设备mlu270

代码：

__nram__ float row1[MAX_IMAGE_W*8]; //MAX_IMAGE_W=1920
__nram__ float row2[MAX_IMAGE_W*8];
__memcpy(row1, srcimg1+ pw * taskId *8, pw *8 * sizeof(float), GDRAM2NRAM);
__memcpy(row2, srcimg2+ pw * taskId *8, pw *8 * sizeof(float), GDRAM2NRAM);

8改成4就可以正常运行，不改就会报错

cnas error: Generating `/tmp/sift-62NmMc.s` file Failed, Exceeds NRAM limit. try use -O1 or above option
cncc: error: unable to execute command: Aborted (core dumped)
cncc: error: cnas command failed due to signal (use -v to see invocation)
cncc v2.15.2 clang version 5.0.1 (ssh://git@gitlab.software.cambricon.com:2289/neuware/clang.git 3d566217fdd7a60b7bc1c738917807010694db2f) (ssh://git@gitlab.software.cambricon.com:2289/neuware/llvm.git 88ed2768195ca38bca670ffe5c414c227721d92f) ( d on LLVM 5.0.1)
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/local/neuware/bin
cncc: note: diagnostic msg: PLEASE submit a bug report to http://llvm.org/bugs/ and include the crash backtrace, preprocessed source, and associated run .
cncc: note: diagnostic msg:

按理说这里才使用了120k nram内存，不应该超空间才对

另外我在使用__bang_transpose时遇到了相似的问题。其输入输出也是在几十k的时候就会报超空间.

请问这些问题如何解决，是我的nram上空间被其他东西占用了，还是说有某些设置限制了nram最大使用空间？