Rename VBROADCASTSDrm into VBROADCASTSDYrm to match the naming convention.
Allow the folding of vbroadcastRR to vbroadcastRM, where the memory operand is a spill slot.
PR12782.
Together with Michael Kuperstein <michael.m.kuperstein@intel.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@160230 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp
index 53d5ee3..9f5f66e 100644
--- a/lib/Target/X86/X86InstrInfo.cpp
+++ b/lib/Target/X86/X86InstrInfo.cpp
@@ -539,6 +539,8 @@
{ X86::VSQRTPSr_Int, X86::VSQRTPSm_Int, TB_ALIGN_16 },
{ X86::VUCOMISDrr, X86::VUCOMISDrm, 0 },
{ X86::VUCOMISSrr, X86::VUCOMISSrm, 0 },
+ { X86::VBROADCASTSSrr, X86::VBROADCASTSSrm, TB_NO_REVERSE },
+
// AVX 256-bit foldable instructions
{ X86::VMOVAPDYrr, X86::VMOVAPDYrm, TB_ALIGN_32 },
{ X86::VMOVAPSYrr, X86::VMOVAPSYrm, TB_ALIGN_32 },
@@ -547,6 +549,7 @@
{ X86::VMOVUPSYrr, X86::VMOVUPSYrm, 0 },
{ X86::VPERMILPDYri, X86::VPERMILPDYmi, TB_ALIGN_32 },
{ X86::VPERMILPSYri, X86::VPERMILPSYmi, TB_ALIGN_32 },
+
// AVX2 foldable instructions
{ X86::VPABSBrr256, X86::VPABSBrm256, TB_ALIGN_32 },
{ X86::VPABSDrr256, X86::VPABSDrm256, TB_ALIGN_32 },
@@ -562,6 +565,8 @@
{ X86::VSQRTPDYr_Int, X86::VSQRTPDYm_Int, TB_ALIGN_32 },
{ X86::VSQRTPSYr, X86::VSQRTPSYm, TB_ALIGN_32 },
{ X86::VSQRTPSYr_Int, X86::VSQRTPSYm_Int, TB_ALIGN_32 },
+ { X86::VBROADCASTSSYrr, X86::VBROADCASTSSYrm, TB_NO_REVERSE },
+ { X86::VBROADCASTSDYrr, X86::VBROADCASTSDYrm, TB_NO_REVERSE },
};
for (unsigned i = 0, e = array_lengthof(OpTbl1); i != e; ++i) {
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 5a224af..e4caace 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -7274,7 +7274,7 @@
int_x86_avx_vbroadcast_ss_256>;
}
let ExeDomain = SSEPackedDouble in
-def VBROADCASTSDrm : avx_broadcast<0x19, "vbroadcastsd", VR256, f64mem,
+def VBROADCASTSDYrm : avx_broadcast<0x19, "vbroadcastsd", VR256, f64mem,
int_x86_avx_vbroadcast_sd_256>;
def VBROADCASTF128 : avx_broadcast<0x1A, "vbroadcastf128", VR256, f128mem,
int_x86_avx_vbroadcastf128_pd_256>;
@@ -7753,11 +7753,11 @@
def : Pat<(v8i32 (X86VBroadcast (loadi32 addr:$src))),
(VBROADCASTSSYrm addr:$src)>;
def : Pat<(v4i64 (X86VBroadcast (loadi64 addr:$src))),
- (VBROADCASTSDrm addr:$src)>;
+ (VBROADCASTSDYrm addr:$src)>;
def : Pat<(v8f32 (X86VBroadcast (loadf32 addr:$src))),
(VBROADCASTSSYrm addr:$src)>;
def : Pat<(v4f64 (X86VBroadcast (loadf64 addr:$src))),
- (VBROADCASTSDrm addr:$src)>;
+ (VBROADCASTSDYrm addr:$src)>;
def : Pat<(v4f32 (X86VBroadcast (loadf32 addr:$src))),
(VBROADCASTSSrm addr:$src)>;
def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),