[mlir][openacc][flang] Simplify gang, vector and worker representation (

#77667) The IR representation for gang, vector and worker has grown with the support for device_type. This patch simplify the IR representation for gang, vector and worker information on the acc.loop operation. When the only the keyword is present without any values, the information is printed at the same place than when there is values. The device_type is omitted if there is no values and it is equal to None. Otherwise the full information is displayed. First the keyword only device_type information and then the values with their device_type.
llvm · Jan 11, 2024 · 40f5f90 · 40f5f90
1 parent 37c1a5e
commit 40f5f90
Show file tree

Hide file tree

Showing 7 changed files with 287 additions and 102 deletions.
diff --git a/flang/test/Lower/OpenACC/acc-kernels-loop.f90 b/flang/test/Lower/OpenACC/acc-kernels-loop.f90
@@ -497,10 +497,10 @@ subroutine acc_kernels_loop
   END DO
 
 ! CHECK:      acc.kernels {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop gang {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {gang = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.terminator
 ! CHECK-NEXT: }{{$}}
 
@@ -550,10 +550,10 @@ subroutine acc_kernels_loop
     a(i) = b(i)
   END DO
 ! CHECK:      acc.kernels {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop vector {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {vector = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.terminator
 ! CHECK-NEXT: }{{$}}
 
@@ -591,10 +591,10 @@ subroutine acc_kernels_loop
   END DO
 
 ! CHECK:      acc.kernels {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop worker {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {worker = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.terminator
 ! CHECK-NEXT: }{{$}}
 

diff --git a/flang/test/Lower/OpenACC/acc-loop.f90 b/flang/test/Lower/OpenACC/acc-loop.f90
@@ -67,10 +67,10 @@ program acc_loop
     a(i) = b(i)
   END DO
 
-!CHECK:      acc.loop {
+!CHECK:      acc.loop gang {
 !CHECK:        fir.do_loop
 !CHECK:        acc.yield
-!CHECK-NEXT: } attributes {gang = [#acc.device_type<none>]}{{$}}
+!CHECK-NEXT: }{{$}}
 
   !$acc loop gang(num: 8)
   DO i = 1, n
@@ -109,10 +109,10 @@ program acc_loop
     a(i) = b(i)
   END DO
 
-!CHECK:      acc.loop {
+!CHECK:      acc.loop vector {
 !CHECK:        fir.do_loop
 !CHECK:        acc.yield
-!CHECK-NEXT: } attributes {vector = [#acc.device_type<none>]}{{$}}
+!CHECK-NEXT: }{{$}}
 
   !$acc loop vector(128)
   DO i = 1, n
@@ -141,10 +141,10 @@ program acc_loop
     a(i) = b(i)
   END DO
 
-!CHECK:      acc.loop {
+!CHECK:      acc.loop worker {
 !CHECK:        fir.do_loop
 !CHECK:        acc.yield
-!CHECK-NEXT: } attributes {worker = [#acc.device_type<none>]}{{$}}
+!CHECK-NEXT: }{{$}}
 
   !$acc loop worker(128)
   DO i = 1, n
@@ -320,4 +320,10 @@ program acc_loop
 ! CHECK: acc.loop
 ! CHECK: fir.do_loop
 
+  !$acc loop gang device_type(nvidia) gang(8)
+  DO i = 1, n
+  END DO
+
+! CHECK: acc.loop gang([#acc.device_type<none>], {num=%c8{{.*}} : i32} [#acc.device_type<nvidia>])
+
 end program
diff --git a/flang/test/Lower/OpenACC/acc-parallel-loop.f90 b/flang/test/Lower/OpenACC/acc-parallel-loop.f90
@@ -512,10 +512,10 @@ subroutine acc_parallel_loop
   END DO
 
 ! CHECK:      acc.parallel {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop gang {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {gang = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.yield
 ! CHECK-NEXT: }{{$}}
 
@@ -565,10 +565,10 @@ subroutine acc_parallel_loop
     a(i) = b(i)
   END DO
 ! CHECK:      acc.parallel {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop vector {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {vector = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.yield
 ! CHECK-NEXT: }{{$}}
 
@@ -606,10 +606,10 @@ subroutine acc_parallel_loop
   END DO
 
 ! CHECK:      acc.parallel {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop worker {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {worker = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.yield
 ! CHECK-NEXT: }{{$}}
 

diff --git a/flang/test/Lower/OpenACC/acc-serial-loop.f90 b/flang/test/Lower/OpenACC/acc-serial-loop.f90
@@ -447,10 +447,10 @@ subroutine acc_serial_loop
   END DO
 
 ! CHECK:      acc.serial {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop gang {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {gang = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.yield
 ! CHECK-NEXT: }{{$}}
 
@@ -500,10 +500,10 @@ subroutine acc_serial_loop
     a(i) = b(i)
   END DO
 ! CHECK:      acc.serial {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop vector {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {vector = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.yield
 ! CHECK-NEXT: }{{$}}
 
@@ -541,10 +541,10 @@ subroutine acc_serial_loop
   END DO
 
 ! CHECK:      acc.serial {
-! CHECK:        acc.loop {
+! CHECK:        acc.loop worker {
 ! CHECK:          fir.do_loop
 ! CHECK:          acc.yield
-! CHECK-NEXT:   } attributes {worker = [#acc.device_type<none>]}{{$}}
+! CHECK-NEXT:   }{{$}}
 ! CHECK:        acc.yield
 ! CHECK-NEXT: }{{$}}
 

diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -1483,7 +1483,7 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
     Example:
 
     ```mlir
-    acc.loop {
+    acc.loop gang vector {
       scf.for %arg3 = %c0 to %c10 step %c1 {
         scf.for %arg4 = %c0 to %c10 step %c1 {
           scf.for %arg5 = %c0 to %c10 step %c1 {
@@ -1492,10 +1492,7 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
         }
       }
       acc.yield
-    } attributes {
-      collapse = [3], gang = [#acc.device_type<none>],
-      vector = [#acc.device_type<none>]
-    }
+    } attributes { collapse = [3] }
     ```
   }];
 
@@ -1613,13 +1610,14 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
   let hasCustomAssemblyFormat = 1;
   let assemblyFormat = [{
     oilist(
-        `gang` `` `(` custom<GangClause>($gangOperands, type($gangOperands),
+        `gang` `` custom<GangClause>($gangOperands, type($gangOperands),
             $gangOperandsArgType, $gangOperandsDeviceType,
-            $gangOperandsSegments) `)`
-      | `worker` `` `(` custom<DeviceTypeOperands>($workerNumOperands,
-            type($workerNumOperands), $workerNumOperandsDeviceType) `)`
-      | `vector` `` `(` custom<DeviceTypeOperands>($vectorOperands,
-            type($vectorOperands), $vectorOperandsDeviceType) `)`
+            $gangOperandsSegments, $gang)
+      | `worker` `` custom<DeviceTypeOperandsWithKeywordOnly>(
+            $workerNumOperands, type($workerNumOperands),
+            $workerNumOperandsDeviceType, $worker)
+      | `vector` `` custom<DeviceTypeOperandsWithKeywordOnly>($vectorOperands,
+            type($vectorOperands), $vectorOperandsDeviceType, $vector)
       | `private` `(` custom<SymOperandList>(
             $privateOperands, type($privateOperands), $privatizations) `)`
       | `tile` `(` custom<DeviceTypeOperandsWithSegment>($tileOperands,