[llvm] ea27dd7 - [VectorCombine] Add tests which require DT to use info from assumes.

Florian Hahn via llvm-commits llvm-commits at lists.llvm.org
Tue Sep 21 05:07:29 PDT 2021


Author: Florian Hahn
Date: 2021-09-21T13:07:06+01:00
New Revision: ea27dd74972e95e513fefcf96067522364f4e3d7

URL: https://github.com/llvm/llvm-project/commit/ea27dd74972e95e513fefcf96067522364f4e3d7
DIFF: https://github.com/llvm/llvm-project/commit/ea27dd74972e95e513fefcf96067522364f4e3d7.diff

LOG: [VectorCombine] Add tests which require DT to use info from assumes.

Added: 
    

Modified: 
    llvm/test/Transforms/VectorCombine/AArch64/load-extract-insert-store-scalarization.ll
    llvm/test/Transforms/VectorCombine/AArch64/load-extractelement-scalarization.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/Transforms/VectorCombine/AArch64/load-extract-insert-store-scalarization.ll b/llvm/test/Transforms/VectorCombine/AArch64/load-extract-insert-store-scalarization.ll
index ca44e2cad20ca..bb70c87a53c52 100644
--- a/llvm/test/Transforms/VectorCombine/AArch64/load-extract-insert-store-scalarization.ll
+++ b/llvm/test/Transforms/VectorCombine/AArch64/load-extract-insert-store-scalarization.ll
@@ -58,6 +58,98 @@ entry:
   ret void
 }
 
