further reduce unecessary test values

Author: alexbaden

test/TritonIntelGPU/optimize-block-io-encoding.mlir

@@ -20,22 +20,12 @@ module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 32 : i32, "ttg.th
     %c64_i32 = arith.constant 64 : i32
     %c5120_i32 = arith.constant 5120 : i32
     %cst = arith.constant dense<0.000000e+00> : tensor<256x256xf32, #blocked>
-    %0 = tt.get_program_id x : i32
-    %1 = arith.divsi %0, %c64_i32 : i32
-    %2 = arith.muli %1, %c4_i32 : i32
-    %3 = arith.subi %c4_i32, %2 : i32
-    %4 = arith.minsi %3, %c4_i32 : i32
-    %5 = arith.remsi %0, %4 : i32
-    %6 = arith.addi %2, %5 : i32
-    %7 = arith.remsi %0, %c64_i32 : i32
-    %8 = arith.divsi %7, %4 : i32
-    %9 = arith.muli %6, %c256_i32 : i32
+
     // CHECK: %[[MAKE_TENSOR_PTR_A:.*]] = tt.make_tensor_ptr {{.*}} : <tensor<256x32xf16, #[[$SUBGROUP_BLOCK_A]]>>
-    %10 = tt.make_tensor_ptr %arg0, [%c1024_i64, %c5120_i64], [%c5120_i64, %c1_i64], [%9, %c0_i32] {order = array<i32: 1, 0>} : <tensor<256x32xf16, #blocked1>>
-    %11 = arith.muli %8, %c256_i32 : i32
+    %10 = tt.make_tensor_ptr %arg0, [%c1024_i64, %c5120_i64], [%c5120_i64, %c1_i64], [%c256_i32, %c0_i32] {order = array<i32: 1, 0>} : <tensor<256x32xf16, #blocked1>>
     // CHECK: %[[MAKE_TENSOR_PTR_B:.*]] = tt.make_tensor_ptr {{.*}} : <tensor<32x256xf16, #[[$SUBGROUP_BLOCK_B]]>>
-    %12 = tt.make_tensor_ptr %arg1, [%c5120_i64, %c4096_i64], [%c4096_i64, %c1_i64], [%c0_i32, %11] {order = array<i32: 1, 0>} : <tensor<32x256xf16, #blocked2>>
-    // CHECK: scf.for {{.*}} iter_args({{.*}} = {{.*}}, %[[ARG5:.*]] = %[[MAKE_TENSOR_PTR_A]], %[[ARG6:.*]] = %[[MAKE_TENSOR_PTR_B]])
+    %12 = tt.make_tensor_ptr %arg1, [%c5120_i64, %c4096_i64], [%c4096_i64, %c1_i64], [%c0_i32, %c256_i32] {order = array<i32: 1, 0>} : <tensor<32x256xf16, #blocked2>>
+    // CHECK: %[[RES:.*]]:3 = scf.for {{.*}} iter_args({{.*}} = {{.*}}, %[[ARG5:.*]] = %[[MAKE_TENSOR_PTR_A]], %[[ARG6:.*]] = %[[MAKE_TENSOR_PTR_B]])
     %13:3 = scf.for %arg3 = %c0_i32 to %c5120_i32 step %c32_i32 iter_args(%arg4 = %cst, %arg5 = %10, %arg6 = %12) -> (tensor<256x256xf32, #blocked>, !tt.ptr<tensor<256x32xf16, #blocked1>>, !tt.ptr<tensor<32x256xf16, #blocked2>>)  : i32 {
       %17 = tt.load %arg5 {boundaryCheck = array<i32: 0, 1>, ttig.block_io = "row_major"} : !tt.ptr<tensor<256x32xf16, #blocked1>>
       // CHECK: %[[A_LOAD:.*]] = tt.load %[[ARG5]] {boundaryCheck = array<i32: 0, 1>, ttig.block_io = "row_major"} : !tt.ptr<tensor<256x32xf16, #[[$SUBGROUP_BLOCK_A]]>>
@@ -58,7 +48,8 @@ module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 32 : i32, "ttg.th
       // CHECK: scf.yield {{.*}}, %[[ADVANCE_A]], %[[ADVANCE_B]]
       scf.yield %25, %26, %27 : tensor<256x256xf32, #blocked>, !tt.ptr<tensor<256x32xf16, #blocked1>>, !tt.ptr<tensor<32x256xf16, #blocked2>>
     }
-    %14 = tt.make_tensor_ptr %arg2, [%c1024_i64, %c4096_i64], [%c4096_i64, %c1_i64], [%9, %11] {order = array<i32: 1, 0>} : <tensor<256x256xf16, #blocked2>>
+    %14 = tt.make_tensor_ptr %arg2, [%c1024_i64, %c4096_i64], [%c4096_i64, %c1_i64], [%c0_i32, %c256_i32] {order = array<i32: 1, 0>} : <tensor<256x256xf16, #blocked2>>
+    // CHECK aritch.truncf %[[RES]]#0 : tensor<256x256xf32, #blocked> to tensor<256x256xf16, #blocked>
     %15 = arith.truncf %13#0 : tensor<256x256xf32, #blocked> to tensor<256x256xf16, #blocked>
     %16 = ttg.convert_layout %15 : tensor<256x256xf16, #blocked> -> tensor<256x256xf16, #blocked2>
     tt.store %14, %16 {boundaryCheck = array<i32: 0, 1>} : !tt.ptr<tensor<256x256xf16, #blocked2>>