[llvm] c1dc44f - [AArch64] add tests for store merge of truncs; NFC
Sanjay Patel via llvm-commits
llvm-commits at lists.llvm.org
Sat Aug 22 11:55:01 PDT 2020
Author: Sanjay Patel
Date: 2020-08-22T14:54:40-04:00
New Revision: c1dc44f914e250979cf95dd486a3b87d2d692bd4
URL: https://github.com/llvm/llvm-project/commit/c1dc44f914e250979cf95dd486a3b87d2d692bd4
DIFF: https://github.com/llvm/llvm-project/commit/c1dc44f914e250979cf95dd486a3b87d2d692bd4.diff
LOG: [AArch64] add tests for store merge of truncs; NFC
Added:
llvm/test/CodeGen/AArch64/merge-trunc-store.ll
Modified:
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/AArch64/merge-trunc-store.ll b/llvm/test/CodeGen/AArch64/merge-trunc-store.ll
new file mode 100644
index 000000000000..9d9ea3ec8951
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/merge-trunc-store.ll
@@ -0,0 +1,774 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s --check-prefixes=CHECK,LE
+; RUN: llc < %s -mtriple=aarch64_be-- | FileCheck %s --check-prefixes=CHECK,BE
+
+define void @le_i16_to_i8(i16 %x, i8* %p0) {
+; LE-LABEL: le_i16_to_i8:
+; LE: // %bb.0:
+; LE-NEXT: strh w0, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: le_i16_to_i8:
+; BE: // %bb.0:
+; BE-NEXT: rev w8, w0
+; BE-NEXT: lsr w8, w8, #16
+; BE-NEXT: strh w8, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i16 %x, 8
+ %t0 = trunc i16 %x to i8
+ %t1 = trunc i16 %sh1 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ store i8 %t0, i8* %p0, align 1
+ store i8 %t1, i8* %p1, align 1
+ ret void
+}
+
+define void @le_i16_to_i8_order(i16 %x, i8* %p0) {
+; LE-LABEL: le_i16_to_i8_order:
+; LE: // %bb.0:
+; LE-NEXT: strh w0, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: le_i16_to_i8_order:
+; BE: // %bb.0:
+; BE-NEXT: rev w8, w0
+; BE-NEXT: lsr w8, w8, #16
+; BE-NEXT: strh w8, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i16 %x, 8
+ %t0 = trunc i16 %x to i8
+ %t1 = trunc i16 %sh1 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ store i8 %t1, i8* %p1, align 1
+ store i8 %t0, i8* %p0, align 1
+ ret void
+}
+
+define void @be_i16_to_i8_offset(i16 %x, i8* %p0) {
+; LE-LABEL: be_i16_to_i8_offset:
+; LE: // %bb.0:
+; LE-NEXT: rev w8, w0
+; LE-NEXT: lsr w8, w8, #16
+; LE-NEXT: sturh w8, [x1, #11]
+; LE-NEXT: ret
+;
+; BE-LABEL: be_i16_to_i8_offset:
+; BE: // %bb.0:
+; BE-NEXT: sturh w0, [x1, #11]
+; BE-NEXT: ret
+ %sh1 = lshr i16 %x, 8
+ %t0 = trunc i16 %x to i8
+ %t1 = trunc i16 %sh1 to i8
+ %p11 = getelementptr inbounds i8, i8* %p0, i64 11
+ %p12 = getelementptr inbounds i8, i8* %p0, i64 12
+ store i8 %t0, i8* %p12, align 1
+ store i8 %t1, i8* %p11, align 1
+ ret void
+}
+
+define void @be_i16_to_i8_order(i16 %x, i8* %p0) {
+; LE-LABEL: be_i16_to_i8_order:
+; LE: // %bb.0:
+; LE-NEXT: rev w8, w0
+; LE-NEXT: lsr w8, w8, #16
