[OPTIMIZER] Coalesce pass no longer takes a num-warps argument (#99)

Improved design to avoid inconsistent `num-warps` value between the pass and the parent module of the operation it processes.

include/triton/Dialect/TritonGPU/IR/TritonGPUDialect.td

@@ -18,6 +18,18 @@ def TritonGPU_Dialect : Dialect {
     "triton::TritonDialect",
     "mlir::gpu::GPUDialect"
   ];
+
+  let extraClassDeclaration = [{
+    static std::string getNumWarpsAttrName() { return "triton_gpu.num-warps"; }
+    static int getNumWarps(ModuleOp mod) { 
+      if(!mod->hasAttr("triton_gpu.num-warps"))
+        llvm::report_fatal_error(
+            "TritonGPU module should contain a triton_gpu.num-warps attribute");
+      return mod->getAttr("triton_gpu.num-warps").cast<IntegerAttr>().getInt();
+    }
+  }];
+  
+
 }
 
 #endif

include/triton/Dialect/TritonGPU/Transforms/Passes.td

@@ -33,12 +33,6 @@ def TritonGPUCoalesce: Pass<"tritongpu-coalesce", "mlir::ModuleOp"> {
   let constructor = "mlir::createTritonGPUCoalescePass()";
 
   let dependentDialects = ["mlir::triton::gpu::TritonGPUDialect"];
-
-  let options = [
-    Option<"numWarps", "num-warps",
-          "int32_t", /*default*/"4",
-          "number of warps">
-  ];
 }
 
 def TritonGPUCombineOps : Pass<"tritongpu-combine", "mlir::ModuleOp"> {

lib/Conversion/TritonGPUToLLVM/TritonGPUToLLVM.cpp

@@ -795,24 +795,6 @@ private:
   AxisInfoAnalysis &AxisAnalysisPass;
 };
 
-// Extract numWarps information from TritonGPU module, return 0 if failed.
-// This is a naive implementation, it assumes that all the blocked layout should
-// have the same numWarps setting in a module, it just find a blocked layout
-// encoding and return the warpsPerCTA field.
-int extractNumWarps(mlir::ModuleOp module) {
-  int numWarps{};
-  if (module->hasAttr(AttrNumWarpsName))
-    numWarps = module->getAttr(AttrNumWarpsName)
-                   .dyn_cast<IntegerAttr>()
-                   .getValue()
-                   .getZExtValue();
-  else
-    llvm::report_fatal_error(
-        "TritonGPU module should contain a triton_gpu.num-warps attribute");
-
-  return numWarps;
-}
-
 struct BroadcastOpConversion
     : public ConvertTritonGPUOpToLLVMPattern<triton::BroadcastOp> {
   using ConvertTritonGPUOpToLLVMPattern<
@@ -1319,7 +1301,7 @@ public:
 
     RewritePatternSet patterns(context);
 
-    int numWarps = extractNumWarps(mod);
+    int numWarps = triton::gpu::TritonGPUDialect::getNumWarps(mod);
 
     auto axisAnalysis = runAxisAnalysis(mod);
 

lib/Dialect/TritonGPU/Transforms/Coalesce.cpp

@@ -11,7 +11,8 @@ using namespace mlir::triton;
 
 struct CoalescePass : public TritonGPUCoalesceBase<CoalescePass> {
 
-  Attribute getCoalescedEncoding(AxisInfoAnalysis &axisInfo, Value ptr) {
+  Attribute getCoalescedEncoding(AxisInfoAnalysis &axisInfo, Value ptr,
+                                 int numWarps) {
     auto origType = ptr.getType().cast<RankedTensorType>();
     // Get the shape of the tensor.
     size_t rank = origType.getRank();
@@ -36,14 +37,13 @@ struct CoalescePass : public TritonGPUCoalesceBase<CoalescePass> {
     std::iota(dims.begin(), dims.end(), 0);
     // create encoding
     Attribute encoding = triton::gpu::BlockedEncodingAttr::get(
-        &getContext(), origType.getShape(), sizePerThread, order,
-        this->numWarps);
+        &getContext(), origType.getShape(), sizePerThread, order, numWarps);
     return encoding;
   }
 
   std::function<Type(Type)> getTypeConverter(AxisInfoAnalysis &axisInfo,
-                                             Value ptr) {
-    Attribute encoding = getCoalescedEncoding(axisInfo, ptr);
+                                             Value ptr, int numWarps) {
+    Attribute encoding = getCoalescedEncoding(axisInfo, ptr, numWarps);
     return [encoding](Type _type) {
       RankedTensorType type = _type.cast<RankedTensorType>();
       return RankedTensorType::get(type.getShape(), type.getElementType(),
@@ -57,8 +57,11 @@ struct CoalescePass : public TritonGPUCoalesceBase<CoalescePass> {
     RankedTensorType ty = ptr.getType().template dyn_cast<RankedTensorType>();
     if (!ty)
       return;
+    auto mod = op->getParentOfType<ModuleOp>();
+    int numWarps = triton::gpu::TritonGPUDialect::getNumWarps(mod);
+
     AxisInfo info = axisInfo.lookupLatticeElement(ptr)->getValue();
-    auto convertType = getTypeConverter(axisInfo, ptr);
+    auto convertType = getTypeConverter(axisInfo, ptr, numWarps);
     // convert operands
     SmallVector<Value, 4> newArgs;
     for (auto v : op->getOperands())

test/TritonGPU/coalesce.mlir

@@ -4,6 +4,8 @@
 #blocked1 = #triton_gpu.blocked<{sizePerThread = [1, 1], threadsPerWarp = [32, 1], warpsPerCTA = [4, 1], order = [0, 1]}>
 #blocked2 = #triton_gpu.blocked<{sizePerThread = [1, 1], threadsPerWarp = [1, 32], warpsPerCTA = [1, 4], order = [0, 1]}>
 
+module attributes {"triton_gpu.num-warps" = 4 : i32} {
+
 
 // CHECK: [[row_layout:#.*]] = #triton_gpu.blocked<{sizePerThread = [1, 4], threadsPerWarp = [2, 16], warpsPerCTA = [1, 4], order = [1, 0]}>
 // CHECK: [[col_layout:#.*]] = #triton_gpu.blocked<{sizePerThread = [4, 1], threadsPerWarp = [16, 2], warpsPerCTA = [4, 1], order = [0, 1]}>
@@ -44,3 +46,5 @@ func @transpose(%arg0: !tt.ptr<f32> {tt.divisibility = 16 : i32},
   tt.store %18, %19, %cst : tensor<64x64xf32, #blocked1>
   return
 }
+
+}