; ; MMX32 iDCT algorithm (IEEE-1180 compliant) :: idct_mmx32() ; ; MPEG2AVI ; -------- ; v0.16B33 initial release ; ; This was one of the harder pieces of work to code. ; Intel's app-note focuses on the numerical issues of the algorithm, but ; assumes the programmer is familiar with IDCT mathematics, leaving the ; form of the complete function up to the programmer's imagination. ; ; ALGORITHM OVERVIEW ; ------------------ ; I played around with the code for quite a few hours. I came up ; with *A* working IDCT algorithm, however I'm not sure whether my routine ; is "the correct one." But rest assured, my code passes all six IEEE ; accuracy tests with plenty of margin. ; ; My IDCT algorithm consists of 4 steps: ; ; 1) IDCT-row transformation (using the IDCT-row function) on all 8 rows ; This yields an intermediate 8x8 matrix. ; ; 2) intermediate matrix transpose (mandatory) ; ; 3) IDCT-row transformation (2nd time) on all 8 rows of the intermediate ; matrix. The output is the final-result, in transposed form. ; ; 4) post-transformation matrix transpose ; (not necessary if the input-data is already transposed, this could ; be done during the MPEG "zig-zag" scan, but since my algorithm ; requires at least one transpose operation, why not re-use the ; transpose-code.) ; ; Although the (1st) and (3rd) steps use the SAME row-transform operation, ; the (3rd) step uses different shift&round constants (explained later.) ; ; Also note that the intermediate transpose (2) would not be neccessary, ; if the subsequent operation were a iDCT-column transformation. Since ; we only have the iDCT-row transform, we transpose the intermediate ; matrix and use the iDCT-row transform a 2nd time. ; ; I had to change some constants/variables for my method to work : ; ; As given by Intel, the #defines for SHIFT_INV_COL and RND_INV_COL are ; wrong. Not surprising since I'm not using a true column-transform ; operation, but the row-transform operation (as mentioned earlier.) ; round_inv_col[], which is given as "4 short" values, should have the ; same dimensions as round_inv_row[]. The corrected variables are ; shown. ; ; Intel's code defines a different table for each each row operation. ; The tables given are 0/4, 1/7, 2/6, and 5/3. My code only uses row#0. ; Using the other rows messes up the overall transform. ; ; IMPLEMENTATION DETAILs ; ---------------------- ; ; I divided the algorithm's work into two subroutines, ; 1) idct_mmx32_rows() - transforms 8 rows, then transpose ; 2) idct_mmx32_cols() - transforms 8 rows, then transpose ; yields final result ("drop-in" direct replacement for INT32 IDCT) ; ; The 2nd function is a clone of the 1st, with changes made only to the ; shift&rounding instructions. ; ; In the 1st function (rows), the shift & round instructions use ; SHIFT_INV_ROW & round_inv_row[] (renamed to r_inv_row[]) ; ; In the 2nd function (cols)-> r_inv_col[], and ; SHIFT_INV_COL & round_inv_col[] (renamed to r_inv_col[]) ; ; Each function contains an integrated transpose-operator, which comes ; AFTER the primary transformation operation. In the future, I'll optimize ; the code to do more of the transpose-work "in-place". Right now, I've ; left the code as two subroutines and a main calling function, so other ; people can read the code more easily. ; ; liaor@umcc.ais.org http:;members.tripod.com/~liaor ; ;;; ;;; A.Stevens Jul 2000 easy-peasy quick port to nasm ;;; Isn't open source a sensible idea... ;;; ;============================================================================= ; ; AP-922 http:;developer.intel.com/vtune/cbts/strmsimd ; These examples contain code fragments for first stage iDCT 8x8 ; (for rows) and first stage DCT 8x8 (for columns) ; ;============================================================================ %define INP eax ; pointer to (short *blk) %define OUT ecx ; pointer to output (temporary store space qwTemp[]) %define TABLE ebx ; pointer to idct_tab_01234567[] %define round_inv_row edx %define round_inv_col edx %define ROW_STRIDE 16 ; for 8x8 matrix transposer %define BITS_INV_ACC 4 ; 4 or 5 for IEEE %define SHIFT_INV_ROW (16 - BITS_INV_ACC) %define SHIFT_INV_COL (1 + BITS_INV_ACC +14 ) ; changed from Intel's val) ;; ;; Variables and tables defined in C for convenience ;; extern idct_r_inv_row ; 2 DWORDSs extern idct_r_inv_col ; " extern idct_r_inv_corr ; " extern idct_tab_01234567 ; Catenated table of coefficients ;; ;; private variables and functions ;; SECTION .bss align 16 ; qwTemp: resw 64 ; temporary storage space, 8x8 of shorts SECTION .text ;; static void idct_mmx( short *blk global idct_mmx idct_mmx: push ebp ; save frame pointer mov ebp, esp ; link push ebx push ecx push edx push edi ;; ;; transform all 8 rows of 8x8 iDCT block ;; ; this subroutine performs two operations ; 1) iDCT row transform ; for( i = 0; i < 8; ++ i) ; DCT_8_INV_ROW_1( blk[i*8], qwTemp[i] ); ; ; 2) transpose the matrix (which was stored in qwTemp[]) ; qwTemp[] -> [8x8 matrix transpose] -> blk[] mov INP, [ebp+8] ; INP = blk mov edi, 0x00; ; x = 0 lea TABLE,[idct_tab_01234567]; ; row 0 ; lea OUT, [qwTemp]; mov OUT, [ebp+12]; lea round_inv_row, [idct_r_inv_row] jmp lpa ; for ( x = 0; x < 8; ++x ) ; transform one row per iteration align 32 lpa: movq mm0, [INP] ; 0 ; x3 x2 x1 x0 movq mm1, [INP+8] ; 1 ; x7 x6 x5 x4 movq mm2, mm0 ; ; 2 ; x3 x2 x1 x0 movq mm3, [TABLE] ; 3 ; w06 w04 w02 w00 punpcklwd mm0, mm1 ; x5 x1 x4 x0 ; ---------- movq mm5, mm0 ; ; 5 ; x5 x1 x4 x0 punpckldq mm0, mm0 ; ; x4 x0 x4 x0 movq mm4, [TABLE+8] ; ; 4 ; w07 w05 w03 w01 punpckhwd mm2, mm1 ; ; 1 ; x7 x3 x6 x2 pmaddwd mm3, mm0 ; ; x4*w06+x0*w04 x4*w02+x0*w00 movq mm6, mm2 ; ; 6 ; x7 x3 x6 x2 movq mm1, [TABLE+32] ;; 1 ; w22 w20 w18 w16 punpckldq mm2, mm2 ; ; x6 x2 x6 x2 pmaddwd