+; LE-NEXT: strh w8, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: be_i16_to_i8_order:
+; BE: // %bb.0:
+; BE-NEXT: strh w0, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i16 %x, 8
+ %t0 = trunc i16 %x to i8
+ %t1 = trunc i16 %sh1 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ store i8 %t1, i8* %p0, align 1
+ store i8 %t0, i8* %p1, align 1
+ ret void
+}
+
+define void @le_i32_to_i8(i32 %x, i8* %p0) {
+; LE-LABEL: le_i32_to_i8:
+; LE: // %bb.0:
+; LE-NEXT: str w0, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: le_i32_to_i8:
+; BE: // %bb.0:
+; BE-NEXT: rev w8, w0
+; BE-NEXT: str w8, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i32 %x, 8
+ %sh2 = lshr i32 %x, 16
+ %sh3 = lshr i32 %x, 24
+ %t0 = trunc i32 %x to i8
+ %t1 = trunc i32 %sh1 to i8
+ %t2 = trunc i32 %sh2 to i8
+ %t3 = trunc i32 %sh3 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ store i8 %t0, i8* %p0, align 1
+ store i8 %t1, i8* %p1, align 1
+ store i8 %t2, i8* %p2, align 1
+ store i8 %t3, i8* %p3, align 1
+ ret void
+}
+
+define void @le_i32_to_i8_order(i32 %x, i8* %p0) {
+; LE-LABEL: le_i32_to_i8_order:
+; LE: // %bb.0:
+; LE-NEXT: str w0, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: le_i32_to_i8_order:
+; BE: // %bb.0:
+; BE-NEXT: rev w8, w0
+; BE-NEXT: str w8, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i32 %x, 8
+ %sh2 = lshr i32 %x, 16
+ %sh3 = lshr i32 %x, 24
+ %t0 = trunc i32 %x to i8
+ %t1 = trunc i32 %sh1 to i8
+ %t2 = trunc i32 %sh2 to i8
+ %t3 = trunc i32 %sh3 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ store i8 %t3, i8* %p3, align 1
+ store i8 %t1, i8* %p1, align 1
+ store i8 %t0, i8* %p0, align 1
+ store i8 %t2, i8* %p2, align 1
+ ret void
+}
+
+define void @be_i32_to_i8(i32 %x, i8* %p0) {
+; LE-LABEL: be_i32_to_i8:
+; LE: // %bb.0:
+; LE-NEXT: rev w8, w0
+; LE-NEXT: str w8, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: be_i32_to_i8:
+; BE: // %bb.0:
+; BE-NEXT: str w0, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i32 %x, 8
+ %sh2 = lshr i32 %x, 16
+ %sh3 = lshr i32 %x, 24
+ %t0 = trunc i32 %x to i8
+ %t1 = trunc i32 %sh1 to i8
+ %t2 = trunc i32 %sh2 to i8
+ %t3 = trunc i32 %sh3 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ store i8 %t0, i8* %p3, align 1
+ store i8 %t1, i8* %p2, align 1
+ store i8 %t2, i8* %p1, align 1
+ store i8 %t3, i8* %p0, align 1
+ ret void
+}
+
+define void @be_i32_to_i8_order(i32 %x, i8* %p0) {
+; LE-LABEL: be_i32_to_i8_order:
+; LE: // %bb.0:
+; LE-NEXT: rev w8, w0
+; LE-NEXT: str w8, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: be_i32_to_i8_order:
+; BE: // %bb.0:
+; BE-NEXT: str w0, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i32 %x, 8
+ %sh2 = lshr i32 %x, 16
+ %sh3 = lshr i32 %x, 24
+ %t0 = trunc i32 %x to i8
+ %t1 = trunc i32 %sh1 to i8
