On Mit, 2013-01-16 at 12:16 +0100, Christian König wrote: > On 16.01.2013 10:42, Michel Dänzer wrote: > > From: Michel Dänzer <michel.daen...@amd.com> > > > > > > Signed-off-by: Michel Dänzer <michel.daen...@amd.com> > > --- > > > > This is a prerequisite for patch 3 (and probably more similar changes in the > > future). Unfortunately, it introduces a couple of (non-fatal) error messages > > such as below during the R600 target build. Tom or anyone, any ideas for > > silencing these? > > See below. > > > > > anonymous.105: (intrinsic_w_chain:v4f32 2651:iPTR, > > (imm:i32):$writemask, VReg_128:v4f32:$coord, SReg_256:v8i32:$rsrc, > > SReg_128:v4i32:$sampler, (imm:i32):$textureTarget) > > Included from /home/daenzer/src/llvm-git/llvm/lib/Target/R600/AMDGPU.td:40: > > Included from > > /home/daenzer/src/llvm-git/llvm/lib/Target/R600/AMDGPUInstructions.td:189: > > Included from > > /home/daenzer/src/llvm-git/llvm/lib/Target/R600/SIInstrInfo.td:591: > > /home/daenzer/src/llvm-git/llvm/lib/Target/R600/SIInstructions.td:1188:1: > > error: In anonymous.105: Pattern has dead named input: $textureTarget > > def : Pat < > > ^ > > > > lib/Target/R600/SIInstructions.td | 6 +++--- > > lib/Target/R600/SIIntrinsics.td | 2 +- > > 2 files changed, 4 insertions(+), 4 deletions(-) > > > > diff --git a/lib/Target/R600/SIInstructions.td > > b/lib/Target/R600/SIInstructions.td > > index cac42da..01c6ec1 100644 > > --- a/lib/Target/R600/SIInstructions.td > > +++ b/lib/Target/R600/SIInstructions.td > > @@ -1186,21 +1186,21 @@ def : Pat < > > > > /* int_SI_sample */ > > def : Pat < > > - (int_SI_sample imm:$writemask, VReg_128:$coord, SReg_256:$rsrc, > > SReg_128:$sampler), > > + (int_SI_sample imm:$writemask, VReg_128:$coord, SReg_256:$rsrc, > > SReg_128:$sampler, imm:$textureTarget), > > (IMAGE_SAMPLE imm:$writemask, 0, 0, 0, 0, 0, 0, 0, VReg_128:$coord, > > SReg_256:$rsrc, SReg_128:$sampler) > > >; > > Not 100% sure, but try declaring it like this: > > def : Pat < > - (int_SI_sample imm:$writemask, VReg_128:$coord, SReg_256:$rsrc, > SReg_128:$sampler), > + (int_SI_sample imm:$writemask, VReg_128:$coord, SReg_256:$rsrc, > SReg_128:$sampler, 0), > (IMAGE_SAMPLE imm:$writemask, 0, 0, 0, 0, 0, 0, 0, VReg_128:$coord, > SReg_256:$rsrc, SReg_128:$sampler) > >;
That results in failure to select the intrinsic at runtime. I figured out how to leave the parameter unnamed, which prevents the messages. See the v2 patches. Thanks anyway, -- Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Debian, X and DRI developer _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
