[参考译文] J722SXH01EVM：TISP + Linux 远程处理器集成：DmaUtilsAutoInc3d_wait()中的无限循环

https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1419572/j722sxh01evm-tisp-linux-remote-proc-integration-infinite-loop-in-dmautilsautoinc3d_wait

工具与软件：

您好！

我修改了mcu_plus_sdk_j722s_10_00_00_25/examples/drivers/ipc/ipc_rpmsg_echo_linux c75ss0-0_freertos/ti-c7000目标示例、使其链接到 TISP 库中、并 tisp/test/misc/TISP_blockCopy_superNode/TISP_blockCopy_superNode_test.cpp test. 可以使用 Linux remoteproc 框架加载另一个具有所生成二进制文件的精简版源文件、如果将 TISP::execute::graph(mySuperNodeVec);调用注释掉、则运行正常、否则 DSP 最终将永远循环进入 mcu_plus_sdk_j722s_10_00_00_25/source/drivers/dmautils/src/dmautils_autoincrement_3d.c::DmaUtilsAutoInc3d_wait().

的符号  mcu_plus_sdk_j722s_10_00_00_25/source/drivers/dmautils/src/dmautils_autoincrement_3d.c::DmaUtilsAutoInc3d_wait():

 waitWord = dmautilsContext -> channelContext[channelId] -> waitWord;
 eflRegisterVal = __get_indexed(__EFR, 0);
 while ((eflRegisterVal & waitWord) != waitWord) {
     eflRegisterVal = __get_indexed(__EFR, 0);
 }

waitWord is 0x1 and eflRegisterVal 始终读为0。 在此之前、DMA 传输设置功能不会报告任何错误。 

Any suggestion as to what might be causing this? Any help would be very apreciated, thank you.





#ifdef __cplusplus
extern "C" {
#endif
#include <kernel/dpl/DebugP.h>
#ifdef __cplusplus
}
#endif

#include "TISP_superNode.hpp"
//#include "TISP_test.h"
#include "TISP_types.hpp"
#include "TISP_util.hpp"
#include <fftlib.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <tisp.hpp>

#include <drivers/hw_include/csl_clec.h>

#define WIDTH (16)
#define HEIGHT (16)

#define BLOCK_WIDTH (8)
#define BLOCK_HEIGHT (2)

#define TOTAL_DMA_CHANNELS (2)

#define NUM_REPS (1024)

#define DDR_SIZE (4 * 1024 * 1024) // 4 MB

Udma_DrvHandle TISP::SuperNode::appDrvHandle;

static __attribute__((aligned(128))) uint8_t ddrBuffer[DDR_SIZE];

#define RESTRICT restrict

void TISP_blockCopy_test_cn(float *RESTRICT pX,
                            float *RESTRICT pY,
                            size_t          width,
                            size_t          height,
                            size_t          inPitch,
                            size_t          outPitch)
{
   for (int32_t i = 0; i < height; i++) {
      for (int32_t j = 0; j < width; j++) {
         pY[i * outPitch + j] = pX[i * inPitch + j];
      }
   }
}

#ifdef __cplusplus
extern "C" {
#endif
int32_t tisp_test_main()
{
   DebugP_log((char *)"Init local vars\n");

   struct Udma_DrvObj appUdmaDrvObj;

   TISP::SuperNode::appDrvHandle = TISP::SuperNode::superNode::setUdmaDrvHandle(appUdmaDrvObj);

   /* DebugP_log((char *)"\nTISP demo with FFTLIB_1dBatched\n"); */

   uint32_t blockWidth  = BLOCK_WIDTH;
   uint32_t blockHeight = BLOCK_HEIGHT;

   /* size_t dataMemSize = width * height * sizeof(float); */

   uint32_t blockSize = blockHeight * blockWidth * sizeof(float);

   float *pInputBlock  = (float *) memalign(128, blockSize * 2);
   float *pOutputBlock = (float *) memalign(128, blockSize * 2);

   DebugP_log((char *)"Init mem to copy\n");

   uint32_t width  = WIDTH;
   uint32_t height = HEIGHT;

   float *input   = (float *) ddrBuffer;
   float *inputCn = input + width * height;

   for (uint32_t i = 0; i < width * height; i++) {
      /* input[i] = (float)rand() / (float)(RAND_MAX / (2 * 8)) - 8; */
      input[i]   = i;
      inputCn[i] = input[i];
   }

   DebugP_log((char *)"\n--------------------------------------------------");
   DebugP_log((char *)"\n        Memcopy Demo using TISP                 ");
   DebugP_log((char *)"\n--------------------------------------------------\n");

   float *output  = inputCn + width * height;
   // float *refOut = output + width * height;


   DebugP_log((char *)"\nCreating Graph:\n");
   DebugP_log((char *)"\nDDR-->L2-->blockCopy-->L2-->DDR\n");

   auto k0 = new TISP::misc::blockCopy<float, TISP::DMA_IN_OUT>(pInputBlock, pOutputBlock, blockWidth, blockHeight);

   TISP::opVec myOpVec;

   myOpVec.push_back(k0);

   auto s0 = new TISP::SuperNode::i2d_to_o2d<float, float>(myOpVec, input, output, pInputBlock, pOutputBlock, width,
                                                           height, blockWidth, blockHeight, width * sizeof(float),
                                                           width * sizeof(float), (TOTAL_DMA_CHANNELS));

   uint64_t startTime, endTime;
   startTime = (__get_GTSC(1), __STSC);

   TISP::SuperNode::superNodeVec mySuperNodeVec;
   mySuperNodeVec.push_back(s0);

   DebugP_log((char *)"\nExecuting graph\n");

   /* s0->exec(); */
   // TISP::execute::graph(mySuperNodeVec); // Uncommenting this line reesults in an infinite loop

   endTime            = (__get_GTSC(1), __STSC);
   uint64_t totalTime = (endTime - startTime);
   DebugP_log((char *)"Cycles - Using Opvec = %llu\n", totalTime);
   DebugP_log((char *)"Bandwidth: %.2f GB/sec\n", ((float) WIDTH * HEIGHT * sizeof(float)) / (totalTime));

   DebugP_log((char *)"\nFreeing resources...\n");

   free(pInputBlock);
   free(pOutputBlock);

   delete (k0);
   /* delete (k1); */
   delete (s0);

   DebugP_log((char *)"\nDemo complete!");
   return 1;
}
#ifdef __cplusplus
}
#endif