Re: [Mesa-dev] [PATCH 1/7] i965: Introduce the BROADCAST pseudo-opcode.

2015-04-29 Thread Matt Turner
I replied with a bunch of style nits, and perhaps a suggestion to combine the pass added in 4/7 with opt_algebraic, pending Ken's thoughts. In the case we want to combine 4/7 with opt_algebraic, I don't think I need to see the final result -- I trust you can manage. :) The series is Reviewed-by:

Re: [Mesa-dev] [PATCH 1/7] i965: Introduce the BROADCAST pseudo-opcode.

2015-04-29 Thread Matt Turner
On Fri, Feb 20, 2015 at 11:48 AM, Francisco Jerez wrote: > The BROADCAST instruction picks the channel from its first source > given by an index passed in as second source. This will be used in > situations where all channels from the same SIMD thread have to agree > on the value of something, e.

Re: [Mesa-dev] [PATCH 1/7] i965: Introduce the BROADCAST pseudo-opcode.

2015-02-20 Thread Francisco Jerez
Ian Romanick writes: > I want to be sure I understand what this series is doing and why it's > necessary. > > Sampler arrays and UBO arrays can, with some restrictions, be indexed > dynamically. The primary restriction is that any active SIMD channels > that access the array must access the same

Re: [Mesa-dev] [PATCH 1/7] i965: Introduce the BROADCAST pseudo-opcode.

2015-02-20 Thread Ian Romanick
I want to be sure I understand what this series is doing and why it's necessary. Sampler arrays and UBO arrays can, with some restrictions, be indexed dynamically. The primary restriction is that any active SIMD channels that access the array must access the same index. When you get to a particu

[Mesa-dev] [PATCH 1/7] i965: Introduce the BROADCAST pseudo-opcode.

2015-02-20 Thread Francisco Jerez
The BROADCAST instruction picks the channel from its first source given by an index passed in as second source. This will be used in situations where all channels from the same SIMD thread have to agree on the value of something, e.g. a surface binding table index. --- src/mesa/drivers/dri/i965/b