[X86] Remove some InstRWs for plain store instructions on Sandy Bridge.

We were forcing the latency of these instructions to 5 cycles, but every other scheduler model had them as 1 cycle. I'm sure I didn't get everything, but this gets a big portion.

llvm-svn: 329339

diff --git a/llvm/lib/Target/X86/X86SchedBroadwell.td b/llvm/lib/Target/X86/X86SchedBroadwell.td
index d3fb25d..ea4dda7 100755 (executable)
--- a/llvm/lib/Target/X86/X86SchedBroadwell.td
+++ b/llvm/lib/Target/X86/X86SchedBroadwell.td
@@ -663,9 +663,6 @@ def: InstRW<[BWWriteResGroup10], (instregex "FBSTPm",
                                             "MMX_MOVD64mr",
                                             "MMX_MOVNTQmr",
                                             "MMX_MOVQ64mr",
-                                            "MOV(16|32|64)mr",
-                                            "MOV8mi",
-                                            "MOV8mr",
                                             "MOVNTI_64mr",
                                             "MOVNTImr",
                                             "ST_FP32m",

diff --git a/llvm/lib/Target/X86/X86SchedHaswell.td b/llvm/lib/Target/X86/X86SchedHaswell.td
index 8007f0c..2858209 100644 (file)
--- a/llvm/lib/Target/X86/X86SchedHaswell.td
+++ b/llvm/lib/Target/X86/X86SchedHaswell.td
@@ -633,9 +633,6 @@ def: InstRW<[HWWriteResGroup1], (instregex "FBSTPm",
                                            "MMX_MOVD64mr",
                                            "MMX_MOVNTQmr",
                                            "MMX_MOVQ64mr",
-                                           "MOV(16|32|64)mr",
-                                           "MOV8mi",
-                                           "MOV8mr",
                                            "MOVNTI_64mr",
                                            "MOVNTImr",
                                            "ST_FP32m",

diff --git a/llvm/lib/Target/X86/X86SchedSandyBridge.td b/llvm/lib/Target/X86/X86SchedSandyBridge.td
index 1252835..0adf9ea 100644 (file)
--- a/llvm/lib/Target/X86/X86SchedSandyBridge.td
+++ b/llvm/lib/Target/X86/X86SchedSandyBridge.td
@@ -925,30 +925,8 @@ def SBWriteResGroup33 : SchedWriteRes<[SBPort4,SBPort23]> {
   let NumMicroOps = 2;
   let ResourceCycles = [1,1];
 }
-def: InstRW<[SBWriteResGroup33], (instregex "MOV(8|16|32|64)mr",
-                                            "MOVNTI_64mr",
-                                            "MOVNTImr",
-                                            "PUSH64i8",
-                                            "PUSH(16|32|64)r",
-                                            "VEXTRACTF128mr",
-                                            "(V?)MOVAPD(Y?)mr",
-                                            "(V?)MOVAPS(Y?)mr",
-                                            "(V?)MOVDQA(Y?)mr",
-                                            "(V?)MOVDQU(Y?)mr",
-                                            "(V?)MOVHPDmr",
-                                            "(V?)MOVHPSmr",
-                                            "(V?)MOVLPDmr",
-                                            "(V?)MOVLPSmr",
-                                            "(V?)MOVNTDQ(Y?)mr",
-                                            "(V?)MOVNTPD(Y?)mr",
-                                            "(V?)MOVNTPS(Y?)mr",
-                                            "(V?)MOVPDI2DImr",
-                                            "(V?)MOVPQI2QImr",
-                                            "(V?)MOVPQIto64mr",
-                                            "(V?)MOVSDmr",
-                                            "(V?)MOVSSmr",
-                                            "(V?)MOVUPD(Y?)mr",
-                                            "(V?)MOVUPS(Y?)mr")>;
+def: InstRW<[SBWriteResGroup33], (instregex "PUSH64i8",
+                                            "PUSH(16|32|64)r")>;
 
 def SBWriteResGroup34 : SchedWriteRes<[SBPort0,SBPort15]> {
   let Latency = 7;
@@ -998,7 +976,7 @@ def: InstRW<[SBWriteResGroup37], (instregex "VMASKMOVPD(Y?)mr",
                                             "VMASKMOVPS(Y?)mr")>;
 
 def SBWriteResGroup38 : SchedWriteRes<[SBPort4,SBPort23,SBPort05]> {
-  let Latency = 5;
+  let Latency = 2;
   let NumMicroOps = 3;
   let ResourceCycles = [1,1,1];
 }
@@ -1018,8 +996,7 @@ def SBWriteResGroup40 : SchedWriteRes<[SBPort4,SBPort23,SBPort015]> {
   let NumMicroOps = 3;
   let ResourceCycles = [1,1,1];
 }
-def: InstRW<[SBWriteResGroup40], (instregex "MOV8mi",
-                                            "STOSB",
+def: InstRW<[SBWriteResGroup40], (instregex "STOSB",
                                             "STOSL",
                                             "STOSQ",
                                             "STOSW")>;
@@ -1039,7 +1016,7 @@ def SBWriteResGroup42 : SchedWriteRes<[SBPort05,SBPort015]> {
 def: InstRW<[SBWriteResGroup42], (instregex "CMPXCHG(8|16|32|64)rr")>;
 
 def SBWriteResGroup43 : SchedWriteRes<[SBPort4,SBPort23,SBPort05]> {
-  let Latency = 5;
+  let Latency = 3;
   let NumMicroOps = 4;
   let ResourceCycles = [1,1,2];
 }

diff --git a/llvm/lib/Target/X86/X86SchedSkylakeClient.td b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
index 407b1cc..3c26858 100644 (file)
--- a/llvm/lib/Target/X86/X86SchedSkylakeClient.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
@@ -667,8 +667,6 @@ def: InstRW<[SKLWriteResGroup11], (instregex "FBSTPm",
                                              "MMX_MOVD64mr",
                                              "MMX_MOVNTQmr",
                                              "MMX_MOVQ64mr",
-                                             "MOV(8|16|32|64)mr",
-                                             "MOV8mi",
                                              "MOVNTI_64mr",
                                              "MOVNTImr",
                                              "ST_FP32m",

