Spaces:
Sleeping
Sleeping
File size: 11,991 Bytes
5aefcf4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 |
Ticket Name: TDA2: VCOP register allocation failed error while compiling
Query Text:
Part Number: TDA2 Hi, I have written a code in Kernel C for an algorithm. I am trying to compile the code through CCS7.4, but I get the following error while compiling: Can anyone please help me understand why this error is generated? Regards, Shruti
Responses:
Shruti, This means that your loop needs more registers than available register in VCOP. Without looking at the kernel code its difficult to comment more. If you can share your kernel then we can have look at it and comment more. If not then you should check the generated assembly code and see the registers used by the VCOP. Regards, Anshu
Shruti, Were you able to make progress on this issue? If yes, please close this thread. If not then let us know about it. Regards, Anshu
Hi Anshu, Sorry for the late response. I tried looking into the assembly code, but not found any lead. I am attaching my kernel C code here for reference. Can you please have a look at it? sobel.txt /*==========================================================================*/
/* Copyright (C) 2009-2013 Texas Instruments Incorporated. */
/* All Rights Reserved */
/*==========================================================================*/
/*----------------------------------------------------------------------------*/
/* NAME: eve_sobel_edge_detect */
/* */
/* */
/* DESCRIPTION: */
/* */
/* The function "eve_sobel_edge_detect" takes input image */
/* inptr, and runs the sobel operator on it and writes it out to */
/* the output array "optr". The input and output arrays are 2D arrays */
/* of the form [width * height]. */
/* */
/* The function has the following prototype: */
/* */
/* void eve_sobel_edge_detect */
/* ( */
/* __vptr_uint8 inptr, // input data pointer */
/* __vptr_uint8 optr, // output data pointer */
/* unsigned int width, // width of input image */
/* unsigned int height // height of input image */
/* ) */
/* */
/*----------------------------------------------------------------------------*/
#define ELEMSZ sizeof(*inptr)
#define VECTORSZ (VCOP_SIMD_WIDTH*ELEMSZ)
#define ELEMSZ_OUT sizeof(*optr)
#define VECTORSZ_OUT (VCOP_SIMD_WIDTH * ELEMSZ_OUT)
#define VCOP_2SIMD_WIDTH (2 * VCOP_SIMD_WIDTH)
void vcop_detectEdge(
__vptr_uint8 inptr,
__vptr_uint8 optr,
__vptr_uint8 intsumx,
unsigned int width,
unsigned int height)
{
__vector Vin, Vin1;
__vector Vout, Vout1;
__vector Vreg1, Vreg11;
__vector Vreg2, Vreg12;
__vector Vreg3, Vreg13;
__vector Vreg4, Vreg14;
__vector Vreg5, Vreg15;
__vector Vsumx, Vsumx1;
__vector Vsumy, Vsumy1;
__vector Vshift = -1;
__vector Vin01, Vin02;
__vector Vin03, Vin13;
__vector Vin04, Vin14;
for(int Irow = 0; Irow < (height - 2); Irow++)
{
for(int Icol = 0; Icol < (width / VCOP_2SIMD_WIDTH); Icol++)
{
__agen addr;
addr = Irow * (width * ELEMSZ) + Icol* (VCOP_SIMD_WIDTH* ELEMSZ);
(Vin, Vin1) = inptr[addr].deinterleave();
(Vreg1, Vreg11) = (inptr + (2*width) +(2 * sizeof(*inptr)))[addr].deinterleave();
Vreg1 = Vreg1 - Vin;
Vreg11 = Vreg11 - Vin1;
(Vreg2, Vreg12) = (inptr + (2*width))[addr].deinterleave();
