blank result from CUDA about hiperc HOT 1 CLOSED

Testing the convolution kernel:

• Created a pattern variable, scramble = double(src_row * src_col) / (nx * ny)```
• Substituting C[j][i] = scramble + value with scramble = double(tx+ty)/(blockDim.x*blockDim.y) produces a tiled gradient pattern.
  • This tends to exonerate the mask; rather, it appears the input data is set to 0 everywhere.
• In the initialization of N_ds, setting N_ds[ty][tx] = scramble instead of 0 produces a scrambled pattern.
  • This suggests the initialization conditions for N_ds are incorrect: it should be mostly C_0, not mostly 0.
• Focus on correct indexing.
• c goes negative when the tile size is less than 32, and greater than 1 otherwise.

from hiperc.