aboutsummaryrefslogtreecommitdiff
path: root/llvm/test/CodeGen/PowerPC
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/PowerPC')
-rw-r--r--llvm/test/CodeGen/PowerPC/bittest.ll193
-rw-r--r--llvm/test/CodeGen/PowerPC/p10-spill-crun.ll2
-rw-r--r--llvm/test/CodeGen/PowerPC/vec_insert_elt.ll58
-rw-r--r--llvm/test/CodeGen/PowerPC/vector-reduce-add.ll22
4 files changed, 228 insertions, 47 deletions
diff --git a/llvm/test/CodeGen/PowerPC/bittest.ll b/llvm/test/CodeGen/PowerPC/bittest.ll
new file mode 100644
index 0000000..cba56e3
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/bittest.ll
@@ -0,0 +1,193 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -verify-machineinstrs < %s -O3 -mcpu=ppc -mtriple powerpc-ibm-aix \
+; RUN: -ppc-asm-full-reg-names | FileCheck %s
+
+define i32 @foo(i32 noundef signext %x) {
+; CHECK-LABEL: foo:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: mflr r0
+; CHECK-NEXT: stwu r1, -64(r1)
+; CHECK-NEXT: stw r0, 72(r1)
+; CHECK-NEXT: cmpwi r3, 8
+; CHECK-NEXT: stw r31, 60(r1) # 4-byte Folded Spill
+; CHECK-NEXT: mr r31, r3
+; CHECK-NEXT: li r3, 0
+; CHECK-NEXT: ble cr0, L..BB0_4
+; CHECK-NEXT: # %bb.1: # %entry
+; CHECK-NEXT: cmpwi r31, 11
+; CHECK-NEXT: bge cr0, L..BB0_7
+; CHECK-NEXT: # %bb.2: # %entry
+; CHECK-NEXT: cmplwi r31, 9
+; CHECK-NEXT: beq cr0, L..BB0_9
+; CHECK-NEXT: # %bb.3: # %entry
+; CHECK-NEXT: cmplwi r31, 10
+; CHECK-NEXT: beq cr0, L..BB0_11
+; CHECK-NEXT: b L..BB0_13
+; CHECK-NEXT: L..BB0_4: # %entry
+; CHECK-NEXT: cmplwi r31, 4
+; CHECK-NEXT: beq cr0, L..BB0_12
+; CHECK-NEXT: # %bb.5: # %entry
+; CHECK-NEXT: cmplwi r31, 7
+; CHECK-NEXT: beq cr0, L..BB0_11
+; CHECK-NEXT: # %bb.6: # %entry
+; CHECK-NEXT: cmplwi r31, 8
+; CHECK-NEXT: beq cr0, L..BB0_10
+; CHECK-NEXT: b L..BB0_13
+; CHECK-NEXT: L..BB0_7: # %entry
+; CHECK-NEXT: beq cr0, L..BB0_10
+; CHECK-NEXT: # %bb.8: # %entry
+; CHECK-NEXT: cmplwi r31, 12
+; CHECK-NEXT: bne cr0, L..BB0_13
+; CHECK-NEXT: L..BB0_9: # %sw.bb2
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: bl .foo3[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: b L..BB0_13
+; CHECK-NEXT: L..BB0_10: # %sw.bb1
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: bl .foo2[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: b L..BB0_13
+; CHECK-NEXT: L..BB0_11: # %sw.bb
