[X86] Match vpmullq latency to uops.info. Correct port usage for 512-bit memory form

[craig.topper]

Summary

uops.info says these should be 15 cycle instructions. Uops.info also shows the 512-bit form uses port 0 and 5 for both register and memory. We had memory using 0 and 1.

[RKSimon]

LGTM - Agner mostly agrees (but misses the port05 for the Z512 case)

llvm/lib/Target/X86/X86SchedSkylakeServer.td

 def SKXWriteResGroup171 : SchedWriteRes<[SKXPort06,SKXPort0156]> {
   let Latency = 11;
   let NumMicroOps = 11;
   let ResourceCycles = [2,9];
 }
 def: InstRW<[SKXWriteResGroup171], (instrs LOOPE, LOOPNE)>;
 
 def SKXWriteResGroup174 : SchedWriteRes<[SKXPort01]> {
-  let Latency = 12;
+  let Latency = 15;
   let NumMicroOps = 3;
   let ResourceCycles = [3];
 }
 def: InstRW<[SKXWriteResGroup174], (instregex "VPMULLQ(Z128|Z256)rr")>;
 
 def SKXWriteResGroup174z : SchedWriteRes<[SKXPort05]> {
-  let Latency = 12;
+  let Latency = 15;
   let NumMicroOps = 3;
   let ResourceCycles = [3];
 }
 def: InstRW<[SKXWriteResGroup174z], (instregex "VPMULLQZrr")>;
 
 def SKXWriteResGroup175 : SchedWriteRes<[SKXPort5,SKXPort23]> {
   let Latency = 12;
   let NumMicroOps = 3;

 
 def SKXWriteResGroup202 : SchedWriteRes<[SKXPort0,SKXPort1,SKXPort5,SKXPort6,SKXPort05,SKXPort0156]> {
   let Latency = 17;
   let NumMicroOps = 15;
   let ResourceCycles = [2,1,2,4,2,4];
 }
 def: InstRW<[SKXWriteResGroup202], (instrs XCH_F)>;
 
-def SKXWriteResGroup205 : SchedWriteRes<[SKXPort23,SKXPort015]> {
-  let Latency = 18;
+def SKXWriteResGroup205 : SchedWriteRes<[SKXPort23,SKXPort01]> {
+  let Latency = 21;
   let NumMicroOps = 4;
   let ResourceCycles = [1,3];
 }
 def: InstRW<[SKXWriteResGroup205], (instregex "VPMULLQZ128rm(b?)")>;
 
 def SKXWriteResGroup207 : SchedWriteRes<[SKXPort5,SKXPort6,SKXPort06,SKXPort0156]> {
   let Latency = 18;
   let NumMicroOps = 8;

 
 def SKXWriteResGroup209 : SchedWriteRes<[SKXPort0,SKXPort23,SKXFPDivider]> {
   let Latency = 19;
   let NumMicroOps = 2;
   let ResourceCycles = [1,1,4];
 }
 def : SchedAlias<WriteFDiv64Ld,  SKXWriteResGroup209>; // TODO - convert to ZnWriteResFpuPair
 
-def SKXWriteResGroup211 : SchedWriteRes<[SKXPort23,SKXPort015]> {
-  let Latency = 19;
+def SKXWriteResGroup211 : SchedWriteRes<[SKXPort23,SKXPort01]> {
+  let Latency = 22;
+  let NumMicroOps = 4;
+  let ResourceCycles = [1,3];
+}
+def: InstRW<[SKXWriteResGroup211], (instregex "VPMULLQZ256rm(b?)")>;
+
+def SKXWriteResGroup211_1 : SchedWriteRes<[SKXPort23,SKXPort05]> {
+  let Latency = 22;
   let NumMicroOps = 4;
   let ResourceCycles = [1,3];
 }
-def: InstRW<[SKXWriteResGroup211], (instregex "VPMULLQZ256rm(b?)",
-                                              "VPMULLQZrm(b?)")>;
+def: InstRW<[SKXWriteResGroup211_1], (instregex "VPMULLQZrm(b?)")>;
 
