Making sure forge runs on Blackhole

[gfengTT]

This issue will be updated with the steps I take to build and run forge on blackhole, and problem that I run into.

[gfengTT]

Update Nov 4:

pytest -sv forge/test/test_api.py passes on yyzo-bh-07.

Changes needed:

1. export ARCH_NAME=blackhole
2. Small change of code in metal:

   
   diff --git a/tt_metal/hw/inc/blackhole/tensix_types.h b/tt_metal/hw/inc/blackhole/tensix_types.h
   index b64779765a..955677d9a9 100644
   --- a/tt_metal/hw/inc/blackhole/tensix_types.h
   +++ b/tt_metal/hw/inc/blackhole/tensix_types.h
   @@ -57,7 +57,7 @@ typedef struct {
    uint32_t reserved_3 : 32;
   } packer_config_t;  // 16B

-typedef struct { +struct fifo_ctl_t { uint32_t rd_ptr; uint32_t wr_ptr; uint32_t rsvd0; @@ -67,7 +67,7 @@ typedef struct { return fmt::format("Fifo Control: rd_ptr(0x{:08x}) wr_ptr(0x{:08x})", rd_ptr, wr_ptr); }

endif

-} fifo_ctl_t; +};

typedef struct { uint32_t val[4];

[staylorTT]

can we use the

to track the sub items needed to enable this?

[zzigler-tt]

@gfengTT Can you please add the following labels - priority, Blackhole and Forge. Thanks

[gfengTT]

@staylorTT Sure. The issue I described requires a fix on the metal side. Should I create an issue there? Do you know who I should assign it to? Just realized I don't have push permissions on metal.

[staylorTT]

When you made the changes in your previous comment did this pass locally? Is this a good sign of life that the test can pass using BH?

As far as sending the change, I believe because metal is public you could fork and open an MR with the changes? I believe @abhullar-tt might be able to help with this

[gfengTT]

Yes. Compilation failed without the change, and passed with it. In fact the same change has been made to wormhole.