Kernel Raising

[wsmoses]

#set1 = affine_set<(d0, d1) : (d0 + d1 * 16 - (d0 floordiv 16) * 16 - (d1 floordiv 12) * 192 >= 0, -d0 - d1 * 16 + (d0 floordiv 16) * 16 + (d1 floordiv 12) * 192 + 179 >= 0, d0 floordiv 16 + (d1 floordiv 12) * 16 - ((d1 floordiv 12) floordiv 6) * 96 >= 0, -(d0 floordiv 16) - (d1 floordiv 12) * 16 + ((d1 floordiv 12) floordiv 6) * 96 + 84 >= 0)>
module {
  func.func private @"##call__Z37gpu_above_freezing_ocean_temperature_16CompilerMetadataI10StaticSizeI13_180__85__20_E12DynamicCheckvv7NDRangeILi3ES0_I11_12__6__20_ES0_I11_16__16__1_EvvEE11OffsetArrayI7Float64Li3E13CuTracedArrayIS9_Li3ELi1E13_194__99__34_EESC_14LinearLiquidusIS9_E#1329$par72"(%arg0: memref<34x99x194xf64, 1>, %arg1: memref<34x99x194xf64, 1>) {
    %cst = arith.constant 5.400000e-02 : f64
    affine.parallel (%arg2, %arg3) = (0, 0) to (1440, 256) {
      affine.if #set1(%arg3, %arg2) {
        %0 = affine.load %arg1[%arg2 floordiv 72 + 7, %arg3 floordiv 16 + (%arg2 floordiv 12) * 16 - (%arg2 floordiv 72) * 96 + 7, %arg3 + %arg2 * 16 - (%arg3 floordiv 16) * 16 - (%arg2 floordiv 12) * 192 + 7] : memref<34x99x194xf64, 1>
        %1 = affine.load %arg0[%arg2 floordiv 72 + 7, %arg3 floordiv 16 + (%arg2 floordiv 12) * 16 - (%arg2 floordiv 72) * 96 + 7, %arg3 + %arg2 * 16 - (%arg3 floordiv 16) * 16 - (%arg2 floordiv 12) * 192 + 7] : memref<34x99x194xf64, 1>
        %2 = arith.mulf %0, %cst {fastmathFlags = #llvm.fastmath<none>} : f64
        %3 = arith.negf %2 : f64
        %4 = arith.cmpf olt, %1, %3 {fastmathFlags = #llvm.fastmath<none>} : f64
        %5 = arith.select %4, %3, %1 : f64
        affine.store %5, %arg0[%arg2 floordiv 72 + 7, %arg3 floordiv 16 + (%arg2 floordiv 12) * 16 - (%arg2 floordiv 72) * 96 + 7, %arg3 + %arg2 * 16 - (%arg3 floordiv 16) * 16 - (%arg2 floordiv 12) * 192 + 7] : memref<34x99x194xf64, 1>
      }     
    }       
    return  
  } 
}

[wsmoses]

Current raising status (and things that fail to raise):

funk.mlir.txt
yamps.mlir.txt
pipeline.sh.txt