Rather than up to dozens of individual files exposing part of each
copy engine's configuration, have one file which exposes a unified
view of the full topology. Example new output on RTX 2080 Ti:
$ cat /proc/gpu0/copy_topology
GRCE0 -> LCE04
GRCE1 -> LCE03
LCE02 -> PCE02
LCE03 -> PCE03
LCE04 -> PCE01
Old output:
$ tail -n 1 /proc/gpu0/lce_for_pce*
==> /proc/gpu0/lce_for_pce0 <==
0xf
==> /proc/gpu0/lce_for_pce1 <==
0x4
==> /proc/gpu0/lce_for_pce2 <==
0x2
==> /proc/gpu0/lce_for_pce3 <==
0x3
$ tail -n 1 /proc/gpu1/shared_lce_for_grce*
==> /proc/gpu0/shared_lce_for_grce0 <==
0x4
==> /proc/gpu0/shared_lce_for_grce1 <==
0x3
Specifically:
- Add `copy_topology` API
- Remove `shared_lce_for_grce#` and `lce_for_pce#` APIs
- Move logic from `nvdebug_entry.c` to `copy_topology_procfs.c`
- Do not print PCE or Shared LCE configuration if flagged absent
- Refer to LCE0 and LCE1 as GRCE0 and GRCE1
- Print by LCE ID, which is move helpful when attempting to trace
how a given copy runlist maps to a physical copy engine.
- Document two errata with CE registers
Tested working on Pascal Integrated, Pascal, Volta Integrated
Volta, Turing, and Ampere Integrated on Linux 4.9 through 5.10.
id='n94' href='#n94'>94
95
96
97
98
99
/* compile with
rm liblagrcv.a
gcc -c lagrcv.cpp
ar rc liblagrcv.a lagrcv.o
ranlib liblagrcv.a
mex7 calcTextureMex.cc -L/home/ikkjin/LagrMatlab/opencv/matlab -llagrcv -I/home/ikkjin/LagrMatlab/opencv/matlab/
*/
#include "mex.h"
#include "lagrcv.h"
#include <stdio.h>
#include <math.h>
#ifndef MAX_LEVEL
# define MAX_LEVEL 5
#endif
// TODO: add number of corners parameter
void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) {
// usage: [newFeaturePnt validFlag ] =
// calcOptFlowPyrLKMex(Ipyr, IdxPyr, IdyPyr, Jpyr, featurePnt, winSize, accuracy_th, max_iter)
// Ipyr, IdxPyr, IdyPyr, Jpyr: levelx1 size cell.
// featurePnt 2xn int
// winSize c_level int
// c_xx c_xy c_yy: image size double*
// image must be double
const mxArray* imgArray;
int* imdims;
const int *curImgDims;
const int *nFeatures;
//double **iP, **iDxP, **iDyP, **jP;
double *iP[MAX_LEVEL], *iDxP[MAX_LEVEL], *iDyP[MAX_LEVEL], *jP[MAX_LEVEL];
double *fPnt, *newFPnt;
char* valid;
double accuracy_th;
int winSize, max_iter;
const int *cellDims = mxGetDimensions(prhs[0]);
if (nrhs > 6) {
accuracy_th=(double)mxGetScalar(prhs[6]);
max_iter=(int)mxGetScalar(prhs[7]);
}
winSize = (int)mxGetScalar(prhs[5]);
fPnt=(double*)mxGetPr(prhs[4]);
nFeatures=mxGetDimensions(prhs[4]);
imdims=(int*)malloc(sizeof(int)*cellDims[0]*2);
//iP=(double**)malloc(sizeof(double*)*cellDims[0]);
//iDxP=(double**)malloc(sizeof(double*)*cellDims[0]);
//iDyP=(double**)malloc(sizeof(double*)*cellDims[0]);
//jP=(double**)malloc(sizeof(double*)*cellDims[0]);
for(int i=0; i<cellDims[0]; i++){
//imgArray=mxGetCell(prhs[0],i);
//curImgDims = mxGetDimensions(imgArray);
curImgDims = mxGetDimensions(mxGetCell(prhs[0],i));
imdims[i*2+0]= curImgDims[0];
imdims[i*2+1]= curImgDims[1];
//imdims[i][1] = curImgDims[1];
iP[i]= (double*)mxGetPr(mxGetCell(prhs[0],i));
iDxP[i]= (double*)mxGetPr(mxGetCell(prhs[1],i));
iDyP[i]= (double*)mxGetPr(mxGetCell(prhs[2],i));
jP[i]= (double*)mxGetPr(mxGetCell(prhs[3],i));
}
plhs[0] = mxCreateNumericMatrix(nFeatures[0], nFeatures[1], mxDOUBLE_CLASS, mxREAL);
plhs[1] = mxCreateNumericMatrix(1, nFeatures[1], mxUINT8_CLASS, mxREAL);
newFPnt = (double*)mxGetPr(plhs[0]);
valid = (char*)mxGetPr(plhs[1]);
//idx convert from matlab to c
for(int i=0; i<nFeatures[1]; i++){
fPnt[i*2]=fPnt[i*2]-1;
fPnt[i*2+1]=fPnt[i*2+1]-1;
valid[i]=1;
}
if(nrhs>6){
calcPyrLKTrack( iP, iDxP, iDyP, jP, imdims, cellDims[0], fPnt, nFeatures[1], winSize,
newFPnt, valid, accuracy_th, max_iter);
}else{
calcPyrLKTrack( iP, iDxP, iDyP, jP, imdims, cellDims[0], fPnt, nFeatures[1], winSize,
newFPnt, valid);
}
//idx convert from matlab to c
for(int i=0; i<nFeatures[1]*2; i++){
fPnt[i]=fPnt[i]+1;
newFPnt[i]=newFPnt[i]+1;
}
free(imdims);
//free(iP);
//free(iDxP);
//free(iDyP);
//free(jP);
}
|