diff --git a/llvm/lib/Target/X86/X86SchedSkylakeServer.td b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
index aab1b39..9d8a4c5 100755 (executable)
--- a/llvm/lib/Target/X86/X86SchedSkylakeServer.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
@@ -1331,9 +1331,6 @@ def: InstRW<[SKXWriteResGroup11], (instregex "FBSTPm",
                                              "MMX_MOVD64mr",
                                              "MMX_MOVNTQmr",
                                              "MMX_MOVQ64mr",
-                                             "MOV(16|32|64)mr",
-                                             "MOV8mi",
-                                             "MOV8mr",
                                              "MOVAPDmr",
                                              "MOVAPSmr",
                                              "MOVDQAmr",

diff --git a/llvm/test/CodeGen/X86/avx-schedule.ll b/llvm/test/CodeGen/X86/avx-schedule.ll
index b11a9bb..eff1d6c 100644 (file)
--- a/llvm/test/CodeGen/X86/avx-schedule.ll
+++ b/llvm/test/CodeGen/X86/avx-schedule.ll
@@ -1662,14 +1662,14 @@ define <4 x float> @test_extractf128(<8 x float> %a0, <8 x float> %a1, <4 x floa
 ; GENERIC-LABEL: test_extractf128:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT:    vextractf128 $1, %ymm1, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    vzeroupper # sched: [100:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_extractf128:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT:    vextractf128 $1, %ymm1, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    vzeroupper # sched: [100:0.33]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
@@ -2526,14 +2526,14 @@ define <4 x double> @test_movapd(<4 x double> *%a0, <4 x double> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    vmovapd (%rdi), %ymm0 # sched: [7:0.50]
 ; GENERIC-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovapd %ymm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovapd %ymm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movapd:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovapd (%rdi), %ymm0 # sched: [7:0.50]
 ; SANDY-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovapd %ymm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovapd %ymm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movapd:
@@ -2588,14 +2588,14 @@ define <8 x float> @test_movaps(<8 x float> *%a0, <8 x float> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    vmovaps (%rdi), %ymm0 # sched: [7:0.50]
 ; GENERIC-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovaps %ymm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovaps %ymm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movaps:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovaps (%rdi), %ymm0 # sched: [7:0.50]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovaps %ymm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovaps %ymm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movaps:
@@ -2816,7 +2816,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
 ; GENERIC-LABEL: test_movntdq:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    #APP
-; GENERIC-NEXT:    vmovntdq %ymm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    #NO_APP
 ; GENERIC-NEXT:    vzeroupper # sched: [100:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
@@ -2824,7 +2824,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
 ; SANDY-LABEL: test_movntdq:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    #APP
-; SANDY-NEXT:    vmovntdq %ymm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    #NO_APP
 ; SANDY-NEXT:    vzeroupper # sched: [100:0.33]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
@@ -2883,13 +2883,13 @@ define <4 x double> @test_movntpd(<4 x double> %a0, <4 x double> *%a1) {
 ; GENERIC-LABEL: test_movntpd:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovntpd %ymm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movntpd:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovntpd %ymm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movntpd:
@@ -2936,13 +2936,13 @@ define <8 x float> @test_movntps(<8 x float> %a0, <8 x float> *%a1) {
 ; GENERIC-LABEL: test_movntps:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovntps %ymm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovntps %ymm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movntps:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovntps %ymm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovntps %ymm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movntps:
@@ -3116,7 +3116,7 @@ define <4 x double> @test_movupd(<4 x double> *%a0, <4 x double> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    vmovupd (%rdi), %ymm0 # sched: [7:0.50]
 ; GENERIC-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovupd %ymm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovupd %ymm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movupd:
@@ -3124,8 +3124,8 @@ define <4 x double> @test_movupd(<4 x double> *%a0, <4 x double> *%a1) {
 ; SANDY-NEXT:    vmovups (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
 ; SANDY-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    vmovupd %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [1:1.00]
+; SANDY-NEXT:    vmovupd %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movupd:
@@ -3180,7 +3180,7 @@ define <8 x float> @test_movups(<8 x float> *%a0, <8 x float> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    vmovups (%rdi), %ymm0 # sched: [7:0.50]
 ; GENERIC-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovups %ymm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovups %ymm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movups:
@@ -3188,8 +3188,8 @@ define <8 x float> @test_movups(<8 x float> *%a0, <8 x float> *%a1) {
 ; SANDY-NEXT:    vmovups (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    vmovups %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [1:1.00]
+; SANDY-NEXT:    vmovups %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movups:

