[llvm] 646478f - [AArch64] Scalarize i128 add/sub/mul/and/or/xor vectors

David Green via llvm-commits llvm-commits at lists.llvm.org
Fri Aug 23 02:53:37 PDT 2024


Author: David Green
Date: 2024-08-23T10:53:31+01:00
New Revision: 646478f38b03cbc861ae17533c641c2a944118b3

URL: https://github.com/llvm/llvm-project/commit/646478f38b03cbc861ae17533c641c2a944118b3
DIFF: https://github.com/llvm/llvm-project/commit/646478f38b03cbc861ae17533c641c2a944118b3.diff

LOG: [AArch64] Scalarize i128 add/sub/mul/and/or/xor vectors

This mirrors what we do for SDAG, scalarizing i128 vectors with
add/sub/mul/and/or/xor operators.

Added: 
    

Modified: 
    llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
    llvm/test/CodeGen/AArch64/add.ll
    llvm/test/CodeGen/AArch64/andorxor.ll
    llvm/test/CodeGen/AArch64/mul.ll
    llvm/test/CodeGen/AArch64/sub.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
index 35d73d36df46fe..7eaf6a84bd204f 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
@@ -149,6 +149,7 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)
             return Query.Types[0].getNumElements() <= 16;
           },
           0, s8)
+      .scalarizeIf(scalarOrEltWiderThan(0, 64), 0)
       .moreElementsToNextPow2(0);
 
   getActionDefinitionsBuilder({G_SHL, G_ASHR, G_LSHR})

diff  --git a/llvm/test/CodeGen/AArch64/add.ll b/llvm/test/CodeGen/AArch64/add.ll
index 39d1933f0e7b97..ee15445a7bbd62 100644
--- a/llvm/test/CodeGen/AArch64/add.ll
+++ b/llvm/test/CodeGen/AArch64/add.ll
@@ -1,10 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc -mtriple=aarch64-none-eabi -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
-; RUN: llc -mtriple=aarch64-none-eabi -global-isel -global-isel-abort=2 -verify-machineinstrs %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
-
-; CHECK-GI:       warning: Instruction selection used fallback path for v2i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for v3i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for v4i128
+; RUN: llc -mtriple=aarch64-none-eabi -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 define i8 @i8(i8 %a, i8 %b) {
 ; CHECK-LABEL: i8:
@@ -480,21 +476,37 @@ entry:
 }
 
 define <4 x i128> @v4i128(<4 x i128> %d, <4 x i128> %e) {
-; CHECK-LABEL: v4i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    ldp x11, x10, [sp, #16]
-; CHECK-NEXT:    ldp x13, x12, [sp, #32]
-; CHECK-NEXT:    adds x0, x0, x8
-; CHECK-NEXT:    adc x1, x1, x9
-; CHECK-NEXT:    ldp x8, x9, [sp, #48]
-; CHECK-NEXT:    adds x2, x2, x11
-; CHECK-NEXT:    adc x3, x3, x10
-; CHECK-NEXT:    adds x4, x4, x13
-; CHECK-NEXT:    adc x5, x5, x12
-; CHECK-NEXT:    adds x6, x6, x8
-; CHECK-NEXT:    adc x7, x7, x9
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: v4i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x8, x9, [sp]
+; CHECK-SD-NEXT:    ldp x11, x10, [sp, #16]
+; CHECK-SD-NEXT:    ldp x13, x12, [sp, #32]
+; CHECK-SD-NEXT:    adds x0, x0, x8
+; CHECK-SD-NEXT:    adc x1, x1, x9
+; CHECK-SD-NEXT:    ldp x8, x9, [sp, #48]
+; CHECK-SD-NEXT:    adds x2, x2, x11
+; CHECK-SD-NEXT:    adc x3, x3, x10
+; CHECK-SD-NEXT:    adds x4, x4, x13
+; CHECK-SD-NEXT:    adc x5, x5, x12
+; CHECK-SD-NEXT:    adds x6, x6, x8
+; CHECK-SD-NEXT:    adc x7, x7, x9
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: v4i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    ldp x12, x13, [sp, #32]
+; CHECK-GI-NEXT:    adds x0, x0, x8
+; CHECK-GI-NEXT:    adc x1, x1, x9
+; CHECK-GI-NEXT:    ldp x8, x9, [sp, #48]
+; CHECK-GI-NEXT:    adds x2, x2, x10
+; CHECK-GI-NEXT:    adc x3, x3, x11
+; CHECK-GI-NEXT:    adds x4, x4, x12
+; CHECK-GI-NEXT:    adc x5, x5, x13
+; CHECK-GI-NEXT:    adds x6, x6, x8
+; CHECK-GI-NEXT:    adc x7, x7, x9
+; CHECK-GI-NEXT:    ret
 entry:
   %s = add <4 x i128> %d, %e
   ret <4 x i128> %s

