[Mesa-dev] [PATCH] ac/nir: fix translation of nir_op_frcp for doubles

[Samuel Pitoiset]

Reviewed-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>

On 01/11/2018 11:14 AM, Timothy Arceri wrote:
> Without this we end up with the llvm error message:
> 
> "Both operands to a binary operator are not of the same type!"
> ---
> 
>   This is an additional fix on top of this series:
> 
>   https://patchwork.freedesktop.org/series/36313/
> 
>   src/amd/common/ac_nir_to_llvm.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
> index f1472677e1..b868f6613d 100644
> --- a/src/amd/common/ac_nir_to_llvm.c
> +++ b/src/amd/common/ac_nir_to_llvm.c
> @@ -1715,7 +1715,8 @@ static void visit_alu(struct ac_nir_context *ctx, const nir_alu_instr *instr)
>   		break;
>   	case nir_op_frcp:
>   		src[0] = ac_to_float(&ctx->ac, src[0]);
> -		result = ac_build_fdiv(&ctx->ac, ctx->ac.f32_1, src[0]);
> +		result = ac_build_fdiv(&ctx->ac, instr->dest.dest.ssa.bit_size == 32 ? ctx->ac.f32_1 : ctx->ac.f64_1,
> +				       src[0]);
>   		break;
>   	case nir_op_iand:
>   		result = LLVMBuildAnd(ctx->ac.builder, src[0], src[1], "");
>