caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
a58dcc5e08665f2d58a28c9d4510cf94de6ed3bf
clang-p2996/llvm/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll
Florian Hahn 51afb10174 [LV] Create block in mask up-front if needed. (#76635) 
At the moment, block and edge masks are created on demand, which means
that they are inserted at the point where they are demanded and then
cached. It is possible that the mask for a block is looked up later at a
point that's not dominated by the point where the mask has been
inserted.

To avoid this, create masks up front on entry to the corresponding basic
block and leave it to VPlan simplification to remove unneeded masks.

Note that we need to create masks for all blocks, if any of the blocks
in the loop needs predication, as computing the mask of a block depends
on the masks of its predecessor.

Needed for #76090.

https://github.com/llvm/llvm-project/pull/76635
2024-01-09 10:50:08 +00:00

192 lines
9.1 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -S -passes=loop-vectorize -mtriple=x86_64-apple-darwin %s | FileCheck %s --check-prefixes=CHECK,SSE
; RUN: opt -S -passes=loop-vectorize -mtriple=x86_64-apple-darwin -mattr=+avx %s | FileCheck %s --check-prefixes=CHECK,AVX
; Two mostly identical functions. The only difference is the presence of
; fast-math flags on the second. The loop is a pretty simple reduction:
; for (int i = 0; i < 32; ++i)
; if (arr[i] != 42)
; tot += arr[i];
define double @sumIfScalar(ptr nocapture readonly %arr) {
; CHECK-LABEL: @sumIfScalar(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
; CHECK-NEXT: [[TOT:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TOT_NEXT:%.*]], [[NEXT_ITER]] ]
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr double, ptr [[ARR:%.*]], i32 [[I]]
; CHECK-NEXT: [[NEXTVAL:%.*]] = load double, ptr [[ADDR]], align 8
; CHECK-NEXT: [[TST:%.*]] = fcmp une double [[NEXTVAL]], 4.200000e+01
; CHECK-NEXT: br i1 [[TST]], label [[DO_ADD:%.*]], label [[NO_ADD:%.*]]
; CHECK: do.add:
; CHECK-NEXT: [[TOT_NEW:%.*]] = fadd double [[TOT]], [[NEXTVAL]]
; CHECK-NEXT: br label [[NEXT_ITER]]
; CHECK: no.add:
; CHECK-NEXT: br label [[NEXT_ITER]]
; CHECK: next.iter:
; CHECK-NEXT: [[TOT_NEXT]] = phi double [ [[TOT]], [[NO_ADD]] ], [ [[TOT_NEW]], [[DO_ADD]] ]
; CHECK-NEXT: [[I_NEXT]] = add i32 [[I]], 1
; CHECK-NEXT: [[AGAIN:%.*]] = icmp ult i32 [[I_NEXT]], 32
; CHECK-NEXT: br i1 [[AGAIN]], label [[LOOP]], label [[DONE:%.*]]
; CHECK: done:
; CHECK-NEXT: [[TOT_NEXT_LCSSA:%.*]] = phi double [ [[TOT_NEXT]], [[NEXT_ITER]] ]
; CHECK-NEXT: ret double [[TOT_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%i = phi i32 [0, %entry], [%i.next, %next.iter]
%tot = phi double [0.0, %entry], [%tot.next, %next.iter]
%addr = getelementptr double, ptr %arr, i32 %i
%nextval = load double, ptr %addr
%tst = fcmp une double %nextval, 42.0
br i1 %tst, label %do.add, label %no.add
do.add:
%tot.new = fadd double %tot, %nextval
br label %next.iter
no.add:
br label %next.iter
next.iter:
%tot.next = phi double [%tot, %no.add], [%tot.new, %do.add]
%i.next = add i32 %i, 1
