diff --git a/Include/Hallocy/Core/Memory.h b/Include/Hallocy/Core/Memory.h index b41c05c..b001d22 100644 --- a/Include/Hallocy/Core/Memory.h +++ b/Include/Hallocy/Core/Memory.h @@ -30,5 +30,6 @@ HallocyError hallocy_set_memory(void *destination, int value, const size_t size); HallocyError hallocy_copy_memory(void *destination, void *source, const size_t size); +HallocyError hallocy_move_memory(void *destination, void *source, const size_t size); #endif \ No newline at end of file diff --git a/Src/Core/Memory.c b/Src/Core/Memory.c index 0710701..fd2e7e6 100644 --- a/Src/Core/Memory.c +++ b/Src/Core/Memory.c @@ -281,5 +281,165 @@ HallocyError hallocy_copy_memory(void *destination, void *source, const size_t s source_bytes += 1; } + return HALLOCY_ERROR_NONE; +} + +HallocyError hallocy_move_memory(void *destination, void *source, const size_t size) { + if (!((void*)((size_t)source + size) > destination && source < destination)) { + return hallocy_copy_memory(destination, source, size); + } + + if (destination == NULL || source == NULL) { + return HALLOCY_ERROR_INVALID_POINTER; + } + + unsigned char *end_address = (unsigned char*)destination; + + unsigned char *destination_bytes = (unsigned char*)destination + size; + unsigned char *source_bytes = (unsigned char*)source + size; + + switch (hallocy_is_simd_supported()) { + #if defined(_M_ARM64) || defined(__aarch64__) || defined(__arm__) + case HALLOCY_SIMD_NEON: { + if ((size_t)destination_bytes % 16 == (size_t)source_bytes % 16) { + while (((size_t)destination_bytes % 16) != 0 && destination_bytes != end_address) { + destination_bytes -= 1; + source_bytes -= 1; + *destination_bytes = *source_bytes; + } + } + + uint8x16_t simd_value; + while (end_address - destination_bytes >= 16) { + destination_bytes -= 16; + source_bytes -= 16; + + simd_value = vdupq_n_u8(source_bytes); + vst1q_u8(destination_bytes, simd_value); + } + break; + } + #else + case HALLOCY_SIMD_AVX512: { + if ((size_t)destination_bytes % 64 == (size_t)source_bytes % 64) { + while (((size_t)destination_bytes % 64) != 0 && destination_bytes != end_address) { + destination_bytes -= 1; + source_bytes -= 1; + *destination_bytes = *source_bytes; + } + + __m512i simd_value; + while (end_address - destination_bytes >= 64) { + destination_bytes -= 64; + source_bytes -= 64; + + simd_value = _mm512_load_si512((__m512i*)source_bytes); + _mm512_store_si512((__m512i*)destination, simd_value); + } + } else { + __m512i simd_value; + while (end_address - destination_bytes >= 64) { + destination_bytes -= 64; + source_bytes -= 64; + + simd_value = _mm512_loadu_si512((__m512i*)source_bytes); + _mm512_storeu_si512((__m512i*)destination, simd_value); + } + } + break; + } + + case HALLOCY_SIMD_AVX2: + case HALLOCY_SIMD_AVX: { + if ((size_t)destination_bytes % 32 == (size_t)source_bytes % 64) { + while (((size_t)destination_bytes % 32) != 0 && destination_bytes != end_address) { + destination_bytes -= 1; + source_bytes -= 1; + + *destination_bytes = *source_bytes; + } + + __m256i simd_value; + while (end_address - destination_bytes >= 32) { + destination_bytes -= 32; + source_bytes -= 32; + + simd_value = _mm256_load_si256((__m256i*)source_bytes); + _mm256_store_si256((__m256i*)destination, simd_value); + } + } else { + __m256i simd_value; + while (end_address - destination_bytes >= 32) { + destination_bytes -= 32; + source_bytes -= 32; + + simd_value = _mm256_loadu_si256((__m256i*)source_bytes); + _mm256_storeu_si256((__m256i*)destination, simd_value); + } + } + break; + } + + case HALLOCY_SIMD_SSE2: + case HALLOCY_SIMD_SSE: { + if ((size_t)destination_bytes % 16 == (size_t)source_bytes % 64) { + while (((size_t)destination_bytes % 16) != 0 && destination_bytes != end_address) { + destination_bytes -= 1; + source_bytes -= 1; + *destination_bytes = *source_bytes; + } + + __m128i simd_value; + while (end_address - destination_bytes >= 16) { + destination_bytes -= 16; + source_bytes -= 16; + + simd_value = _mm_load_si128((__m128i*)source_bytes); + _mm_store_si128((__m128i*)destination, simd_value); + } + } else { + __m128i simd_value; + while (end_address - destination_bytes >= 16) { + destination_bytes -= 16; + source_bytes -= 16; + + simd_value = _mm_loadu_si128((__m128i*)source_bytes); + _mm_storeu_si128((__m128i*)destination, simd_value); + } + } + break; + } + #endif + default: { + size_t word_size = sizeof(size_t); + if ((size_t)destination_bytes % word_size == (size_t)source_bytes % word_size) { + while (((size_t)destination_bytes % word_size) != 0 && destination_bytes != end_address) { + destination_bytes -= 1; + source_bytes -= 1; + *destination_bytes = *source_bytes; + } + + size_t *destination_word = (size_t*)destination_bytes; + size_t *source_word = (size_t*)source_bytes; + while ((unsigned char*)(destination_word - 1) > end_address) { + destination_word -= 1; + source_word -= 1; + *destination_word = *source_word; + } + + source_bytes = (unsigned char*)source_word; + destination_bytes = (unsigned char*)destination_word; + } + + break; + } + } + + while (destination_bytes != end_address) { + destination_bytes -= 1; + source_bytes -= 1; + *destination_bytes = *source_bytes; + } + return HALLOCY_ERROR_NONE; } \ No newline at end of file diff --git a/Tests/Main.c b/Tests/Main.c index 4967b57..09645e3 100644 --- a/Tests/Main.c +++ b/Tests/Main.c @@ -49,7 +49,14 @@ printf("%s\n", text); char *text_copy = (char*)hallocy_malloc(36); - hallocy_copy_memory(text_copy, text, 36); + hallocy_copy_memory(text_copy, text, 35); + printf("%s\n", text_copy); + + hallocy_set_memory(text_copy + 10, 'L', 25); + text_copy[14] = 'r'; + text_copy[24] = 'r'; + printf("%s\n", text_copy); + hallocy_move_memory(text_copy + 5, text_copy + 10, 15); printf("%s\n", text_copy); hallocy_free(text);