diff options
Diffstat (limited to 'core/io/marshalls.h')
-rw-r--r-- | core/io/marshalls.h | 21 |
1 files changed, 0 insertions, 21 deletions
diff --git a/core/io/marshalls.h b/core/io/marshalls.h index 1ba786d5d9..279f782b1a 100644 --- a/core/io/marshalls.h +++ b/core/io/marshalls.h @@ -41,21 +41,17 @@ */ union MarshallFloat { - uint32_t i; ///< int float f; ///< float }; union MarshallDouble { - uint64_t l; ///< long long double d; ///< double }; static inline unsigned int encode_uint16(uint16_t p_uint, uint8_t *p_arr) { - for (int i = 0; i < 2; i++) { - *p_arr = p_uint & 0xFF; p_arr++; p_uint >>= 8; @@ -65,9 +61,7 @@ static inline unsigned int encode_uint16(uint16_t p_uint, uint8_t *p_arr) { } static inline unsigned int encode_uint32(uint32_t p_uint, uint8_t *p_arr) { - for (int i = 0; i < 4; i++) { - *p_arr = p_uint & 0xFF; p_arr++; p_uint >>= 8; @@ -77,7 +71,6 @@ static inline unsigned int encode_uint32(uint32_t p_uint, uint8_t *p_arr) { } static inline unsigned int encode_float(float p_float, uint8_t *p_arr) { - MarshallFloat mf; mf.f = p_float; encode_uint32(mf.i, p_arr); @@ -86,9 +79,7 @@ static inline unsigned int encode_float(float p_float, uint8_t *p_arr) { } static inline unsigned int encode_uint64(uint64_t p_uint, uint8_t *p_arr) { - for (int i = 0; i < 8; i++) { - *p_arr = p_uint & 0xFF; p_arr++; p_uint >>= 8; @@ -98,7 +89,6 @@ static inline unsigned int encode_uint64(uint64_t p_uint, uint8_t *p_arr) { } static inline unsigned int encode_double(double p_double, uint8_t *p_arr) { - MarshallDouble md; md.d = p_double; encode_uint64(md.l, p_arr); @@ -107,13 +97,10 @@ static inline unsigned int encode_double(double p_double, uint8_t *p_arr) { } static inline int encode_cstring(const char *p_string, uint8_t *p_data) { - int len = 0; while (*p_string) { - if (p_data) { - *p_data = (uint8_t)*p_string; p_data++; } @@ -127,11 +114,9 @@ static inline int encode_cstring(const char *p_string, uint8_t *p_data) { } static inline uint16_t decode_uint16(const uint8_t *p_arr) { - uint16_t u = 0; for (int i = 0; i < 2; i++) { - uint16_t b = *p_arr; b <<= (i * 8); u |= b; @@ -142,11 +127,9 @@ static inline uint16_t decode_uint16(const uint8_t *p_arr) { } static inline uint32_t decode_uint32(const uint8_t *p_arr) { - uint32_t u = 0; for (int i = 0; i < 4; i++) { - uint32_t b = *p_arr; b <<= (i * 8); u |= b; @@ -157,18 +140,15 @@ static inline uint32_t decode_uint32(const uint8_t *p_arr) { } static inline float decode_float(const uint8_t *p_arr) { - MarshallFloat mf; mf.i = decode_uint32(p_arr); return mf.f; } static inline uint64_t decode_uint64(const uint8_t *p_arr) { - uint64_t u = 0; for (int i = 0; i < 8; i++) { - uint64_t b = (*p_arr) & 0xFF; b <<= (i * 8); u |= b; @@ -179,7 +159,6 @@ static inline uint64_t decode_uint64(const uint8_t *p_arr) { } static inline double decode_double(const uint8_t *p_arr) { - MarshallDouble md; md.l = decode_uint64(p_arr); return md.d; |