%again = icmp ult i32 %i.next, 32
br i1 %again, label %loop, label %done
done:
ret double %tot.next
}
define double @sumIfVector(ptr nocapture readonly %arr) {
; SSE-LABEL: @sumIfVector(
; SSE-NEXT: entry:
; SSE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SSE: vector.ph:
; SSE-NEXT: br label [[VECTOR_BODY:%.*]]
; SSE: vector.body:
; SSE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SSE-NEXT: [[VEC_PHI:%.*]] = phi <2 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
; SSE-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0
; SSE-NEXT: [[TMP1:%.*]] = getelementptr double, ptr [[ARR:%.*]], i32 [[TMP0]]
; SSE-NEXT: [[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i32 0
; SSE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8
; SSE-NEXT: [[TMP3:%.*]] = fcmp fast une <2 x double> [[WIDE_LOAD]], <double 4.200000e+01, double 4.200000e+01>
; SSE-NEXT: [[TMP5:%.*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
; SSE-NEXT: [[TMP4:%.*]] = fadd fast <2 x double> [[VEC_PHI]], [[WIDE_LOAD]]
; SSE-NEXT: [[PREDPHI]] = select <2 x i1> [[TMP3]], <2 x double> [[TMP4]], <2 x double> [[VEC_PHI]]
; SSE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; SSE-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
; SSE-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; SSE: middle.block:
; SSE-NEXT: [[TMP7:%.*]] = call fast double @llvm.vector.reduce.fadd.v2f64(double -0.000000e+00, <2 x double> [[PREDPHI]])
; SSE-NEXT: br i1 true, label [[DONE:%.*]], label [[SCALAR_PH]]
; SSE: scalar.ph:
; SSE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; SSE-NEXT: [[BC_MERGE_RDX:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; SSE-NEXT: br label [[LOOP:%.*]]
; SSE: loop:
; SSE-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
; SSE-NEXT: [[TOT:%.*]] = phi double [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TOT_NEXT:%.*]], [[NEXT_ITER]] ]
; SSE-NEXT: [[ADDR:%.*]] = getelementptr double, ptr [[ARR]], i32 [[I]]
; SSE-NEXT: [[NEXTVAL:%.*]] = load double, ptr [[ADDR]], align 8
; SSE-NEXT: [[TST:%.*]] = fcmp fast une double [[NEXTVAL]], 4.200000e+01
; SSE-NEXT: br i1 [[TST]], label [[DO_ADD:%.*]], label [[NO_ADD:%.*]]
; SSE: do.add:
; SSE-NEXT: [[TOT_NEW:%.*]] = fadd fast double [[TOT]], [[NEXTVAL]]
; SSE-NEXT: br label [[NEXT_ITER]]
; SSE: no.add:
; SSE-NEXT: br label [[NEXT_ITER]]
; SSE: next.iter:
; SSE-NEXT: [[TOT_NEXT]] = phi double [ [[TOT]], [[NO_ADD]] ], [ [[TOT_NEW]], [[DO_ADD]] ]
; SSE-NEXT: [[I_NEXT]] = add i32 [[I]], 1
; SSE-NEXT: [[AGAIN:%.*]] = icmp ult i32 [[I_NEXT]], 32
; SSE-NEXT: br i1 [[AGAIN]], label [[LOOP]], label [[DONE]], !llvm.loop [[LOOP2:![0-9]+]]
; SSE: done:
; SSE-NEXT: [[TOT_NEXT_LCSSA:%.*]] = phi double [ [[TOT_NEXT]], [[NEXT_ITER]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; SSE-NEXT: ret double [[TOT_NEXT_LCSSA]]
;
; AVX-LABEL: @sumIfVector(
; AVX-NEXT: entry:
; AVX-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; AVX: vector.ph:
; AVX-NEXT: br label [[VECTOR_BODY:%.*]]
; AVX: vector.body:
; AVX-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; AVX-NEXT: [[VEC_PHI:%.*]] = phi <4 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
