SAVE4x12\r
\r
/* here for the prefetch of next b source block */\r
- /* the increment should be proportional to the ratio of GEMM_Q/GEMM_P */\r
+ /* the increment should be proportional to GEMM_Q/GEMM_P */\r
\r
salq $3, K\r
#ifdef WINDOWS_ABI /* GEMM_P == GEMM_Q * 4 */\r
SAVE4x12\r
\r
/* here for the prefetch of next b source block */\r
- /* the increment should be proportional to the ratio of GEMM_Q/GEMM_P */\r
+ /* the increment should be proportional to GEMM_Q/GEMM_P */\r
\r
salq $3, K\r
#ifdef WINDOWS_ABI /* GEMM_P == GEMM_Q * 4 */\r
prefetcht2 (B)\r
prefetcht2 (B, K, 8)\r
- addq $64, B\r
+ addq $64, B /* increment */\r
#else /* GEMM_P == GEMM_Q * 2 under linux x86_64 */\r
prefetcht2 (B)\r
prefetcht2 (B, K, 8)\r
prefetcht2 64(B)\r
prefetcht2 64(B, K, 8)\r
- addq $128, B\r
+ addq $128, B /* increment */\r
#endif\r
sarq $3, K\r
\r