v0.2.42

Reactant v0.2.42

Diff since v0.2.41

Merged pull requests:

• feat: cost_analysis (#901) (@avik-pal)
• feat: JLL changes needed for cost analysis (#904) (@avik-pal)
• Update libtpu (#906) (@wsmoses)
• feat: Google Cloud setup (#907) (@avik-pal)
• Format code of branch "main" (#908) (@github-actions[bot])
• Add flag to toggle openmp usage in jit lowering (#909) (@wsmoses)
• Add global option to disable attributor in sroa-wrapper passes (#910) (@giordano)
• Don't jit execute kernel during precompilation (#911) (@wsmoses)
• feat: broadcasting of concrete arrays (#913) (@avik-pal)
• Revert "fix: apply init values after reduce" (#915) (@giordano)
• Add default trunc (#917) (@wsmoses)
• CUDA: support tracedrnumber argument (#918) (@wsmoses)
• fix: sharding improvements for GB (#919) (@avik-pal)
• @trace if handle ternary problem (#920) (@jumerckx)
• Release v0.2.42 (#921) (@giordano)
• Add isless (#923) (@wsmoses)
• Fix irrational promotion (#924) (@wsmoses)
• feat(JLL): changes for #926 (#927) (@avik-pal)

Closed issues:

• Implement cost_analysis (#899)
• Broadcast of concreterarray requires scalar indexing (#912)
• @trace if error type Symbol has no field head (#916)
• Sharding + NoSharding with Argument Mutation (#925)