[llvm] [LV]: Improve accuracy of calculating remaining iterations of MainLoopVF (PR #156723)
Paul Walker via llvm-commits
llvm-commits at lists.llvm.org
Wed Sep 10 05:10:12 PDT 2025
================
@@ -490,65 +471,71 @@ define void @trip_count_vscale(ptr noalias %a, ptr noalias %b) vscale_range(1, 1
; CHECK-NEXT: iter.check:
; CHECK-NEXT: [[V:%.*]] = tail call i64 @llvm.vscale.i64()
; CHECK-NEXT: [[N:%.*]] = mul nuw nsw i64 [[V]], 1033
+; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 1
; CHECK-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK: vector.main.loop.iter.check:
-; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT: [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 3
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP3:%.*]] = shl nuw i64 [[TMP2]], 3
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP3]]
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
-; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
-; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP5]]
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT: [[TMP6:%.*]] = shl nuw i64 [[TMP5]], 2
-; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[TMP4]], i64 [[TMP6]]
-; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP4]], align 4
-; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
-; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT: [[TMP10:%.*]] = shl nuw i64 [[TMP9]], 2
-; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i64 [[TMP10]]
-; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
-; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
-; CHECK-NEXT: [[TMP12:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
-; CHECK-NEXT: [[TMP13:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
-; CHECK-NEXT: [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT: [[TMP15:%.*]] = shl nuw i64 [[TMP14]], 2
-; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i64 [[TMP15]]
-; CHECK-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP8]], align 4
-; CHECK-NEXT: store <vscale x 4 x float> [[TMP13]], ptr [[TMP16]], align 4
-; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP3]]
-; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP8:%.*]] = shl nuw i64 [[TMP7]], 2
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
+; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP12:%.*]] = shl nuw i64 [[TMP11]], 2
+; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i64 [[TMP12]]
+; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
+; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT: [[TMP14:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
+; CHECK-NEXT: [[TMP15:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
+; CHECK-NEXT: [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP17:%.*]] = shl nuw i64 [[TMP16]], 2
+; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i64 [[TMP17]]
+; CHECK-NEXT: store <vscale x 4 x float> [[TMP14]], ptr [[TMP10]], align 4
+; CHECK-NEXT: store <vscale x 4 x float> [[TMP15]], ptr [[TMP18]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK: vec.epilog.iter.check:
; CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
-; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2
-; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK-NEXT: [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP21:%.*]] = shl nuw i64 [[TMP20]], 1
+; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], [[TMP21]]
+; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]], !prof [[PROF14:![0-9]+]]
; CHECK: vec.epilog.ph:
; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
-; CHECK-NEXT: [[N_MOD_VF4:%.*]] = urem i64 [[N]], 2
+; CHECK-NEXT: [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 2
+; CHECK-NEXT: [[N_MOD_VF4:%.*]] = urem i64 [[N]], [[TMP23]]
; CHECK-NEXT: [[N_VEC5:%.*]] = sub i64 [[N]], [[N_MOD_VF4]]
; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK: vec.epilog.vector.body:
; CHECK-NEXT: [[INDEX6:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX6]]
-; CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <2 x float>, ptr [[TMP18]], align 4
-; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX6]]
-; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <2 x float>, ptr [[TMP19]], align 4
-; CHECK-NEXT: [[TMP20:%.*]] = fmul <2 x float> [[WIDE_LOAD7]], [[WIDE_LOAD8]]
-; CHECK-NEXT: store <2 x float> [[TMP20]], ptr [[TMP19]], align 4
-; CHECK-NEXT: [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 2
-; CHECK-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC5]]
-; CHECK-NEXT: br i1 [[TMP21]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX6]]
+; CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <vscale x 2 x float>, ptr [[TMP24]], align 4
+; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX6]]
+; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <vscale x 2 x float>, ptr [[TMP25]], align 4
+; CHECK-NEXT: [[TMP26:%.*]] = fmul <vscale x 2 x float> [[WIDE_LOAD7]], [[WIDE_LOAD8]]
+; CHECK-NEXT: store <vscale x 2 x float> [[TMP26]], ptr [[TMP25]], align 4
+; CHECK-NEXT: [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], [[TMP23]]
+; CHECK-NEXT: [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC5]]
+; CHECK-NEXT: br i1 [[TMP27]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
----------------
paulwalker-arm wrote:
This looks worse? perhaps an artefact of what I mentioned above?
The loop has a trip count of `1033*vscale`, the main VF is `4*vscale`, so after the vector loop there will be `(1033-(1033/4))*vscale ==> vscale` iterations remaining, yet the epilogue loop now has a VF of `2*vscale` and thus will never execute?
https://github.com/llvm/llvm-project/pull/156723
More information about the llvm-commits
mailing list