diff options
Diffstat (limited to 'cachelab-handout/trans.c')
| -rw-r--r-- | cachelab-handout/trans.c | 47 |
1 files changed, 8 insertions, 39 deletions
diff --git a/cachelab-handout/trans.c b/cachelab-handout/trans.c index 8ce61ac..8f347aa 100644 --- a/cachelab-handout/trans.c +++ b/cachelab-handout/trans.c @@ -120,34 +120,14 @@ void transpose_submit(int M, int N, int A[N][M], int B[M][N]) } // A00 -> B00 - j2 = 0; - t4 = 0; - for (i2 = 0; i2 < CASE_64_64_BLOCK_SIZE; i2++) - { - t0 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 0]; - t1 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 1]; - t2 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 2]; - t3 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 3]; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 0][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t0; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 1][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t1; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 2][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t2; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 3][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t3; - } + TRANSPOSE_4_BY_4(i*8 + 0*4, j*8 + 0*4, + j*8 + 0*4, i*8 + 0*4, i2); + // A01 -> temp B01 - j2 = 0; - t4 = 4; - for (i2 = 0; i2 < CASE_64_64_BLOCK_SIZE; i2++) - { - t0 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 0]; - t1 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 1]; - t2 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 2]; - t3 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 3]; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + 0][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + i2] = t0; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + 1][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + i2] = t1; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + 2][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + i2] = t2; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + 3][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + i2] = t3; - } + TRANSPOSE_4_BY_4(i*8 + 0*4, j*8 + 1*4, + j*8 + 0*4, i*8 + 1*4, i2); + // B01 -> temp var, A10 -> B01, temp var -> B10 j2 = 4, t4 = 0; @@ -193,19 +173,8 @@ void transpose_submit(int M, int N, int A[N][M], int B[M][N]) } // A11 -> B11 - j2 = 4; - t4 = 4; - for (i2 = 0; i2 < CASE_64_64_BLOCK_SIZE; i2++) - { - t0 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 0]; - t1 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 1]; - t2 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 2]; - t3 = A[i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2][j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 3]; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 0][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t0; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 1][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t1; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 2][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t2; - B[j * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + t4 + 3][i * CASE_64_64_MAGIC_NUMBER * CASE_64_64_BLOCK_SIZE + j2 + i2] = t3; - } + TRANSPOSE_4_BY_4(i*8 + 1*4, j*8 + 1*4, + j*8 + 1*4, i*8 + 1*4, i2); } } } |