+ %t2 = trunc i32 %sh2 to i8
+ %t3 = trunc i32 %sh3 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ store i8 %t3, i8* %p0, align 1
+ store i8 %t2, i8* %p1, align 1
+ store i8 %t0, i8* %p3, align 1
+ store i8 %t1, i8* %p2, align 1
+ ret void
+}
+
+define void @le_i32_to_i16(i32 %x, i16* %p0) {
+; CHECK-LABEL: le_i32_to_i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr w8, w0, #16
+; CHECK-NEXT: strh w0, [x1]
+; CHECK-NEXT: strh w8, [x1, #2]
+; CHECK-NEXT: ret
+ %sh1 = lshr i32 %x, 16
+ %t0 = trunc i32 %x to i16
+ %t1 = trunc i32 %sh1 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ store i16 %t0, i16* %p0, align 2
+ store i16 %t1, i16* %p1, align 2
+ ret void
+}
+
+define void @le_i32_to_i16_order(i32 %x, i16* %p0) {
+; CHECK-LABEL: le_i32_to_i16_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr w8, w0, #16
+; CHECK-NEXT: strh w8, [x1, #2]
+; CHECK-NEXT: strh w0, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i32 %x, 16
+ %t0 = trunc i32 %x to i16
+ %t1 = trunc i32 %sh1 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ store i16 %t1, i16* %p1, align 2
+ store i16 %t0, i16* %p0, align 2
+ ret void
+}
+
+define void @be_i32_to_i16(i32 %x, i16* %p0) {
+; CHECK-LABEL: be_i32_to_i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr w8, w0, #16
+; CHECK-NEXT: strh w0, [x1, #2]
+; CHECK-NEXT: strh w8, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i32 %x, 16
+ %t0 = trunc i32 %x to i16
+ %t1 = trunc i32 %sh1 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ store i16 %t0, i16* %p1, align 2
+ store i16 %t1, i16* %p0, align 2
+ ret void
+}
+
+define void @be_i32_to_i16_order(i32 %x, i16* %p0) {
+; CHECK-LABEL: be_i32_to_i16_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr w8, w0, #16
+; CHECK-NEXT: strh w8, [x1]
+; CHECK-NEXT: strh w0, [x1, #2]
+; CHECK-NEXT: ret
+ %sh1 = lshr i32 %x, 16
+ %t0 = trunc i32 %x to i16
+ %t1 = trunc i32 %sh1 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ store i16 %t1, i16* %p0, align 2
+ store i16 %t0, i16* %p1, align 2
+ ret void
+}
+
+define void @le_i64_to_i8(i64 %x, i8* %p0) {
+; LE-LABEL: le_i64_to_i8:
+; LE: // %bb.0:
+; LE-NEXT: str x0, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: le_i64_to_i8:
+; BE: // %bb.0:
+; BE-NEXT: rev x8, x0
+; BE-NEXT: str x8, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i64 %x, 8
+ %sh2 = lshr i64 %x, 16
+ %sh3 = lshr i64 %x, 24
+ %sh4 = lshr i64 %x, 32
+ %sh5 = lshr i64 %x, 40
+ %sh6 = lshr i64 %x, 48
+ %sh7 = lshr i64 %x, 56
+ %t0 = trunc i64 %x to i8
+ %t1 = trunc i64 %sh1 to i8
+ %t2 = trunc i64 %sh2 to i8
+ %t3 = trunc i64 %sh3 to i8
+ %t4 = trunc i64 %sh4 to i8
+ %t5 = trunc i64 %sh5 to i8
+ %t6 = trunc i64 %sh6 to i8
+ %t7 = trunc i64 %sh7 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ %p4 = getelementptr inbounds i8, i8* %p0, i64 4
+ %p5 = getelementptr inbounds i8, i8* %p0, i64 5
+ %p6 = getelementptr inbounds i8, i8* %p0, i64 6
+ %p7 = getelementptr inbounds i8, i8* %p0, i64 7
+ store i8 %t0, i8* %p0, align 1
+ store i8 %t1, i8* %p1, align 1
+ store i8 %t2, i8* %p2, align 1
+ store i8 %t3, i8* %p3, align 1
+ store i8 %t4, i8* %p4, align 1
+ store i8 %t5, i8* %p5, align 1
+ store i8 %t6, i8* %p6, align 1
+ store i8 %t7, i8* %p7, align 1
+ ret void
+}
+
+define void @le_i64_to_i8_order(i64 %x, i8* %p0) {
+; LE-LABEL: le_i64_to_i8_order:
+; LE: // %bb.0:
+; LE-NEXT: str x0, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: le_i64_to_i8_order:
+; BE: // %bb.0:
+; BE-NEXT: rev x8, x0
+; BE-NEXT: str x8, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i64 %x, 8
+ %sh2 = lshr i64 %x, 16
+ %sh3 = lshr i64 %x, 24
+ %sh4 = lshr i64 %x, 32
+ %sh5 = lshr i64 %x, 40
+ %sh6 = lshr i64 %x, 48
+ %sh7 = lshr i64 %x, 56
+ %t0 = trunc i64 %x to i8
+ %t1 = trunc i64 %sh1 to i8
+ %t2 = trunc i64 %sh2 to i8
+ %t3 = trunc i64 %sh3 to i8
+ %t4 = trunc i64 %sh4 to i8
+ %t5 = trunc i64 %sh5 to i8
+ %t6 = trunc i64 %sh6 to i8
+ %t7 = trunc i64 %sh7 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ %p4 = getelementptr inbounds i8, i8* %p0, i64 4
+ %p5 = getelementptr inbounds i8, i8* %p0, i64 5
+ %p6 = getelementptr inbounds i8, i8* %p0, i64 6
+ %p7 = getelementptr inbounds i8, i8* %p0, i64 7
+ store i8 %t5, i8* %p5, align 1
+ store i8 %t0, i8* %p0, align 1
+ store i8 %t3, i8* %p3, align 1
+ store i8 %t7, i8* %p7, align 1
+ store i8 %t1, i8* %p1, align 1
+ store i8 %t6, i8* %p6, align 1
+ store i8 %t2, i8* %p2, align 1
+ store i8 %t4, i8* %p4, align 1
+ ret void
+}
+
+define void @be_i64_to_i8(i64 %x, i8* %p0) {
+; LE-LABEL: be_i64_to_i8:
+; LE: // %bb.0:
+; LE-NEXT: rev x8, x0
+; LE-NEXT: str x8, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: be_i64_to_i8:
+; BE: // %bb.0:
+; BE-NEXT: str x0, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i64 %x, 8
+ %sh2 = lshr i64 %x, 16
+ %sh3 = lshr i64 %x, 24
+ %sh4 = lshr i64 %x, 32
+ %sh5 = lshr i64 %x, 40
+ %sh6 = lshr i64 %x, 48
+ %sh7 = lshr i64 %x, 56
+ %t0 = trunc i64 %x to i8
+ %t1 = trunc i64 %sh1 to i8
+ %t2 = trunc i64 %sh2 to i8
+ %t3 = trunc i64 %sh3 to i8
+ %t4 = trunc i64 %sh4 to i8
+ %t5 = trunc i64 %sh5 to i8
+ %t6 = trunc i64 %sh6 to i8
+ %t7 = trunc i64 %sh7 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ %p4 = getelementptr inbounds i8, i8* %p0, i64 4
+ %p5 = getelementptr inbounds i8, i8* %p0, i64 5
+ %p6 = getelementptr inbounds i8, i8* %p0, i64 6
+ %p7 = getelementptr inbounds i8, i8* %p0, i64 7
+ store i8 %t0, i8* %p7, align 1
+ store i8 %t1, i8* %p6, align 1
+ store i8 %t2, i8* %p5, align 1
+ store i8 %t3, i8* %p4, align 1
+ store i8 %t4, i8* %p3, align 1
+ store i8 %t5, i8* %p2, align 1
+ store i8 %t6, i8* %p1, align 1
+ store i8 %t7, i8* %p0, align 1
+ ret void
+}
+
+define void @be_i64_to_i8_order(i64 %x, i8* %p0) {
+; LE-LABEL: be_i64_to_i8_order:
+; LE: // %bb.0:
+; LE-NEXT: rev x8, x0
+; LE-NEXT: str x8, [x1]
+; LE-NEXT: ret
+;
+; BE-LABEL: be_i64_to_i8_order:
+; BE: // %bb.0:
+; BE-NEXT: str x0, [x1]
+; BE-NEXT: ret
+ %sh1 = lshr i64 %x, 8
+ %sh2 = lshr i64 %x, 16
+ %sh3 = lshr i64 %x, 24
+ %sh4 = lshr i64 %x, 32
+ %sh5 = lshr i64 %x, 40
+ %sh6 = lshr i64 %x, 48
+ %sh7 = lshr i64 %x, 56
+ %t0 = trunc i64 %x to i8
+ %t1 = trunc i64 %sh1 to i8
+ %t2 = trunc i64 %sh2 to i8
+ %t3 = trunc i64 %sh3 to i8
+ %t4 = trunc i64 %sh4 to i8
+ %t5 = trunc i64 %sh5 to i8
+ %t6 = trunc i64 %sh6 to i8
+ %t7 = trunc i64 %sh7 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ %p4 = getelementptr inbounds i8, i8* %p0, i64 4
+ %p5 = getelementptr inbounds i8, i8* %p0, i64 5
+ %p6 = getelementptr inbounds i8, i8* %p0, i64 6
+ %p7 = getelementptr inbounds i8, i8* %p0, i64 7
+ store i8 %t7, i8* %p0, align 1
+ store i8 %t6, i8* %p1, align 1
+ store i8 %t5, i8* %p2, align 1
+ store i8 %t4, i8* %p3, align 1
+ store i8 %t3, i8* %p4, align 1
+ store i8 %t2, i8* %p5, align 1
+ store i8 %t1, i8* %p6, align 1
+ store i8 %t0, i8* %p7, align 1
+ ret void
+}
+
+define void @le_i64_to_i16(i64 %x, i16* %p0) {
+; CHECK-LABEL: le_i64_to_i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #16
+; CHECK-NEXT: lsr x9, x0, #32
+; CHECK-NEXT: lsr x10, x0, #48
+; CHECK-NEXT: strh w0, [x1]
+; CHECK-NEXT: strh w8, [x1, #2]
+; CHECK-NEXT: strh w9, [x1, #4]
+; CHECK-NEXT: strh w10, [x1, #6]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 16
+ %sh2 = lshr i64 %x, 32
+ %sh3 = lshr i64 %x, 48
+ %t0 = trunc i64 %x to i16
+ %t1 = trunc i64 %sh1 to i16
+ %t2 = trunc i64 %sh2 to i16
+ %t3 = trunc i64 %sh3 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ %p2 = getelementptr inbounds i16, i16* %p0, i64 2
+ %p3 = getelementptr inbounds i16, i16* %p0, i64 3
+ store i16 %t0, i16* %p0, align 2
+ store i16 %t1, i16* %p1, align 2
+ store i16 %t2, i16* %p2, align 2
+ store i16 %t3, i16* %p3, align 2
+ ret void
+}
+
+define void @le_i64_to_i16_order(i64 %x, i16* %p0) {
+; CHECK-LABEL: le_i64_to_i16_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #16
+; CHECK-NEXT: lsr x9, x0, #32
+; CHECK-NEXT: lsr x10, x0, #48
+; CHECK-NEXT: strh w0, [x1]
+; CHECK-NEXT: strh w8, [x1, #2]
+; CHECK-NEXT: strh w10, [x1, #6]
+; CHECK-NEXT: strh w9, [x1, #4]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 16
+ %sh2 = lshr i64 %x, 32
+ %sh3 = lshr i64 %x, 48
+ %t0 = trunc i64 %x to i16
+ %t1 = trunc i64 %sh1 to i16
+ %t2 = trunc i64 %sh2 to i16
+ %t3 = trunc i64 %sh3 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ %p2 = getelementptr inbounds i16, i16* %p0, i64 2
+ %p3 = getelementptr inbounds i16, i16* %p0, i64 3
+ store i16 %t1, i16* %p1, align 2
+ store i16 %t3, i16* %p3, align 2
+ store i16 %t0, i16* %p0, align 2
+ store i16 %t2, i16* %p2, align 2
+ ret void
+}
+
+define void @be_i64_to_i16(i64 %x, i16* %p0) {
+; CHECK-LABEL: be_i64_to_i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #16
+; CHECK-NEXT: lsr x9, x0, #32
+; CHECK-NEXT: lsr x10, x0, #48
+; CHECK-NEXT: strh w0, [x1, #6]
+; CHECK-NEXT: strh w8, [x1, #4]
+; CHECK-NEXT: strh w9, [x1, #2]
+; CHECK-NEXT: strh w10, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 16
+ %sh2 = lshr i64 %x, 32
+ %sh3 = lshr i64 %x, 48
+ %t0 = trunc i64 %x to i16
+ %t1 = trunc i64 %sh1 to i16
+ %t2 = trunc i64 %sh2 to i16
+ %t3 = trunc i64 %sh3 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ %p2 = getelementptr inbounds i16, i16* %p0, i64 2
+ %p3 = getelementptr inbounds i16, i16* %p0, i64 3
+ store i16 %t0, i16* %p3, align 2
+ store i16 %t1, i16* %p2, align 2
+ store i16 %t2, i16* %p1, align 2
+ store i16 %t3, i16* %p0, align 2
+ ret void
+}
+
+define void @be_i64_to_i16_order(i64 %x, i16* %p0) {
+; CHECK-LABEL: be_i64_to_i16_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #16
+; CHECK-NEXT: lsr x9, x0, #32
+; CHECK-NEXT: lsr x10, x0, #48
+; CHECK-NEXT: strh w0, [x1, #6]
+; CHECK-NEXT: strh w10, [x1]
+; CHECK-NEXT: strh w9, [x1, #2]
+; CHECK-NEXT: strh w8, [x1, #4]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 16
+ %sh2 = lshr i64 %x, 32
+ %sh3 = lshr i64 %x, 48
+ %t0 = trunc i64 %x to i16
+ %t1 = trunc i64 %sh1 to i16
+ %t2 = trunc i64 %sh2 to i16
+ %t3 = trunc i64 %sh3 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ %p2 = getelementptr inbounds i16, i16* %p0, i64 2
+ %p3 = getelementptr inbounds i16, i16* %p0, i64 3
+ store i16 %t0, i16* %p3, align 2
+ store i16 %t3, i16* %p0, align 2
+ store i16 %t2, i16* %p1, align 2
+ store i16 %t1, i16* %p2, align 2
+ ret void
+}
+
+define void @le_i64_to_i32(i64 %x, i32* %p0) {
+; CHECK-LABEL: le_i64_to_i32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #32
+; CHECK-NEXT: stp w0, w8, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 32
+ %t0 = trunc i64 %x to i32
+ %t1 = trunc i64 %sh1 to i32
+ %p1 = getelementptr inbounds i32, i32* %p0, i64 1
+ store i32 %t0, i32* %p0, align 4
+ store i32 %t1, i32* %p1, align 4
+ ret void
+}
+
+define void @le_i64_to_i32_order(i64 %x, i32* %p0) {
+; CHECK-LABEL: le_i64_to_i32_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #32
+; CHECK-NEXT: stp w0, w8, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 32
+ %t0 = trunc i64 %x to i32
+ %t1 = trunc i64 %sh1 to i32
+ %p1 = getelementptr inbounds i32, i32* %p0, i64 1
+ store i32 %t1, i32* %p1, align 4
+ store i32 %t0, i32* %p0, align 4
+ ret void
+}
+
+define void @be_i64_to_i32(i64 %x, i32* %p0) {
+; CHECK-LABEL: be_i64_to_i32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #32
+; CHECK-NEXT: stp w8, w0, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 32
+ %t0 = trunc i64 %x to i32
+ %t1 = trunc i64 %sh1 to i32
+ %p1 = getelementptr inbounds i32, i32* %p0, i64 1
+ store i32 %t0, i32* %p1, align 4
+ store i32 %t1, i32* %p0, align 4
+ ret void
+}
+
+define void @be_i64_to_i32_order(i64 %x, i32* %p0) {
+; CHECK-LABEL: be_i64_to_i32_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #32
+; CHECK-NEXT: stp w8, w0, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 32
+ %t0 = trunc i64 %x to i32
+ %t1 = trunc i64 %sh1 to i32
+ %p1 = getelementptr inbounds i32, i32* %p0, i64 1
+ store i32 %t1, i32* %p0, align 4
+ store i32 %t0, i32* %p1, align 4
+ ret void
+}
+
+define void @i64_to_i32_wrong_addr(i64 %x, i32* %p0) {
+; CHECK-LABEL: i64_to_i32_wrong_addr:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #32
+; CHECK-NEXT: str w8, [x1, #12]
+; CHECK-NEXT: str w0, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 32
+ %t0 = trunc i64 %x to i32
+ %t1 = trunc i64 %sh1 to i32
+ %p3 = getelementptr inbounds i32, i32* %p0, i64 3
+ store i32 %t1, i32* %p3, align 4
+ store i32 %t0, i32* %p0, align 4
+ ret void
+}
+
+define void @i64_to_i16_wrong_order(i64 %x, i16* %p0) {
+; CHECK-LABEL: i64_to_i16_wrong_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #16
+; CHECK-NEXT: lsr x9, x0, #32
+; CHECK-NEXT: lsr x10, x0, #48
+; CHECK-NEXT: strh w10, [x1, #6]
+; CHECK-NEXT: strh w8, [x1, #4]
+; CHECK-NEXT: strh w9, [x1, #2]
+; CHECK-NEXT: strh w0, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 16
+ %sh2 = lshr i64 %x, 32
+ %sh3 = lshr i64 %x, 48
+ %t0 = trunc i64 %x to i16
+ %t1 = trunc i64 %sh1 to i16
+ %t2 = trunc i64 %sh2 to i16
+ %t3 = trunc i64 %sh3 to i16
+ %p1 = getelementptr inbounds i16, i16* %p0, i64 1
+ %p2 = getelementptr inbounds i16, i16* %p0, i64 2
+ %p3 = getelementptr inbounds i16, i16* %p0, i64 3
+ store i16 %t3, i16* %p3, align 2
+ store i16 %t1, i16* %p2, align 2
+ store i16 %t2, i16* %p1, align 2
+ store i16 %t0, i16* %p0, align 2
+ ret void
+}
+
+define void @i32_to_i8_incomplete(i32 %x, i8* %p0) {
+; CHECK-LABEL: i32_to_i8_incomplete:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr w8, w0, #16
+; CHECK-NEXT: lsr w9, w0, #24
+; CHECK-NEXT: strb w0, [x1]
+; CHECK-NEXT: strb w8, [x1, #2]
+; CHECK-NEXT: strb w9, [x1, #3]
+; CHECK-NEXT: ret
+ %sh1 = lshr i32 %x, 8
+ %sh2 = lshr i32 %x, 16
+ %sh3 = lshr i32 %x, 24
+ %t0 = trunc i32 %x to i8
+ %t1 = trunc i32 %sh1 to i8
+ %t2 = trunc i32 %sh2 to i8
+ %t3 = trunc i32 %sh3 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ store i8 %t0, i8* %p0, align 1
+ store i8 %t2, i8* %p2, align 1
+ store i8 %t3, i8* %p3, align 1
+ ret void
+}
+