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: bl .foo1[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: b L..BB0_13
+; CHECK-NEXT: L..BB0_12: # %sw.bb3
+; CHECK-NEXT: li r3, 4
+; CHECK-NEXT: bl .foo4[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: li r3, 4
+; CHECK-NEXT: L..BB0_13: # %return
+; CHECK-NEXT: lwz r31, 60(r1) # 4-byte Folded Reload
+; CHECK-NEXT: addi r1, r1, 64
+; CHECK-NEXT: lwz r0, 8(r1)
+; CHECK-NEXT: mtlr r0
+; CHECK-NEXT: blr
+entry:
+ switch i32 %x, label %return [
+ i32 7, label %sw.bb
+ i32 10, label %sw.bb
+ i32 8, label %sw.bb1
+ i32 11, label %sw.bb1
+ i32 9, label %sw.bb2
+ i32 12, label %sw.bb2
+ i32 4, label %sw.bb3
+ ]
+
+sw.bb: ; preds = %entry, %entry
+ tail call void @foo1(i32 noundef signext %x)
+ br label %return
+
+sw.bb1: ; preds = %entry, %entry
+ tail call void @foo2(i32 noundef signext %x)
+ br label %return
+
+sw.bb2: ; preds = %entry, %entry
+ tail call void @foo3(i32 noundef signext %x)
+ br label %return
+
+sw.bb3: ; preds = %entry
+ tail call void @foo4(i32 noundef signext 4)
+ br label %return
+
+return: ; preds = %sw.bb, %sw.bb1, %sw.bb2, %sw.bb3, %entry
+ %retval.0 = phi i32 [ 0, %entry ], [ 4, %sw.bb3 ], [ %x, %sw.bb2 ], [ %x, %sw.bb1 ], [ %x, %sw.bb ]
+ ret i32 %retval.0
+}
+
+define i32 @goo(i32 noundef signext %x) {
+; CHECK-LABEL: goo:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: mflr r0
+; CHECK-NEXT: stwu r1, -64(r1)
+; CHECK-NEXT: stw r0, 72(r1)
+; CHECK-NEXT: cmplwi r3, 12
+; CHECK-NEXT: stw r31, 60(r1) # 4-byte Folded Spill
+; CHECK-NEXT: mr r31, r3
+; CHECK-NEXT: bgt cr0, L..BB1_7
+; CHECK-NEXT: # %bb.1: # %entry
+; CHECK-NEXT: li r3, 1
+; CHECK-NEXT: slw r3, r3, r31
+; CHECK-NEXT: andi. r4, r3, 5632
+; CHECK-NEXT: bne cr0, L..BB1_4
+; CHECK-NEXT: # %bb.2: # %entry
+; CHECK-NEXT: andi. r3, r3, 2304
+; CHECK-NEXT: beq cr0, L..BB1_5
+; CHECK-NEXT: # %bb.3: # %sw.bb1
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: bl .foo2[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: b L..BB1_9
+; CHECK-NEXT: L..BB1_4: # %sw.bb2
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: bl .foo3[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: b L..BB1_9
+; CHECK-NEXT: L..BB1_5: # %entry
+; CHECK-NEXT: cmplwi r31, 7
+; CHECK-NEXT: bne cr0, L..BB1_7
+; CHECK-NEXT: # %bb.6: # %sw.bb
+; CHECK-NEXT: li r3, 7
+; CHECK-NEXT: li r31, 7