 def SKXWriteResGroup215 : SchedWriteRes<[SKXPort0]> {
   let Latency = 20;
   let NumMicroOps = 1;
   let ResourceCycles = [1];
 }
 def: InstRW<[SKXWriteResGroup215], (instregex "DIV_(FPrST0|FST0r|FrST0)")>;
 

llvm/test/tools/llvm-mca/X86/SkylakeServer/resources-avx512dq.s

 # CHECK-NEXT:  1      4     1.00                        vfpclasssd  $171, %xmm16, %k1
 # CHECK-NEXT:  2      9     1.00    *                   vfpclasssd  $171, (%rax), %k1
 # CHECK-NEXT:  1      4     1.00                        vfpclasssd  $171, %xmm16, %k1 {%k2}
 # CHECK-NEXT:  2      9     1.00    *                   vfpclasssd  $171, (%rax), %k1 {%k2}
 # CHECK-NEXT:  1      4     1.00                        vfpclassss  $171, %xmm16, %k1
 # CHECK-NEXT:  2      9     1.00    *                   vfpclassss  $171, (%rax), %k1
 # CHECK-NEXT:  1      4     1.00                        vfpclassss  $171, %xmm16, %k1 {%k2}
 # CHECK-NEXT:  2      9     1.00    *                   vfpclassss  $171, (%rax), %k1 {%k2}
-# CHECK-NEXT:  3      12    1.50                        vpmullq %zmm16, %zmm17, %zmm19
-# CHECK-NEXT:  4      19    1.00    *                   vpmullq (%rax), %zmm17, %zmm19
-# CHECK-NEXT:  3      12    1.50                        vpmullq %zmm16, %zmm17, %zmm19 {%k1}
-# CHECK-NEXT:  4      19    1.00    *                   vpmullq (%rax), %zmm17, %zmm19 {%k1}
-# CHECK-NEXT:  3      12    1.50                        vpmullq %zmm16, %zmm17, %zmm19 {%k1} {z}
-# CHECK-NEXT:  4      19    1.00    *                   vpmullq (%rax), %zmm17, %zmm19 {%k1} {z}
+# CHECK-NEXT:  3      15    1.50                        vpmullq %zmm16, %zmm17, %zmm19
+# CHECK-NEXT:  4      22    1.50    *                   vpmullq (%rax), %zmm17, %zmm19
+# CHECK-NEXT:  3      15    1.50                        vpmullq %zmm16, %zmm17, %zmm19 {%k1}
+# CHECK-NEXT:  4      22    1.50    *                   vpmullq (%rax), %zmm17, %zmm19 {%k1}
+# CHECK-NEXT:  3      15    1.50                        vpmullq %zmm16, %zmm17, %zmm19 {%k1} {z}
+# CHECK-NEXT:  4      22    1.50    *                   vpmullq (%rax), %zmm17, %zmm19 {%k1} {z}
 # CHECK-NEXT:  1      1     0.50                        vxorpd  %zmm16, %zmm17, %zmm19
 # CHECK-NEXT:  2      8     0.50    *                   vxorpd  (%rax), %zmm17, %zmm19
 # CHECK-NEXT:  2      8     0.50    *                   vxorpd  (%rax){1to8}, %zmm17, %zmm19
 # CHECK-NEXT:  1      1     0.50                        vxorpd  %zmm16, %zmm17, %zmm19 {%k1}
 # CHECK-NEXT:  2      8     0.50    *                   vxorpd  (%rax), %zmm17, %zmm19 {%k1}
 # CHECK-NEXT:  2      8     0.50    *                   vxorpd  (%rax){1to8}, %zmm17, %zmm19 {%k1}
 # CHECK-NEXT:  1      1     0.50                        vxorpd  %zmm16, %zmm17, %zmm19 {%k1} {z}
 # CHECK-NEXT:  2      8     0.50    *                   vxorpd  (%rax), %zmm17, %zmm19 {%k1} {z}

