Jason Ekstrand <ja...@jlekstrand.net> writes:
> On Wed, Jan 21, 2015 at 5:26 PM, Eric Anholt <e...@anholt.net> wrote:
>
>> There aren't many users yet, but I wanted to do this from my scalarizing
>> pass.
>> ---
>> src/glsl/nir/nir.c | 18 ++++++++++++++++++
>> src/glsl/nir/nir.h | 5 ++++-
>> src/glsl/nir/nir_lower_vec_to_movs.c | 7 ++-----
>> src/glsl/nir/nir_opt_peephole_select.c | 5 +----
>> 4 files changed, 25 insertions(+), 10 deletions(-)
>>
>> diff --git a/src/glsl/nir/nir.c b/src/glsl/nir/nir.c
>> index 16ad2da..e414df9 100644
>> --- a/src/glsl/nir/nir.c
>> +++ b/src/glsl/nir/nir.c
>> @@ -175,6 +175,24 @@ nir_dest nir_dest_copy(nir_dest dest, void *mem_ctx)
>> return ret;
>> }
>>
>> +void
>> +nir_alu_src_copy(nir_alu_src *dest, nir_alu_src *src, void *mem_ctx)
>> +{
>>
>
> We already have nir_src_copy which returns a nir_src instead of taking a
> pointer. TBH, I'm not sure which I prefer, but it would be good to be
> consistent. Thoughts?
Yeah, I was thinking that as the struct gets bigger, passing it around
on the stack gets worse. For API consistency, I think that would mean
pointers for both.
>> diff --git a/src/glsl/nir/nir.h b/src/glsl/nir/nir.h
>> index 8dc5222..7f0aa36 100644
>> --- a/src/glsl/nir/nir.h
>> +++ b/src/glsl/nir/nir.h
>> @@ -569,7 +569,10 @@ typedef struct {
>> unsigned write_mask : 4; /* ignored if dest.is_ssa is true */
>> } nir_alu_dest;
>>
>> -#define OPCODE(name, num_inputs, output_size, output_type, \
>> +void nir_alu_src_copy(nir_alu_src *dest, nir_alu_src *src, void *mem_ctx);
>> +void nir_alu_dest_copy(nir_alu_dest *dest, nir_alu_dest *src, void
>> *mem_ctx);
>> +
>> +#define OPCODE(name, num_inputs, output_size, output_type, \
>>
>
> Accidental whitespace change?
Not sure how that happened, will fix.
signature.asc
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
