[PATCH] D155688: [PATCH] [llvm] [InstCombine] Reassociate loop invariant GEP index calculations.
d-smirnov updated this revision to Diff 557128. d-smirnov added a comment. Amended Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/CodeGen/Hexagon/autohvx/vector-align-tbaa.ll llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT:[[TMP14:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT:[[TMP15:%.*]] = add i64 [[TMP14]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP16:%.*]] = getelementptr i64,
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Reassociate loop invariant GEP index calculations.
d-smirnov updated this revision to Diff 557138. d-smirnov marked 3 inline comments as done. d-smirnov added a comment. Reordered and removed extra check Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/CodeGen/Hexagon/autohvx/vector-align-tbaa.ll llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT:[[TMP14:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT:[[TMP15:%.*]] = add i64 [[TMP
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Reassociate loop invariant GEP index calculations.
d-smirnov marked an inline comment as done. d-smirnov added a comment. updated Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov updated this revision to Diff 557187. d-smirnov added a comment. comment updated Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/CodeGen/Hexagon/autohvx/vector-align-tbaa.ll llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT:[[TMP14:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT:[[TMP15:%.*]] = add i64 [[TMP14]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP16:%.*]] = getelement
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov added a comment. @nikic Updated. Please review Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov updated this revision to Diff 557537. d-smirnov added a comment. amended Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/CodeGen/Hexagon/autohvx/vector-align-tbaa.ll llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT:[[TMP14:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT:[[TMP15:%.*]] = add i64 [[TMP14]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP16:%.*]] = getelementptr i64,
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov updated this revision to Diff 557538. d-smirnov marked 2 inline comments as done. d-smirnov added a comment. amended Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/CodeGen/Hexagon/autohvx/vector-align-tbaa.ll llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT:[[TMP14:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT:[[TMP15:%.*]] = add i64 [[TMP14]], [[ROWS_IV]] -; CHECK
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov marked an inline comment as done. d-smirnov added a comment. @nikic Amended. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov updated this revision to Diff 557609. d-smirnov added a comment. Updated Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/CodeGen/Hexagon/autohvx/vector-align-tbaa.ll llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT:[[TMP14:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT:[[TMP15:%.*]] = add i64 [[TMP14]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP16:%.*]] = getelementptr i64,
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov marked an inline comment as done. d-smirnov added a comment. Updated Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Canonicalise ADD+GEP
d-smirnov added a comment. We have some improvements with the patch, most notable: 549.fotonik_3d improves about 6%. @nikic Should we revert the patch and try another location for it (in LICM pass, as you previously suggested)? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Reassociate loop invariant GEP index calculations.
d-smirnov updated this revision to Diff 556881. d-smirnov added a comment. Herald added subscribers: cfe-commits, wangpc, zzheng. Herald added a project: clang. Relaxed checks Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/Transforms/InstCombine/gep-combine-loop-invariant.ll llvm/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-vector-reverse.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT:[
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Reassociate loop invariant GEP index calculations.
d-smirnov updated this revision to Diff 556884. d-smirnov added a comment. updated Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/gep-combine-loop-invariant.ll llvm/test/Transforms/InstCombine/lshr-phi.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-vector-reverse.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Reassociate loop invariant GEP index calculations.
d-smirnov updated this revision to Diff 556894. d-smirnov added a comment. unit test fixed Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 Files: clang/test/CodeGen/SystemZ/systemz-inline-asm-03.c clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp llvm/lib/Transforms/InstCombine/InstructionCombining.cpp llvm/test/Transforms/InstCombine/align-addr.ll llvm/test/Transforms/InstCombine/gep-combine-loop-invariant.ll llvm/test/Transforms/InstCombine/lshr-phi.ll llvm/test/Transforms/InstCombine/mem-par-metadata-memcpy.ll llvm/test/Transforms/InstCombine/memrchr-4.ll llvm/test/Transforms/InstCombine/shift.ll llvm/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-vector-reverse.ll llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll llvm/test/Transforms/LoopVectorize/induction.ll llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll llvm/test/Transforms/LoopVectorize/runtime-check.ll llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll Index: llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll === --- llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT:[[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT:[[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT:[[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT:[[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT:[[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT:store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT:[[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT:store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT:br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT:[[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT:[[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT:[[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT:[[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT:[[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT:[[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT:[[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT:[[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT:[[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT:[[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT:[[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT:[[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT:[[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT:[[COL_LOA
[PATCH] D155688: [PATCH] [llvm] [InstCombine] Reassociate loop invariant GEP index calculations.
d-smirnov added a comment. @nikic Could you check out the updated code to make sure we're on the right track before I try to fix the rest of the unit tests? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155688/new/ https://reviews.llvm.org/D155688 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