diff --git a/llvm/test/CodeGen/X86/avx512-schedule.ll b/llvm/test/CodeGen/X86/avx512-schedule.ll
index b26b528..41ea3c3 100755 (executable)
--- a/llvm/test/CodeGen/X86/avx512-schedule.ll
+++ b/llvm/test/CodeGen/X86/avx512-schedule.ll
@@ -1937,7 +1937,7 @@ define void @f32tof64_loadstore() {
 ; GENERIC:       # %bb.0: # %entry
 ; GENERIC-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [6:0.50]
 ; GENERIC-NEXT:    vcvtss2sd %xmm0, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT:    vmovsd %xmm0, -{{[0-9]+}}(%rsp) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovsd %xmm0, -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: f32tof64_loadstore:
@@ -1960,7 +1960,7 @@ define void @f64tof32_loadstore() nounwind uwtable {
 ; GENERIC:       # %bb.0: # %entry
 ; GENERIC-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero sched: [6:0.50]
 ; GENERIC-NEXT:    vcvtsd2ss %xmm0, %xmm0, %xmm0 # sched: [4:1.00]
-; GENERIC-NEXT:    vmovss %xmm0, -{{[0-9]+}}(%rsp) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovss %xmm0, -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: f64tof32_loadstore:
@@ -5934,7 +5934,7 @@ define <4 x i32> @mov_test4(i32* %x) {
 define void @mov_test5(float %x, float* %y) {
 ; GENERIC-LABEL: mov_test5:
 ; GENERIC:       # %bb.0:
-; GENERIC-NEXT:    vmovss %xmm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovss %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: mov_test5:
@@ -5948,7 +5948,7 @@ define void @mov_test5(float %x, float* %y) {
 define void @mov_test6(double %x, double* %y) {
 ; GENERIC-LABEL: mov_test6:
 ; GENERIC:       # %bb.0:
-; GENERIC-NEXT:    vmovsd %xmm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovsd %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: mov_test6:
@@ -6943,8 +6943,8 @@ define i8 @zext_test3(<16 x i32> %a, <16 x i32> %b) {
 define i8 @conv1(<8 x i1>* %R) {
 ; GENERIC-LABEL: conv1:
 ; GENERIC:       # %bb.0: # %entry
-; GENERIC-NEXT:    movb $-1, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT:    movb $-2, -{{[0-9]+}}(%rsp) # sched: [5:1.00]
+; GENERIC-NEXT:    movb $-1, (%rdi) # sched: [1:1.00]
+; GENERIC-NEXT:    movb $-2, -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; GENERIC-NEXT:    movb $-2, %al # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
@@ -7512,7 +7512,7 @@ define void @f1(i32 %c) {
 ; GENERIC:       # %bb.0: # %entry
 ; GENERIC-NEXT:    movzbl {{.*}}(%rip), %edi # sched: [5:0.50]
 ; GENERIC-NEXT:    xorl $1, %edi # sched: [1:0.33]
-; GENERIC-NEXT:    movb %dil, {{.*}}(%rip) # sched: [5:1.00]
+; GENERIC-NEXT:    movb %dil, {{.*}}(%rip) # sched: [1:1.00]
 ; GENERIC-NEXT:    jmp f2 # TAILCALL
 ;
 ; SKX-LABEL: f1:
@@ -7536,7 +7536,7 @@ define void @store_i16_i1(i16 %x, i1 *%y) {
 ; GENERIC-LABEL: store_i16_i1:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    andl $1, %edi # sched: [1:0.33]
-; GENERIC-NEXT:    movb %dil, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movb %dil, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: store_i16_i1:
@@ -7553,7 +7553,7 @@ define void @store_i8_i1(i8 %x, i1 *%y) {
 ; GENERIC-LABEL: store_i8_i1:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    andl $1, %edi # sched: [1:0.33]
-; GENERIC-NEXT:    movb %dil, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movb %dil, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: store_i8_i1:
@@ -8698,7 +8698,7 @@ define <16 x float> @broadcast_ss_spill(float %x) {
 ; GENERIC-NEXT:    subq $24, %rsp # sched: [1:0.33]
 ; GENERIC-NEXT:    .cfi_def_cfa_offset 32
 ; GENERIC-NEXT:    vaddss %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill sched: [5:1.00]
+; GENERIC-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill sched: [1:1.00]
 ; GENERIC-NEXT:    callq func_f32
 ; GENERIC-NEXT:    vbroadcastss (%rsp), %zmm0 # 16-byte Folded Reload sched: [5:1.00]
 ; GENERIC-NEXT:    addq $24, %rsp # sched: [1:0.33]
@@ -8728,7 +8728,7 @@ define <8 x double> @broadcast_sd_spill(double %x) {
 ; GENERIC-NEXT:    subq $24, %rsp # sched: [1:0.33]
 ; GENERIC-NEXT:    .cfi_def_cfa_offset 32
 ; GENERIC-NEXT:    vaddsd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovapd %xmm0, (%rsp) # 16-byte Spill sched: [5:1.00]
+; GENERIC-NEXT:    vmovapd %xmm0, (%rsp) # 16-byte Spill sched: [1:1.00]
 ; GENERIC-NEXT:    callq func_f64
 ; GENERIC-NEXT:    vbroadcastsd (%rsp), %zmm0 # 16-byte Folded Reload sched: [5:1.00]
 ; GENERIC-NEXT:    addq $24, %rsp # sched: [1:0.33]

diff --git a/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll b/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll
index a2aa23b..3752ebd 100644 (file)
--- a/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll
+++ b/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll
@@ -15,23 +15,22 @@ define void @test_extractelement_legalization_storereuse(<4 x i32> %a, i32* noca
 ; CHECK-NEXT:    pushl %esi
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; CHECK-NEXT:    paddd (%ecx), %xmm0
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; CHECK-NEXT:    movdqa %xmm0, (%ecx)
-; CHECK-NEXT:    movl (%ecx), %esi
-; CHECK-NEXT:    movl 4(%ecx), %edi
-; CHECK-NEXT:    shll $4, %edx
-; CHECK-NEXT:    movl 8(%ecx), %ebx
-; CHECK-NEXT:    movl 12(%ecx), %ecx
-; CHECK-NEXT:    movl %esi, 12(%eax,%edx)
-; CHECK-NEXT:    movl %edi, (%eax,%edx)
-; CHECK-NEXT:    movl %ebx, 8(%eax,%edx)
-; CHECK-NEXT:    movl %ecx, 4(%eax,%edx)
+; CHECK-NEXT:    paddd (%edx), %xmm0
+; CHECK-NEXT:    movdqa %xmm0, (%edx)
+; CHECK-NEXT:    movl (%edx), %esi
+; CHECK-NEXT:    movl 4(%edx), %edi
+; CHECK-NEXT:    shll $4, %ecx
+; CHECK-NEXT:    movl 8(%edx), %ebx
+; CHECK-NEXT:    movl 12(%edx), %edx
+; CHECK-NEXT:    movl %esi, 12(%eax,%ecx)
+; CHECK-NEXT:    movl %edi, (%eax,%ecx)
+; CHECK-NEXT:    movl %ebx, 8(%eax,%ecx)
+; CHECK-NEXT:    movl %edx, 4(%eax,%ecx)
 ; CHECK-NEXT:    popl %esi
 ; CHECK-NEXT:    popl %edi
 ; CHECK-NEXT:    popl %ebx
 ; CHECK-NEXT:    retl
-; CHECK-NEXT:    ## -- End function
 entry:
   %0 = bitcast i32* %y to <4 x i32>*
   %1 = load <4 x i32>, <4 x i32>* %0, align 16