+declare i1 @cond()
+
+define void @load_extract_insert_store_var_idx_assume_valid_in_dominating_block(i64 %idx.1, i64 %idx.2, <225 x double>* %A, i1 %c.1) {
+; CHECK-LABEL: @load_extract_insert_store_var_idx_assume_valid_in_dominating_block(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP_1:%.*]] = icmp ult i64 [[IDX_1:%.*]], 225
+; CHECK-NEXT:    call void @llvm.assume(i1 [[CMP_1]])
+; CHECK-NEXT:    [[CMP_2:%.*]] = icmp ult i64 [[IDX_2:%.*]], 225
+; CHECK-NEXT:    call void @llvm.assume(i1 [[CMP_2]])
+; CHECK-NEXT:    br i1 [[C_1:%.*]], label [[LOOP:%.*]], label [[EXIT:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LV:%.*]] = load <225 x double>, <225 x double>* [[A:%.*]], align 8
+; CHECK-NEXT:    [[EXT_0:%.*]] = extractelement <225 x double> [[LV]], i64 [[IDX_1]]
+; CHECK-NEXT:    [[MUL:%.*]] = fmul double 2.000000e+01, [[EXT_0]]
+; CHECK-NEXT:    [[EXT_1:%.*]] = extractelement <225 x double> [[LV]], i64 [[IDX_2]]
+; CHECK-NEXT:    [[SUB:%.*]] = fsub double [[EXT_1]], [[MUL]]
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <225 x double> [[LV]], double [[SUB]], i64 [[IDX_1]]
+; CHECK-NEXT:    store <225 x double> [[INS]], <225 x double>* [[A]], align 8
+; CHECK-NEXT:    [[C_2:%.*]] = call i1 @cond()
+; CHECK-NEXT:    br i1 [[C_2]], label [[LOOP]], label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  %cmp.1 = icmp ult i64 %idx.1, 225
+  call void @llvm.assume(i1 %cmp.1)
+  %cmp.2 = icmp ult i64 %idx.2, 225
+  call void @llvm.assume(i1 %cmp.2)
+  br i1 %c.1, label %loop, label %exit
+
+loop:
+  %lv = load <225 x double>, <225 x double>* %A, align 8
+  %ext.0 = extractelement <225 x double> %lv, i64 %idx.1
+  %mul = fmul double 20.0, %ext.0
+  %ext.1 = extractelement <225 x double> %lv, i64 %idx.2
+  %sub = fsub double %ext.1, %mul
+  %ins = insertelement <225 x double> %lv, double %sub, i64 %idx.1
+  store <225 x double> %ins, <225 x double>* %A, align 8
+  %c.2 = call i1 @cond()
+  br i1 %c.2, label %loop, label %exit
+
+exit:
+  ret void
+}
+
+define void @load_extract_insert_store_var_idx_assume_valid_in_non_dominating_block(i64 %idx.1, i64 %idx.2, <225 x double>* %A, i1 %c.1, i1 %c.2) {
+; CHECK-LABEL: @load_extract_insert_store_var_idx_assume_valid_in_non_dominating_block(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[C_1:%.*]], label [[ASSUME_BLOCK:%.*]], label [[LOOP:%.*]]
+; CHECK:       assume_block:
+; CHECK-NEXT:    [[CMP_1:%.*]] = icmp ult i64 [[IDX_1:%.*]], 225
+; CHECK-NEXT:    call void @llvm.assume(i1 [[CMP_1]])
+; CHECK-NEXT:    [[CMP_2:%.*]] = icmp ult i64 [[IDX_2:%.*]], 225
+; CHECK-NEXT:    call void @llvm.assume(i1 [[CMP_2]])
+; CHECK-NEXT:    br i1 [[C_2:%.*]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LV:%.*]] = load <225 x double>, <225 x double>* [[A:%.*]], align 8
+; CHECK-NEXT:    [[EXT_0:%.*]] = extractelement <225 x double> [[LV]], i64 [[IDX_1]]
+; CHECK-NEXT:    [[MUL:%.*]] = fmul double 2.000000e+01, [[EXT_0]]
+; CHECK-NEXT:    [[EXT_1:%.*]] = extractelement <225 x double> [[LV]], i64 [[IDX_2]]
+; CHECK-NEXT:    [[SUB:%.*]] = fsub double [[EXT_1]], [[MUL]]
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <225 x double> [[LV]], double [[SUB]], i64 [[IDX_1]]
+; CHECK-NEXT:    store <225 x double> [[INS]], <225 x double>* [[A]], align 8
+; CHECK-NEXT:    [[C_3:%.*]] = call i1 @cond()
+; CHECK-NEXT:    br i1 [[C_3]], label [[LOOP]], label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br i1 %c.1, label %assume_block, label %loop
+
+assume_block:
+  %cmp.1 = icmp ult i64 %idx.1, 225
+  call void @llvm.assume(i1 %cmp.1)
+  %cmp.2 = icmp ult i64 %idx.2, 225
+  call void @llvm.assume(i1 %cmp.2)
+  br i1 %c.2, label %loop, label %exit
+
+loop:
+  %lv = load <225 x double>, <225 x double>* %A, align 8
+  %ext.0 = extractelement <225 x double> %lv, i64 %idx.1
+  %mul = fmul double 20.0, %ext.0
+  %ext.1 = extractelement <225 x double> %lv, i64 %idx.2
+  %sub = fsub double %ext.1, %mul
+  %ins = insertelement <225 x double> %lv, double %sub, i64 %idx.1
+  store <225 x double> %ins, <225 x double>* %A, align 8
+  %c.3 = call i1 @cond()
+  br i1 %c.3, label %loop, label %exit
+
+exit:
+  ret void
+}
 
 define void @load_extract_insert_store_var_idx_no_assume_valid(i64 %idx.1, i64 %idx.2, <225 x double>* %A) {
 ; CHECK-LABEL: @load_extract_insert_store_var_idx_no_assume_valid(

diff  --git a/llvm/test/Transforms/VectorCombine/AArch64/load-extractelement-scalarization.ll b/llvm/test/Transforms/VectorCombine/AArch64/load-extractelement-scalarization.ll
index ecd1762bb23b8..7818e9be7bd4f 100644
--- a/llvm/test/Transforms/VectorCombine/AArch64/load-extractelement-scalarization.ll
+++ b/llvm/test/Transforms/VectorCombine/AArch64/load-extractelement-scalarization.ll
@@ -105,6 +105,79 @@ entry:
   ret i32 %r
 }
 
