[X86][SSE] Fix load folding for (V)CVTDQ2PD

This only requires a 64-bit memory source, not the whole 128-bits. But the 128-bit case is still supported via X86InstrInfo::foldMemoryOperandImpl

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@297523 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 823467d..f128128 100644 (file)
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -2165,7 +2165,7 @@ let hasSideEffects = 0, mayLoad = 1 in
 def VCVTDQ2PDrm  : S2SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
                         "vcvtdq2pd\t{$src, $dst|$dst, $src}",
                         [(set VR128:$dst,
-                          (v2f64 (X86VSintToFP (bc_v4i32 (loadv2i64 addr:$src)))))]>,
+                          (v2f64 (X86VSintToFP (bc_v4i32 (v2i64 (X86vzload addr:$src))))))]>,
                         VEX, Sched<[WriteCvtI2FLd]>, VEX_WIG;
 def VCVTDQ2PDrr  : S2SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                         "vcvtdq2pd\t{$src, $dst|$dst, $src}",
@@ -2188,7 +2188,7 @@ let hasSideEffects = 0, mayLoad = 1 in
 def CVTDQ2PDrm  : S2SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
                        "cvtdq2pd\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst,
-                         (v2f64 (X86VSintToFP (bc_v4i32 (loadv2i64 addr:$src)))))],
+                         (v2f64 (X86VSintToFP (bc_v4i32 (v2i64 (X86vzload addr:$src))))))],
                        IIC_SSE_CVT_PD_RR>, Sched<[WriteCvtI2FLd]>;
 def CVTDQ2PDrr  : S2SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                        "cvtdq2pd\t{$src, $dst|$dst, $src}",

diff --git a/test/CodeGen/X86/known-signbits-vector.ll b/test/CodeGen/X86/known-signbits-vector.ll
index 3a57c06..12dfdc0 100644 (file)
--- a/test/CodeGen/X86/known-signbits-vector.ll
+++ b/test/CodeGen/X86/known-signbits-vector.ll
@@ -5,8 +5,7 @@
 define <2 x double> @signbits_sext_v2i64_sitofp_v2f64(i32 %a0, i32 %a1) nounwind {
 ; X32-LABEL: signbits_sext_v2i64_sitofp_v2f64:
 ; X32:       # BB#0:
-; X32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; X32-NEXT:    vcvtdq2pd %xmm0, %xmm0
+; X32-NEXT:    vcvtdq2pd {{[0-9]+}}(%esp), %xmm0
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: signbits_sext_v2i64_sitofp_v2f64:

diff --git a/test/CodeGen/X86/mmx-cvt.ll b/test/CodeGen/X86/mmx-cvt.ll
index 884c4a8..71daf7b 100644 (file)
--- a/test/CodeGen/X86/mmx-cvt.ll
+++ b/test/CodeGen/X86/mmx-cvt.ll
@@ -262,8 +262,7 @@ define <2 x double> @sitofp_v2i32_v2f64(<1 x i64>*) nounwind {
 ; X86-NEXT:    movq (%eax), %mm0
 ; X86-NEXT:    paddd %mm0, %mm0
 ; X86-NEXT:    movq %mm0, (%esp)
-; X86-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
-; X86-NEXT:    cvtdq2pd %xmm0, %xmm0
+; X86-NEXT:    cvtdq2pd (%esp), %xmm0
 ; X86-NEXT:    movl %ebp, %esp
 ; X86-NEXT:    popl %ebp
 ; X86-NEXT:    retl