(Vreg3, Vreg13) = (inptr + (2*sizeof(*inptr)))[addr].deinterleave();
Vreg3 = Vreg3 - Vreg2;
Vreg13 = Vreg13 - Vreg12;
(Vreg4, Vreg14) = (inptr + width)[addr].deinterleave();
(Vreg5, Vreg15) = (inptr + width + (2 * sizeof(*inptr)))[addr].deinterleave();
// Vtemp1 = Vreg4 + Vreg4;
Vreg14 += Vreg14;
Vsumx = Vreg5 + Vreg5;
Vsumx -= Vreg4;// - Vreg4
Vsumx -= Vreg4;
Vreg15 += Vreg15;
//Vsumx = Vtemp2 - Vtemp1;
Vsumx += Vreg1 + Vreg3;
Vsumx1 = Vreg15 - Vreg14;
Vsumx1 = Vreg11 + Vreg13;
//Vsumx1 = Vtemp12 - Vtemp11;
//Vsumx += Vdata2;
Vsumx = abs(Vsumx);
Vsumx1 = abs(Vsumx1);
(intsumx + width + sizeof(*inptr))[addr].interleave() = (Vsumx, Vsumx1);
//(optr + width + sizeof(*inptr))[addr] = Vsumx;
}
}
for(int Irow = 0; Irow < (height - 2); Irow++)
{
for(int Icol = 0; Icol < (width / VCOP_2SIMD_WIDTH); Icol++)
{
__agen addr;
addr = Irow * (width * ELEMSZ) + Icol* (2 * VCOP_SIMD_WIDTH * ELEMSZ);
(Vsumx, Vsumx1) = (intsumx + width + sizeof(*inptr))[addr].deinterleave();
(Vin, Vin1) = inptr[addr].deinterleave();
(Vreg1, Vreg11) = (inptr + (2*width) +(2 * sizeof(*inptr)))[addr].deinterleave();
Vreg1 = Vreg1 - Vin;
Vreg11 = Vreg11 - Vin1;
(Vreg2, Vreg12) = (inptr + (2*width))[addr].deinterleave();
(Vreg3, Vreg13) = (inptr + (2*sizeof(*inptr)))[addr].deinterleave();
Vreg3 = Vreg3 - Vreg2;
Vreg13 = Vreg13 - Vreg12;
(Vreg4, Vreg14) = (inptr + (sizeof(*inptr)))[addr].deinterleave();
(Vreg5, Vreg15) = (inptr + (2*width) + (sizeof(*inptr)))[addr].deinterleave();
Vreg4 += Vreg4;
Vreg14 += Vreg14;
Vreg5 += Vreg5;
Vreg15 += Vreg15;
Vreg4 -= Vreg5;
Vreg14 -= Vreg15;
Vreg1 -= Vreg3;
Vreg1-= Vreg4;
Vreg11 -= Vreg13;
Vreg11 -= Vreg14;
Vreg1 = abs(Vreg1);
Vreg11 = abs(Vreg11);
Vout = Vsumx + Vreg1;
Vout1 = Vsumx1 + Vreg11;
Vout = Vout << Vshift;
Vout1 = Vout1 << Vshift;
// Vsumy = Vout << Vshift;
//Vsumy1 = Vout1 << Vshift;
//(optr + width + sizeof(*inptr))[addr].interleave() = (Vsumy, Vsumy1);
(optr + width + sizeof(*inptr))[addr].interleave() = (Vout, Vout1);
}
}
for(int col= 0;col < height-1; col++)
{
__agen AddrIn = col* width *ELEMSZ;
__agen AddrOut = col*width *ELEMSZ_OUT;
Vin01= (optr + (width+1) * sizeof(*optr))[AddrIn].onept();
Vin02= (optr+(2*width-2)*sizeof(*optr))[AddrIn].onept();
(optr + width * sizeof(*optr))[AddrOut].onept() = Vin01;
(optr+(2*width)-sizeof(*optr))[AddrOut].onept() = Vin02;
}
for(int col = 0; col < width/VCOP_2SIMD_WIDTH; col++)
{
__agen Addr1= col*(2 * VECTORSZ_OUT);
(Vin03, Vin13) = (optr+width)[Addr1].deinterleave();
(Vin04, Vin14) = (optr+width *(height-2))[Addr1].deinterleave();
(optr)[Addr1].interleave() = (Vin03, Vin13);
(optr+width*(height-1))[Addr1].interleave() = (Vin04, Vin14);
}
#if 0
for(int Irow = 0; Irow < (height); Irow++)
{
for(int Icol = 0; Icol < (width / VCOP_SIMD_WIDTH); Icol++)
{
__agen addr;
addr = Irow * (width * ELEMSZ) + Icol* (VCOP_SIMD_WIDTH* ELEMSZ);
Vin = inptr[addr];
Vreg1 = (inptr +(2 * sizeof(*inptr)))[addr];
Vdata1 = Vreg1 - Vin;
Vreg2 = (inptr + (width*2))[addr];
Vreg3 = (inptr + (width*2) + (2 * sizeof(*inptr)))[addr];
Vdata2 = Vreg3 - Vreg2;
Vreg4 = (inptr + width)[addr];
Vreg5 = (inptr + width + (2 * sizeof(*inptr)))[addr];
Vtemp1 = Vreg4 + Vreg4;
Vtemp2 = Vreg5 + Vreg5;
Vsumx = Vtemp2 - Vtemp1;
Vsumx += Vdata1 + Vdata2;
//Vsumx += Vdata2;