diff --git a/llvm/test/CodeGen/X86/fp128-i128.ll b/llvm/test/CodeGen/X86/fp128-i128.ll
index f611737..5c28535 100644 (file)
--- a/llvm/test/CodeGen/X86/fp128-i128.ll
+++ b/llvm/test/CodeGen/X86/fp128-i128.ll
@@ -50,8 +50,8 @@ define void @TestUnionLD1(fp128 %s, i64 %n) #0 {
 ; CHECK-NEXT:    andq %rdi, %rcx
 ; CHECK-NEXT:    movabsq $-281474976710656, %rdx # imm = 0xFFFF000000000000
 ; CHECK-NEXT:    andq -{{[0-9]+}}(%rsp), %rdx
-; CHECK-NEXT:    movq %rax, -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    orq %rcx, %rdx
+; CHECK-NEXT:    movq %rax, -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
 ; CHECK-NEXT:    jmp foo # TAILCALL
@@ -105,11 +105,11 @@ define fp128 @TestI128_1(fp128 %x) #0 {
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    subq $40, %rsp
 ; CHECK-NEXT:    movaps %xmm0, {{[0-9]+}}(%rsp)
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; CHECK-NEXT:    movabsq $9223372036854775807, %rcx # imm = 0x7FFFFFFFFFFFFFFF
-; CHECK-NEXT:    andq {{[0-9]+}}(%rsp), %rcx
-; CHECK-NEXT:    movq %rcx, {{[0-9]+}}(%rsp)
-; CHECK-NEXT:    movq %rax, (%rsp)
+; CHECK-NEXT:    movabsq $9223372036854775807, %rax # imm = 0x7FFFFFFFFFFFFFFF
+; CHECK-NEXT:    andq {{[0-9]+}}(%rsp), %rax
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rcx
+; CHECK-NEXT:    movq %rax, {{[0-9]+}}(%rsp)
+; CHECK-NEXT:    movq %rcx, (%rsp)
 ; CHECK-NEXT:    movaps (%rsp), %xmm0
 ; CHECK-NEXT:    movaps {{.*}}(%rip), %xmm1
 ; CHECK-NEXT:    callq __lttf2
@@ -336,11 +336,11 @@ define void @TestCopySign({ fp128, fp128 }* noalias nocapture sret %agg.result,
 ; CHECK-NEXT:    movq %rdi, %rbx
 ; CHECK-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm0
 ; CHECK-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm1
-; CHECK-NEXT:    movaps %xmm1, {{[0-9]+}}(%rsp) # 16-byte Spill
+; CHECK-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
 ; CHECK-NEXT:    callq __gttf2
 ; CHECK-NEXT:    movl %eax, %ebp
-; CHECK-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
+; CHECK-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    movaps %xmm0, %xmm1
 ; CHECK-NEXT:    callq __subtf3
 ; CHECK-NEXT:    testl %ebp, %ebp
@@ -355,8 +355,8 @@ define void @TestCopySign({ fp128, fp128 }* noalias nocapture sret %agg.result,
 ; CHECK-NEXT:    movaps (%rsp), %xmm2 # 16-byte Reload
 ; CHECK-NEXT:  .LBB10_3: # %cleanup
 ; CHECK-NEXT:    movaps {{.*}}(%rip), %xmm1
-; CHECK-NEXT:    andps {{[0-9]+}}(%rsp), %xmm1 # 16-byte Folded Reload
 ; CHECK-NEXT:    andps {{.*}}(%rip), %xmm0
+; CHECK-NEXT:    andps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; CHECK-NEXT:    orps %xmm1, %xmm0
 ; CHECK-NEXT:    movaps %xmm2, (%rbx)
 ; CHECK-NEXT:    movaps %xmm0, 16(%rbx)

diff --git a/llvm/test/CodeGen/X86/memcpy-2.ll b/llvm/test/CodeGen/X86/memcpy-2.ll
index 6deeaa6..a6d285b 100644 (file)
--- a/llvm/test/CodeGen/X86/memcpy-2.ll
+++ b/llvm/test/CodeGen/X86/memcpy-2.ll
@@ -27,8 +27,8 @@ entry:
 
 ; SSE1-LABEL: t1:
 ; SSE1: movaps _.str, %xmm0
-; SSE1: movb $0, 24(%esp)
 ; SSE1: movaps %xmm0
+; SSE1: movb $0, 24(%esp)
 ; SSE1: movl $0, 20(%esp)
 ; SSE1: movl $0, 16(%esp)
 

diff --git a/llvm/test/CodeGen/X86/misched-matrix.ll b/llvm/test/CodeGen/X86/misched-matrix.ll
index 495ca71..7ab995c 100644 (file)
--- a/llvm/test/CodeGen/X86/misched-matrix.ll
+++ b/llvm/test/CodeGen/X86/misched-matrix.ll
@@ -17,9 +17,9 @@
 ;
 ; TOPDOWN-LABEL: %for.body
 ; TOPDOWN: movl %{{.*}}, (
-; TOPDOWN-NOT: imull {{[0-9]*}}(
+; TOPDOWN: imull {{[0-9]*}}(
 ; TOPDOWN: movl %{{.*}}, 4(
-; TOPDOWN-NOT: imull {{[0-9]*}}(
+; TOPDOWN: imull {{[0-9]*}}(
 ; TOPDOWN: movl %{{.*}}, 8(
 ; TOPDOWN: movl %{{.*}}, 12(
 ; TOPDOWN-LABEL: %for.end

diff --git a/llvm/test/CodeGen/X86/mmx-schedule.ll b/llvm/test/CodeGen/X86/mmx-schedule.ll
index a68db2d..2309468 100644 (file)
--- a/llvm/test/CodeGen/X86/mmx-schedule.ll
+++ b/llvm/test/CodeGen/X86/mmx-schedule.ll
@@ -642,7 +642,7 @@ define i32 @test_movd(x86_mmx %a0, i32 %a1, i32 *%a2) {
 ; GENERIC-NEXT:    paddd %mm2, %mm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    movd %mm2, %ecx # sched: [1:0.33]
 ; GENERIC-NEXT:    movd %mm0, %eax # sched: [1:0.33]
-; GENERIC-NEXT:    movl %ecx, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movl %ecx, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movd:
@@ -675,7 +675,7 @@ define i32 @test_movd(x86_mmx %a0, i32 %a1, i32 *%a2) {
 ; SANDY-NEXT:    paddd %mm2, %mm0 # sched: [3:1.00]
 ; SANDY-NEXT:    movd %mm2, %ecx # sched: [1:0.33]
 ; SANDY-NEXT:    movd %mm0, %eax # sched: [1:0.33]
-; SANDY-NEXT:    movl %ecx, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    movl %ecx, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_movd:

