O循环展开4次( (straight forward way) Loop: LD Fo,O(R1)stall ADDD F4F0F2 stal stall SD 0(R1),F4 g drop sUBl BNEZ 3456789 LD F6, 8(R1)stall ADDD F8.F6F2 stallstall SD -8(R1,F8 drop sUBI BNEZ LD F10,-16(R1) stall ADDD F12F10F2 stall stall SD -16(R1),F12 drop sUBl BNEZ 10 LD F14, 24(R1)stall ADDD F16F14.F2 stall stall 12 SD -24(R1),F16 34 SUB R1R132 stall alter to 4*8 BNEZ R1 LOOP 15 NOP Rewrite loop to minimize stalls? 15+4x(1+2)+1=28 cycles, or 7 per iteration Assumes R1 is multiple of 4Rewrite loop to minimize stalls? 1 Loop: LD F0,0(R1) stall 2 ADDD F4,F0,F2 stall stall 3 SD 0(R1),F4 ;drop SUBI & BNEZ 4 LD F6,-8(R1) stall 5 ADDD F8,F6,F2 stall stall 6 SD -8(R1),F8 ;drop SUBI & BNEZ 7 LD F10,-16(R1) stall 8 ADDD F12,F10,F2 stall stall 9 SD -16(R1),F12 ;drop SUBI & BNEZ 10 LD F14,-24(R1) stall 11 ADDD F16,F14,F2 stall stall 12 SD -24(R1),F16 13 SUBI R1,R1,#32 stall ;alter to 4*8 14 BNEZ R1,LOOP 15 NOP 15 + 4 x (1+2) + 1 = 28 cycles, or 7 per iteration Assumes R1 is multiple of 4 2021/2/7 14 循环展开4次(straight forward way)