+; CHECK-NEXT: bl .foo1[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: b L..BB1_9
+; CHECK-NEXT: L..BB1_7: # %entry
+; CHECK-NEXT: cmplwi r31, 4
+; CHECK-NEXT: li r31, 0
+; CHECK-NEXT: bne cr0, L..BB1_9
+; CHECK-NEXT: # %bb.8: # %sw.bb3
+; CHECK-NEXT: li r3, 4
+; CHECK-NEXT: li r31, 4
+; CHECK-NEXT: bl .foo4[PR]
+; CHECK-NEXT: nop
+; CHECK-NEXT: L..BB1_9: # %return
+; CHECK-NEXT: mr r3, r31
+; CHECK-NEXT: lwz r31, 60(r1) # 4-byte Folded Reload
+; CHECK-NEXT: addi r1, r1, 64
+; CHECK-NEXT: lwz r0, 8(r1)
+; CHECK-NEXT: mtlr r0
+; CHECK-NEXT: blr
+entry:
+ switch i32 %x, label %return [
+ i32 7, label %sw.bb
+ i32 8, label %sw.bb1
+ i32 11, label %sw.bb1
+ i32 9, label %sw.bb2
+ i32 10, label %sw.bb2
+ i32 12, label %sw.bb2
+ i32 4, label %sw.bb3
+ ]
+
+sw.bb: ; preds = %entry
+ tail call void @foo1(i32 noundef signext 7)
+ br label %return
+
+sw.bb1: ; preds = %entry, %entry
+ tail call void @foo2(i32 noundef signext %x)
+ br label %return
+
+sw.bb2: ; preds = %entry, %entry, %entry
+ tail call void @foo3(i32 noundef signext %x)
+ br label %return
+
+sw.bb3: ; preds = %entry
+ tail call void @foo4(i32 noundef signext 4)
+ br label %return
+
+return: ; preds = %sw.bb, %sw.bb1, %sw.bb2, %sw.bb3, %entry
+ %retval.0 = phi i32 [ 0, %entry ], [ 4, %sw.bb3 ], [ %x, %sw.bb2 ], [ %x, %sw.bb1 ], [ 7, %sw.bb ]
+ ret i32 %retval.0
+}
+
+declare void @foo1(i32 noundef signext)
+
+declare void @foo2(i32 noundef signext)
+
+declare void @foo3(i32 noundef signext)
+
+declare void @foo4(i32 noundef signext)
diff --git a/llvm/test/CodeGen/PowerPC/p10-spill-crun.ll b/llvm/test/CodeGen/PowerPC/p10-spill-crun.ll
index 4ca2dc5db..eba9faa 100644
--- a/llvm/test/CodeGen/PowerPC/p10-spill-crun.ll
+++ b/llvm/test/CodeGen/PowerPC/p10-spill-crun.ll
@@ -234,8 +234,8 @@ define dso_local void @P10_Spill_CR_UN(ptr %arg, ptr %arg1, i32 %arg2) local_unn
; CHECK-BE-NEXT: # %bb.4: # %bb37
; CHECK-BE-NEXT: bc 4, 4*cr5+lt, .LBB0_14
; CHECK-BE-NEXT: .LBB0_5: # %bb42
-; CHECK-BE-NEXT: addi r3, r3, global_1@toc@l
; CHECK-BE-NEXT: li r4, 0
+; CHECK-BE-NEXT: addi r3, r3, global_1@toc@l
; CHECK-BE-NEXT: cmpwi r28, 0
; CHECK-BE-NEXT: isel r3, r3, r4, 4*cr2+gt
; CHECK-BE-NEXT: crnot 4*cr2+lt, eq
diff --git a/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll b/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll
index 291a9c1..b006c78 100644