 # CHECK-NEXT: [5]   - SKXPort3
 # CHECK-NEXT: [6]   - SKXPort4
 # CHECK-NEXT: [7]   - SKXPort5
 # CHECK-NEXT: [8]   - SKXPort6
 # CHECK-NEXT: [9]   - SKXPort7
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]
-# CHECK-NEXT:  -      -     55.50  7.00   37.50  37.50   -     81.50   -      -
+# CHECK-NEXT:  -      -     57.00  4.00   37.50  37.50   -     83.00   -      -
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    Instructions:
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -     kaddb %k0, %k1, %k2
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -     kaddw %k0, %k1, %k2
 # CHECK-NEXT:  -      -     1.00    -      -      -      -      -      -      -     kandb %k0, %k1, %k2
 # CHECK-NEXT:  -      -     1.00    -      -      -      -      -      -      -     kandnb  %k0, %k1, %k2
 # CHECK-NEXT:  -      -     1.00    -      -      -      -      -      -      -     korb  %k0, %k1, %k2

 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclasssd  $171, (%rax), %k1
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -     vfpclasssd  $171, %xmm16, %k1 {%k2}
 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclasssd  $171, (%rax), %k1 {%k2}
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -     vfpclassss  $171, %xmm16, %k1
 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclassss  $171, (%rax), %k1
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -     vfpclassss  $171, %xmm16, %k1 {%k2}
 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclassss  $171, (%rax), %k1 {%k2}
 # CHECK-NEXT:  -      -     1.50    -      -      -      -     1.50    -      -     vpmullq %zmm16, %zmm17, %zmm19
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %zmm17, %zmm19
+# CHECK-NEXT:  -      -     1.50    -     0.50   0.50    -     1.50    -      -     vpmullq (%rax), %zmm17, %zmm19
 # CHECK-NEXT:  -      -     1.50    -      -      -      -     1.50    -      -     vpmullq %zmm16, %zmm17, %zmm19 {%k1}
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %zmm17, %zmm19 {%k1}
+# CHECK-NEXT:  -      -     1.50    -     0.50   0.50    -     1.50    -      -     vpmullq (%rax), %zmm17, %zmm19 {%k1}
 # CHECK-NEXT:  -      -     1.50    -      -      -      -     1.50    -      -     vpmullq %zmm16, %zmm17, %zmm19 {%k1} {z}
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %zmm17, %zmm19 {%k1} {z}
+# CHECK-NEXT:  -      -     1.50    -     0.50   0.50    -     1.50    -      -     vpmullq (%rax), %zmm17, %zmm19 {%k1} {z}
 # CHECK-NEXT:  -      -     0.50    -      -      -      -     0.50    -      -     vxorpd  %zmm16, %zmm17, %zmm19
 # CHECK-NEXT:  -      -     0.50    -     0.50   0.50    -     0.50    -      -     vxorpd  (%rax), %zmm17, %zmm19
 # CHECK-NEXT:  -      -     0.50    -     0.50   0.50    -     0.50    -      -     vxorpd  (%rax){1to8}, %zmm17, %zmm19
 # CHECK-NEXT:  -      -     0.50    -      -      -      -     0.50    -      -     vxorpd  %zmm16, %zmm17, %zmm19 {%k1}
 # CHECK-NEXT:  -      -     0.50    -     0.50   0.50    -     0.50    -      -     vxorpd  (%rax), %zmm17, %zmm19 {%k1}
 # CHECK-NEXT:  -      -     0.50    -     0.50   0.50    -     0.50    -      -     vxorpd  (%rax){1to8}, %zmm17, %zmm19 {%k1}
 # CHECK-NEXT:  -      -     0.50    -      -      -      -     0.50    -      -     vxorpd  %zmm16, %zmm17, %zmm19 {%k1} {z}
 # CHECK-NEXT:  -      -     0.50    -     0.50   0.50    -     0.50    -      -     vxorpd  (%rax), %zmm17, %zmm19 {%k1} {z}