diff --git a/llvm/test/CodeGen/X86/pr31045.ll b/llvm/test/CodeGen/X86/pr31045.ll
index f628363..c87b4b3 100644 (file)
--- a/llvm/test/CodeGen/X86/pr31045.ll
+++ b/llvm/test/CodeGen/X86/pr31045.ll
@@ -21,26 +21,25 @@ define void @_Z1av() local_unnamed_addr #0 {
 ; CHECK-NEXT:    movl struct_obj_3+{{.*}}(%rip), %eax
 ; CHECK-NEXT:    movsbl {{.*}}(%rip), %ecx
 ; CHECK-NEXT:    movzbl {{.*}}(%rip), %edx
-; CHECK-NEXT:    movzbl {{.*}}(%rip), %esi
 ; CHECK-NEXT:    andl $1, %eax
-; CHECK-NEXT:    leal (%rax,%rax), %edi
-; CHECK-NEXT:    subl %ecx, %edi
-; CHECK-NEXT:    subl %edx, %edi
-; CHECK-NEXT:    movl %edi, %ecx
-; CHECK-NEXT:    notl %ecx
-; CHECK-NEXT:    movzbl %cl, %ecx
-; CHECK-NEXT:    movw %cx, struct_obj_12+{{.*}}(%rip)
+; CHECK-NEXT:    leal (%rax,%rax), %esi
+; CHECK-NEXT:    subl %ecx, %esi
+; CHECK-NEXT:    subl %edx, %esi
 ; CHECK-NEXT:    xorl %ecx, %ecx
 ; CHECK-NEXT:    testb %al, %al
 ; CHECK-NEXT:    cmovel %eax, %ecx
+; CHECK-NEXT:    movzbl {{.*}}(%rip), %edx
 ; CHECK-NEXT:    andl struct_obj_8+{{.*}}(%rip), %ecx
 ; CHECK-NEXT:    andl $1, %ecx
 ; CHECK-NEXT:    negl %ecx
-; CHECK-NEXT:    andl %esi, %ecx
+; CHECK-NEXT:    andl %edx, %ecx
 ; CHECK-NEXT:    negl %ecx
 ; CHECK-NEXT:    andl %eax, %ecx
 ; CHECK-NEXT:    negl %ecx
-; CHECK-NEXT:    testl %ecx, %edi
+; CHECK-NEXT:    testl %ecx, %esi
+; CHECK-NEXT:    notl %esi
+; CHECK-NEXT:    movzbl %sil, %eax
+; CHECK-NEXT:    movw %ax, struct_obj_12+{{.*}}(%rip)
 ; CHECK-NEXT:    setne {{.*}}(%rip)
 ; CHECK-NEXT:    retq
 entry:

diff --git a/llvm/test/CodeGen/X86/pr34080.ll b/llvm/test/CodeGen/X86/pr34080.ll
index e0b0974..4eb156b 100644 (file)
--- a/llvm/test/CodeGen/X86/pr34080.ll
+++ b/llvm/test/CodeGen/X86/pr34080.ll
@@ -54,10 +54,10 @@ define void @_Z1fe(x86_fp80 %z) local_unnamed_addr #0 {
 ; SSE2-SCHEDULE-NEXT:    movq %rsp, %rbp
 ; SSE2-SCHEDULE-NEXT:    .cfi_def_cfa_register %rbp
 ; SSE2-SCHEDULE-NEXT:    fnstcw -4(%rbp)
-; SSE2-SCHEDULE-NEXT:    fldt 16(%rbp)
 ; SSE2-SCHEDULE-NEXT:    movzwl -4(%rbp), %eax
 ; SSE2-SCHEDULE-NEXT:    movw $3199, -4(%rbp) ## imm = 0xC7F
 ; SSE2-SCHEDULE-NEXT:    fldcw -4(%rbp)
+; SSE2-SCHEDULE-NEXT:    fldt 16(%rbp)
 ; SSE2-SCHEDULE-NEXT:    movw %ax, -4(%rbp)
 ; SSE2-SCHEDULE-NEXT:    fistl -8(%rbp)
 ; SSE2-SCHEDULE-NEXT:    fldcw -4(%rbp)
@@ -65,12 +65,12 @@ define void @_Z1fe(x86_fp80 %z) local_unnamed_addr #0 {
 ; SSE2-SCHEDULE-NEXT:    movsd %xmm0, -64(%rbp)
 ; SSE2-SCHEDULE-NEXT:    movsd %xmm0, -32(%rbp)
 ; SSE2-SCHEDULE-NEXT:    fsubl -32(%rbp)
-; SSE2-SCHEDULE-NEXT:    flds {{.*}}(%rip)
 ; SSE2-SCHEDULE-NEXT:    fnstcw -2(%rbp)
-; SSE2-SCHEDULE-NEXT:    fmul %st(0), %st(1)
+; SSE2-SCHEDULE-NEXT:    flds {{.*}}(%rip)
 ; SSE2-SCHEDULE-NEXT:    movzwl -2(%rbp), %eax
 ; SSE2-SCHEDULE-NEXT:    movw $3199, -2(%rbp) ## imm = 0xC7F
 ; SSE2-SCHEDULE-NEXT:    fldcw -2(%rbp)
+; SSE2-SCHEDULE-NEXT:    fmul %st(0), %st(1)
 ; SSE2-SCHEDULE-NEXT:    movw %ax, -2(%rbp)
 ; SSE2-SCHEDULE-NEXT:    fxch %st(1)
 ; SSE2-SCHEDULE-NEXT:    fistl -12(%rbp)

diff --git a/llvm/test/CodeGen/X86/schedule-x86-64-shld.ll b/llvm/test/CodeGen/X86/schedule-x86-64-shld.ll
index b8bf0fa..3bc4362 100644 (file)
--- a/llvm/test/CodeGen/X86/schedule-x86-64-shld.ll
+++ b/llvm/test/CodeGen/X86/schedule-x86-64-shld.ll
@@ -403,7 +403,7 @@ define void @lshift_mem_b(i64 %b) nounwind readnone {
 ; GENERIC:       # %bb.0: # %entry
 ; GENERIC-NEXT:    movq {{.*}}(%rip), %rax # sched: [5:0.50]
 ; GENERIC-NEXT:    shrdq $54, %rdi, %rax # sched: [2:0.67]
-; GENERIC-NEXT:    movq %rax, {{.*}}(%rip) # sched: [5:1.00]
+; GENERIC-NEXT:    movq %rax, {{.*}}(%rip) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; BTVER2-LABEL: lshift_mem_b:
@@ -437,7 +437,7 @@ define void @lshift_mem_b_optsize(i64 %b) nounwind readnone optsize {
 ; GENERIC:       # %bb.0: # %entry
 ; GENERIC-NEXT:    movq {{.*}}(%rip), %rax # sched: [5:0.50]
 ; GENERIC-NEXT:    shrdq $54, %rdi, %rax # sched: [2:0.67]
-; GENERIC-NEXT:    movq %rax, {{.*}}(%rip) # sched: [5:1.00]
+; GENERIC-NEXT:    movq %rax, {{.*}}(%rip) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; BTVER2-LABEL: lshift_mem_b_optsize:

