add B operand mismatch test

Author: alexbaden

test/TritonIntelGPU/tritonintelgpu-invalid.mlir

@@ -162,6 +162,7 @@ module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 8 : i32, "ttg.thr
 
 // -----
 
+// COM: A operand mismatch
 #dpas = #ttig.dpas<{repeatCount = 8, systolicDepth = 8, executionSize = 16, opsPerChan = 2, threadsPerWarp = 16, warpsPerCTA = [2, 2], repCluster = [1, 1]}>
 #dot_operand_a = #ttg.dot_op<{opIdx=0, parent=#dpas, kWidth=1}>
 #dot_operand_b = #ttg.dot_op<{opIdx=1, parent=#dpas, kWidth=2}>
@@ -181,6 +182,26 @@ module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 4 : i32} {
 
 // -----
 
+// COM: B operand mismatch
+#dpas = #ttig.dpas<{repeatCount = 8, systolicDepth = 8, executionSize = 16, opsPerChan = 2, threadsPerWarp = 16, warpsPerCTA = [2, 2], repCluster = [1, 1]}>
+#dot_operand_a = #ttg.dot_op<{opIdx=0, parent=#dpas, kWidth=1}>
+#dot_operand_b = #ttg.dot_op<{opIdx=1, parent=#dpas, kWidth=2}>
+#smem = #ttg.shared_memory
+module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 4 : i32} {
+  // CHECK-LABEL: matmul_tf32dot
+  tt.func @matmul_tf32dot(%ptr:!tt.ptr<f32>,
+  %a_mat:tensor<32x16xf16, #dot_operand_a>, %b_mat:tensor<16x32xf32, #dot_operand_b>) {
+    %cst = arith.constant dense<0.000000e+00> : tensor<32x32xf32, #dpas>
+
+    // expected-error @+1 {{Layout has opsPerChannel = 2 but tensor element type is 'f32'. Expected 16 bit type.}}
+    %28 = tt.dot %a_mat, %b_mat, %cst, inputPrecision = tf32 : tensor<32x16xf16, #dot_operand_a> * tensor<16x32xf32, #dot_operand_b> -> tensor<32x32xf32, #dpas>
+
+    tt.return
+  }
+}
+
+// -----
+
 #dpas = #ttig.dpas<{repeatCount = 8, systolicDepth = 8, executionSize = 16, opsPerChan = 1, threadsPerWarp = 16, warpsPerCTA = [2, 2], repCluster = [1, 1]}>
 #dot_operand_a = #ttg.dot_op<{opIdx=0, parent=#dpas, kWidth=1}>
 // expected-error @below {{ttg.dot_op kWidth parameter must match the parent's opsPerChannel}}