 deletions
@@ -512,7 +512,7 @@ static inline int hsum_i32_4(const __m128i a) { return _mm_cvtsi128_si32(_mm_add_epi32(sum64, hi32)); } -#if __AVX2__ || __AVX512F__ +#if defined(__AVX2__) || defined(__AVX512F__) // spread 32 bits to 32 bytes { 0x00, 0xFF } static inline __m256i bytes_from_bits_32(const uint8_t * x) { uint32_t x32; @@ -688,7 +688,7 @@ static inline float hsum_float_4x4(const __m128 a, const __m128 b, const __m128 #endif // __AVX__ || __AVX2__ || __AVX512F__ #endif // defined(__AVX__) || defined(__AVX2__) || defined(__AVX512F__) || defined(__SSSE3__) -#if __ARM_NEON +#if defined(__ARM_NEON) #if !defined(__aarch64__) diff --git a/llama-util.h b/llama-util.h index 88ec28d..a79c5da 100644 --- a/llama-util.h +++ b/llama-util.h @@ -101,12 +101,12 @@ struct llama_file { LLAMA_ASSERT(ret == 0); // same } - void read_raw(void * ptr, size_t size) { - if (size == 0) { + void read_raw(void * ptr, size_t len) const { + if (len == 0) { return; } errno = 0; - std::size_t ret = std::fread(ptr, size, 1, fp); + std::size_t ret = std::fread(ptr, len, 1, fp); if (ferror(fp)) { throw std::runtime_error(format("read error: %s", strerror(errno))); } @@ -127,12 +127,12 @@ struct llama_file { return std::string(chars.data(), len); } - void write_raw(const void * ptr, size_t size) { - if (size == 0) { + void write_raw(const void * ptr, size_t len) const { + if (len == 0) { return; } errno = 0; - size_t ret = std::fwrite(ptr, size, 1, fp); + size_t ret = std::fwrite(ptr, len, 1, fp); if (ret != 1) { throw std::runtime_error(format("write error: %s", strerror(errno))); } @@ -267,9 +267,9 @@ struct llama_mlock { } } - void init(void * addr) { - LLAMA_ASSERT(this->addr == NULL && this->size == 0); - this->addr = addr; + void init(void * ptr) { + LLAMA_ASSERT(addr == NULL && size == 0); + addr = ptr; } void grow_to(size_t target_size) { @@ -340,14 +340,14 @@ struct llama_mlock { return (size_t) si.dwPageSize; } - bool raw_lock(void * addr, size_t size) { + bool raw_lock(void * ptr, size_t len) { for (int tries = 1; ; tries++) { - if (VirtualLock(addr, size)) { + if (VirtualLock(ptr, len)) { return true; } if (tries == 2) { fprintf(stderr, "warning: failed to VirtualLock %zu-byte buffer (after previously locking %zu bytes): %s\n", - size, this->size, llama_format_win_err(GetLastError()).c_str()); + len, size, llama_format_win_err(GetLastError()).c_str()); return false; } @@ -363,7 +363,7 @@ struct llama_mlock { // is equal to the number of pages in its minimum working set minus // a small overhead." // Hopefully a megabyte is enough overhead: - size_t increment = size + 1048576; + size_t increment = len + 1048576; // The minimum must be <= the maximum, so we need to increase both: min_ws_size += increment; max_ws_size += increment; @@ -375,8 +375,8 @@ struct llama_mlock { } } - void raw_unlock(void * addr, size_t size) { - if (!VirtualUnlock(addr, size)) { + void raw_unlock(void * ptr, size_t len) { + if (!VirtualUnlock(ptr, len)) { fprintf(stderr, "warning: failed to VirtualUnlock buffer: %s\n", llama_format_win_err(GetLastError()).c_str()); } @@ -388,12 +388,12 @@ struct llama_mlock { return (size_t) 65536; } - bool raw_lock(const void * addr, size_t size) { + bool raw_lock(const void * addr, size_t len) { fprintf(stderr, "warning: mlock not supported on this system\n"); return false; } - void raw_unlock(const void * addr, size_t size) {} + void raw_unlock(const void * addr, size_t len) {} #endif }; @@ -404,10 +404,10 @@ struct llama_buffer { llama_buffer() = default; - void resize(size_t size) { + void resize(size_t len) { delete[] addr; - addr = new uint8_t[size]; - this->size = size; + addr = new uint8_t[len]; + size = len; } ~llama_buffer() { @@ -45,6 +45,7 @@ enum e_model { MODEL_65B, }; + static const size_t MB = 1024*1024; // computed for n_ctx == 2048 @@ -110,7 +111,7 @@ struct llama_hparams { enum llama_ftype ftype = LLAMA_FTYPE_MOSTLY_F16; bool operator!=(const llama_hparams & other) const { - return memcmp(this, &other, sizeof(llama_hparams)); + return static_cast<bool>(memcmp(this, &other, sizeof(llama_hparams))); } }; @@ -502,7 +503,7 @@ struct llama_file_loader { if (file_version >= LLAMA_FILE_VERSION_GGJT_V1) { // skip to the next multiple of 32 bytes - file.seek(-file.tell() & 31, SEEK_CUR); + file.seek(-static_cast<ptrdiff_t>(file.tell()) & 31, SEEK_CUR); } shard.file_idx = file_idx; shard.file_off = file.tell(); @@ -577,7 +578,7 @@ struct llama_file_saver { file.write_u32(new_type); file.write_raw(tensor.ne.data(), sizeof(tensor.ne[0]) * tensor.ne.size()); file.write_raw(tensor.name.data(), tensor.name.size()); - file.seek(-file.tell() & 31, SEEK_CUR); + file.seek(-static_cast<ptrdiff_t>(file.tell()) & 31, SEEK_CUR); LLAMA_ASSERT(new_size == llama_calc_tensor_size(tensor.ne, new_type)); file.write_raw(new_data, new_size); } |