diff --git a/llvm/test/CodeGen/X86/schedule-x86_64.ll b/llvm/test/CodeGen/X86/schedule-x86_64.ll
index 9e3ce64..3792ab8 100644 (file)
--- a/llvm/test/CodeGen/X86/schedule-x86_64.ll
+++ b/llvm/test/CodeGen/X86/schedule-x86_64.ll
@@ -7669,8 +7669,8 @@ define void @test_movnti(i32 %a0, i32 *%a1, i64 %a2, i64 *%a3) optsize {
 ; GENERIC-LABEL: test_movnti:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    #APP
-; GENERIC-NEXT:    movntil %edi, (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    movntiq %rdx, (%rcx) # sched: [5:1.00]
+; GENERIC-NEXT:    movntil %edi, (%rsi) # sched: [1:1.00]
+; GENERIC-NEXT:    movntiq %rdx, (%rcx) # sched: [1:1.00]
 ; GENERIC-NEXT:    #NO_APP
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
@@ -7693,8 +7693,8 @@ define void @test_movnti(i32 %a0, i32 *%a1, i64 %a2, i64 *%a3) optsize {
 ; SANDY-LABEL: test_movnti:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    #APP
-; SANDY-NEXT:    movntil %edi, (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    movntiq %rdx, (%rcx) # sched: [5:1.00]
+; SANDY-NEXT:    movntil %edi, (%rsi) # sched: [1:1.00]
+; SANDY-NEXT:    movntiq %rdx, (%rcx) # sched: [1:1.00]
 ; SANDY-NEXT:    #NO_APP
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
@@ -13540,22 +13540,22 @@ define void @test_setcc(i8 %a0, i8 *%a1) optsize {
 ; GENERIC-NEXT:    setge %dil # sched: [1:0.50]
 ; GENERIC-NEXT:    setle %dil # sched: [1:0.50]
 ; GENERIC-NEXT:    setg %dil # sched: [1:0.50]
-; GENERIC-NEXT:    seto (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setno (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setb (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setae (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    sete (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setne (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setbe (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    seta (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    sets (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setns (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setp (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setnp (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setl (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setge (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setle (%rsi) # sched: [5:1.00]
-; GENERIC-NEXT:    setg (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    seto (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setno (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setb (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setae (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    sete (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setne (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setbe (%rsi) # sched: [3:1.00]
+; GENERIC-NEXT:    seta (%rsi) # sched: [3:1.00]
+; GENERIC-NEXT:    sets (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setns (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setp (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setnp (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setl (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setge (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setle (%rsi) # sched: [2:1.00]
+; GENERIC-NEXT:    setg (%rsi) # sched: [2:1.00]
 ; GENERIC-NEXT:    #NO_APP
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
@@ -13654,22 +13654,22 @@ define void @test_setcc(i8 %a0, i8 *%a1) optsize {
 ; SANDY-NEXT:    setge %dil # sched: [1:0.50]
 ; SANDY-NEXT:    setle %dil # sched: [1:0.50]
 ; SANDY-NEXT:    setg %dil # sched: [1:0.50]
-; SANDY-NEXT:    seto (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setno (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setb (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setae (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    sete (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setne (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setbe (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    seta (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    sets (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setns (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setp (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setnp (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setl (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setge (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setle (%rsi) # sched: [5:1.00]
-; SANDY-NEXT:    setg (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    seto (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setno (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setb (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setae (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    sete (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setne (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setbe (%rsi) # sched: [3:1.00]
+; SANDY-NEXT:    seta (%rsi) # sched: [3:1.00]
+; SANDY-NEXT:    sets (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setns (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setp (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setnp (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setl (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setge (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setle (%rsi) # sched: [2:1.00]
+; SANDY-NEXT:    setg (%rsi) # sched: [2:1.00]
 ; SANDY-NEXT:    #NO_APP
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;

diff --git a/llvm/test/CodeGen/X86/sse-schedule.ll b/llvm/test/CodeGen/X86/sse-schedule.ll
index 61f3091..51133eb 100644 (file)
--- a/llvm/test/CodeGen/X86/sse-schedule.ll
+++ b/llvm/test/CodeGen/X86/sse-schedule.ll
@@ -1939,7 +1939,7 @@ define float @test_divss(float %a0, float %a1, float *%a2) {
 define void @test_ldmxcsr(i32 %a0) {
 ; GENERIC-LABEL: test_ldmxcsr:
 ; GENERIC:       # %bb.0:
-; GENERIC-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [5:1.00]
+; GENERIC-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; GENERIC-NEXT:    ldmxcsr -{{[0-9]+}}(%rsp) # sched: [5:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
@@ -1957,13 +1957,13 @@ define void @test_ldmxcsr(i32 %a0) {
 ;
 ; SANDY-SSE-LABEL: test_ldmxcsr:
 ; SANDY-SSE:       # %bb.0:
-; SANDY-SSE-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    ldmxcsr -{{[0-9]+}}(%rsp) # sched: [5:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_ldmxcsr:
 ; SANDY:       # %bb.0:
-; SANDY-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [5:1.00]
+; SANDY-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; SANDY-NEXT:    vldmxcsr -{{[0-9]+}}(%rsp) # sched: [5:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
@@ -2487,7 +2487,7 @@ define void @test_movaps(<4 x float> *%a0, <4 x float> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movaps (%rdi), %xmm0 # sched: [6:0.50]
 ; GENERIC-NEXT:    addps %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    movaps %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movaps %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movaps:
@@ -2508,14 +2508,14 @@ define void @test_movaps(<4 x float> *%a0, <4 x float> *%a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movaps (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-SSE-NEXT:    addps %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movaps %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movaps %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movaps:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovaps (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovaps %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovaps %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movaps:
@@ -2712,7 +2712,7 @@ define void @test_movhps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) {
 ; GENERIC-NEXT:    movhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:1.00]
 ; GENERIC-NEXT:    addps %xmm0, %xmm1 # sched: [3:1.00]
 ; GENERIC-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1] sched: [1:1.00]
-; GENERIC-NEXT:    movlps %xmm1, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movlps %xmm1, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movhps:
@@ -2736,7 +2736,7 @@ define void @test_movhps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) {
 ; SANDY-SSE-NEXT:    movhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:1.00]
 ; SANDY-SSE-NEXT:    addps %xmm0, %xmm1 # sched: [3:1.00]
 ; SANDY-SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1] sched: [1:1.00]
-; SANDY-SSE-NEXT:    movlps %xmm1, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movlps %xmm1, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movhps:
@@ -2959,7 +2959,7 @@ define void @test_movlps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00]
 ; GENERIC-NEXT:    addps %xmm0, %xmm1 # sched: [3:1.00]
-; GENERIC-NEXT:    movlps %xmm1, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movlps %xmm1, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movlps:
@@ -2980,14 +2980,14 @@ define void @test_movlps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00]
 ; SANDY-SSE-NEXT:    addps %xmm0, %xmm1 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movlps %xmm1, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movlps %xmm1, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movlps:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00]
 ; SANDY-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovlps %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovlps %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movlps:
@@ -3178,7 +3178,7 @@ declare i32 @llvm.x86.sse.movmsk.ps(<4 x float>) nounwind readnone
 define void @test_movntps(<4 x float> %a0, <4 x float> *%a1) {
 ; GENERIC-LABEL: test_movntps:
 ; GENERIC:       # %bb.0:
-; GENERIC-NEXT:    movntps %xmm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movntps %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movntps:
@@ -3199,12 +3199,12 @@ define void @test_movntps(<4 x float> %a0, <4 x float> *%a1) {
 ;
 ; SANDY-SSE-LABEL: test_movntps:
 ; SANDY-SSE:       # %bb.0:
-; SANDY-SSE-NEXT:    movntps %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movntps %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movntps:
 ; SANDY:       # %bb.0:
-; SANDY-NEXT:    vmovntps %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovntps %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movntps:
@@ -3275,7 +3275,7 @@ define void @test_movss_mem(float* %a0, float* %a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [6:0.50]
 ; GENERIC-NEXT:    addss %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    movss %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movss %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movss_mem:
@@ -3296,14 +3296,14 @@ define void @test_movss_mem(float* %a0, float* %a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [6:0.50]
 ; SANDY-SSE-NEXT:    addss %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movss %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movss %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movss_mem:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [6:0.50]
 ; SANDY-NEXT:    vaddss %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovss %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovss %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movss_mem:
@@ -3495,7 +3495,7 @@ define void @test_movups(<4 x float> *%a0, <4 x float> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movups (%rdi), %xmm0 # sched: [6:0.50]
 ; GENERIC-NEXT:    addps %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    movups %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movups %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movups:
@@ -3516,14 +3516,14 @@ define void @test_movups(<4 x float> *%a0, <4 x float> *%a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movups (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-SSE-NEXT:    addps %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movups %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movups %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movups:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovups (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovups %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovups %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movups:

diff --git a/llvm/test/CodeGen/X86/sse2-schedule.ll b/llvm/test/CodeGen/X86/sse2-schedule.ll
index f3fc04e..890c747 100644 (file)
--- a/llvm/test/CodeGen/X86/sse2-schedule.ll
+++ b/llvm/test/CodeGen/X86/sse2-schedule.ll
@@ -4104,7 +4104,7 @@ define void @test_movapd(<2 x double> *%a0, <2 x double> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movapd (%rdi), %xmm0 # sched: [6:0.50]
 ; GENERIC-NEXT:    addpd %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    movapd %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movapd %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movapd:
@@ -4125,14 +4125,14 @@ define void @test_movapd(<2 x double> *%a0, <2 x double> *%a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movapd (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-SSE-NEXT:    addpd %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movapd %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movapd %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movapd:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovapd (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovapd %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovapd %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movapd:
@@ -4229,7 +4229,7 @@ define void @test_movdqa(<2 x i64> *%a0, <2 x i64> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movdqa (%rdi), %xmm0 # sched: [6:0.50]
 ; GENERIC-NEXT:    paddq %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT:    movdqa %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movdqa %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movdqa:
@@ -4250,14 +4250,14 @@ define void @test_movdqa(<2 x i64> *%a0, <2 x i64> *%a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movdqa (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-SSE-NEXT:    paddq %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-SSE-NEXT:    movdqa %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movdqa %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movdqa:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovdqa (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vmovdqa %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovdqa %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movdqa:
@@ -4354,7 +4354,7 @@ define void @test_movdqu(<2 x i64> *%a0, <2 x i64> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movdqu (%rdi), %xmm0 # sched: [6:0.50]
 ; GENERIC-NEXT:    paddq %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT:    movdqu %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movdqu %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movdqu:
@@ -4375,14 +4375,14 @@ define void @test_movdqu(<2 x i64> *%a0, <2 x i64> *%a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movdqu (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-SSE-NEXT:    paddq %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-SSE-NEXT:    movdqu %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movdqu %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movdqu:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovdqu (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vmovdqu %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovdqu %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movdqu:
@@ -4482,7 +4482,7 @@ define i32 @test_movd(<4 x i32> %a0, i32 %a1, i32 *%a2) {
 ; GENERIC-NEXT:    paddd %xmm0, %xmm1 # sched: [1:0.50]
 ; GENERIC-NEXT:    paddd %xmm0, %xmm2 # sched: [1:0.50]
 ; GENERIC-NEXT:    movd %xmm2, %eax # sched: [2:1.00]
-; GENERIC-NEXT:    movd %xmm1, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movd %xmm1, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movd:
@@ -4512,7 +4512,7 @@ define i32 @test_movd(<4 x i32> %a0, i32 %a1, i32 *%a2) {
 ; SANDY-SSE-NEXT:    paddd %xmm0, %xmm1 # sched: [1:0.50]
 ; SANDY-SSE-NEXT:    paddd %xmm0, %xmm2 # sched: [1:0.50]
 ; SANDY-SSE-NEXT:    movd %xmm2, %eax # sched: [2:1.00]
-; SANDY-SSE-NEXT:    movd %xmm1, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movd %xmm1, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movd:
@@ -4522,7 +4522,7 @@ define i32 @test_movd(<4 x i32> %a0, i32 %a1, i32 *%a2) {
 ; SANDY-NEXT:    vpaddd %xmm1, %xmm0, %xmm1 # sched: [1:0.50]
 ; SANDY-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # sched: [1:0.50]
 ; SANDY-NEXT:    vmovd %xmm0, %eax # sched: [2:1.00]
-; SANDY-NEXT:    vmovd %xmm1, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovd %xmm1, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movd:
@@ -4663,7 +4663,7 @@ define i64 @test_movd_64(<2 x i64> %a0, i64 %a1, i64 *%a2) {
 ; GENERIC-NEXT:    paddq %xmm0, %xmm1 # sched: [1:0.50]
 ; GENERIC-NEXT:    paddq %xmm0, %xmm2 # sched: [1:0.50]
 ; GENERIC-NEXT:    movq %xmm2, %rax # sched: [2:1.00]
-; GENERIC-NEXT:    movq %xmm1, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movq %xmm1, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movd_64:
@@ -4693,7 +4693,7 @@ define i64 @test_movd_64(<2 x i64> %a0, i64 %a1, i64 *%a2) {
 ; SANDY-SSE-NEXT:    paddq %xmm0, %xmm1 # sched: [1:0.50]
 ; SANDY-SSE-NEXT:    paddq %xmm0, %xmm2 # sched: [1:0.50]
 ; SANDY-SSE-NEXT:    movq %xmm2, %rax # sched: [2:1.00]
-; SANDY-SSE-NEXT:    movq %xmm1, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movq %xmm1, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movd_64:
@@ -4703,7 +4703,7 @@ define i64 @test_movd_64(<2 x i64> %a0, i64 %a1, i64 *%a2) {
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm1 # sched: [1:0.50]
 ; SANDY-NEXT:    vpaddq %xmm2, %xmm0, %xmm0 # sched: [1:0.50]
 ; SANDY-NEXT:    vmovq %xmm0, %rax # sched: [2:1.00]
-; SANDY-NEXT:    vmovq %xmm1, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovq %xmm1, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movd_64:
@@ -4841,7 +4841,7 @@ define void @test_movhpd(<2 x double> %a0, <2 x double> %a1, x86_mmx *%a2) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:1.00]
 ; GENERIC-NEXT:    addpd %xmm0, %xmm1 # sched: [3:1.00]
-; GENERIC-NEXT:    movhpd %xmm1, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movhpd %xmm1, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movhpd:
@@ -4862,14 +4862,14 @@ define void @test_movhpd(<2 x double> %a0, <2 x double> %a1, x86_mmx *%a2) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:1.00]
