Re: [Mesa-dev] [PATCH 05/21] i965/fs: Implement SIMD16 integer multiplies on Gen 7.

2014-12-23 Thread Matt Turner
On 12/22/14, Ben Widawsky wrote: > From: Matt Turner > > In order to do a full 32x32 integer multiply using the MUL/MACH macro, the > operation must be split into 2 SIMD8 operations. This is required even if > you > don't care about the high bits. My interpretation of the requirement is that > th

[Mesa-dev] [PATCH 05/21] i965/fs: Implement SIMD16 integer multiplies on Gen 7.

2014-12-22 Thread Ben Widawsky
From: Matt Turner In order to do a full 32x32 integer multiply using the MUL/MACH macro, the operation must be split into 2 SIMD8 operations. This is required even if you don't care about the high bits. My interpretation of the requirement is that the accumulator simply doesn't have enough bits t