mm4, mm2 ; ; x6*w07+x2*w05 x6*w03+x2*w01 punpckhdq mm5, mm5 ; ; x5 x1 x5 x1 pmaddwd mm0, [TABLE+16] ;; x4*w14+x0*w12 x4*w10+x0*w08 punpckhdq mm6, mm6 ; ; x7 x3 x7 x3 movq mm7, [TABLE+40] ;; 7 ; w23 w21 w19 w17 pmaddwd mm1, mm5 ; ; x5*w22+x1*w20 x5*w18+x1*w16 paddd mm3, [round_inv_row];; +rounder pmaddwd mm7, mm6 ; ; x7*w23+x3*w21 x7*w19+x3*w17 pmaddwd mm2, [TABLE+24] ;; x6*w15+x2*w13 x6*w11+x2*w09 paddd mm3, mm4 ; ; 4 ; a1=sum(even1) a0=sum(even0) pmaddwd mm5, [TABLE+48] ;; x5*w30+x1*w28 x5*w26+x1*w24 movq mm4, mm3 ; ; 4 ; a1 a0 pmaddwd mm6, [TABLE+56] ;; x7*w31+x3*w29 x7*w27+x3*w25 paddd mm1, mm7 ; ; 7 ; b1=sum(odd1) b0=sum(odd0) paddd mm0, [round_inv_row];; +rounder psubd mm3, mm1 ; ; a1-b1 a0-b0 psrad mm3, SHIFT_INV_ROW ; ; y6=a1-b1 y7=a0-b0 paddd mm1, mm4 ; ; 4 ; a1+b1 a0+b0 paddd mm0, mm2 ; ; 2 ; a3=sum(even3) a2=sum(even2) psrad mm1, SHIFT_INV_ROW ; ; y1=a1+b1 y0=a0+b0 paddd mm5, mm6 ; ; 6 ; b3=sum(odd3) b2=sum(odd2) movq mm4, mm0 ; ; 4 ; a3 a2 paddd mm0, mm5 ; ; a3+b3 a2+b2 psubd mm4, mm5 ; ; 5 ; a3-b3 a2-b2 add INP, 16; ; increment INPUT pointer -> row 1 psrad mm4, SHIFT_INV_ROW ; ; y4=a3-b3 y5=a2-b2 ; add TABLE, 0; ; TABLE += 64 -> row 1 psrad mm0, SHIFT_INV_ROW ; ; y3=a3+b3 y2=a2+b2 ; movq mm2, [INP] ; ; row+1; 0; x3 x2 x1 x0 packssdw mm4, mm3 ; ; 3 ; y6 y7 y4 y5 packssdw mm1, mm0 ; ; 0 ; y3 y2 y1 y0 movq mm7, mm4 ; ; 7 ; y6 y7 y4 y5 ; movq mm0, mm2 ; ; row+1; 2 ; x3 x2 x1 x0 psrld mm4, 16 ; ; 0 y6 0 y4 movq [OUT], mm1 ; ; 1 ; save y3 y2 y1 y0 pslld mm7, 16 ; ; y7 0 y5 0 ; movq mm1, [INP+8] ; ; row+1; 1 ; x7 x6 x5 x4 por mm7, mm4 ; ; 4 ; y7 y6 y5 y4 movq mm3, [TABLE] ; ; 3 ; w06 w04 w02 w00 ; punpcklwd mm0, mm1 ; ; row+1; x5 x1 x4 x0 ; begin processing row 1 movq [OUT+8], mm7 ; ; 7 ; save y7 y6 y5 y4 add edi, 0x01; add OUT, 16; ; increment OUTPUT pointer -> row 1 cmp edi, 0x08; jl near lpa; ; end for ( x = 0; x < 8; ++x ) ; done with the iDCT row-transformation ; now we have to transpose the output 8x8 matrix ; 8x8 (OUT) -> 8x8't' (IN) ; the transposition is implemented as 4 sub-operations. ; 1) transpose upper-left quad ; 2) transpose lower-right quad ; 3) transpose lower-left quad ; 4) transpose upper-right quad ; mm0 = 1st row [ A B C D ] row1 ; mm1 = 2nd row [ E F G H ] 2 ; mm2 = 3rd row [ I J K L ] 3 ; mm3 = 4th row [ M N O P ] 4 ; 1) transpose upper-left quad ; lea OUT, [qwTemp]; mov OUT, [ebp+12]; movq mm0, [OUT + ROW_STRIDE * 0 ] movq mm1, [OUT + ROW_STRIDE * 1 ] movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE * 2 ] punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] movq mm3, [OUT + ROW_STRIDE * 3] punpckhwd mm4, mm1; ; mm4 = [ 2 6 3 7] movq mm6, mm2; punpcklwd mm2, mm3; ; mm2 = [ 8 12 9 13] punpckhwd mm6, mm3; ; mm6 = 10 14 11 15] movq mm1, mm0; ; mm1 = [ 0 4 1 5] mov INP, [ebp+8]; ; load input address punpckldq mm0, mm2; ; final result mm0 = row1 [0 4 8 12] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] movq [ INP + ROW_STRIDE * 0 ], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] ; begin reading next quadrant (lower-right) movq mm0, [OUT + ROW_STRIDE*4 + 8]; punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE * 2], mm4; ; store row 3 movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq [ INP +ROW_STRIDE * 1], mm1; ; store row 2 movq mm1, [OUT + ROW_STRIDE*5 + 8] movq [ INP +ROW_STRIDE * 3], mm3; ; store row 4 punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] ; 2) transpose lower-right quadrant ; movq mm0, [OUT + ROW_STRIDE*4 + 8] ; movq mm1, [OUT + ROW_STRIDE*5 + 8] ; movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE*6 + 8] ; punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] punpckhwd mm4, mm1; ; mm4 = [ 2 6 3 7] movq mm3, [OUT + ROW_STRIDE*7 + 8] movq mm6, mm2; punpcklwd mm2, mm3; ; mm2 = [ 8 12 9 13] movq mm1, mm0; ; mm1 = [ 0 4 1 5] punpckhwd mm6, mm3; ; mm6 = 10 14 11 15] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckldq mm0, mm2; ; final result mm0 = row1 [0 4 8 12] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] ; ; slot movq [ INP + ROW_STRIDE*4 + 8], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] movq mm0, [OUT + ROW_STRIDE * 4 ] punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE*6 + 8], mm4; ; store row 3 movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq [ INP +ROW_STRIDE*5 + 8], mm1; ; store row 2 ; ; slot movq mm1, [OUT + ROW_STRIDE * 5 ] ; ; slot movq [ INP +ROW_STRIDE*7 + 8], mm3; ; store row 4 punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] ; 3) transpose lower-left ; movq mm0, [OUT + ROW_STRIDE * 4 ] ; movq mm1, [OUT + ROW_STRIDE * 5 ] ; movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE * 6 ] ; punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] punpckhwd mm4, mm1; ; mm4 = [ 2 6 3 7] movq mm3, [OUT + ROW_STRIDE * 7 ] movq mm6, mm2; punpcklwd mm2, mm3; ; mm2 = [ 8 12 9 13] movq mm1, mm0; ; mm1 = [ 0 4 1 5] punpckhwd mm6, mm3; ; mm6 = 10 14 11 15] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckldq mm0, mm2; ; final result mm0 = row1 [0 4 8 12] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] ; ; slot movq [ INP + ROW_STRIDE * 0 + 8 ], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] ; begin reading next quadrant (upper-right) movq mm0, [OUT + ROW_STRIDE*0 + 8]; punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE * 2 + 8], mm4; ; store row 3 movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq [ INP +ROW_STRIDE * 1 + 8 ], mm1; ; store row 2 movq mm1, [OUT + ROW_STRIDE*1 + 8] movq [ INP +ROW_STRIDE * 3 + 8], mm3; ; store row 4 punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] ; 2) transpose lower-right quadrant ; movq mm0, [OUT + ROW_STRIDE*4 + 8] ; movq mm1, [OUT + ROW_STRIDE*5 + 8] ; movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE*2 + 8] ; punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] punpckhwd mm4, mm1; ; mm4 = [ 2 6 3 7] movq mm3, [OUT + ROW_STRIDE*3 + 8] movq mm6, mm2; punpcklwd mm2, mm3; ; mm2 = [ 8 12 9 13] movq mm1, mm0; ; mm1 = [ 0 4 1 5] punpckhwd mm6, mm3; ; mm6 = 10 14 11 15] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckldq mm0, mm2; ; final result mm0 = row1 [0 4 8 12] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] ; ; slot movq [ INP + ROW_STRIDE*4 ], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] movq [ INP +ROW_STRIDE*5 ], mm1; ; store row 2 punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE*6 ], mm4; ; store row 3 ; ; slot movq [ INP +ROW_STRIDE*7 ], mm3; ; store row 4 ; Conceptually this is the column transform. ; Actually, the matrix is transformed ; row by row. This function is identical to idct_mmx32_rows(), ; except for the SHIFT amount and ROUND_INV amount. ; this subroutine performs two operations ; 1) iDCT row transform ; for( i = 0; i < 8; ++ i) ; DCT_8_INV_ROW_1( blk[i*8], qwTemp[i] ); ; ; 2) transpose the matrix (which was stored in qwTemp[]) ; qwTemp[] -> [8x8 matrix transpose] -> blk[] mov INP, [ebp+8]; ; ; row 0 mov edi, 0x00; ; x = 0 lea TABLE, [idct_tab_01234567]; ; row 0 ; lea OUT, [qwTemp]; mov OUT, [ebp+12]; ; mov OUT, INP; ; algorithm writes data in-place -> row 0 lea round_inv_col, [idct_r_inv_col] jmp acc_idct_colloop1 ; for ( x = 0; x < 8; ++x ) ; transform one row per iteration align 32 acc_idct_colloop1: movq mm0, [INP] ; ; 0 ; x3 x2 x1 x0 movq mm1, [INP+8] ; ; 1 ; x7 x6 x5 x4 movq mm2, mm0 ; ; 2 ; x3 x2 x1 x0 movq mm3, [TABLE] ; ; 3 ; w06 w04 w02 w00 punpcklwd mm0, mm1 ; ; x5 x1 x4 x0 ; ---------- movq mm5, mm0 ; ; 5 ; x5 x1 x4 x0 punpckldq mm0, mm0 ; ; x4 x0 x4 x0 movq mm4, [TABLE+8] ; ; 4 ; w07 w05 w03 w01 punpckhwd mm2, mm1 ; ; 1 ; x7 x3 x6 x2 pmaddwd mm3, mm0 ; ; x4*w06+x0*w04 x4*w02+x0*w00 movq mm6, mm2 ; ; 6 ; x7 x3 x6 x2 movq mm1, [TABLE+32] ;; 1 ; w22 w20 w18 w16 punpckldq mm2, mm2 ; ; x6 x2 x6 x2 pmaddwd mm4, mm2 ; ; x6*w07+x2*w05 x6*w03+x2*w01 punpckhdq mm5, mm5 ; ; x5 x1 x5 x1 pmaddwd mm0, [TABLE+16] ;; x4*w14+x0*w12 x4*w10+x0*w08 punpckhdq mm6, mm6 ; ; x7 x3 x7 x3 movq mm7, [TABLE+40] ;; 7 ; w23 w21 w19 w17 pmaddwd mm1, mm5 ; ; x5*w22+x1*w20 x5*w18+x1*w16 paddd mm3, [round_inv_col] ;; +rounder pmaddwd mm7, mm6 ; ; x7*w23+x3*w21 x7*w19+x3*w17 pmaddwd mm2, [TABLE+24] ;; x6*w15+x2*w13 x6*w11+x2*w09 paddd mm3, mm4 ; ; 4 ; a1=sum(even1) a0=sum(even0) pmaddwd mm5, [TABLE+48] ;; x5*w30+x1*w28 x5*w26+x1*w24 movq mm4, mm3 ; ; 4 ; a1 a0 pmaddwd mm6, [TABLE+56] ;; x7*w31+x3*w29 x7*w27+x3*w25 paddd mm1, mm7 ; ; 7 ; b1=sum(odd1) b0=sum(odd0) paddd mm0, [round_inv_col] ;; +rounder psubd mm3, mm1 ; ; a1-b1 a0-b0 psrad mm3, SHIFT_INV_COL; ; y6=a1-b1 y7=a0-b0 paddd mm1, mm4 ; ; 4 ; a1+b1 a0+b0 paddd mm0, mm2 ; ; 2 ; a3=sum(even3) a2=sum(even2) psrad mm1, SHIFT_INV_COL; ; y1=a1+b1 y0=a0+b0 