Lines Matching +full:matrix +full:- +full:multiply +full:- +full:accumulate +full:- +full:instruction

6  * This software is provided 'as-is', without any express or implied warranty.
43 	 * clang (at least 3.9.[0-1]) pessimizes "rm" (y) and "m" (y)  in _mm_crc32_u8()
46 	 * the latter.  This costs a register and an instruction but  in _mm_crc32_u8()
69 /* CRC-32C (iSCSI) polynomial in reversed bit order. */
73  * Block sizes for three-way parallel crc computation.  LONG and SHORT must
90  * Multiply a matrix times a vector over the Galois field of two elements,
111  * Multiply a matrix by itself over GF(2).  Both mat and square must have 32
133 	uint32_t odd[32];       /* odd-power-of-two zeros operator */  in crc32c_zeros_op()
138 	odd[0] = POLY;              /* CRC-32C polynomial */  in crc32c_zeros_op()
153 	 * bits), in even -- next square puts operator for two zero bytes in  in crc32c_zeros_op()
165 	/* answer ended up in odd -- copy to even */  in crc32c_zeros_op()
172  * for that length, byte-by-byte on the operand.
216 /* Compute CRC-32C using the Intel hardware instruction. */
236 	while (len && ((uintptr_t)next & (align - 1)) != 0) {  in sse42_crc32c()
239 		len--;  in sse42_crc32c()
245 	 * crc instructions, each on LONG bytes -- this is optimized for the  in sse42_crc32c()
271 		/*-  in sse42_crc32c()
273 		 * loop.  'crc' is used to accumulate crc0 and crc1  in sse42_crc32c()
296 		 * can run in 24 cycles, so the 3-way blocking is worse  in sse42_crc32c()
319 		len -= LONG * 3;  in sse42_crc32c()
354 		len -= SHORT * 3;  in sse42_crc32c()
359 	end = next + (len - (len & (align - 1)));  in sse42_crc32c()
368 	len &= (align - 1);  in sse42_crc32c()
374 		len--;  in sse42_crc32c()