https://github.com/goldsteinn updated https://github.com/llvm/llvm-project/pull/67166
>From 3982b40f82c6cb91c9bec407cea6b7a4cdc4220e Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:21 -0500 Subject: [PATCH 1/9] [InstSimplify] Add tests for simplify `llvm.ptrmask`; NFC Differential Revision: https://reviews.llvm.org/D156632 --- llvm/test/Transforms/InstSimplify/ptrmask.ll | 143 +++++++++++++++++++ 1 file changed, 143 insertions(+) create mode 100644 llvm/test/Transforms/InstSimplify/ptrmask.ll diff --git a/llvm/test/Transforms/InstSimplify/ptrmask.ll b/llvm/test/Transforms/InstSimplify/ptrmask.ll new file mode 100644 index 000000000000000..0af7186978ae5bc --- /dev/null +++ b/llvm/test/Transforms/InstSimplify/ptrmask.ll @@ -0,0 +1,143 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2 +; RUN: opt < %s -passes=instsimplify -S | FileCheck %s + +declare ptr @llvm.ptrmask.p0.i32(ptr, i32) +declare ptr @llvm.ptrmask.p0.i64(ptr, i64) + +define ptr @ptrmask_simplify_poison_mask(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_poison_mask +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 poison) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 poison) + ret ptr %r +} + +define ptr @ptrmask_simplify_undef_mask(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_undef_mask +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 undef) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 undef) + ret ptr %r +} + +define ptr @ptrmask_simplify_0_mask(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_0_mask +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 0) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 0) + ret ptr %r +} + +define ptr @ptrmask_simplify_1s_mask(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_1s_mask +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -1) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -1) + ret ptr %r +} + +define ptr @ptrmask_simplify_1s_mask_i32_fail(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_1s_mask_i32_fail +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -1) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -1) + ret ptr %r +} + +define ptr @ptrmask_simplify_poison_ptr(i64 %m) { +; CHECK-LABEL: define ptr @ptrmask_simplify_poison_ptr +; CHECK-SAME: (i64 [[M:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr poison, i64 [[M]]) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i64(ptr poison, i64 %m) + ret ptr %r +} + +define ptr @ptrmask_simplify_undef_ptr(i32 %m) { +; CHECK-LABEL: define ptr @ptrmask_simplify_undef_ptr +; CHECK-SAME: (i32 [[M:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr undef, i32 [[M]]) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i32(ptr undef, i32 %m) + ret ptr %r +} + +define ptr @ptrmask_simplify_null_ptr(i64 %m) { +; CHECK-LABEL: define ptr @ptrmask_simplify_null_ptr +; CHECK-SAME: (i64 [[M:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr null, i64 [[M]]) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i64(ptr null, i64 %m) + ret ptr %r +} + +define ptr @ptrmask_simplify_ptrmask(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_ptrmask +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[M:%.*]] = ptrtoint ptr [[P]] to i64 +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M]]) +; CHECK-NEXT: ret ptr [[R]] +; + %m = ptrtoint ptr %p to i64 + %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m) + ret ptr %r +} + +define ptr @ptrmask_simplify_ptrmask_i32_fail(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_ptrmask_i32_fail +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[M:%.*]] = ptrtoint ptr [[P]] to i32 +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 [[M]]) +; CHECK-NEXT: ret ptr [[R]] +; + %m = ptrtoint ptr %p to i32 + %r = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 %m) + ret ptr %r +} + +define ptr @ptrmask_simplify_aligned_unused(ptr align 64 %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_aligned_unused +; CHECK-SAME: (ptr align 64 [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) + ret ptr %r +} + +define ptr @ptrmask_simplify_aligned_unused_i32_fail(ptr align 64 %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_aligned_unused_i32_fail +; CHECK-SAME: (ptr align 64 [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -64) + ret ptr %r +} + +define ptr @ptrmask_simplify_known_unused(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_simplify_known_unused +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[PM0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) +; CHECK-NEXT: [[PGEP:%.*]] = getelementptr i8, ptr [[PM0]], i64 32 +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PGEP]], i64 -32) +; CHECK-NEXT: ret ptr [[R]] +; + %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) + %pgep = getelementptr i8, ptr %pm0, i64 32 + %r = call ptr @llvm.ptrmask.p0.i64(ptr %pgep, i64 -32) + ret ptr %r +} >From 661d6461750e9e49cd9629df3bed20385948f0e8 Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:27 -0500 Subject: [PATCH 2/9] [InstSimplify] Add some basic simplifications for `llvm.ptrmask` Mostly the same as `and`. We also have a check for a useless `llvm.ptrmask` if the ptr is already known aligned. Differential Revision: https://reviews.llvm.org/D156633 --- llvm/lib/Analysis/InstructionSimplify.cpp | 42 +++++++++++++++++++ .../InferAddressSpaces/AMDGPU/ptrmask.ll | 5 +-- .../test/Transforms/InstCombine/align-addr.ll | 8 +--- llvm/test/Transforms/InstSimplify/ptrmask.ll | 22 ++++------ 4 files changed, 53 insertions(+), 24 deletions(-) diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp index d8aa614cae53b10..1049845d97844a7 100644 --- a/llvm/lib/Analysis/InstructionSimplify.cpp +++ b/llvm/lib/Analysis/InstructionSimplify.cpp @@ -6397,6 +6397,48 @@ static Value *simplifyBinaryIntrinsic(Function *F, Value *Op0, Value *Op1, return Constant::getNullValue(ReturnType); break; } + case Intrinsic::ptrmask: { + // Fail loudly in case this is ever changed. + // TODO: If vector types are supported the logic that checks if the mask is + // useless should be updated to use generic constants. + assert(!Op0->getType()->isVectorTy() && !Op1->getType()->isVectorTy() && + "These simplifications where written at a time when ptrmask did not " + "support vector types and may not work for vectors"); + + // NOTE: We can't apply these simplifications based on the value of Op1 + // because we need to preserve provenance. + if (isa<PoisonValue>(Op0)) + return Op0; + + if (Q.isUndefValue(Op0)) + return Constant::getNullValue(Op0->getType()); + + if (match(Op0, m_Zero())) + return Constant::getNullValue(Op0->getType()); + + if (Op1->getType()->getScalarSizeInBits() == + Q.DL.getPointerTypeSizeInBits(Op0->getType())) { + if (match(Op1, m_PtrToInt(m_Specific(Op0)))) + return Op0; + + // TODO: We may have attributes assosiated with the return value of the + // llvm.ptrmask intrinsic that will be lost when we just return the + // operand. We should try to preserve them. + if (match(Op1, m_AllOnes())) + return Op0; + + const APInt *C; + if (match(Op1, m_APInt(C))) { + KnownBits PtrKnown = + computeKnownBits(Op0, Q.DL, /*Depth*/ 0, Q.AC, Q.CxtI, Q.DT); + // See if we only masking off bits we know are already zero due to + // alignment. + if ((*C | PtrKnown.Zero).isAllOnes()) + return Op0; + } + } + break; + } case Intrinsic::smax: case Intrinsic::smin: case Intrinsic::umax: diff --git a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/ptrmask.ll b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/ptrmask.ll index c9db0656c6b7dd4..561a5ff35ba1082 100644 --- a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/ptrmask.ll +++ b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/ptrmask.ll @@ -320,8 +320,7 @@ define i8 @ptrmask_cast_local_to_flat_const_mask_fffffffffffffffe(ptr addrspace( define i8 @ptrmask_cast_local_to_flat_const_mask_ffffffffffffffff(ptr addrspace(3) %src.ptr) { ; CHECK-LABEL: @ptrmask_cast_local_to_flat_const_mask_ffffffffffffffff( -; CHECK-NEXT: [[TMP1:%.*]] = call ptr addrspace(3) @llvm.ptrmask.p3.i32(ptr addrspace(3) [[SRC_PTR:%.*]], i32 -1) -; CHECK-NEXT: [[LOAD:%.*]] = load i8, ptr addrspace(3) [[TMP1]], align 1 +; CHECK-NEXT: [[LOAD:%.*]] = load i8, ptr addrspace(3) [[SRC_PTR:%.*]], align 1 ; CHECK-NEXT: ret i8 [[LOAD]] ; %cast = addrspacecast ptr addrspace(3) %src.ptr to ptr @@ -333,7 +332,7 @@ define i8 @ptrmask_cast_local_to_flat_const_mask_ffffffffffffffff(ptr addrspace( ; Make sure non-constant masks can also be handled. define i8 @ptrmask_cast_local_to_flat_load_range_mask(ptr addrspace(3) %src.ptr, ptr addrspace(1) %mask.ptr) { ; CHECK-LABEL: @ptrmask_cast_local_to_flat_load_range_mask( -; CHECK-NEXT: [[LOAD_MASK:%.*]] = load i64, ptr addrspace(1) [[MASK_PTR:%.*]], align 8, !range !0 +; CHECK-NEXT: [[LOAD_MASK:%.*]] = load i64, ptr addrspace(1) [[MASK_PTR:%.*]], align 8, !range [[RNG0:![0-9]+]] ; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[LOAD_MASK]] to i32 ; CHECK-NEXT: [[TMP2:%.*]] = call ptr addrspace(3) @llvm.ptrmask.p3.i32(ptr addrspace(3) [[SRC_PTR:%.*]], i32 [[TMP1]]) ; CHECK-NEXT: [[LOAD:%.*]] = load i8, ptr addrspace(3) [[TMP2]], align 1 diff --git a/llvm/test/Transforms/InstCombine/align-addr.ll b/llvm/test/Transforms/InstCombine/align-addr.ll index 23f620310d7c26c..b8e11f1b47165e8 100644 --- a/llvm/test/Transforms/InstCombine/align-addr.ll +++ b/llvm/test/Transforms/InstCombine/align-addr.ll @@ -180,11 +180,9 @@ define <16 x i8> @ptrmask_align8_ptr_align1(ptr align 1 %ptr) { } ; Underlying alignment already the same as forced alignment by ptrmask -; TODO: Should be able to drop the ptrmask define <16 x i8> @ptrmask_align8_ptr_align8(ptr align 8 %ptr) { ; CHECK-LABEL: @ptrmask_align8_ptr_align8( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -8) -; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 +; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[PTR:%.*]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; %aligned = call ptr @llvm.ptrmask.p0.i64(ptr %ptr, i64 -8) @@ -193,11 +191,9 @@ define <16 x i8> @ptrmask_align8_ptr_align8(ptr align 8 %ptr) { } ; Underlying alignment greater than alignment forced by ptrmask -; TODO: Should be able to drop the ptrmask define <16 x i8> @ptrmask_align8_ptr_align16(ptr align 16 %ptr) { ; CHECK-LABEL: @ptrmask_align8_ptr_align16( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -8) -; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 +; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[PTR:%.*]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; %aligned = call ptr @llvm.ptrmask.p0.i64(ptr %ptr, i64 -8) diff --git a/llvm/test/Transforms/InstSimplify/ptrmask.ll b/llvm/test/Transforms/InstSimplify/ptrmask.ll index 0af7186978ae5bc..6c219e8587f0a0d 100644 --- a/llvm/test/Transforms/InstSimplify/ptrmask.ll +++ b/llvm/test/Transforms/InstSimplify/ptrmask.ll @@ -37,8 +37,7 @@ define ptr @ptrmask_simplify_0_mask(ptr %p) { define ptr @ptrmask_simplify_1s_mask(ptr %p) { ; CHECK-LABEL: define ptr @ptrmask_simplify_1s_mask ; CHECK-SAME: (ptr [[P:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -1) -; CHECK-NEXT: ret ptr [[R]] +; CHECK-NEXT: ret ptr [[P]] ; %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -1) ret ptr %r @@ -57,8 +56,7 @@ define ptr @ptrmask_simplify_1s_mask_i32_fail(ptr %p) { define ptr @ptrmask_simplify_poison_ptr(i64 %m) { ; CHECK-LABEL: define ptr @ptrmask_simplify_poison_ptr ; CHECK-SAME: (i64 [[M:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr poison, i64 [[M]]) -; CHECK-NEXT: ret ptr [[R]] +; CHECK-NEXT: ret ptr poison ; %r = call ptr @llvm.ptrmask.p0.i64(ptr poison, i64 %m) ret ptr %r @@ -67,8 +65,7 @@ define ptr @ptrmask_simplify_poison_ptr(i64 %m) { define ptr @ptrmask_simplify_undef_ptr(i32 %m) { ; CHECK-LABEL: define ptr @ptrmask_simplify_undef_ptr ; CHECK-SAME: (i32 [[M:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr undef, i32 [[M]]) -; CHECK-NEXT: ret ptr [[R]] +; CHECK-NEXT: ret ptr null ; %r = call ptr @llvm.ptrmask.p0.i32(ptr undef, i32 %m) ret ptr %r @@ -77,8 +74,7 @@ define ptr @ptrmask_simplify_undef_ptr(i32 %m) { define ptr @ptrmask_simplify_null_ptr(i64 %m) { ; CHECK-LABEL: define ptr @ptrmask_simplify_null_ptr ; CHECK-SAME: (i64 [[M:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr null, i64 [[M]]) -; CHECK-NEXT: ret ptr [[R]] +; CHECK-NEXT: ret ptr null ; %r = call ptr @llvm.ptrmask.p0.i64(ptr null, i64 %m) ret ptr %r @@ -87,9 +83,7 @@ define ptr @ptrmask_simplify_null_ptr(i64 %m) { define ptr @ptrmask_simplify_ptrmask(ptr %p) { ; CHECK-LABEL: define ptr @ptrmask_simplify_ptrmask ; CHECK-SAME: (ptr [[P:%.*]]) { -; CHECK-NEXT: [[M:%.*]] = ptrtoint ptr [[P]] to i64 -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M]]) -; CHECK-NEXT: ret ptr [[R]] +; CHECK-NEXT: ret ptr [[P]] ; %m = ptrtoint ptr %p to i64 %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m) @@ -111,8 +105,7 @@ define ptr @ptrmask_simplify_ptrmask_i32_fail(ptr %p) { define ptr @ptrmask_simplify_aligned_unused(ptr align 64 %p) { ; CHECK-LABEL: define ptr @ptrmask_simplify_aligned_unused ; CHECK-SAME: (ptr align 64 [[P:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) -; CHECK-NEXT: ret ptr [[R]] +; CHECK-NEXT: ret ptr [[P]] ; %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) ret ptr %r @@ -133,8 +126,7 @@ define ptr @ptrmask_simplify_known_unused(ptr %p) { ; CHECK-SAME: (ptr [[P:%.*]]) { ; CHECK-NEXT: [[PM0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) ; CHECK-NEXT: [[PGEP:%.