diff  --git a/llvm/test/CodeGen/AArch64/andorxor.ll b/llvm/test/CodeGen/AArch64/andorxor.ll
index efa4be707ceda9..1176c98ce44e34 100644
--- a/llvm/test/CodeGen/AArch64/andorxor.ll
+++ b/llvm/test/CodeGen/AArch64/andorxor.ll
@@ -1,16 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc -mtriple=aarch64-none-eabi -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
-; RUN: llc -mtriple=aarch64-none-eabi -global-isel -global-isel-abort=2 -verify-machineinstrs %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
-
-; CHECK-GI:       warning: Instruction selection used fallback path for and_v2i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for or_v2i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for xor_v2i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for and_v3i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for or_v3i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for xor_v3i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for and_v4i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for or_v4i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for xor_v4i128
+; RUN: llc -mtriple=aarch64-none-eabi -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 define i8 @and_i8(i8 %a, i8 %b) {
 ; CHECK-LABEL: and_i8:
@@ -1369,153 +1359,261 @@ entry:
 }
 
 define <2 x i128> @and_v2i128(<2 x i128> %d, <2 x i128> %e) {
-; CHECK-LABEL: and_v2i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    and x2, x2, x6
-; CHECK-NEXT:    and x0, x0, x4
-; CHECK-NEXT:    and x1, x1, x5
-; CHECK-NEXT:    and x3, x3, x7
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: and_v2i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    and x2, x2, x6
+; CHECK-SD-NEXT:    and x0, x0, x4
+; CHECK-SD-NEXT:    and x1, x1, x5
+; CHECK-SD-NEXT:    and x3, x3, x7
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_v2i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    and x0, x0, x4
+; CHECK-GI-NEXT:    and x1, x1, x5
+; CHECK-GI-NEXT:    and x2, x2, x6
+; CHECK-GI-NEXT:    and x3, x3, x7
+; CHECK-GI-NEXT:    ret
 entry:
   %s = and <2 x i128> %d, %e
   ret <2 x i128> %s
 }
 
 define <2 x i128> @or_v2i128(<2 x i128> %d, <2 x i128> %e) {
-; CHECK-LABEL: or_v2i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    orr x2, x2, x6
-; CHECK-NEXT:    orr x0, x0, x4
-; CHECK-NEXT:    orr x1, x1, x5
-; CHECK-NEXT:    orr x3, x3, x7
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: or_v2i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    orr x2, x2, x6
+; CHECK-SD-NEXT:    orr x0, x0, x4
+; CHECK-SD-NEXT:    orr x1, x1, x5
+; CHECK-SD-NEXT:    orr x3, x3, x7
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: or_v2i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    orr x0, x0, x4
+; CHECK-GI-NEXT:    orr x1, x1, x5
+; CHECK-GI-NEXT:    orr x2, x2, x6
+; CHECK-GI-NEXT:    orr x3, x3, x7
+; CHECK-GI-NEXT:    ret
 entry:
   %s = or <2 x i128> %d, %e
   ret <2 x i128> %s
 }
 
 define <2 x i128> @xor_v2i128(<2 x i128> %d, <2 x i128> %e) {
-; CHECK-LABEL: xor_v2i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    eor x2, x2, x6
-; CHECK-NEXT:    eor x0, x0, x4
-; CHECK-NEXT:    eor x1, x1, x5
-; CHECK-NEXT:    eor x3, x3, x7
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: xor_v2i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    eor x2, x2, x6
+; CHECK-SD-NEXT:    eor x0, x0, x4
+; CHECK-SD-NEXT:    eor x1, x1, x5
+; CHECK-SD-NEXT:    eor x3, x3, x7
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: xor_v2i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    eor x0, x0, x4
+; CHECK-GI-NEXT:    eor x1, x1, x5
+; CHECK-GI-NEXT:    eor x2, x2, x6
+; CHECK-GI-NEXT:    eor x3, x3, x7
+; CHECK-GI-NEXT:    ret
 entry:
   %s = xor <2 x i128> %d, %e
   ret <2 x i128> %s
 }
 