+define void @i64_to_i8_incomplete(i64 %x, i8* %p0) {
+; CHECK-LABEL: i64_to_i8_incomplete:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr x8, x0, #8
+; CHECK-NEXT: lsr x9, x0, #16
+; CHECK-NEXT: lsr x10, x0, #32
+; CHECK-NEXT: lsr x11, x0, #40
+; CHECK-NEXT: lsr x12, x0, #48
+; CHECK-NEXT: lsr x13, x0, #56
+; CHECK-NEXT: strb w13, [x1]
+; CHECK-NEXT: strb w12, [x1, #1]
+; CHECK-NEXT: strb w11, [x1, #2]
+; CHECK-NEXT: strb w10, [x1, #3]
+; CHECK-NEXT: strb w9, [x1, #5]
+; CHECK-NEXT: strb w8, [x1, #6]
+; CHECK-NEXT: strb w0, [x1, #7]
+; CHECK-NEXT: ret
+ %sh1 = lshr i64 %x, 8
+ %sh2 = lshr i64 %x, 16
+ %sh3 = lshr i64 %x, 24
+ %sh4 = lshr i64 %x, 32
+ %sh5 = lshr i64 %x, 40
+ %sh6 = lshr i64 %x, 48
+ %sh7 = lshr i64 %x, 56
+ %t0 = trunc i64 %x to i8
+ %t1 = trunc i64 %sh1 to i8
+ %t2 = trunc i64 %sh2 to i8
+ %t3 = trunc i64 %sh3 to i8
+ %t4 = trunc i64 %sh4 to i8
+ %t5 = trunc i64 %sh5 to i8
+ %t6 = trunc i64 %sh6 to i8
+ %t7 = trunc i64 %sh7 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ %p4 = getelementptr inbounds i8, i8* %p0, i64 4
+ %p5 = getelementptr inbounds i8, i8* %p0, i64 5
+ %p6 = getelementptr inbounds i8, i8* %p0, i64 6
+ %p7 = getelementptr inbounds i8, i8* %p0, i64 7
+ store i8 %t7, i8* %p0, align 1
+ store i8 %t6, i8* %p1, align 1
+ store i8 %t5, i8* %p2, align 1
+ store i8 %t4, i8* %p3, align 1
+ store i8 %t2, i8* %p5, align 1
+ store i8 %t1, i8* %p6, align 1
+ store i8 %t0, i8* %p7, align 1
+ ret void
+}
+
+define void @i32_to_i16_wrong_addr(i32 %x, i16* %p0) {
+; CHECK-LABEL: i32_to_i16_wrong_addr:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr w8, w0, #16
+; CHECK-NEXT: strh w8, [x1, #4]
+; CHECK-NEXT: strh w0, [x1]
+; CHECK-NEXT: ret
+ %sh1 = lshr i32 %x, 16
+ %t0 = trunc i32 %x to i16
+ %t1 = trunc i32 %sh1 to i16
+ %p2 = getelementptr inbounds i16, i16* %p0, i64 2
+ store i16 %t1, i16* %p2, align 2
+ store i16 %t0, i16* %p0, align 2
+ ret void
+}
+
+define void @i32_to_i8_wrong_order(i32 %x, i8* %p0) {
+; CHECK-LABEL: i32_to_i8_wrong_order:
+; CHECK: // %bb.0:
+; CHECK-NEXT: lsr w8, w0, #8
+; CHECK-NEXT: lsr w9, w0, #16
+; CHECK-NEXT: lsr w10, w0, #24
+; CHECK-NEXT: strb w0, [x1, #3]
+; CHECK-NEXT: strb w10, [x1, #1]
+; CHECK-NEXT: strb w9, [x1]
+; CHECK-NEXT: strb w8, [x1, #2]
+; CHECK-NEXT: ret
+ %sh1 = lshr i32 %x, 8
+ %sh2 = lshr i32 %x, 16
+ %sh3 = lshr i32 %x, 24
+ %t0 = trunc i32 %x to i8
+ %t1 = trunc i32 %sh1 to i8
+ %t2 = trunc i32 %sh2 to i8
+ %t3 = trunc i32 %sh3 to i8
+ %p1 = getelementptr inbounds i8, i8* %p0, i64 1
+ %p2 = getelementptr inbounds i8, i8* %p0, i64 2
+ %p3 = getelementptr inbounds i8, i8* %p0, i64 3
+ store i8 %t3, i8* %p1, align 1
+ store i8 %t2, i8* %p0, align 1
+ store i8 %t0, i8* %p3, align 1
+ store i8 %t1, i8* %p2, align 1
+ ret void
+}
More information about the llvm-commits
mailing list