*]] = getelementptr i8, ptr [[PM0]], i64 32 -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PGEP]], i64 -32) -; CHECK-NEXT: ret ptr [[R]] +; CHECK-NEXT: ret ptr [[PGEP]] ; %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) %pgep = getelementptr i8, ptr %pm0, i64 32 >From 33cc86bb26344964fa37499978b0e05092707cd6 Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:32 -0500 Subject: [PATCH 3/9] [InstCombine] Add tests for combining `llvm.ptrmask`; NFC Differential Revision: https://reviews.llvm.org/D156634 --- .../InstCombine/consecutive-ptrmask.ll | 38 +++++ llvm/test/Transforms/InstCombine/ptrmask.ll | 154 ++++++++++++++++++ 2 files changed, 192 insertions(+) create mode 100644 llvm/test/Transforms/InstCombine/ptrmask.ll diff --git a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll index 904c758b99306f4..adf044da92d2bdf 100644 --- a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll @@ -65,3 +65,41 @@ define ptr @fold_2x_fail_type_mismatch2(ptr %p, i64 %m0, i32 %m1) { %p1 = call ptr @llvm.ptrmask.p0.i32(ptr %p0, i32 %m1) ret ptr %p1 } + + +define ptr @fold_2x_type_mismatch_const0(ptr %p, i32 %m1) { +; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const0 +; CHECK-SAME: (ptr [[P:%.*]], i32 [[M1:%.*]]) { +; CHECK-NEXT: [[P0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -128) +; CHECK-NEXT: [[P1:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 [[M1]]) +; CHECK-NEXT: ret ptr [[P1]] +; + %p0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -128) + %p1 = call ptr @llvm.ptrmask.p0.i32(ptr %p0, i32 %m1) + ret ptr %p1 +} + +define ptr @fold_2x_type_mismatch_const1(ptr %p, i64 %m0) { +; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const1 +; CHECK-SAME: (ptr [[P:%.*]], i64 [[M0:%.*]]) { +; CHECK-NEXT: [[P0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M0]]) +; CHECK-NEXT: [[P1:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 -2) +; CHECK-NEXT: ret ptr [[P1]] +; + %p0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m0) + %p1 = call ptr @llvm.ptrmask.p0.i32(ptr %p0, i32 -2) + ret ptr %p1 +} + + +define ptr @fold_2x_type_mismatch_const2(ptr %p) { +; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const2 +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[P0:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -4) +; CHECK-NEXT: [[P1:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 -31) +; CHECK-NEXT: ret ptr [[P1]] +; + %p0 = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -4) + %p1 = call ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 -31) + ret ptr %p1 +} diff --git a/llvm/test/Transforms/InstCombine/ptrmask.ll b/llvm/test/Transforms/InstCombine/ptrmask.ll new file mode 100644 index 000000000000000..b4ea72b8fb8b65d --- /dev/null +++ b/llvm/test/Transforms/InstCombine/ptrmask.ll @@ -0,0 +1,154 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2 +; RUN: opt < %s -passes=instcombine -S | FileCheck %s + +declare ptr @llvm.ptrmask.p0.i32(ptr, i32) +declare ptr @llvm.ptrmask.p0.i64(ptr, i64) + +define ptr @ptrmask_combine_consecutive_preserve_attrs(ptr %p0, i64 %m1) { +; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs +; CHECK-SAME: (ptr [[P0:%.*]], i64 [[M1:%.*]]) { +; CHECK-NEXT: [[TMP1:%.*]] = and i64 [[M1]], 224 +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 [[TMP1]]) +; CHECK-NEXT: ret ptr [[R]] +; + %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 224) + %r = call noalias ptr @llvm.ptrmask.p0.i64(ptr %pm0, i64 %m1) + ret ptr %r +} + +define ptr @ptrmask_combine_consecutive_preserve_attrs_fail(ptr %p0, i64 %m0) { +; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs_fail +; CHECK-SAME: (ptr [[P0:%.*]], i64 [[M0:%.*]]) { +; CHECK-NEXT: [[TMP1:%.*]] = and i64 [[M0]], 193 +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 [[TMP1]]) +; CHECK-NEXT: ret ptr [[R]] +; + %pm0 = call noalias ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 %m0) + %r = call ptr @llvm.ptrmask.p0.i64(ptr %pm0, i64 193) + ret ptr %r +} + +define ptr @ptrmask_combine_consecutive_preserve_attrs_todo0(ptr %p0) { +; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs_todo0 +; CHECK-SAME: (ptr [[P0:%.*]]) { +; CHECK-NEXT: [[PM0:%.*]] = call noalias ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 224) +; CHECK-NEXT: ret ptr [[PM0]] +; + %pm0 = call noalias ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 224) + %r = call ptr @llvm.ptrmask.p0.i64(ptr %pm0, i64 224) + ret ptr %r +} + +define ptr @ptrmask_combine_consecutive_preserve_attrs_todo1(ptr %p0) { +; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs_todo1 +; CHECK-SAME: (ptr [[P0:%.*]]) { +; CHECK-NEXT: [[PM0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 224) +; CHECK-NEXT: ret ptr [[PM0]] +; + %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 224) + %r = call noalias ptr @llvm.ptrmask.p0.i64(ptr %pm0, i64 224) + ret ptr %r +} + +define ptr @ptrmask_combine_consecutive_preserve_attrs_todo2(ptr %p0) { +; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs_todo2 +; CHECK-SAME: (ptr [[P0:%.*]]) { +; CHECK-NEXT: [[PM0:%.*]] = call noalias ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 224) +; CHECK-NEXT: ret ptr [[PM0]] +; + %pm0 = call noalias ptr @llvm.ptrmask.p0.i32(ptr %p0, i32 224) + %r = call ptr @llvm.ptrmask.p0.i64(ptr %pm0, i64 224) + ret ptr %r +} + +define ptr @ptrmask_combine_add_nonnull(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_combine_add_nonnull +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[PM0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) +; CHECK-NEXT: [[PGEP:%.*]] = getelementptr i8, ptr [[PM0]], i64 33 +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PGEP]], i64 -16) +; CHECK-NEXT: ret ptr [[R]] +; + %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) + %pgep = getelementptr i8, ptr %pm0, i64 33 + %r = call ptr @llvm.ptrmask.p0.i64(ptr %pgep, i64 -16) + ret ptr %r +} + +define ptr @ptrmask_combine_add_alignment(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_combine_add_alignment +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) + ret ptr %r +} + +define ptr @ptrmask_combine_add_alignment2(ptr align 32 %p) { +; CHECK-LABEL: define ptr @ptrmask_combine_add_alignment2 +; CHECK-SAME: (ptr align 32 [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -64) + ret ptr %r +} + +define ptr @ptrmask_combine_improve_alignment(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_combine_improve_alignment +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call align 32 ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -64) + ret ptr %r +} + +define ptr @ptrmask_combine_improve_alignment_fail(ptr %p) { +; CHECK-LABEL: define ptr @ptrmask_combine_improve_alignment_fail +; CHECK-SAME: (ptr [[P:%.*]]) { +; CHECK-NEXT: [[R:%.*]] = call align 128 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) +; CHECK-NEXT: ret ptr [[R]] +; + %r = call align 128 ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -64) + ret ptr %r +} + +define i64 @ptrtoint_of_ptrmask(ptr %p, i64 %m) { +; CHECK-LABEL: define i64 @ptrtoint_of_ptrmask +; CHECK-SAME: (ptr [[P:%.*]], i64 [[M:%.*]]) { +; CHECK-NEXT: [[PM:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M]]) +; CHECK-NEXT: [[R:%.*]] = ptrtoint ptr [[PM]] to i64 +; CHECK-NEXT: ret i64 [[R]] +; + %pm = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m) + %r = ptrtoint ptr %pm to i64 + ret i64 %r +} + +; This succeeds because (ptrtoint i32) gets folded to (trunc i32 (ptrtoint i64)) +define i32 @ptrtoint_of_ptrmask2(ptr %p, i64 %m) { +; CHECK-LABEL: define i32 @ptrtoint_of_ptrmask2 +; CHECK-SAME: (ptr [[P:%.*]], i64 [[M:%.*]]) { +; CHECK-NEXT: [[PM:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M]]) +; CHECK-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[PM]] to i64 +; CHECK-NEXT: [[R:%.*]] = trunc i64 [[TMP1]] to i32 +; CHECK-NEXT: ret i32 [[R]] +; + %pm = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m) + %r = ptrtoint ptr %pm to i32 + ret i32 %r +} + +define i64 @ptrtoint_of_ptrmask_fail(ptr %p, i32 %m) { +; CHECK-LABEL: define i64 @ptrtoint_of_ptrmask_fail +; CHECK-SAME: (ptr [[P:%.*]], i32 [[M:%.*]]) { +; CHECK-NEXT: [[PM:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 [[M]]) +; CHECK-NEXT: [[R:%.*]] = ptrtoint ptr [[PM]] to i64 +; CHECK-NEXT: ret i64 [[R]] +; + %pm = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 %m) + %r = ptrtoint ptr %pm to i64 + ret i64 %r +} >From 45ae0b080750b7cc6b3ae1c86a4d55e402dda092 Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:38 -0500 Subject: [PATCH 4/9] [InstCombine] Make combines on `llvm.ptrmask` fail loudly if we have vec types; NFC Currently vec types are unsupported by `llvm.ptrmask`. Should this change its probably better to fail loudly rather hope the code ends up working properly for vec types. Differential Revision: https://reviews.llvm.org/D156635 --- llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp index c6100f24b0507de..4b9b401f3ab070d 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -1964,6 +1964,15 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) { break; } case Intrinsic::ptrmask: { + Value *Op0 = II->getArgOperand(0); + Value *Op1 = II->getArgOperand(1); + // Fail loudly in case this is ever changed. + // TODO: If vector types are supported the merging of (ptrmask (ptrmask)) + // need to ensure we don't merge a vectype with non-vec type. + assert(!Op0->getType()->isVectorTy() && !Op1->getType()->isVectorTy() && + "These combines where written at a time when ptrmask did not " + "support vector types and may not work for vectors"); + Value *InnerPtr, *InnerMask; if (match(II->getArgOperand(0), m_OneUse(m_Intrinsic<Intrinsic::ptrmask>(m_Value(InnerPtr), >From f725023017914d9bfa431648bbd34999ece4a73e Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:43 -0500 Subject: [PATCH 5/9] [InstCombine] Deduce `align` and `nonnull` return attributes for `llvm.ptrmask` We can deduce the former based on the mask / incoming pointer alignment. We can set the latter based if know the result in non-zero (this is essentially just caching our analysis result). Differential Revision: https://reviews.llvm.org/D156636 --- clang/test/CodeGen/arm64_32-vaarg.c | 2 +- .../InstCombine/InstCombineCalls.cpp | 22 +++++++++++++++++++ .../test/Transforms/InstCombine/align-addr.ll | 12 +++++----- .../InstCombine/consecutive-ptrmask.ll | 10 ++++----- llvm/test/Transforms/InstCombine/ptrmask.ll | 18 +++++++-------- 5 files changed, 43 insertions(+), 21 deletions(-) diff --git a/clang/test/CodeGen/arm64_32-vaarg.c b/clang/test/CodeGen/arm64_32-vaarg.c index 9fbcf88ecfdcc33..3f1f4443436da15 100644 --- a/clang/test/CodeGen/arm64_32-vaarg.c +++ b/clang/test/CodeGen/arm64_32-vaarg.c @@ -29,7 +29,7 @@ long long test_longlong(OneLongLong input, va_list *mylist) { // CHECK-LABEL: define{{.*}} i64 @test_longlong(i64 %input // CHECK: [[STARTPTR:%.*]] = load ptr, ptr %mylist // CHECK: [[ALIGN_TMP:%.+]] = getelementptr inbounds i8, ptr [[STARTPTR]], i32 7 - // CHECK: [[ALIGNED_ADDR:%.+]] = tail call ptr @llvm.ptrmask.p0.i32(ptr nonnull [[ALIGN_TMP]], i32 -8) + // CHECK: [[ALIGNED_ADDR:%.+]] = tail call align 8 ptr @llvm.ptrmask.p0.i32(ptr nonnull [[ALIGN_TMP]], i32 -8) // CHECK: [[NEXT:%.*]] = getelementptr inbounds i8, ptr [[ALIGNED_ADDR]], i32 8 // CHECK: store ptr [[NEXT]], ptr %mylist diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp index 4b9b401f3ab070d..58b69bfa8e49b4f 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -1985,6 +1985,28 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) { {InnerPtr, NewMask})); } } + bool Changed = false; + KnownBits Known = computeKnownBits(II, /*Depth*/ 0, II); + // See if we can deduce non-null. + if (!CI.hasRetAttr(Attribute::NonNull) && + (Known.isNonZero() || + isKnownNonZero(II, DL, /*Depth*/ 0, &AC, II, &DT))) { + CI.addRetAttr(Attribute::NonNull); + Changed = true; + } + + // Known bits will capture if we had alignment information assosiated with + // the pointer argument. + if (Known.countMinTrailingZeros() > Log2(CI.getRetAlign().valueOrOne())) { + if (CI.hasRetAttr(Attribute::Alignment)) + CI.removeRetAttr(Attribute::Alignment); + CI.addRetAttr( + Attribute::get(CI.getContext(), Attribute::Alignment, + uint64_t(1) << Known.countMinTrailingZeros())); + Changed = true; + } + if (Changed) + return &CI; break; } case Intrinsic::uadd_with_overflow: diff --git a/llvm/test/Transforms/InstCombine/align-addr.ll b/llvm/test/Transforms/InstCombine/align-addr.ll index b8e11f1b47165e8..d0294e08d50f8e7 100644 --- a/llvm/test/Transforms/InstCombine/align-addr.ll +++ b/llvm/test/Transforms/InstCombine/align-addr.ll @@ -134,7 +134,7 @@ define <16 x i8> @ptrmask_align_unknown_ptr_align1(ptr align 1 %ptr, i64 %mask) define <16 x i8> @ptrmask_align_unknown_ptr_align8(ptr align 8 %ptr, i64 %mask) { ; CHECK-LABEL: @ptrmask_align_unknown_ptr_align8( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 [[MASK:%.*]]) +; CHECK-NEXT: [[ALIGNED:%.*]] = call align 8 ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 [[MASK:%.*]]) ; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; @@ -146,7 +146,7 @@ define <16 x i8> @ptrmask_align_unknown_ptr_align8(ptr align 8 %ptr, i64 %mask) ; Increase load align from 1 to 2 define <16 x i8> @ptrmask_align2_ptr_align1(ptr align 1 %ptr) { ; CHECK-LABEL: @ptrmask_align2_ptr_align1( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -2) +; CHECK-NEXT: [[ALIGNED:%.*]] = call align 2 ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -2) ; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; @@ -158,7 +158,7 @@ define <16 x i8> @ptrmask_align2_ptr_align1(ptr align 1 %ptr) { ; Increase load align from 1 to 4 define <16 x i8> @ptrmask_align4_ptr_align1(ptr align 1 %ptr) { ; CHECK-LABEL: @ptrmask_align4_ptr_align1( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -4) +; CHECK-NEXT: [[ALIGNED:%.*]] = call align 4 ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -4) ; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; @@ -170,7 +170,7 @@ define <16 x i8> @ptrmask_align4_ptr_align1(ptr align 1 %ptr) { ; Increase load align from 1 to 8 define <16 x i8> @ptrmask_align8_ptr_align1(ptr align 1 %ptr) { ; CHECK-LABEL: @ptrmask_align8_ptr_align1( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -8) +; CHECK-NEXT: [[ALIGNED:%.*]] = call align 8 ptr @llvm.ptrmask.p0.i64(ptr [[PTR:%.*]], i64 -8) ; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; @@ -205,7 +205,7 @@ define <16 x i8> @ptrmask_align8_ptr_align16(ptr align 16 %ptr) { ; than the pointer size. define <16 x i8> @ptrmask_align8_ptr_align1_smallmask(ptr align 1 %ptr) { ; CHECK-LABEL: @ptrmask_align8_ptr_align1_smallmask( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[PTR:%.*]], i32 -8) +; CHECK-NEXT: [[ALIGNED:%.*]] = call align 8 ptr @llvm.ptrmask.p0.i32(ptr [[PTR:%.*]], i32 -8) ; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; @@ -218,7 +218,7 @@ define <16 x i8> @ptrmask_align8_ptr_align1_smallmask(ptr align 1 %ptr) { ; than the pointer size. define <16 x i8> @ptrmask_align8_ptr_align1_bigmask(ptr align 1 %ptr) { ; CHECK-LABEL: @ptrmask_align8_ptr_align1_bigmask( -; CHECK-NEXT: [[ALIGNED:%.*]] = call ptr @llvm.ptrmask.p0.i128(ptr [[PTR:%.*]], i128 -8) +; CHECK-NEXT: [[ALIGNED:%.*]] = call align 8 ptr @llvm.ptrmask.p0.i128(ptr [[PTR:%.*]], i128 -8) ; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; diff --git a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll index adf044da92d2bdf..a4f75774baad7f9 100644 --- a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll @@ -70,8 +70,8 @@ define ptr @fold_2x_fail_type_mismatch2(ptr %p, i64 %m0, i32 %m1) { define ptr @fold_2x_type_mismatch_const0(ptr %p, i32 %m1) { ; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const0 ; CHECK-SAME: (ptr [[P:%.*]], i32 [[M1:%.*]]) { -; CHECK-NEXT: [[P0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -128) -; CHECK-NEXT: [[P1:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 [[M1]]) +; CHECK-NEXT: [[P0:%.*]] = call align 128 ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -128) +; CHECK-NEXT: [[P1:%.*]] = call align 128 ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 [[M1]]) ; CHECK-NEXT: ret ptr [[P1]] ; %p0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -128) @@ -83,7 +83,7 @@ define ptr @fold_2x_type_mismatch_const1(ptr %p, i64 %m0) { ; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const1 ; CHECK-SAME: (ptr [[P:%.*]], i64 [[M0:%.*]]) { ; CHECK-NEXT: [[P0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M0]]) -; CHECK-NEXT: [[P1:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 -2) +; CHECK-NEXT: [[P1:%.*]] = call align 2 ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 -2) ; CHECK-NEXT: ret ptr [[P1]] ; %p0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m0) @@ -95,8 +95,8 @@ define ptr @fold_2x_type_mismatch_const1(ptr %p, i64 %m0) { define ptr @fold_2x_type_mismatch_const2(ptr %p) { ; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const2 ; CHECK-SAME: (ptr [[P:%.*]]) { -; CHECK-NEXT: [[P0:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -4) -; CHECK-NEXT: [[P1:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 -31) +; CHECK-NEXT: [[P0:%.*]] = call align 4 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -4) +; CHECK-NEXT: [[P1:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 -31) ; CHECK-NEXT: ret ptr [[P1]] ; %p0 = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -4) diff --git a/llvm/test/Transforms/InstCombine/ptrmask.ll b/llvm/test/Transforms/InstCombine/ptrmask.ll index b4ea72b8fb8b65d..2fd055650a863b4 100644 --- a/llvm/test/Transforms/InstCombine/ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/ptrmask.ll @@ -8,7 +8,7 @@ define ptr @ptrmask_combine_consecutive_preserve_attrs(ptr %p0, i64 %m1) { ; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs ; CHECK-SAME: (ptr [[P0:%.*]], i64 [[M1:%.*]]) { ; CHECK-NEXT: [[TMP1:%.*]] = and i64 [[M1]], 224 -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 [[TMP1]]) +; CHECK-NEXT: [[R:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 [[TMP1]]) ; CHECK-NEXT: ret ptr [[R]] ; %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 224) @@ -31,7 +31,7 @@ define ptr @ptrmask_combine_consecutive_preserve_attrs_fail(ptr %p0, i64 %m0) { define ptr @ptrmask_combine_consecutive_preserve_attrs_todo0(ptr %p0) { ; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs_todo0 ; CHECK-SAME: (ptr [[P0:%.*]]) { -; CHECK-NEXT: [[PM0:%.*]] = call noalias ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 224) +; CHECK-NEXT: [[PM0:%.*]] = call noalias align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 224) ; CHECK-NEXT: ret ptr [[PM0]] ; %pm0 = call noalias ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 224) @@ -42,7 +42,7 @@ define ptr @ptrmask_combine_consecutive_preserve_attrs_todo0(ptr %p0) { define ptr @ptrmask_combine_consecutive_preserve_attrs_todo1(ptr %p0) { ; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs_todo1 ; CHECK-SAME: (ptr [[P0:%.*]]) { -; CHECK-NEXT: [[PM0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 224) +; CHECK-NEXT: [[PM0:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 224) ; CHECK-NEXT: ret ptr [[PM0]] ; %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 224) @@ -53,7 +53,7 @@ define ptr @ptrmask_combine_consecutive_preserve_attrs_todo1(ptr %p0) { define ptr @ptrmask_combine_consecutive_preserve_attrs_todo2(ptr %p0) { ; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs_todo2 ; CHECK-SAME: (ptr [[P0:%.*]]) { -; CHECK-NEXT: [[PM0:%.*]] = call noalias ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 224) +; CHECK-NEXT: [[PM0:%.*]] = call noalias align 32 ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 224) ; CHECK-NEXT: ret ptr [[PM0]] ; %pm0 = call noalias ptr @llvm.ptrmask.p0.i32(ptr %p0, i32 224) @@ -64,9 +64,9 @@ define ptr @ptrmask_combine_consecutive_preserve_attrs_todo2(ptr %p0) { define ptr @ptrmask_combine_add_nonnull(ptr %p) { ; CHECK-LABEL: define ptr @ptrmask_combine_add_nonnull ; CHECK-SAME: (ptr [[P:%.*]]) { -; CHECK-NEXT: [[PM0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) +; CHECK-NEXT: [[PM0:%.*]] = call align 64 ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) ; CHECK-NEXT: [[PGEP:%.*]] = getelementptr i8, ptr [[PM0]], i64 33 -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[PGEP]], i64 -16) +; CHECK-NEXT: [[R:%.*]] = call nonnull align 32 ptr @llvm.ptrmask.p0.i64(ptr [[PGEP]], i64 -16) ; CHECK-NEXT: ret ptr [[R]] ; %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) @@ -78,7 +78,7 @@ define ptr @ptrmask_combine_add_nonnull(ptr %p) { define ptr @ptrmask_combine_add_alignment(ptr %p) { ; CHECK-LABEL: define ptr @ptrmask_combine_add_alignment ; CHECK-SAME: (ptr [[P:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) +; CHECK-NEXT: [[R:%.*]] = call align 64 ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) ; CHECK-NEXT: ret ptr [[R]] ; %r = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) @@ -88,7 +88,7 @@ define ptr @ptrmask_combine_add_alignment(ptr %p) { define ptr @ptrmask_combine_add_alignment2(ptr align 32 %p) { ; CHECK-LABEL: define ptr @ptrmask_combine_add_alignment2 ; CHECK-SAME: (ptr align 32 [[P:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) +; CHECK-NEXT: [[R:%.*]] = call align 64 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) ; CHECK-NEXT: ret ptr [[R]] ; %r = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -64) @@ -98,7 +98,7 @@ define ptr @ptrmask_combine_add_alignment2(ptr align 32 %p) { define ptr @ptrmask_combine_improve_alignment(ptr %p) { ; CHECK-LABEL: define ptr @ptrmask_combine_improve_alignment ; CHECK-SAME: (ptr [[P:%.*]]) { -; CHECK-NEXT: [[R:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) +; CHECK-NEXT: [[R:%.*]] = call align 64 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -64) ; CHECK-NEXT: ret ptr [[R]] ; %r = call align 32 ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -64) >From 0531af1f2470ee79ff27a6912f79e402ee6ef163 Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:48 -0500 Subject: [PATCH 6/9] [InstCombine] Implement `SimplifyDemandedBits` for `llvm.ptrmask` Logic basically copies 'and' but we can't return a constant if the result == `rhs` (mask) so that case is skipped. Differential Revision: https://reviews.llvm.org/D156637 --- .../InstCombine/InstCombineCalls.cpp | 14 ++-- .../InstCombine/InstCombineInternal.h | 1 + .../InstCombineSimplifyDemanded.cpp | 69 +++++++++++++++++-- .../test/Transforms/InstCombine/align-addr.ll | 2 +- .../InstCombine/consecutive-ptrmask.ll | 2 +- llvm/test/Transforms/InstCombine/ptrmask.ll | 2 +- 6 files changed, 78 insertions(+), 12 deletions(-) diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp index 58b69bfa8e49b4f..5c1de2af2282299 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -1964,6 +1964,10 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) { break; } case Intrinsic::ptrmask: { + KnownBits Known(DL.getPointerTypeSizeInBits(II->getType())); + if (SimplifyDemandedInstructionBits(*II, Known)) + return II; + Value *Op0 = II->getArgOperand(0); Value *Op1 = II->getArgOperand(1); // Fail loudly in case this is ever changed. @@ -1985,8 +1989,8 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) { {InnerPtr, NewMask})); } } + bool Changed = false; - KnownBits Known = computeKnownBits(II, /*Depth*/ 0, II); // See if we can deduce non-null. if (!CI.hasRetAttr(Attribute::NonNull) && (Known.isNonZero() || @@ -2000,9 +2004,11 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) { if (Known.countMinTrailingZeros() > Log2(CI.getRetAlign().valueOrOne())) { if (CI.hasRetAttr(Attribute::Alignment)) CI.removeRetAttr(Attribute::Alignment); - CI.addRetAttr( - Attribute::get(CI.getContext(), Attribute::Alignment, - uint64_t(1) << Known.countMinTrailingZeros())); + CI.addRetAttr(Attribute::get( + CI.getContext(), Attribute::Alignment, + uint64_t(1) << (Known.isZero() ? (Known.getBitWidth() - 1) + : Known.countMinTrailingZeros()))); + Changed = true; } if (Changed) diff --git a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h index 40c24d87bfec508..0b03aa48fa0c0cc 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h +++ b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h @@ -526,6 +526,7 @@ class LLVM_LIBRARY_VISIBILITY InstCombinerImpl final /// Tries to simplify operands to an integer instruction based on its /// demanded bits. bool SimplifyDemandedInstructionBits(Instruction &Inst); + bool SimplifyDemandedInstructionBits(Instruction &Inst, KnownBits &Known); Value *SimplifyDemandedVectorElts(Value *V, APInt DemandedElts, APInt &UndefElts, unsigned Depth = 0, diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp index be005e61a8d2d89..7902eb0c6c829cc 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp @@ -48,15 +48,20 @@ static bool ShrinkDemandedConstant(Instruction *I, unsigned OpNo, return true; } +/// Returns the bitwidth of the given scalar or pointer type. For vector types, +/// returns the element type's bitwidth. +static unsigned getBitWidth(Type *Ty, const DataLayout &DL) { + if (unsigned BitWidth = Ty->getScalarSizeInBits()) + return BitWidth; + return DL.getPointerTypeSizeInBits(Ty); +} /// Inst is an integer instruction that SimplifyDemandedBits knows about. See if /// the instruction has any properties that allow us to simplify its operands. -bool InstCombinerImpl::SimplifyDemandedInstructionBits(Instruction &Inst) { - unsigned BitWidth = Inst.getType()->getScalarSizeInBits(); - KnownBits Known(BitWidth); - APInt DemandedMask(APInt::getAllOnes(BitWidth)); - +bool InstCombinerImpl::SimplifyDemandedInstructionBits(Instruction &Inst, + KnownBits &Known) { + APInt DemandedMask(APInt::getAllOnes(Known.getBitWidth())); Value *V = SimplifyDemandedUseBits(&Inst, DemandedMask, Known, 0, &Inst); if (!V) return false; @@ -65,6 +70,13 @@ bool InstCombinerImpl::SimplifyDemandedInstructionBits(Instruction &Inst) { return true; } +/// Inst is an integer instruction that SimplifyDemandedBits knows about. See if +/// the instruction has any properties that allow us to simplify its operands. +bool InstCombinerImpl::SimplifyDemandedInstructionBits(Instruction &Inst) { + KnownBits Known(getBitWidth(Inst.getType(), DL)); + return SimplifyDemandedInstructionBits(Inst, Known); +} + /// This form of SimplifyDemandedBits simplifies the specified instruction /// operand if possible, updating it in place. It returns true if it made any /// change and false otherwise. @@ -898,6 +910,53 @@ Value *InstCombinerImpl::SimplifyDemandedUseBits(Value *V, APInt DemandedMask, } break; } + case Intrinsic::ptrmask: { + // Fail loudly in case this is ever changed. + // Likely not much needs to be changed here to support vector types. + assert(!I->getOperand(0)->getType()->isVectorTy() && + !I->getOperand(1)->getType()->isVectorTy() && + "These simplifications where written at a time when ptrmask did " + "not support vector types and may not work for vectors"); + + unsigned MaskWidth = I->getOperand(1)->getType()->getScalarSizeInBits(); + RHSKnown = KnownBits(MaskWidth); + // If either the LHS or the RHS are Zero, the result is zero. + if (SimplifyDemandedBits(I, 0, DemandedMask, LHSKnown, Depth + 1) || + SimplifyDemandedBits( + I, 1, (DemandedMask & ~LHSKnown.Zero).zextOrTrunc(MaskWidth), + RHSKnown, Depth + 1)) + return I; + + RHSKnown = RHSKnown.zextOrTrunc(BitWidth); + assert(!RHSKnown.hasConflict() && "Bits known to be one AND zero?"); + assert(!LHSKnown.hasConflict() && "Bits known to be one AND zero?"); + + Known = LHSKnown & RHSKnown; + KnownBitsComputed = DemandedMask.isAllOnes(); + + // If the client is only demanding bits we know to be zero, return + // `llvm.ptrmask(p, 0)`. We can't return `null` here due to pointer + // provenance, but making the mask zero will be easily optimizable in + // the backend. + if (DemandedMask.isSubsetOf(Known.Zero)) + return replaceOperand( + *I, 1, Constant::getNullValue(I->getOperand(1)->getType())); + + // Mask in demanded space does nothing. + // TODO: We may have attributes assosiated with the return value of the + // llvm.ptrmask intrinsic that will be lost when we just return the + // operand. We should try to preserve them. + if (DemandedMask.isSubsetOf(RHSKnown.One | LHSKnown.Zero)) + return I->getOperand(0); + + // If the RHS is a constant, see if we can simplify it. + if (ShrinkDemandedConstant( + I, 1, (DemandedMask & ~LHSKnown.Zero).zextOrTrunc(MaskWidth))) + return I; + + break; + } + case Intrinsic::fshr: case Intrinsic::fshl: { const APInt *SA; diff --git a/llvm/test/Transforms/InstCombine/align-addr.ll b/llvm/test/Transforms/InstCombine/align-addr.ll index d0294e08d50f8e7..5753966c124ae89 100644 --- a/llvm/test/Transforms/InstCombine/align-addr.ll +++ b/llvm/test/Transforms/InstCombine/align-addr.ll @@ -218,7 +218,7 @@ define <16 x i8> @ptrmask_align8_ptr_align1_smallmask(ptr align 1 %ptr) { ; than the pointer size. define <16 x i8> @ptrmask_align8_ptr_align1_bigmask(ptr align 1 %ptr) { ; CHECK-LABEL: @ptrmask_align8_ptr_align1_bigmask( -; CHECK-NEXT: [[ALIGNED:%.*]] = call align 8 ptr @llvm.ptrmask.p0.i128(ptr [[PTR:%.*]], i128 -8) +; CHECK-NEXT: [[ALIGNED:%.*]] = call align 8 ptr @llvm.ptrmask.p0.i128(ptr [[PTR:%.*]], i128 18446744073709551608) ; CHECK-NEXT: [[LOAD:%.*]] = load <16 x i8>, ptr [[ALIGNED]], align 1 ; CHECK-NEXT: ret <16 x i8> [[LOAD]] ; diff --git a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll index a4f75774baad7f9..c5ee06c3902fae1 100644 --- a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll @@ -96,7 +96,7 @@ define ptr @fold_2x_type_mismatch_const2(ptr %p) { ; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const2 ; CHECK-SAME: (ptr [[P:%.*]]) { ; CHECK-NEXT: [[P0:%.*]] = call align 4 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -4) -; CHECK-NEXT: [[P1:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 -31) +; CHECK-NEXT: [[P1:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 4294967264) ; CHECK-NEXT: ret ptr [[P1]] ; %p0 = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -4) diff --git a/llvm/test/Transforms/InstCombine/ptrmask.ll b/llvm/test/Transforms/InstCombine/ptrmask.ll index 2fd055650a863b4..91fbd373f57b372 100644 --- a/llvm/test/Transforms/InstCombine/ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/ptrmask.ll @@ -66,7 +66,7 @@ define ptr @ptrmask_combine_add_nonnull(ptr %p) { ; CHECK-SAME: (ptr [[P:%.*]]) { ; CHECK-NEXT: [[PM0:%.*]] = call align 64 ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -64) ; CHECK-NEXT: [[PGEP:%.*]] = getelementptr i8, ptr [[PM0]], i64 33 -; CHECK-NEXT: [[R:%.*]] = call nonnull align 32 ptr @llvm.ptrmask.p0.i64(ptr [[PGEP]], i64 -16) +; CHECK-NEXT: [[R:%.*]] = call nonnull align 32 ptr @llvm.ptrmask.p0.i64(ptr [[PGEP]], i64 -32) ; CHECK-NEXT: ret ptr [[R]] ; %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -64) >From 02753bb4d76baaff2599e08af9a53168156ab037 Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:53 -0500 Subject: [PATCH 7/9] [InstCombine] Preserve return attributes when merging `llvm.ptrmask` If we have assosiated attributes i.e `([ret_attrs] (ptrmask (ptrmask p0, m0), m1))` we should preserve `[ret_attrs]` when combining the two `llvm.ptrmask`s. Differential Revision: https://reviews.llvm.org/D156638 --- .../InstCombine/InstCombineCalls.cpp | 23 +++++++++++-------- llvm/test/Transforms/InstCombine/ptrmask.ll | 2 +- 2 files changed, 14 insertions(+), 11 deletions(-) diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp index 5c1de2af2282299..8fa61e3804933f2 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -1978,19 +1978,22 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) { "support vector types and may not work for vectors"); Value *InnerPtr, *InnerMask; - if (match(II->getArgOperand(0), - m_OneUse(m_Intrinsic<Intrinsic::ptrmask>(m_Value(InnerPtr), - m_Value(InnerMask))))) { - if (II->getArgOperand(1)->getType() == InnerMask->getType()) { - Value *NewMask = Builder.CreateAnd(II->getArgOperand(1), InnerMask); - return replaceInstUsesWith( - *II, - Builder.CreateIntrinsic(InnerPtr->getType(), Intrinsic::ptrmask, - {InnerPtr, NewMask})); + bool Changed = false; + // Combine: + // (ptrmask (ptrmask p, A), B) + // -> (ptrmask p, (and A, B)) + if (match(Op0, m_OneUse(m_Intrinsic<Intrinsic::ptrmask>( + m_Value(InnerPtr), m_Value(InnerMask))))) { + if (InnerMask->getType() == Op1->getType()) { + // TODO: If InnerMask == Op1, we could copy attributes from inner + // callsite -> outer callsite. + Value *NewMask = Builder.CreateAnd(Op1, InnerMask); + replaceOperand(CI, 0, InnerPtr); + replaceOperand(CI, 1, NewMask); + Changed = true; } } - bool Changed = false; // See if we can deduce non-null. if (!CI.hasRetAttr(Attribute::NonNull) && (Known.isNonZero() || diff --git a/llvm/test/Transforms/InstCombine/ptrmask.ll b/llvm/test/Transforms/InstCombine/ptrmask.ll index 91fbd373f57b372..7b4b534b5b2f920 100644 --- a/llvm/test/Transforms/InstCombine/ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/ptrmask.ll @@ -8,7 +8,7 @@ define ptr @ptrmask_combine_consecutive_preserve_attrs(ptr %p0, i64 %m1) { ; CHECK-LABEL: define ptr @ptrmask_combine_consecutive_preserve_attrs ; CHECK-SAME: (ptr [[P0:%.*]], i64 [[M1:%.*]]) { ; CHECK-NEXT: [[TMP1:%.*]] = and i64 [[M1]], 224 -; CHECK-NEXT: [[R:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 [[TMP1]]) +; CHECK-NEXT: [[R:%.*]] = call noalias align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 [[TMP1]]) ; CHECK-NEXT: ret ptr [[R]] ; %pm0 = call ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 224) >From b7e7fe2724a2d7542f739f320f5503045839144b Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:21:59 -0500 Subject: [PATCH 8/9] [InstCombine] Merge consecutive `llvm.ptrmask` with different mask types if a mask is constant. We can `zext` / `trunc` a constant mask for free, so if one of the masks if constant, we can proceed even if they are not the same type. Differential Revision: https://reviews.llvm.org/D156639 --- .../InstCombine/InstCombineCalls.cpp | 20 +++++++++++++++++++ .../InstCombine/consecutive-ptrmask.ll | 11 +++++----- 2 files changed, 25 insertions(+), 6 deletions(-) diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp index 8fa61e3804933f2..e0a8b30b34a8ff0 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -1984,10 +1984,30 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) { // -> (ptrmask p, (and A, B)) if (match(Op0, m_OneUse(m_Intrinsic<Intrinsic::ptrmask>( m_Value(InnerPtr), m_Value(InnerMask))))) { + // See if combining the two masks is free. + bool OkayToMerge = InnerMask->getType() == Op1->getType(); + bool NeedsNew = false; + if (!OkayToMerge) { + if (match(InnerMask, m_ImmConstant())) { + InnerMask = Builder.CreateZExtOrTrunc(InnerMask, Op1->getType()); + OkayToMerge = true; + } else if (match(Op1, m_ImmConstant())) { + Op1 = Builder.CreateZExtOrTrunc(Op1, InnerMask->getType()); + OkayToMerge = true; + // Need to create a new one here, as the intrinsic id needs to change. + NeedsNew = true; + } + } if (InnerMask->getType() == Op1->getType()) { // TODO: If InnerMask == Op1, we could copy attributes from inner // callsite -> outer callsite. Value *NewMask = Builder.CreateAnd(Op1, InnerMask); + if (NeedsNew) + return replaceInstUsesWith( + *II, + Builder.CreateIntrinsic(InnerPtr->getType(), Intrinsic::ptrmask, + {InnerPtr, NewMask})); + replaceOperand(CI, 0, InnerPtr); replaceOperand(CI, 1, NewMask); Changed = true; diff --git a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll index c5ee06c3902fae1..0723fe2fd4239c4 100644 --- a/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/consecutive-ptrmask.ll @@ -70,8 +70,8 @@ define ptr @fold_2x_fail_type_mismatch2(ptr %p, i64 %m0, i32 %m1) { define ptr @fold_2x_type_mismatch_const0(ptr %p, i32 %m1) { ; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const0 ; CHECK-SAME: (ptr [[P:%.*]], i32 [[M1:%.*]]) { -; CHECK-NEXT: [[P0:%.*]] = call align 128 ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 -128) -; CHECK-NEXT: [[P1:%.*]] = call align 128 ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 [[M1]]) +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[M1]], -128 +; CHECK-NEXT: [[P1:%.*]] = call align 128 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 [[TMP1]]) ; CHECK-NEXT: ret ptr [[P1]] ; %p0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 -128) @@ -82,8 +82,8 @@ define ptr @fold_2x_type_mismatch_const0(ptr %p, i32 %m1) { define ptr @fold_2x_type_mismatch_const1(ptr %p, i64 %m0) { ; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const1 ; CHECK-SAME: (ptr [[P:%.*]], i64 [[M0:%.*]]) { -; CHECK-NEXT: [[P0:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M0]]) -; CHECK-NEXT: [[P1:%.*]] = call align 2 ptr @llvm.ptrmask.p0.i32(ptr [[P0]], i32 -2) +; CHECK-NEXT: [[TMP1:%.*]] = and i64 [[M0]], 4294967294 +; CHECK-NEXT: [[P1:%.*]] = call align 2 ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[TMP1]]) ; CHECK-NEXT: ret ptr [[P1]] ; %p0 = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m0) @@ -95,8 +95,7 @@ define ptr @fold_2x_type_mismatch_const1(ptr %p, i64 %m0) { define ptr @fold_2x_type_mismatch_const2(ptr %p) { ; CHECK-LABEL: define ptr @fold_2x_type_mismatch_const2 ; CHECK-SAME: (ptr [[P:%.*]]) { -; CHECK-NEXT: [[P0:%.*]] = call align 4 ptr @llvm.ptrmask.p0.i32(ptr [[P]], i32 -4) -; CHECK-NEXT: [[P1:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P0]], i64 4294967264) +; CHECK-NEXT: [[P1:%.*]] = call align 32 ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 4294967264) ; CHECK-NEXT: ret ptr [[P1]] ; %p0 = call ptr @llvm.ptrmask.p0.i32(ptr %p, i32 -4) >From a3283a35c51db78b6464f9bb4f088c29ae298ab6 Mon Sep 17 00:00:00 2001 From: Noah Goldstein <goldstein....@gmail.com> Date: Fri, 22 Sep 2023 08:22:01 -0500 Subject: [PATCH 9/9] [InstCombine] Fold `(ptrtoint (ptrmask p0, m0))` -> `(and (ptrtoint p0), m0)` `and` is generally more supported so if we have a `ptrmask` anyways might as well use `and`. Differential Revision: https://reviews.llvm.org/D156640 --- llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp | 9 +++++++++ llvm/test/Transforms/InstCombine/ptrmask.ll | 10 +++++----- 2 files changed, 14 insertions(+), 5 deletions(-) diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp index 8ff61ab36da307b..508609670b9ac88 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp @@ -1955,6 +1955,15 @@ Instruction *InstCombinerImpl::visitPtrToInt(PtrToIntInst &CI) { return CastInst::CreateIntegerCast(P, Ty, /*isSigned=*/false); } + // (ptrtoint (ptrmask P, M)) + // -> (and (ptrtoint P), M) + // This is generally beneficial as `and` is better supported than `ptrmask`. + Value *Ptr, *Mask; + if (match(SrcOp, m_OneUse(m_Intrinsic<Intrinsic::ptrmask>(m_Value(Ptr), + m_Value(Mask)))) && + Mask->getType() == Ty) + return BinaryOperator::CreateAnd(Builder.CreatePtrToInt(Ptr, Ty), Mask); + if (auto *GEP = dyn_cast<GetElementPtrInst>(SrcOp)) { // Fold ptrtoint(gep null, x) to multiply + constant if the GEP has one use. // While this can increase the number of instructions it doesn't actually diff --git a/llvm/test/Transforms/InstCombine/ptrmask.ll b/llvm/test/Transforms/InstCombine/ptrmask.ll index 7b4b534b5b2f920..f4839f546f82746 100644 --- a/llvm/test/Transforms/InstCombine/ptrmask.ll +++ b/llvm/test/Transforms/InstCombine/ptrmask.ll @@ -118,8 +118,8 @@ define ptr @ptrmask_combine_improve_alignment_fail(ptr %p) { define i64 @ptrtoint_of_ptrmask(ptr %p, i64 %m) { ; CHECK-LABEL: define i64 @ptrtoint_of_ptrmask ; CHECK-SAME: (ptr [[P:%.*]], i64 [[M:%.*]]) { -; CHECK-NEXT: [[PM:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M]]) -; CHECK-NEXT: [[R:%.*]] = ptrtoint ptr [[PM]] to i64 +; CHECK-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[P]] to i64 +; CHECK-NEXT: [[R:%.*]] = and i64 [[TMP1]], [[M]] ; CHECK-NEXT: ret i64 [[R]] ; %pm = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m) @@ -131,9 +131,9 @@ define i64 @ptrtoint_of_ptrmask(ptr %p, i64 %m) { define i32 @ptrtoint_of_ptrmask2(ptr %p, i64 %m) { ; CHECK-LABEL: define i32 @ptrtoint_of_ptrmask2 ; CHECK-SAME: (ptr [[P:%.*]], i64 [[M:%.*]]) { -; CHECK-NEXT: [[PM:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P]], i64 [[M]]) -; CHECK-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[PM]] to i64 -; CHECK-NEXT: [[R:%.*]] = trunc i64 [[TMP1]] to i32 +; CHECK-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[P]] to i64 +; CHECK-NEXT: [[TMP2:%.*]] = and i64 [[TMP1]], [[M]] +; CHECK-NEXT: [[R:%.*]] = trunc i64 [[TMP2]] to i32 ; CHECK-NEXT: ret i32 [[R]] ; %pm = call ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m) _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
