as following IR code shows:
// attr [packedB] storage_scope = "global"
allocate packedB[float32x32 * 32 * 1024 * 1]
produce packedB {
parallel (x, 0, 32) {
for (y, 0, 1024) {
packedB[ramp((((x*1024) + y)*32), 1, 32)] = B[ramp(((x + (y*32))*32), 1, 32)]
}
}
}
- what does parallel mean and what’s used for?
- what does ramp mean and what’s used for?
- what does this code section mean?
thank you very much~