Vsumx = abs(Vsumx);
intsumx[addr] = Vsumx;
}
}
for(int Irow = 0; Irow < (height); Irow++)
{
for(int Icol = 0; Icol < (width / VCOP_SIMD_WIDTH); Icol++)
{
__agen addr;
addr = Irow * (width * ELEMSZ) + Icol* (VCOP_SIMD_WIDTH* ELEMSZ);
Vsumx = intsumx[addr];
Vin = inptr[addr];
Vreg1 = (inptr + (width*2))[addr];
Vdata1 = Vreg1 - Vin;
Vreg2 = (inptr + (2*sizeof(*inptr)))[addr];
Vreg3 = (inptr + (width*2) + (2*sizeof(*inptr)))[addr];
Vdata2 = Vreg3 - Vreg2;
Vreg4 = (inptr + (sizeof(*inptr)))[addr];
Vreg5 = (inptr + (width*2) + (sizeof(*inptr)))[addr];
Vtemp1 = Vreg4 + Vreg4;
Vtemp2 = Vreg5 + Vreg5;
Vsumy = Vtemp2 - vtemp1;
Vsumy+=Vdata1;
Vsumy+=Vdata2;
Vsumy = abs(Vsumy);
Vout = Vsumx + Vsumy;
Vout = Vout << Vshift;
outptr[addr] = Vout;
}
}
#endif
}
The code is for the Sobel edge detection algorithm. The commented part is the original algorithm, which is later converted to interleaved format. Regards, Shruti
Shruti, I tried compiling the code you shared and I don't see any build error. Following are my compiler options. Can you make sure the compiler options are the same : C:\ti\ccsv7.3\tools\compiler\arp32_1.0.9\bin\cl-arp32 -g --opt_for_speed=5 -kh --silicon_version=v210 -lu --display_error_number --diag_suppress=26066 --diag_suppress=10063 -IC:\ti\ccsv7.3\tools\compiler\arp32_1.0.9\include -IC:\ti\ccsv7.3\tools\compiler\arp32_1.0.9\include\vcop -I"..\inc" -I"..\..\..\inc" --temp_directory="..\..\..\inc" --obj_directory="..\src_kernelC" ..\src_kernelC\vcop_sobel_xy_kernel.k Regards, Anshu
Hi Anshu, There are few differences in my compiler options. Below is my compiler log: "C:\\ti\\ccsv7\\utils\\bin\\gmake" -k -j 4 all -O 'Building file: "../main.c"' 'Invoking: EVE Compiler' "C:/ti/ccsv7/tools/compiler/arp32_1.0.9/bin/cl-arp32" -v210 -g --include_path="C:/ti/ccsv7/tools/compiler/arp32_1.0.9/include/vcop" --include_path="C:/Users/shruti.joshi/workspace_v7/Sobel_interleave" --include_path="C:/ti/ccsv7/tools/compiler/arp32_1.0.9/include" --diag_warning=225 --display_error_number --diag_wrap=off --preproc_with_compile --preproc_dependency="main.d_raw" "../main.c" 'Finished building: "../main.c"' 'Building file: "../sobel.k"' 'Invoking: EVE Compiler' "C:/ti/ccsv7/tools/compiler/arp32_1.0.9/bin/cl-arp32" -v210 -g --include_path="C:/ti/ccsv7/tools/compiler/arp32_1.0.9/include/vcop" --include_path="C:/Users/shruti.joshi/workspace_v7/Sobel_interleave" --include_path="C:/ti/ccsv7/tools/compiler/arp32_1.0.9/include" --diag_warning=225 --display_error_number --diag_wrap=off --preproc_with_compile --preproc_dependency="sobel.d_raw" "../sobel.k" >> Compilation failure subdir_rules.mk:16: recipe for target 'sobel.obj' failed error: VCOP register allocation failed. Please restructure and/or hand-allocate the kernel function. gmake: *** [sobel.obj] Error 1 gmake: Target 'all' not remade because of errors. **** Build Finished **** I will try to match the configurations with yours and let you know the results. Regards, Shruti
Hi Shruti, Were you able to make these compiler option changes? If yes are you still seeing "VCOP register allocation failed" error? Regards, Anshu
Shruti, I am closing this thread as we haven't heard back from you. If required please re-open this thread again. Regards, Anshu
|