--- a/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll
+++ b/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll
@@ -242,17 +242,14 @@ define <2 x i64> @testDoubleword(<2 x i64> %a, i64 %b, i64 %idx) {
; AIX-P8-32-LABEL: testDoubleword:
; AIX-P8-32: # %bb.0: # %entry
; AIX-P8-32-NEXT: add r6, r6, r6
-; AIX-P8-32-NEXT: addi r5, r1, -32
+; AIX-P8-32-NEXT: addi r5, r1, -16
; AIX-P8-32-NEXT: rlwinm r7, r6, 2, 28, 29
-; AIX-P8-32-NEXT: stxvw4x v2, 0, r5
+; AIX-P8-32-NEXT: stxvd2x v2, 0, r5
; AIX-P8-32-NEXT: stwx r3, r5, r7
-; AIX-P8-32-NEXT: addi r3, r1, -16
-; AIX-P8-32-NEXT: lxvw4x vs0, 0, r5
-; AIX-P8-32-NEXT: addi r5, r6, 1
-; AIX-P8-32-NEXT: rlwinm r5, r5, 2, 28, 29
-; AIX-P8-32-NEXT: stxvw4x vs0, 0, r3
-; AIX-P8-32-NEXT: stwx r4, r3, r5
-; AIX-P8-32-NEXT: lxvw4x v2, 0, r3
+; AIX-P8-32-NEXT: addi r3, r6, 1
+; AIX-P8-32-NEXT: rlwinm r3, r3, 2, 28, 29
+; AIX-P8-32-NEXT: stwx r4, r5, r3
+; AIX-P8-32-NEXT: lxvd2x v2, 0, r5
; AIX-P8-32-NEXT: blr
entry:
%vecins = insertelement <2 x i64> %a, i64 %b, i64 %idx
@@ -426,17 +423,14 @@ define <4 x float> @testFloat2(<4 x float> %a, ptr %b, i32 zeroext %idx1, i32 ze
; AIX-P8-LABEL: testFloat2:
; AIX-P8: # %bb.0: # %entry
; AIX-P8-NEXT: lwz r6, 0(r3)
-; AIX-P8-NEXT: rlwinm r4, r4, 2, 28, 29
-; AIX-P8-NEXT: addi r7, r1, -32
+; AIX-P8-NEXT: lwz r3, 1(r3)
+; AIX-P8-NEXT: addi r7, r1, -16
; AIX-P8-NEXT: stxvw4x v2, 0, r7
; AIX-P8-NEXT: rlwinm r5, r5, 2, 28, 29
+; AIX-P8-NEXT: rlwinm r4, r4, 2, 28, 29
; AIX-P8-NEXT: stwx r6, r7, r4
-; AIX-P8-NEXT: addi r4, r1, -16
-; AIX-P8-NEXT: lxvw4x vs0, 0, r7
-; AIX-P8-NEXT: lwz r3, 1(r3)
-; AIX-P8-NEXT: stxvw4x vs0, 0, r4
-; AIX-P8-NEXT: stwx r3, r4, r5
-; AIX-P8-NEXT: lxvw4x v2, 0, r4
+; AIX-P8-NEXT: stwx r3, r7, r5
+; AIX-P8-NEXT: lxvw4x v2, 0, r7
; AIX-P8-NEXT: blr
entry:
%add.ptr1 = getelementptr inbounds i8, ptr %b, i64 1
@@ -493,38 +487,32 @@ define <4 x float> @testFloat3(<4 x float> %a, ptr %b, i32 zeroext %idx1, i32 ze
;
; AIX-P8-64-LABEL: testFloat3:
; AIX-P8-64: # %bb.0: # %entry
+; AIX-P8-64-NEXT: li r7, 1
; AIX-P8-64-NEXT: lis r6, 1
-; AIX-P8-64-NEXT: rlwinm r4, r4, 2, 28, 29
-; AIX-P8-64-NEXT: addi r7, r1, -32
; AIX-P8-64-NEXT: rlwinm r5, r5, 2, 28, 29
+; AIX-P8-64-NEXT: rlwinm r4, r4, 2, 28, 29
+; AIX-P8-64-NEXT: rldic r7, r7, 36, 27
; AIX-P8-64-NEXT: lwzx r6, r3, r6
+; AIX-P8-64-NEXT: lwzx r3, r3, r7
+; AIX-P8-64-NEXT: addi r7, r1, -16
; AIX-P8-64-NEXT: stxvw4x v2, 0, r7
; AIX-P8-64-NEXT: stwx r6, r7, r4
-; AIX-P8-64-NEXT: li r4, 1
-; AIX-P8-64-NEXT: lxvw4x vs0, 0, r7
-; AIX-P8-64-NEXT: rldic r4, r4, 36, 27
-; AIX-P8-64-NEXT: lwzx r3, r3, r4
-; AIX-P8-64-NEXT: addi r4, r1, -16
-; AIX-P8-64-NEXT: stxvw4x vs0, 0, r4
-; AIX-P8-64-NEXT: stwx r3, r4, r5
-; AIX-P8-64-NEXT: lxvw4x v2, 0, r4
+; AIX-P8-64-NEXT: stwx r3, r7, r5
+; AIX-P8-64-NEXT: lxvw4x v2, 0, r7
; AIX-P8-64-NEXT: blr
;
; AIX-P8-32-LABEL: testFloat3:
; AIX-P8-32: # %bb.0: # %entry
; AIX-P8-32-NEXT: lis r6, 1
-; AIX-P8-32-NEXT: rlwinm r4, r4, 2, 28, 29
-; AIX-P8-32-NEXT: addi r7, r1, -32
; AIX-P8-32-NEXT: rlwinm r5, r5, 2, 28, 29
+; AIX-P8-32-NEXT: rlwinm r4, r4, 2, 28, 29
+; AIX-P8-32-NEXT: addi r7, r1, -16
; AIX-P8-32-NEXT: lwzx r6, r3, r6
+; AIX-P8-32-NEXT: lwz r3, 0(r3)
; AIX-P8-32-NEXT: stxvw4x v2, 0, r7
; AIX-P8-32-NEXT: stwx r6, r7, r4
-; AIX-P8-32-NEXT: addi r4, r1, -16
-; AIX-P8-32-NEXT: lxvw4x vs0, 0, r7
-; AIX-P8-32-NEXT: lwz r3, 0(r3)
-; AIX-P8-32-NEXT: stxvw4x vs0, 0, r4
-; AIX-P8-32-NEXT: stwx r3, r4, r5
-; AIX-P8-32-NEXT: lxvw4x v2, 0, r4
+; AIX-P8-32-NEXT: stwx r3, r7, r5
+; AIX-P8-32-NEXT: lxvw4x v2, 0, r7
; AIX-P8-32-NEXT: blr
entry:
%add.ptr = getelementptr inbounds i8, ptr %b, i64 65536
diff --git a/llvm/test/CodeGen/PowerPC/vector-reduce-add.ll b/llvm/test/CodeGen/PowerPC/vector-reduce-add.ll
index d506d20..e5d305f 100644
--- a/llvm/test/CodeGen/PowerPC/vector-reduce-add.ll
+++ b/llvm/test/CodeGen/PowerPC/vector-reduce-add.ll
@@ -1085,14 +1085,14 @@ define dso_local signext i32 @v16i8tov16i32_sign(<16 x i8> %a) local_unnamed_add
; PWR10BE-NEXT: addis r3, r2, .LCPI17_2@toc@ha
; PWR10BE-NEXT: vperm v3, v2, v2, v3
; PWR10BE-NEXT: addi r3, r3, .LCPI17_2@toc@l
-; PWR10BE-NEXT: vextsb2w v3, v3
; PWR10BE-NEXT: lxv v5, 0(r3)
; PWR10BE-NEXT: addis r3, r2, .LCPI17_3@toc@ha
+; PWR10BE-NEXT: vextsb2w v3, v3
; PWR10BE-NEXT: vperm v4, v2, v2, v4
; PWR10BE-NEXT: addi r3, r3, .LCPI17_3@toc@l
-; PWR10BE-NEXT: vextsb2w v4, v4
; PWR10BE-NEXT: lxv v0, 0(r3)
; PWR10BE-NEXT: li r3, 0
+; PWR10BE-NEXT: vextsb2w v4, v4
; PWR10BE-NEXT: vperm v5, v2, v2, v5
; PWR10BE-NEXT: vadduwm v3, v4, v3
; PWR10BE-NEXT: vextsb2w v5, v5
@@ -1212,9 +1212,9 @@ define dso_local zeroext i32 @v16i8tov16i32_zero(<16 x i8> %a) local_unnamed_add
; PWR10BE-NEXT: addis r3, r2, .LCPI18_3@toc@ha
; PWR10BE-NEXT: vperm v5, v4, v2, v5
; PWR10BE-NEXT: addi r3, r3, .LCPI18_3@toc@l
-; PWR10BE-NEXT: vadduwm v3, v5, v3
; PWR10BE-NEXT: lxv v1, 0(r3)
; PWR10BE-NEXT: li r3, 0
+; PWR10BE-NEXT: vadduwm v3, v5, v3
; PWR10BE-NEXT: vperm v0, v4, v2, v0
; PWR10BE-NEXT: vperm v2, v4, v2, v1
; PWR10BE-NEXT: vadduwm v2, v2, v0
@@ -1568,41 +1568,41 @@ define dso_local i64 @v16i8tov16i64_sign(<16 x i8> %a) local_unnamed_addr #0 {
; PWR10BE-NEXT: addis r3, r2, .LCPI23_0@toc@ha
; PWR10BE-NEXT: xxspltib v1, 255
; PWR10BE-NEXT: addi r3, r3, .LCPI23_0@toc@l
-; PWR10BE-NEXT: vsrq v1, v1, v1
; PWR10BE-NEXT: lxv v3, 0(r3)
; PWR10BE-NEXT: addis r3, r2, .LCPI23_1@toc@ha
+; PWR10BE-NEXT: vsrq v1, v1, v1
; PWR10BE-NEXT: addi r3, r3, .LCPI23_1@toc@l
; PWR10BE-NEXT: vperm v1, v2, v2, v1
; PWR10BE-NEXT: lxv v4, 0(r3)
; PWR10BE-NEXT: addis r3, r2, .LCPI23_2@toc@ha
-; PWR10BE-NEXT: vextsb2d v1, v1
; PWR10BE-NEXT: vperm v3, v2, v2, v3
+; PWR10BE-NEXT: vextsb2d v1, v1
; PWR10BE-NEXT: addi r3, r3, .LCPI23_2@toc@l
-; PWR10BE-NEXT: vextsb2d v3, v3
; PWR10BE-NEXT: lxv v5, 0(r3)
; PWR10BE-NEXT: addis r3, r2, .LCPI23_3@toc@ha
+; PWR10BE-NEXT: vextsb2d v3, v3
; PWR10BE-NEXT: vperm v4, v2, v2, v4
; PWR10BE-NEXT: addi r3, r3, .LCPI23_3@toc@l
-; PWR10BE-NEXT: vextsb2d v4, v4
; PWR10BE-NEXT: lxv v0, 0(r3)
; PWR10BE-NEXT: addis r3, r2, .LCPI23_4@toc@ha
+; PWR10BE-NEXT: vextsb2d v4, v4
; PWR10BE-NEXT: vperm v5, v2, v2, v5
; PWR10BE-NEXT: addi r3, r3, .LCPI23_4@toc@l
-; PWR10BE-NEXT: vextsb2d v5, v5
; PWR10BE-NEXT: lxv v6, 0(r3)
; PWR10BE-NEXT: addis r3, r2, .LCPI23_5@toc@ha
+; PWR10BE-NEXT: vextsb2d v5, v5
; PWR10BE-NEXT: vperm v0, v2, v2, v0
; PWR10BE-NEXT: addi r3, r3, .LCPI23_5@toc@l
-; PWR10BE-NEXT: vextsb2d v0, v0
; PWR10BE-NEXT: lxv v7, 0(r3)
; PWR10BE-NEXT: addis r3, r2, .LCPI23_6@toc@ha
+; PWR10BE-NEXT: vextsb2d v0, v0
; PWR10BE-NEXT: vperm v6, v2, v2, v6
+; PWR10BE-NEXT: addi r3, r3, .LCPI23_6@toc@l
; PWR10BE-NEXT: vaddudm v5, v0, v5
; PWR10BE-NEXT: vaddudm v3, v4, v3
; PWR10BE-NEXT: vaddudm v3, v3, v5
-; PWR10BE-NEXT: addi r3, r3, .LCPI23_6@toc@l
-; PWR10BE-NEXT: vextsb2d v6, v6
; PWR10BE-NEXT: lxv v8, 0(r3)
+; PWR10BE-NEXT: vextsb2d v6, v6
; PWR10BE-NEXT: vperm v7, v2, v2, v7
; PWR10BE-NEXT: vextsb2d v7, v7
; PWR10BE-NEXT: vperm v2, v2, v2, v8
generated by cgit v1.1 at 2025-10-30 00:24:55 +0000