ifdefelse

> try to switch to MAD or MUL operation(s) that has/have full throughput (one per cycle). As I said in #16 I'm pretty sure such an operation doesn't exist across...

Another issue with a chain of FP operations is that FADD with random inputs will normally do nothing. For an FP32 operation the exponent is 8 bits so can have...

Likewise unconstrained random FMUL(A, B) has a ~25% chance of over/under flowing. If exp_a is 127 then any exp >= 1 will cause overflow - 127/256 if exp_a is 126...

I have a private branch with a bunch of extra printf's to generate that output. I'll get that cleaned up and checked in under #ifdef's. We've also been asked to...

You're right that a simple parameter tune could make the ProgPoW hashrate pretty close to Ethash. Ethash does 64 reads each of 128 bytes, reading 8kb of memory DAG total....

Instead of checking in a .diff, can you wrap these changes in a #ifdef?

We're pretty against these micro-optimizations that make the code significantly harder to read and understand but don't necessarily help the saturation of the hardware. This is chasing hashrate - a...

Thanks for the pointers everyone. This repository is only intended for initial research and development. With the spec in a settled state there are a handful of nearly-production-ready repositories as...

The number of requests and bank conflicts are data dependent and change for every hash. A random value from a register is used for the load address so how many...