 ; SANDY-SSE-NEXT:    addpd %xmm0, %xmm1 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movhpd %xmm1, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movhpd %xmm1, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movhpd:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:1.00]
 ; SANDY-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovhpd %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovhpd %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movhpd:
@@ -4969,7 +4969,7 @@ define void @test_movlpd(<2 x double> %a0, <2 x double> %a1, x86_mmx *%a2) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00]
 ; GENERIC-NEXT:    addpd %xmm0, %xmm1 # sched: [3:1.00]
-; GENERIC-NEXT:    movlpd %xmm1, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movlpd %xmm1, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movlpd:
@@ -4990,14 +4990,14 @@ define void @test_movlpd(<2 x double> %a0, <2 x double> %a1, x86_mmx *%a2) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00]
 ; SANDY-SSE-NEXT:    addpd %xmm0, %xmm1 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movlpd %xmm1, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movlpd %xmm1, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movlpd:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00]
 ; SANDY-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovlpd %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovlpd %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movlpd:
@@ -5188,7 +5188,7 @@ define void @test_movntdqa(<2 x i64> %a0, <2 x i64> *%a1) {
 ; GENERIC-LABEL: test_movntdqa:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    paddq %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT:    movntdq %xmm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movntdq %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movntdqa:
@@ -5208,13 +5208,13 @@ define void @test_movntdqa(<2 x i64> %a0, <2 x i64> *%a1) {
 ; SANDY-SSE-LABEL: test_movntdqa:
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    paddq %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-SSE-NEXT:    movntdq %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movntdq %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movntdqa:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vmovntdq %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovntdq %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movntdqa:
@@ -5297,7 +5297,7 @@ define void @test_movntpd(<2 x double> %a0, <2 x double> *%a1) {
 ; GENERIC-LABEL: test_movntpd:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    addpd %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    movntpd %xmm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movntpd %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movntpd:
@@ -5315,13 +5315,13 @@ define void @test_movntpd(<2 x double> %a0, <2 x double> *%a1) {
 ; SANDY-SSE-LABEL: test_movntpd:
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    addpd %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movntpd %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movntpd %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movntpd:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovntpd %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovntpd %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movntpd:
@@ -5405,7 +5405,7 @@ define <2 x i64> @test_movq_mem(<2 x i64> %a0, i64 *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero sched: [6:0.50]
 ; GENERIC-NEXT:    paddq %xmm1, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT:    movq %xmm0, (%rdi) # sched: [5:1.00]
+; GENERIC-NEXT:    movq %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movq_mem:
@@ -5426,14 +5426,14 @@ define <2 x i64> @test_movq_mem(<2 x i64> %a0, i64 *%a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero sched: [6:0.50]
 ; SANDY-SSE-NEXT:    paddq %xmm1, %xmm0 # sched: [1:0.50]
-; SANDY-SSE-NEXT:    movq %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movq %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movq_mem:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero sched: [6:0.50]
 ; SANDY-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SANDY-NEXT:    vmovq %xmm0, (%rdi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovq %xmm0, (%rdi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movq_mem:
@@ -5641,7 +5641,7 @@ define void @test_movsd_mem(double* %a0, double* %a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero sched: [6:0.50]
 ; GENERIC-NEXT:    addsd %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    movsd %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movsd %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movsd_mem:
@@ -5662,14 +5662,14 @@ define void @test_movsd_mem(double* %a0, double* %a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero sched: [6:0.50]
 ; SANDY-SSE-NEXT:    addsd %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movsd %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movsd %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movsd_mem:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero sched: [6:0.50]
 ; SANDY-NEXT:    vaddsd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovsd %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovsd %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movsd_mem:
@@ -5869,7 +5869,7 @@ define void @test_movupd(<2 x double> *%a0, <2 x double> *%a1) {
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    movupd (%rdi), %xmm0 # sched: [6:0.50]
 ; GENERIC-NEXT:    addpd %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT:    movupd %xmm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    movupd %xmm0, (%rsi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_movupd:
@@ -5890,14 +5890,14 @@ define void @test_movupd(<2 x double> *%a0, <2 x double> *%a1) {
 ; SANDY-SSE:       # %bb.0:
 ; SANDY-SSE-NEXT:    movupd (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-SSE-NEXT:    addpd %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-SSE-NEXT:    movupd %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-SSE-NEXT:    movupd %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-SSE-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movupd:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    vmovupd (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT:    vmovupd %xmm0, (%rsi) # sched: [5:1.00]
+; SANDY-NEXT:    vmovupd %xmm0, (%rsi) # sched: [1:1.00]
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-SSE-LABEL: test_movupd: