Re: [Mesa-dev] [PATCH 2/2] radeonsi: Use pointers rather than resource descriptors for shader constants v3

2013-08-06 Thread Laurent Carlier
Le mardi 6 août 2013 01:11:44 vous avez écrit : Le lundi 5 août 2013 14:58:44 Tom Stellard a écrit : From: Tom Stellard thomas.stell...@amd.com The TGSI-LLVM pass for radeonsi preloads constants and relies on LLVM's sinking pass to reduce SGPR usage by lowering constant reads to an

Re: [Mesa-dev] [PATCH 2/2] radeonsi: Use pointers rather than resource descriptors for shader constants v3

2013-08-06 Thread Michel Dänzer
On Mon, 2013-08-05 at 14:58 -0400, Tom Stellard wrote: From: Tom Stellard thomas.stell...@amd.com The TGSI-LLVM pass for radeonsi preloads constants and relies on LLVM's sinking pass to reduce SGPR usage by lowering constant reads to an optimal place in the code. However, LLVM's machine

Re: [Mesa-dev] [PATCH 2/2] radeonsi: Use pointers rather than resource descriptors for shader constants v3

2013-08-06 Thread Tom Stellard
On Tue, Aug 06, 2013 at 12:26:51PM +0200, Michel Dänzer wrote: On Mon, 2013-08-05 at 14:58 -0400, Tom Stellard wrote: From: Tom Stellard thomas.stell...@amd.com The TGSI-LLVM pass for radeonsi preloads constants and relies on LLVM's sinking pass to reduce SGPR usage by lowering constant

[Mesa-dev] [PATCH 2/2] radeonsi: Use pointers rather than resource descriptors for shader constants v3

2013-08-05 Thread Tom Stellard
From: Tom Stellard thomas.stell...@amd.com The TGSI-LLVM pass for radeonsi preloads constants and relies on LLVM's sinking pass to reduce SGPR usage by lowering constant reads to an optimal place in the code. However, LLVM's machine sink pass will not lower instructions that have been selected

Re: [Mesa-dev] [PATCH 2/2] radeonsi: Use pointers rather than resource descriptors for shader constants v3

2013-08-05 Thread Laurent Carlier
Le lundi 5 août 2013 14:58:44 Tom Stellard a écrit : From: Tom Stellard thomas.stell...@amd.com The TGSI-LLVM pass for radeonsi preloads constants and relies on LLVM's sinking pass to reduce SGPR usage by lowering constant reads to an optimal place in the code. However, LLVM's machine sink