Closed matheusd closed 10 months ago
I wonder if the generate portion should just handle the building of the compiled blake3 cuda kernel and then a normal go build -tags cuda to actually build gominer after that?
Then we're back at having the user setup their own CGO_CFLAGS
and CGO_LDFLAGS
by hand, because those are also needed during the go build
stage. Adding #cgo
directives to the source code is not sufficient, because barnex/cuda5
requires cuda.h
, and that isn't available by default on windows without setting up the env vars.
Then we're back at having the user setup their own
CGO_CFLAGS
andCGO_LDFLAGS
by hand, because those are also needed during thego build
stage. Adding#cgo
directives to the source code is not sufficient, becausebarnex/cuda5
requirescuda.h
, and that isn't available by default on windows without setting up the env vars.
Ah, alright. I thought it was only needed for the nvcc bits to build the kernel.
Updated readme with the various sections (view compiled file).
Also, I added a document which describes how to manually setup CUDA builds in windows (view)
Made a first batch of review fixes.
All of the updates thus far look good.
Updated with changes to the debian section.
I'll merge this after the squash.
This adds support to building the CUDA-enabled version of
gominer
on Windows, as well as adding instructions on how to do so.Additionally, it switches from using a Makefile to using
go generate
in order to make the building process more idiomatic to Go projects.