[llvm] r295492 - [X86][SSE] Add (V)MOVD folding pattern with zextloadi64i32 load node.

Simon Pilgrim via llvm-commits llvm-commits at lists.llvm.org
Fri Feb 17 12:43:34 PST 2017


Author: rksimon
Date: Fri Feb 17 14:43:32 2017
New Revision: 295492

URL: http://llvm.org/viewvc/llvm-project?rev=295492&view=rev
Log:
[X86][SSE] Add (V)MOVD folding pattern with zextloadi64i32 load node.

Fixes PRPR31309

Modified:
    llvm/trunk/lib/Target/X86/X86InstrAVX512.td
    llvm/trunk/lib/Target/X86/X86InstrSSE.td
    llvm/trunk/test/CodeGen/X86/merge-consecutive-loads-128.ll

Modified: llvm/trunk/lib/Target/X86/X86InstrAVX512.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrAVX512.td?rev=295492&r1=295491&r2=295492&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrAVX512.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrAVX512.td Fri Feb 17 14:43:32 2017
@@ -3552,6 +3552,8 @@ let Predicates = [HasAVX512] in {
   }
   // AVX 128-bit movd/movq instruction write zeros in the high 128-bit part.
   let AddedComplexity = 20 in {
+    def : Pat<(v2i64 (X86vzmovl (v2i64 (scalar_to_vector (zextloadi64i32 addr:$src))))),
+              (VMOVDI2PDIZrm addr:$src)>;
     def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector (loadi32 addr:$src))))),
               (VMOVDI2PDIZrm addr:$src)>;
     def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),

Modified: llvm/trunk/lib/Target/X86/X86InstrSSE.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrSSE.td?rev=295492&r1=295491&r2=295492&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrSSE.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrSSE.td Fri Feb 17 14:43:32 2017
@@ -4822,6 +4822,8 @@ let Predicates = [UseAVX] in {
   // AVX 128-bit movd/movq instructions write zeros in the high 128-bit part.
   // These instructions also write zeros in the high part of a 256-bit register.
   let AddedComplexity = 20 in {
+    def : Pat<(v2i64 (X86vzmovl (v2i64 (scalar_to_vector (zextloadi64i32 addr:$src))))),
+              (VMOVDI2PDIrm addr:$src)>;
     def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector (loadi32 addr:$src))))),
               (VMOVDI2PDIrm addr:$src)>;
     def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),
@@ -4851,6 +4853,8 @@ let Predicates = [UseSSE2] in {
               (MOV64toPQIrr GR64:$src)>;
   }
   let AddedComplexity = 20 in {
+    def : Pat<(v2i64 (X86vzmovl (v2i64 (scalar_to_vector (zextloadi64i32 addr:$src))))),
+              (MOVDI2PDIrm addr:$src)>;
     def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector (loadi32 addr:$src))))),
               (MOVDI2PDIrm addr:$src)>;
     def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),

Modified: llvm/trunk/test/CodeGen/X86/merge-consecutive-loads-128.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/merge-consecutive-loads-128.ll?rev=295492&r1=295491&r2=295492&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/merge-consecutive-loads-128.ll (original)
+++ llvm/trunk/test/CodeGen/X86/merge-consecutive-loads-128.ll Fri Feb 17 14:43:32 2017
@@ -1137,18 +1137,16 @@ define <4 x float> @merge_4f32_f32_X0YY(
 ; Extension tests.
 ;
 
-; FIXME: PR31309
+; PR31309
 define <4 x i32> @load_i32_zext_i128_v4i32(i32* %ptr) {
 ; SSE-LABEL: load_i32_zext_i128_v4i32:
 ; SSE:       # BB#0:
-; SSE-NEXT:    movl (%rdi), %eax
-; SSE-NEXT:    movd %rax, %xmm0
+; SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_zext_i128_v4i32:
 ; AVX:       # BB#0:
-; AVX-NEXT:    movl (%rdi), %eax
-; AVX-NEXT:    vmovq %rax, %xmm0
+; AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; AVX-NEXT:    retq
 ;
 ; X32-SSE1-LABEL: load_i32_zext_i128_v4i32:




More information about the llvm-commits mailing list