Re: [Mesa-dev] [PATCH 1/3] llvmpipe: use vector loads for (optimized) tri raster funcs

2016-03-14 Thread Roland Scheidegger
Am 15.03.2016 um 02:37 schrieb Stéphane Marchesin: > On Mon, Feb 1, 2016 at 8:14 AM, Brian Paul wrote: >> On 01/31/2016 06:00 PM, srol...@vmware.com wrote: >>> >>> From: Roland Scheidegger >>> >>> When we switched to 64bit rasterization, we could no longer use straight >>> aligned loads for loadi

Re: [Mesa-dev] [PATCH 1/3] llvmpipe: use vector loads for (optimized) tri raster funcs

2016-03-14 Thread Stéphane Marchesin
On Mon, Feb 1, 2016 at 8:14 AM, Brian Paul wrote: > On 01/31/2016 06:00 PM, srol...@vmware.com wrote: >> >> From: Roland Scheidegger >> >> When we switched to 64bit rasterization, we could no longer use straight >> aligned loads for loading the plane data. However, what the code actually >> does

Re: [Mesa-dev] [PATCH 1/3] llvmpipe: use vector loads for (optimized) tri raster funcs

2016-02-01 Thread Brian Paul
On 01/31/2016 06:00 PM, srol...@vmware.com wrote: From: Roland Scheidegger When we switched to 64bit rasterization, we could no longer use straight aligned loads for loading the plane data. However, what the code actually does for loading 3 planes, is 12 scalar loads + 9 unpacks, and then there

[Mesa-dev] [PATCH 1/3] llvmpipe: use vector loads for (optimized) tri raster funcs

2016-01-31 Thread sroland
From: Roland Scheidegger When we switched to 64bit rasterization, we could no longer use straight aligned loads for loading the plane data. However, what the code actually does for loading 3 planes, is 12 scalar loads + 9 unpacks, and then there's another 8 unpacks for the transpose we need (!).