 define <3 x i128> @and_v3i128(<3 x i128> %d, <3 x i128> %e) {
-; CHECK-LABEL: and_v3i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    and x0, x0, x6
-; CHECK-NEXT:    ldp x11, x10, [sp, #16]
-; CHECK-NEXT:    and x1, x1, x7
-; CHECK-NEXT:    and x2, x2, x8
-; CHECK-NEXT:    and x3, x3, x9
-; CHECK-NEXT:    and x4, x4, x11
-; CHECK-NEXT:    and x5, x5, x10
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: and_v3i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x8, x9, [sp]
+; CHECK-SD-NEXT:    and x0, x0, x6
+; CHECK-SD-NEXT:    ldp x11, x10, [sp, #16]
+; CHECK-SD-NEXT:    and x1, x1, x7
+; CHECK-SD-NEXT:    and x2, x2, x8
+; CHECK-SD-NEXT:    and x3, x3, x9
+; CHECK-SD-NEXT:    and x4, x4, x11
+; CHECK-SD-NEXT:    and x5, x5, x10
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_v3i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    and x0, x0, x6
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    and x1, x1, x7
+; CHECK-GI-NEXT:    and x2, x2, x8
+; CHECK-GI-NEXT:    and x3, x3, x9
+; CHECK-GI-NEXT:    and x4, x4, x10
+; CHECK-GI-NEXT:    and x5, x5, x11
+; CHECK-GI-NEXT:    ret
 entry:
   %s = and <3 x i128> %d, %e
   ret <3 x i128> %s
 }
 
 define <3 x i128> @or_v3i128(<3 x i128> %d, <3 x i128> %e) {
-; CHECK-LABEL: or_v3i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    orr x0, x0, x6
-; CHECK-NEXT:    ldp x11, x10, [sp, #16]
-; CHECK-NEXT:    orr x1, x1, x7
-; CHECK-NEXT:    orr x2, x2, x8
-; CHECK-NEXT:    orr x3, x3, x9
-; CHECK-NEXT:    orr x4, x4, x11
-; CHECK-NEXT:    orr x5, x5, x10
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: or_v3i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x8, x9, [sp]
+; CHECK-SD-NEXT:    orr x0, x0, x6
+; CHECK-SD-NEXT:    ldp x11, x10, [sp, #16]
+; CHECK-SD-NEXT:    orr x1, x1, x7
+; CHECK-SD-NEXT:    orr x2, x2, x8
+; CHECK-SD-NEXT:    orr x3, x3, x9
+; CHECK-SD-NEXT:    orr x4, x4, x11
+; CHECK-SD-NEXT:    orr x5, x5, x10
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: or_v3i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    orr x0, x0, x6
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    orr x1, x1, x7
+; CHECK-GI-NEXT:    orr x2, x2, x8
+; CHECK-GI-NEXT:    orr x3, x3, x9
+; CHECK-GI-NEXT:    orr x4, x4, x10
+; CHECK-GI-NEXT:    orr x5, x5, x11
+; CHECK-GI-NEXT:    ret
 entry:
   %s = or <3 x i128> %d, %e
   ret <3 x i128> %s
 }
 
 define <3 x i128> @xor_v3i128(<3 x i128> %d, <3 x i128> %e) {
-; CHECK-LABEL: xor_v3i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    eor x0, x0, x6
-; CHECK-NEXT:    ldp x11, x10, [sp, #16]
-; CHECK-NEXT:    eor x1, x1, x7
-; CHECK-NEXT:    eor x2, x2, x8
-; CHECK-NEXT:    eor x3, x3, x9
-; CHECK-NEXT:    eor x4, x4, x11
-; CHECK-NEXT:    eor x5, x5, x10
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: xor_v3i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x8, x9, [sp]
+; CHECK-SD-NEXT:    eor x0, x0, x6
+; CHECK-SD-NEXT:    ldp x11, x10, [sp, #16]
+; CHECK-SD-NEXT:    eor x1, x1, x7
+; CHECK-SD-NEXT:    eor x2, x2, x8
+; CHECK-SD-NEXT:    eor x3, x3, x9
+; CHECK-SD-NEXT:    eor x4, x4, x11
+; CHECK-SD-NEXT:    eor x5, x5, x10
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: xor_v3i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    eor x0, x0, x6
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    eor x1, x1, x7
+; CHECK-GI-NEXT:    eor x2, x2, x8
+; CHECK-GI-NEXT:    eor x3, x3, x9
+; CHECK-GI-NEXT:    eor x4, x4, x10
+; CHECK-GI-NEXT:    eor x5, x5, x11
+; CHECK-GI-NEXT:    ret
 entry:
   %s = xor <3 x i128> %d, %e
   ret <3 x i128> %s
 }
 
 define <4 x i128> @and_v4i128(<4 x i128> %d, <4 x i128> %e) {
-; CHECK-LABEL: and_v4i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x9, x8, [sp, #32]
-; CHECK-NEXT:    ldp x11, x10, [sp]
-; CHECK-NEXT:    ldp x13, x12, [sp, #16]
-; CHECK-NEXT:    ldp x15, x14, [sp, #48]
-; CHECK-NEXT:    and x4, x4, x9
-; CHECK-NEXT:    and x0, x0, x11
-; CHECK-NEXT:    and x1, x1, x10
-; CHECK-NEXT:    and x5, x5, x8
-; CHECK-NEXT:    and x2, x2, x13
-; CHECK-NEXT:    and x3, x3, x12
-; CHECK-NEXT:    and x6, x6, x15
-; CHECK-NEXT:    and x7, x7, x14
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: and_v4i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x9, x8, [sp, #32]
+; CHECK-SD-NEXT:    ldp x11, x10, [sp]
+; CHECK-SD-NEXT:    ldp x13, x12, [sp, #16]
+; CHECK-SD-NEXT:    ldp x15, x14, [sp, #48]
+; CHECK-SD-NEXT:    and x4, x4, x9
+; CHECK-SD-NEXT:    and x0, x0, x11
+; CHECK-SD-NEXT:    and x1, x1, x10
+; CHECK-SD-NEXT:    and x5, x5, x8
+; CHECK-SD-NEXT:    and x2, x2, x13
+; CHECK-SD-NEXT:    and x3, x3, x12
+; CHECK-SD-NEXT:    and x6, x6, x15
+; CHECK-SD-NEXT:    and x7, x7, x14
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_v4i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    ldp x12, x13, [sp, #32]
+; CHECK-GI-NEXT:    ldp x14, x15, [sp, #48]
+; CHECK-GI-NEXT:    and x0, x0, x8
+; CHECK-GI-NEXT:    and x1, x1, x9
+; CHECK-GI-NEXT:    and x2, x2, x10
+; CHECK-GI-NEXT:    and x3, x3, x11
+; CHECK-GI-NEXT:    and x4, x4, x12
+; CHECK-GI-NEXT:    and x5, x5, x13
+; CHECK-GI-NEXT:    and x6, x6, x14
+; CHECK-GI-NEXT:    and x7, x7, x15
+; CHECK-GI-NEXT:    ret
 entry:
   %s = and <4 x i128> %d, %e
   ret <4 x i128> %s
 }
 
 define <4 x i128> @or_v4i128(<4 x i128> %d, <4 x i128> %e) {
-; CHECK-LABEL: or_v4i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x9, x8, [sp, #32]
-; CHECK-NEXT:    ldp x11, x10, [sp]
-; CHECK-NEXT:    ldp x13, x12, [sp, #16]
-; CHECK-NEXT:    ldp x15, x14, [sp, #48]
-; CHECK-NEXT:    orr x4, x4, x9
-; CHECK-NEXT:    orr x0, x0, x11
-; CHECK-NEXT:    orr x1, x1, x10
-; CHECK-NEXT:    orr x5, x5, x8
-; CHECK-NEXT:    orr x2, x2, x13
-; CHECK-NEXT:    orr x3, x3, x12
-; CHECK-NEXT:    orr x6, x6, x15
-; CHECK-NEXT:    orr x7, x7, x14
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: or_v4i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x9, x8, [sp, #32]
+; CHECK-SD-NEXT:    ldp x11, x10, [sp]
+; CHECK-SD-NEXT:    ldp x13, x12, [sp, #16]
+; CHECK-SD-NEXT:    ldp x15, x14, [sp, #48]
+; CHECK-SD-NEXT:    orr x4, x4, x9
+; CHECK-SD-NEXT:    orr x0, x0, x11
+; CHECK-SD-NEXT:    orr x1, x1, x10
+; CHECK-SD-NEXT:    orr x5, x5, x8
+; CHECK-SD-NEXT:    orr x2, x2, x13
+; CHECK-SD-NEXT:    orr x3, x3, x12
+; CHECK-SD-NEXT:    orr x6, x6, x15
+; CHECK-SD-NEXT:    orr x7, x7, x14
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: or_v4i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    ldp x12, x13, [sp, #32]
+; CHECK-GI-NEXT:    ldp x14, x15, [sp, #48]
+; CHECK-GI-NEXT:    orr x0, x0, x8
+; CHECK-GI-NEXT:    orr x1, x1, x9
+; CHECK-GI-NEXT:    orr x2, x2, x10
+; CHECK-GI-NEXT:    orr x3, x3, x11
+; CHECK-GI-NEXT:    orr x4, x4, x12
+; CHECK-GI-NEXT:    orr x5, x5, x13
+; CHECK-GI-NEXT:    orr x6, x6, x14
+; CHECK-GI-NEXT:    orr x7, x7, x15
+; CHECK-GI-NEXT:    ret
 entry:
   %s = or <4 x i128> %d, %e
   ret <4 x i128> %s
 }
 
 define <4 x i128> @xor_v4i128(<4 x i128> %d, <4 x i128> %e) {
-; CHECK-LABEL: xor_v4i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x9, x8, [sp, #32]
-; CHECK-NEXT:    ldp x11, x10, [sp]
-; CHECK-NEXT:    ldp x13, x12, [sp, #16]
-; CHECK-NEXT:    ldp x15, x14, [sp, #48]
-; CHECK-NEXT:    eor x4, x4, x9
-; CHECK-NEXT:    eor x0, x0, x11
-; CHECK-NEXT:    eor x1, x1, x10
-; CHECK-NEXT:    eor x5, x5, x8
-; CHECK-NEXT:    eor x2, x2, x13
-; CHECK-NEXT:    eor x3, x3, x12
-; CHECK-NEXT:    eor x6, x6, x15
-; CHECK-NEXT:    eor x7, x7, x14
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: xor_v4i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x9, x8, [sp, #32]
+; CHECK-SD-NEXT:    ldp x11, x10, [sp]
+; CHECK-SD-NEXT:    ldp x13, x12, [sp, #16]
+; CHECK-SD-NEXT:    ldp x15, x14, [sp, #48]
+; CHECK-SD-NEXT:    eor x4, x4, x9
+; CHECK-SD-NEXT:    eor x0, x0, x11
+; CHECK-SD-NEXT:    eor x1, x1, x10
+; CHECK-SD-NEXT:    eor x5, x5, x8
+; CHECK-SD-NEXT:    eor x2, x2, x13
+; CHECK-SD-NEXT:    eor x3, x3, x12
+; CHECK-SD-NEXT:    eor x6, x6, x15
+; CHECK-SD-NEXT:    eor x7, x7, x14
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: xor_v4i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    ldp x12, x13, [sp, #32]
+; CHECK-GI-NEXT:    ldp x14, x15, [sp, #48]
+; CHECK-GI-NEXT:    eor x0, x0, x8
+; CHECK-GI-NEXT:    eor x1, x1, x9
+; CHECK-GI-NEXT:    eor x2, x2, x10
+; CHECK-GI-NEXT:    eor x3, x3, x11
+; CHECK-GI-NEXT:    eor x4, x4, x12
+; CHECK-GI-NEXT:    eor x5, x5, x13
+; CHECK-GI-NEXT:    eor x6, x6, x14
+; CHECK-GI-NEXT:    eor x7, x7, x15
+; CHECK-GI-NEXT:    ret
 entry:
   %s = xor <4 x i128> %d, %e
   ret <4 x i128> %s

diff  --git a/llvm/test/CodeGen/AArch64/mul.ll b/llvm/test/CodeGen/AArch64/mul.ll
index d2804329f1e255..02258bc47c54d4 100644
--- a/llvm/test/CodeGen/AArch64/mul.ll
+++ b/llvm/test/CodeGen/AArch64/mul.ll
@@ -1,10 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc -mtriple=aarch64-none-eabi -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
-; RUN: llc -mtriple=aarch64-none-eabi -global-isel -global-isel-abort=2 -verify-machineinstrs %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
-
-; CHECK-GI:       warning: Instruction selection used fallback path for v2i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for v3i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for v4i128
+; RUN: llc -mtriple=aarch64-none-eabi -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 define i8 @i8(i8 %a, i8 %b) {
 ; CHECK-LABEL: i8:
@@ -531,69 +527,139 @@ entry:
 }
 
 define <2 x i128> @v2i128(<2 x i128> %d, <2 x i128> %e) {
-; CHECK-LABEL: v2i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    umulh x8, x2, x6
-; CHECK-NEXT:    umulh x9, x0, x4
-; CHECK-NEXT:    madd x8, x2, x7, x8
-; CHECK-NEXT:    madd x9, x0, x5, x9
-; CHECK-NEXT:    madd x3, x3, x6, x8
-; CHECK-NEXT:    madd x1, x1, x4, x9
-; CHECK-NEXT:    mul x0, x0, x4
-; CHECK-NEXT:    mul x2, x2, x6
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: v2i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    umulh x8, x2, x6
+; CHECK-SD-NEXT:    umulh x9, x0, x4
+; CHECK-SD-NEXT:    madd x8, x2, x7, x8
+; CHECK-SD-NEXT:    madd x9, x0, x5, x9
+; CHECK-SD-NEXT:    madd x3, x3, x6, x8
+; CHECK-SD-NEXT:    madd x1, x1, x4, x9
+; CHECK-SD-NEXT:    mul x0, x0, x4
+; CHECK-SD-NEXT:    mul x2, x2, x6
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: v2i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    mul x9, x0, x5
+; CHECK-GI-NEXT:    mul x12, x2, x7
+; CHECK-GI-NEXT:    mul x8, x0, x4
+; CHECK-GI-NEXT:    umulh x10, x0, x4
+; CHECK-GI-NEXT:    madd x11, x1, x4, x9
+; CHECK-GI-NEXT:    mov x0, x8
+; CHECK-GI-NEXT:    mul x9, x2, x6
+; CHECK-GI-NEXT:    umulh x13, x2, x6
+; CHECK-GI-NEXT:    add x1, x11, x10
+; CHECK-GI-NEXT:    madd x12, x3, x6, x12
+; CHECK-GI-NEXT:    mov x2, x9
+; CHECK-GI-NEXT:    add x3, x12, x13
+; CHECK-GI-NEXT:    ret
 entry:
   %s = mul <2 x i128> %d, %e
   ret <2 x i128> %s
 }
 
 define <3 x i128> @v3i128(<3 x i128> %d, <3 x i128> %e) {
-; CHECK-LABEL: v3i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    umulh x9, x0, x6
-; CHECK-NEXT:    ldp x8, x10, [sp]
-; CHECK-NEXT:    madd x9, x0, x7, x9
-; CHECK-NEXT:    umulh x11, x2, x8
-; CHECK-NEXT:    madd x1, x1, x6, x9
-; CHECK-NEXT:    ldp x9, x12, [sp, #16]
-; CHECK-NEXT:    madd x10, x2, x10, x11
-; CHECK-NEXT:    umulh x13, x4, x9
-; CHECK-NEXT:    madd x3, x3, x8, x10
-; CHECK-NEXT:    madd x11, x4, x12, x13
-; CHECK-NEXT:    mul x0, x0, x6
-; CHECK-NEXT:    madd x5, x5, x9, x11
-; CHECK-NEXT:    mul x2, x2, x8
-; CHECK-NEXT:    mul x4, x4, x9
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: v3i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    umulh x9, x0, x6
+; CHECK-SD-NEXT:    ldp x8, x10, [sp]
+; CHECK-SD-NEXT:    madd x9, x0, x7, x9
+; CHECK-SD-NEXT:    umulh x11, x2, x8
+; CHECK-SD-NEXT:    madd x1, x1, x6, x9
+; CHECK-SD-NEXT:    ldp x9, x12, [sp, #16]
+; CHECK-SD-NEXT:    madd x10, x2, x10, x11
+; CHECK-SD-NEXT:    umulh x13, x4, x9
+; CHECK-SD-NEXT:    madd x3, x3, x8, x10
+; CHECK-SD-NEXT:    madd x11, x4, x12, x13
+; CHECK-SD-NEXT:    mul x0, x0, x6
+; CHECK-SD-NEXT:    madd x5, x5, x9, x11
+; CHECK-SD-NEXT:    mul x2, x2, x8
+; CHECK-SD-NEXT:    mul x4, x4, x9
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: v3i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x10, x13, [sp]
+; CHECK-GI-NEXT:    mul x9, x0, x7
+; CHECK-GI-NEXT:    mul x8, x0, x6
+; CHECK-GI-NEXT:    mul x13, x2, x13
+; CHECK-GI-NEXT:    madd x12, x1, x6, x9
+; CHECK-GI-NEXT:    mul x9, x2, x10
+; CHECK-GI-NEXT:    umulh x14, x2, x10
+; CHECK-GI-NEXT:    madd x10, x3, x10, x13
+; CHECK-GI-NEXT:    ldp x13, x15, [sp, #16]
+; CHECK-GI-NEXT:    mov x2, x9
+; CHECK-GI-NEXT:    umulh x11, x0, x6
+; CHECK-GI-NEXT:    mov x0, x8
+; CHECK-GI-NEXT:    mul x15, x4, x15
+; CHECK-GI-NEXT:    add x3, x10, x14
+; CHECK-GI-NEXT:    umulh x16, x4, x13
+; CHECK-GI-NEXT:    add x1, x12, x11
+; CHECK-GI-NEXT:    madd x15, x5, x13, x15
+; CHECK-GI-NEXT:    mul x4, x4, x13
+; CHECK-GI-NEXT:    add x5, x15, x16
+; CHECK-GI-NEXT:    ret
 entry:
   %s = mul <3 x i128> %d, %e
   ret <3 x i128> %s
 }
 
 define <4 x i128> @v4i128(<4 x i128> %d, <4 x i128> %e) {
-; CHECK-LABEL: v4i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    ldp x11, x12, [sp, #16]
-; CHECK-NEXT:    umulh x10, x0, x8
-; CHECK-NEXT:    umulh x13, x2, x11
-; CHECK-NEXT:    madd x9, x0, x9, x10
-; CHECK-NEXT:    madd x10, x2, x12, x13
-; CHECK-NEXT:    ldp x13, x14, [sp, #48]
-; CHECK-NEXT:    madd x1, x1, x8, x9
-; CHECK-NEXT:    madd x3, x3, x11, x10
-; CHECK-NEXT:    ldp x9, x10, [sp, #32]
-; CHECK-NEXT:    umulh x15, x6, x13
-; CHECK-NEXT:    umulh x12, x4, x9
-; CHECK-NEXT:    mul x0, x0, x8
-; CHECK-NEXT:    madd x10, x4, x10, x12
-; CHECK-NEXT:    madd x12, x6, x14, x15
-; CHECK-NEXT:    madd x5, x5, x9, x10
-; CHECK-NEXT:    madd x7, x7, x13, x12
-; CHECK-NEXT:    mul x2, x2, x11
-; CHECK-NEXT:    mul x4, x4, x9
-; CHECK-NEXT:    mul x6, x6, x13
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: v4i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x8, x9, [sp]
+; CHECK-SD-NEXT:    ldp x11, x12, [sp, #16]
+; CHECK-SD-NEXT:    umulh x10, x0, x8
+; CHECK-SD-NEXT:    umulh x13, x2, x11
+; CHECK-SD-NEXT:    madd x9, x0, x9, x10
+; CHECK-SD-NEXT:    madd x10, x2, x12, x13
+; CHECK-SD-NEXT:    ldp x13, x14, [sp, #48]
+; CHECK-SD-NEXT:    madd x1, x1, x8, x9
+; CHECK-SD-NEXT:    madd x3, x3, x11, x10
+; CHECK-SD-NEXT:    ldp x9, x10, [sp, #32]
+; CHECK-SD-NEXT:    umulh x15, x6, x13
+; CHECK-SD-NEXT:    umulh x12, x4, x9
+; CHECK-SD-NEXT:    mul x0, x0, x8
+; CHECK-SD-NEXT:    madd x10, x4, x10, x12
+; CHECK-SD-NEXT:    madd x12, x6, x14, x15
+; CHECK-SD-NEXT:    madd x5, x5, x9, x10
+; CHECK-SD-NEXT:    madd x7, x7, x13, x12
+; CHECK-SD-NEXT:    mul x2, x2, x11
+; CHECK-SD-NEXT:    mul x4, x4, x9
+; CHECK-SD-NEXT:    mul x6, x6, x13
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: v4i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x9, x10, [sp]
+; CHECK-GI-NEXT:    ldp x15, x16, [sp, #32]
+; CHECK-GI-NEXT:    mul x10, x0, x10
+; CHECK-GI-NEXT:    mul x16, x4, x16
+; CHECK-GI-NEXT:    madd x12, x1, x9, x10
+; CHECK-GI-NEXT:    ldp x10, x13, [sp, #16]
+; CHECK-GI-NEXT:    mul x8, x0, x9
+; CHECK-GI-NEXT:    mul x13, x2, x13
+; CHECK-GI-NEXT:    umulh x11, x0, x9
+; CHECK-GI-NEXT:    mul x9, x2, x10
+; CHECK-GI-NEXT:    umulh x14, x2, x10
+; CHECK-GI-NEXT:    add x1, x12, x11
+; CHECK-GI-NEXT:    madd x13, x3, x10, x13
+; CHECK-GI-NEXT:    mov x2, x9
+; CHECK-GI-NEXT:    mul x10, x4, x15
+; CHECK-GI-NEXT:    umulh x17, x4, x15
+; CHECK-GI-NEXT:    add x3, x13, x14
+; CHECK-GI-NEXT:    madd x15, x5, x15, x16
+; CHECK-GI-NEXT:    ldp x16, x18, [sp, #48]
+; CHECK-GI-NEXT:    mov x4, x10
+; CHECK-GI-NEXT:    mul x18, x6, x18
+; CHECK-GI-NEXT:    umulh x0, x6, x16
+; CHECK-GI-NEXT:    add x5, x15, x17
+; CHECK-GI-NEXT:    madd x18, x7, x16, x18
+; CHECK-GI-NEXT:    mul x6, x6, x16
+; CHECK-GI-NEXT:    add x7, x18, x0
+; CHECK-GI-NEXT:    mov x0, x8
+; CHECK-GI-NEXT:    ret
 entry:
   %s = mul <4 x i128> %d, %e
   ret <4 x i128> %s