+declare i1 @cond()
+
+define i32 @load_extract_idx_var_i64_known_valid_by_assume_in_dominating_block(<4 x i32>* %x, i64 %idx, i1 %c.1) {
+; CHECK-LABEL: @load_extract_idx_var_i64_known_valid_by_assume_in_dominating_block(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i64 [[IDX:%.*]], 4
+; CHECK-NEXT:    call void @llvm.assume(i1 [[CMP]])
+; CHECK-NEXT:    br i1 [[C_1:%.*]], label [[LOOP:%.*]], label [[EXIT:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LV:%.*]] = load <4 x i32>, <4 x i32>* [[X:%.*]], align 16
+; CHECK-NEXT:    call void @maythrow()
+; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x i32> [[LV]], i64 [[IDX]]
+; CHECK-NEXT:    [[C_2:%.*]] = call i1 @cond()
+; CHECK-NEXT:    br i1 [[C_2]], label [[LOOP]], label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[P:%.*]] = phi i32 [ [[R]], [[LOOP]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    ret i32 [[P]]
+;
+entry:
+  %cmp = icmp ult i64 %idx, 4
+  call void @llvm.assume(i1 %cmp)
+  br i1 %c.1, label %loop, label %exit
+
+loop:
+  %lv = load <4 x i32>, <4 x i32>* %x
+  call void @maythrow()
+  %r = extractelement <4 x i32> %lv, i64 %idx
+  %c.2 = call i1 @cond()
+  br i1 %c.2, label %loop, label %exit
+
+exit:
+  %p = phi i32 [ %r, %loop ], [ 0, %entry ]
+  ret i32 %p
+}
+
+define i32 @load_extract_idx_var_i64_known_valid_by_assume_in_non_dominating_block(<4 x i32>* %x, i64 %idx, i1 %c.1, i1 %c.2) {
+; CHECK-LABEL: @load_extract_idx_var_i64_known_valid_by_assume_in_non_dominating_block(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[C_1:%.*]], label [[ASSUME_CHECK:%.*]], label [[LOOP:%.*]]
+; CHECK:       assume_check:
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i64 [[IDX:%.*]], 4
+; CHECK-NEXT:    call void @llvm.assume(i1 [[CMP]])
+; CHECK-NEXT:    br i1 [[C_2:%.*]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LV:%.*]] = load <4 x i32>, <4 x i32>* [[X:%.*]], align 16
+; CHECK-NEXT:    call void @maythrow()
+; CHECK-NEXT:    [[R:%.*]] = extractelement <4 x i32> [[LV]], i64 [[IDX]]
+; CHECK-NEXT:    [[C_3:%.*]] = call i1 @cond()
+; CHECK-NEXT:    br i1 [[C_3]], label [[LOOP]], label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[P:%.*]] = phi i32 [ [[R]], [[LOOP]] ], [ 0, [[ASSUME_CHECK]] ]
+; CHECK-NEXT:    ret i32 0
+;
+entry:
+  br i1 %c.1, label %assume_check, label %loop
+
+assume_check:
+  %cmp = icmp ult i64 %idx, 4
+  call void @llvm.assume(i1 %cmp)
+  br i1 %c.2, label %loop, label %exit
+
+loop:
+  %lv = load <4 x i32>, <4 x i32>* %x
+  call void @maythrow()
+  %r = extractelement <4 x i32> %lv, i64 %idx
+  %c.3 = call i1 @cond()
+  br i1 %c.3, label %loop, label %exit
+
+exit:
+  %p = phi i32 [ %r, %loop ], [ 0, %assume_check ]
+  ret i32 0
+}
+
 define i32 @load_extract_idx_var_i64_not_known_valid_by_assume_after_load(<4 x i32>* %x, i64 %idx) {
 ; CHECK-LABEL: @load_extract_idx_var_i64_not_known_valid_by_assume_after_load(
 ; CHECK-NEXT:  entry:
@@ -604,3 +677,4 @@ define i32 @load_multiple_extracts_with_variable_indices_large_vector_all_valid_
   %res = add i32 %e.0, %e.1
   ret i32 %res
 }
+


        


More information about the llvm-commits mailing list