+ Time (s) vs. N (NxN submatrix, PDL: Double D [1500,1500] matrix) +-----------------------------------------------------------+ 2 |-+ + + + + + + + + +-| | C C | | C | | C | | D D D D C | 1.5 |-+ A C +-| | D D | | C | | C D | | | | B D D | 1 |-+ +-| | C | | C | | D D | | | | A D | 0.5 |-+ C +-| | D | | | | B D | | | | DC | 0 |-+ DD +-| | + + + + + + + + | +-----------------------------------------------------------+ 0 200 400 600 800 1000 1200 1400 sms_WxH_PDL_naive A sms_WxH_pdlpp_4loops B sms_WxH_pdlpp_3loops C sms_WxH_PDL_lags D +------+-------+-------+-------+-------+ | N | A | B | C | D | +------+-------+-------+-------+-------+ | 2 | 0.038 | 0.009 | 0.009 | 0.047 | | 4 | 0.294 | 0.031 | 0.019 | 0.041 | | 6 | 0.572 | 0.072 | 0.019 | 0.044 | | 10 | 1.531 | 0.250 | 0.031 | 0.031 | | 20 | | 1.150 | 0.072 | 0.022 | | 100 | | | 0.494 | 0.394 | | 200 | | | 0.931 | 0.797 | | 300 | | | 1.263 | 1.153 | | 400 | | | 1.544 | 1.431 | | 500 | | | 1.756 | 1.594 | | 600 | | | 1.869 | 1.653 | | 700 | | | 1.872 | 1.641 | | 800 | | | 1.788 | 1.594 | | 900 | | | 1.597 | 1.481 | | 1000 | | | 1.362 | 1.288 | | 1100 | | | 1.144 | 1.081 | | 1200 | | | 0.841 | 0.809 | | 1300 | | | 0.575 | 0.572 | | 1400 | | | 0.291 | 0.272 | +------+-------+-------+-------+-------+