diff  --git a/llvm/test/CodeGen/AArch64/sub.ll b/llvm/test/CodeGen/AArch64/sub.ll
index 0f18ed1006fac5..907605494dfbd0 100644
--- a/llvm/test/CodeGen/AArch64/sub.ll
+++ b/llvm/test/CodeGen/AArch64/sub.ll
@@ -1,10 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc -mtriple=aarch64-none-eabi -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
-; RUN: llc -mtriple=aarch64-none-eabi -global-isel -global-isel-abort=2 -verify-machineinstrs %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
-
-; CHECK-GI:       warning: Instruction selection used fallback path for v2i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for v3i128
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for v4i128
+; RUN: llc -mtriple=aarch64-none-eabi -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 define i8 @i8(i8 %a, i8 %b) {
 ; CHECK-LABEL: i8:
@@ -480,21 +476,37 @@ entry:
 }
 
 define <4 x i128> @v4i128(<4 x i128> %d, <4 x i128> %e) {
-; CHECK-LABEL: v4i128:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    ldp x11, x10, [sp, #16]
-; CHECK-NEXT:    ldp x13, x12, [sp, #32]
-; CHECK-NEXT:    subs x0, x0, x8
-; CHECK-NEXT:    sbc x1, x1, x9
-; CHECK-NEXT:    ldp x8, x9, [sp, #48]
-; CHECK-NEXT:    subs x2, x2, x11
-; CHECK-NEXT:    sbc x3, x3, x10
-; CHECK-NEXT:    subs x4, x4, x13
-; CHECK-NEXT:    sbc x5, x5, x12
-; CHECK-NEXT:    subs x6, x6, x8
-; CHECK-NEXT:    sbc x7, x7, x9
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: v4i128:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    ldp x8, x9, [sp]
+; CHECK-SD-NEXT:    ldp x11, x10, [sp, #16]
+; CHECK-SD-NEXT:    ldp x13, x12, [sp, #32]
+; CHECK-SD-NEXT:    subs x0, x0, x8
+; CHECK-SD-NEXT:    sbc x1, x1, x9
+; CHECK-SD-NEXT:    ldp x8, x9, [sp, #48]
+; CHECK-SD-NEXT:    subs x2, x2, x11
+; CHECK-SD-NEXT:    sbc x3, x3, x10
+; CHECK-SD-NEXT:    subs x4, x4, x13
+; CHECK-SD-NEXT:    sbc x5, x5, x12
+; CHECK-SD-NEXT:    subs x6, x6, x8
+; CHECK-SD-NEXT:    sbc x7, x7, x9
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: v4i128:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    ldp x8, x9, [sp]
+; CHECK-GI-NEXT:    ldp x10, x11, [sp, #16]
+; CHECK-GI-NEXT:    ldp x12, x13, [sp, #32]
+; CHECK-GI-NEXT:    subs x0, x0, x8
+; CHECK-GI-NEXT:    sbc x1, x1, x9
+; CHECK-GI-NEXT:    ldp x8, x9, [sp, #48]
+; CHECK-GI-NEXT:    subs x2, x2, x10
+; CHECK-GI-NEXT:    sbc x3, x3, x11
+; CHECK-GI-NEXT:    subs x4, x4, x12
+; CHECK-GI-NEXT:    sbc x5, x5, x13
+; CHECK-GI-NEXT:    subs x6, x6, x8
+; CHECK-GI-NEXT:    sbc x7, x7, x9
+; CHECK-GI-NEXT:    ret
 entry:
   %s = sub <4 x i128> %d, %e
   ret <4 x i128> %s


        


More information about the llvm-commits mailing list