; AVX-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0
; AVX-NEXT: [[TMP1:%.*]] = getelementptr double, ptr [[ARR:%.*]], i32 [[TMP0]]
; AVX-NEXT: [[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i32 0
; AVX-NEXT: [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 8
; AVX-NEXT: [[TMP3:%.*]] = fcmp fast une <4 x double> [[WIDE_LOAD]], <double 4.200000e+01, double 4.200000e+01, double 4.200000e+01, double 4.200000e+01>
; AVX-NEXT: [[TMP5:%.*]] = xor <4 x i1> [[TMP3]], <i1 true, i1 true, i1 true, i1 true>
; AVX-NEXT: [[TMP4:%.*]] = fadd fast <4 x double> [[VEC_PHI]], [[WIDE_LOAD]]
; AVX-NEXT: [[PREDPHI]] = select <4 x i1> [[TMP3]], <4 x double> [[TMP4]], <4 x double> [[VEC_PHI]]
; AVX-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; AVX-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
; AVX-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; AVX: middle.block:
; AVX-NEXT: [[TMP7:%.*]] = call fast double @llvm.vector.reduce.fadd.v4f64(double -0.000000e+00, <4 x double> [[PREDPHI]])
; AVX-NEXT: br i1 true, label [[DONE:%.*]], label [[SCALAR_PH]]
; AVX: scalar.ph:
; AVX-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; AVX-NEXT: [[BC_MERGE_RDX:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; AVX-NEXT: br label [[LOOP:%.*]]
; AVX: loop:
; AVX-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
; AVX-NEXT: [[TOT:%.*]] = phi double [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TOT_NEXT:%.*]], [[NEXT_ITER]] ]
; AVX-NEXT: [[ADDR:%.*]] = getelementptr double, ptr [[ARR]], i32 [[I]]
; AVX-NEXT: [[NEXTVAL:%.*]] = load double, ptr [[ADDR]], align 8
; AVX-NEXT: [[TST:%.*]] = fcmp fast une double [[NEXTVAL]], 4.200000e+01
; AVX-NEXT: br i1 [[TST]], label [[DO_ADD:%.*]], label [[NO_ADD:%.*]]
; AVX: do.add:
; AVX-NEXT: [[TOT_NEW:%.*]] = fadd fast double [[TOT]], [[NEXTVAL]]
; AVX-NEXT: br label [[NEXT_ITER]]
; AVX: no.add:
; AVX-NEXT: br label [[NEXT_ITER]]
; AVX: next.iter:
; AVX-NEXT: [[TOT_NEXT]] = phi double [ [[TOT]], [[NO_ADD]] ], [ [[TOT_NEW]], [[DO_ADD]] ]
; AVX-NEXT: [[I_NEXT]] = add i32 [[I]], 1
; AVX-NEXT: [[AGAIN:%.*]] = icmp ult i32 [[I_NEXT]], 32
; AVX-NEXT: br i1 [[AGAIN]], label [[LOOP]], label [[DONE]], !llvm.loop [[LOOP2:![0-9]+]]
; AVX: done:
; AVX-NEXT: [[TOT_NEXT_LCSSA:%.*]] = phi double [ [[TOT_NEXT]], [[NEXT_ITER]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; AVX-NEXT: ret double [[TOT_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%i = phi i32 [0, %entry], [%i.next, %next.iter]
%tot = phi double [0.0, %entry], [%tot.next, %next.iter]
%addr = getelementptr double, ptr %arr, i32 %i
%nextval = load double, ptr %addr
%tst = fcmp fast une double %nextval, 42.0
br i1 %tst, label %do.add, label %no.add
do.add:
%tot.new = fadd fast double %tot, %nextval
br label %next.iter
no.add:
br label %next.iter
next.iter:
%tot.next = phi double [%tot, %no.add], [%tot.new, %do.add]
%i.next = add i32 %i, 1
%again = icmp ult i32 %i.next, 32
br i1 %again, label %loop, label %done
done:
ret double %tot.next
}
Reference in New Issue View Git Blame Copy Permalink