paddd mm5, mm6 ; ; 6 ; b3=sum(odd3) b2=sum(odd2) movq mm4, mm0 ; ; 4 ; a3 a2 paddd mm0, mm5 ; ; a3+b3 a2+b2 psubd mm4, mm5 ; ; 5 ; a3-b3 a2-b2 add INP, 16; ; increment INPUT pointer -> row 1 psrad mm4, SHIFT_INV_COL; ; y4=a3-b3 y5=a2-b2 add TABLE, 0; ; TABLE += 64 -> row 1 psrad mm0, SHIFT_INV_COL; ; y3=a3+b3 y2=a2+b2 ; movq mm2, [INP] ; ; row+1; 0; x3 x2 x1 x0 packssdw mm4, mm3 ; ; 3 ; y6 y7 y4 y5 packssdw mm1, mm0 ; ; 0 ; y3 y2 y1 y0 movq mm7, mm4 ; ; 7 ; y6 y7 y4 y5 ; movq mm0, mm2 ; ; row+1; 2 ; x3 x2 x1 x0 ; por mm1, dct_one_corr ; ; correction y2 +0.5 psrld mm4, 16 ; ; 0 y6 0 y4 movq [OUT], mm1 ; ; 1 ; save y3 y2 y1 y0 pslld mm7, 16 ; ; y7 0 y5 0 ; movq mm1, [INP+8] ; ; row+1; 1 ; x7 x6 x5 x4 ; por mm7, dct_one_corr ; ; correction y2 +0.5 por mm7, mm4 ; ; 4 ; y7 y6 y5 y4 ; movq mm3, [TABLE] ; ; 3 ; w06 w04 w02 w00 ; punpcklwd mm0, mm1 ; ; row+1; x5 x1 x4 x0 ; begin processing row 1 movq [OUT+8], mm7 ; ; 7 ; save y7 y6 y5 y4 add edi, 0x01; add OUT, 16; cmp edi, 0x08; ; compare x <> 8 jl near acc_idct_colloop1; ; end for ( x = 0; x < 8; ++x ) ; done with the iDCT column-transformation ; now we have to transpose the output 8x8 matrix ; 8x8 (OUT) -> 8x8't' (IN) ; the transposition is implemented as 4 sub-operations. ; 1) transpose upper-left quad ; 2) transpose lower-right quad ; 3) transpose lower-left quad ; 4) transpose upper-right quad ; mm0 = 1st row [ A B C D ] row1 ; mm1 = 2nd row [ E F G H ] 2 ; mm2 = 3rd row [ I J K L ] 3 ; mm3 = 4th row [ M N O P ] 4 ; 1) transpose upper-left quad ; lea OUT, [qwTemp]; mov OUT, [ebp+12]; movq mm0, [OUT + ROW_STRIDE * 0 ] movq mm1, [OUT + ROW_STRIDE * 1 ] movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE * 2 ] punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] movq mm3, [OUT + ROW_STRIDE * 3] punpckhwd mm4, mm1 ; mm4 = [ 2 6 3 7] movq mm6, mm2 punpcklwd mm2, mm3 ; mm2 = [ 8 12 9 13] punpckhwd mm6, mm3 ; mm6 = 10 14 11 15] movq mm1, mm0 ; mm1 = [ 0 4 1 5] mov INP, [ebp+8] ; load input address punpckldq mm0, mm2 ; final result mm0 = row1 [0 4 8 12] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] movq [ INP + ROW_STRIDE * 0 ], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] ; begin reading next quadrant (lower-right) movq mm0, [OUT + ROW_STRIDE*4 + 8]; punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE * 2], mm4; ; store row 3 movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq [ INP +ROW_STRIDE * 1], mm1; ; store row 2 movq mm1, [OUT + ROW_STRIDE*5 + 8] movq [ INP +ROW_STRIDE * 3], mm3; ; store row 4 punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] ; 2) transpose lower-right quadrant ; movq mm0, [OUT + ROW_STRIDE*4 + 8] ; movq mm1, [OUT + ROW_STRIDE*5 + 8] ; movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE*6 + 8] ; punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] punpckhwd mm4, mm1; ; mm4 = [ 2 6 3 7] movq mm3, [OUT + ROW_STRIDE*7 + 8] movq mm6, mm2; punpcklwd mm2, mm3; ; mm2 = [ 8 12 9 13] movq mm1, mm0; ; mm1 = [ 0 4 1 5] punpckhwd mm6, mm3; ; mm6 = 10 14 11 15] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckldq mm0, mm2; ; final result mm0 = row1 [0 4 8 12] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] ; ; slot movq [ INP + ROW_STRIDE*4 + 8], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] movq mm0, [OUT + ROW_STRIDE * 4 ] punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE*6 + 8], mm4; ; store row 3 movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq [ INP +ROW_STRIDE*5 + 8], mm1; ; store row 2 ; ; slot movq mm1, [OUT + ROW_STRIDE * 5 ] ; ; slot movq [ INP +ROW_STRIDE*7 + 8], mm3; ; store row 4 punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] ; 3) transpose lower-left ; movq mm0, [OUT + ROW_STRIDE * 4 ] ; movq mm1, [OUT + ROW_STRIDE * 5 ] ; movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE * 6 ] ; punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] punpckhwd mm4, mm1; ; mm4 = [ 2 6 3 7] movq mm3, [OUT + ROW_STRIDE * 7 ] movq mm6, mm2; punpcklwd mm2, mm3; ; mm2 = [ 8 12 9 13] movq mm1, mm0; ; mm1 = [ 0 4 1 5] punpckhwd mm6, mm3; ; mm6 = 10 14 11 15] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckldq mm0, mm2; ; final result mm0 = row1 [0 4 8 12] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] ; ; slot movq [ INP + ROW_STRIDE * 0 + 8 ], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] ; begin reading next quadrant (upper-right) movq mm0, [OUT + ROW_STRIDE*0 + 8]; punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE * 2 + 8], mm4; ; store row 3 movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq [ INP +ROW_STRIDE * 1 + 8 ], mm1; ; store row 2 movq mm1, [OUT + ROW_STRIDE*1 + 8] movq [ INP +ROW_STRIDE * 3 + 8], mm3; ; store row 4 punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] ; 2) transpose lower-right quadrant ; movq mm0, [OUT + ROW_STRIDE*4 + 8] ; movq mm1, [OUT + ROW_STRIDE*5 + 8] ; movq mm4, mm0; ; mm4 = copy of row1[A B C D] movq mm2, [OUT + ROW_STRIDE*2 + 8] ; punpcklwd mm0, mm1; ; mm0 = [ 0 4 1 5] punpckhwd mm4, mm1; ; mm4 = [ 2 6 3 7] movq mm3, [OUT + ROW_STRIDE*3 + 8] movq mm6, mm2; punpcklwd mm2, mm3; ; mm2 = [ 8 12 9 13] movq mm1, mm0; ; mm1 = [ 0 4 1 5] punpckhwd mm6, mm3; ; mm6 = 10 14 11 15] movq mm3, mm4; ; mm3 = [ 2 6 3 7] punpckldq mm0, mm2; ; final result mm0 = row1 [0 4 8 12] punpckhdq mm1, mm2; ; mm1 = final result mm1 = row2 [1 5 9 13] ; ; slot movq [ INP + ROW_STRIDE*4 ], mm0; ; store row 1 punpckldq mm4, mm6; ; final result mm4 = row3 [2 6 10 14] movq [ INP +ROW_STRIDE*5 ], mm1; ; store row 2 punpckhdq mm3, mm6; ; final result mm3 = row4 [3 7 11 15] movq [ INP +ROW_STRIDE*6 ], mm4; ; store row 3 ; ; slot movq [ INP +ROW_STRIDE*7 ], mm3; ; store row 4 pop edi pop edx pop ecx pop ebx pop ebp ; restore frame pointer emms ret