dillonhuff
dillonhuff
@rdaly525 @jeffsetter I think we agreed last year that we were going to deprecate the c++ backend, but the tests never got removed. They take a long time and generate...
@rdaly525 @jeffsetter is anyone actually using that functionality or can I remove it? It si the source of significant performance issues.
@rdaly525 I fetched the upstream of coreir, checked out dev and tried to build the testbench with a new floating point test, but when I did ```make test``` I got...
@jeffsetter now unsharp has an error in its compute units: ```bash ERROR in compute unit: hcompute_ratio_stencil in0_blur_unnormalized_stencil[0] -> 63743 in0_blur_unnormalized_stencil[0] -> 1111100011111111 in1_gray_stencil[0] -> 2 in1_gray_stencil[0] -> 0000000000000010 coreir_result: 255...
@jeffsetter another compute unit inconsistency in a different harris variant: ```bash ERROR in compute unit: hcompute_lgxx_stencil_1 in0_lgxx_stencil[0] -> 236 in0_lgxx_stencil[0] -> 0000000011101100 in1_padded16_global_wrapper_stencil[0] -> 41 in1_padded16_global_wrapper_stencil[0] -> 0000000000101001 in1_padded16_global_wrapper_stencil[1] ->...
In prog: harris_sch4_1pp3c compute unit: hcompute_cim_stencil has a mismatch between C++ and coreir
@jeffsetter @rdaly525 I'm guessing this is related to #108 but there is an error in one of the Harris sch4 compute units: ```bash ERROR in compute unit: hcompute_cim_stencil in0_lgxx_stencil[0] ->...
@jeffsetter I have written a compute unit regression test generator and used it to compare the outputs of the coreir and cpp compute units in isolation for each compute unit...
@jeffsetter it seems that the ROM outputs for the 1 pix / 3 cycles and 1 pix / 9 cycles coreir output is still not correct: ```bash %Error: harris_sch4_1pp3c.v:3137:98: Too...