llvm/test/tools/llvm-mca/X86/SkylakeServer/resources-avx512dqvl.s

 # CHECK-NEXT:  2      10    1.00    *                   vfpclasspsx $171, (%rax), %k1 {%k2}
 # CHECK-NEXT:  2      10    1.00    *                   vfpclassps  $171, (%rax){1to4}, %k1 {%k2}
 # CHECK-NEXT:  1      4     1.00                        vfpclassps  $171, %ymm16, %k1
 # CHECK-NEXT:  2      11    1.00    *                   vfpclasspsy $171, (%rax), %k1
 # CHECK-NEXT:  2      11    1.00    *                   vfpclassps  $171, (%rax){1to8}, %k1
 # CHECK-NEXT:  1      4     1.00                        vfpclassps  $171, %ymm16, %k1 {%k2}
 # CHECK-NEXT:  2      11    1.00    *                   vfpclasspsy $171, (%rax), %k1 {%k2}
 # CHECK-NEXT:  2      11    1.00    *                   vfpclassps  $171, (%rax){1to8}, %k1 {%k2}
-# CHECK-NEXT:  3      12    1.50                        vpmullq %xmm16, %xmm17, %xmm19
-# CHECK-NEXT:  4      18    1.00    *                   vpmullq (%rax), %xmm17, %xmm19
-# CHECK-NEXT:  3      12    1.50                        vpmullq %xmm16, %xmm17, %xmm19 {%k1}
-# CHECK-NEXT:  4      18    1.00    *                   vpmullq (%rax), %xmm17, %xmm19 {%k1}
-# CHECK-NEXT:  3      12    1.50                        vpmullq %xmm16, %xmm17, %xmm19 {%k1} {z}
-# CHECK-NEXT:  4      18    1.00    *                   vpmullq (%rax), %xmm17, %xmm19 {%k1} {z}
-# CHECK-NEXT:  3      12    1.50                        vpmullq %ymm16, %ymm17, %ymm19
-# CHECK-NEXT:  4      19    1.00    *                   vpmullq (%rax), %ymm17, %ymm19
-# CHECK-NEXT:  3      12    1.50                        vpmullq %ymm16, %ymm17, %ymm19 {%k1}
-# CHECK-NEXT:  4      19    1.00    *                   vpmullq (%rax), %ymm17, %ymm19 {%k1}
-# CHECK-NEXT:  3      12    1.50                        vpmullq %ymm16, %ymm17, %ymm19 {%k1} {z}
-# CHECK-NEXT:  4      19    1.00    *                   vpmullq (%rax), %ymm17, %ymm19 {%k1} {z}
+# CHECK-NEXT:  3      15    1.50                        vpmullq %xmm16, %xmm17, %xmm19
+# CHECK-NEXT:  4      21    1.50    *                   vpmullq (%rax), %xmm17, %xmm19
+# CHECK-NEXT:  3      15    1.50                        vpmullq %xmm16, %xmm17, %xmm19 {%k1}
+# CHECK-NEXT:  4      21    1.50    *                   vpmullq (%rax), %xmm17, %xmm19 {%k1}
+# CHECK-NEXT:  3      15    1.50                        vpmullq %xmm16, %xmm17, %xmm19 {%k1} {z}
+# CHECK-NEXT:  4      21    1.50    *                   vpmullq (%rax), %xmm17, %xmm19 {%k1} {z}
+# CHECK-NEXT:  3      15    1.50                        vpmullq %ymm16, %ymm17, %ymm19
+# CHECK-NEXT:  4      22    1.50    *                   vpmullq (%rax), %ymm17, %ymm19
+# CHECK-NEXT:  3      15    1.50                        vpmullq %ymm16, %ymm17, %ymm19 {%k1}
+# CHECK-NEXT:  4      22    1.50    *                   vpmullq (%rax), %ymm17, %ymm19 {%k1}
+# CHECK-NEXT:  3      15    1.50                        vpmullq %ymm16, %ymm17, %ymm19 {%k1} {z}
+# CHECK-NEXT:  4      22    1.50    *                   vpmullq (%rax), %ymm17, %ymm19 {%k1} {z}
 # CHECK-NEXT:  1      1     0.33                        vxorpd  %xmm16, %xmm17, %xmm19
 # CHECK-NEXT:  2      7     0.50    *                   vxorpd  (%rax), %xmm17, %xmm19
 # CHECK-NEXT:  2      7     0.50    *                   vxorpd  (%rax){1to2}, %xmm17, %xmm19
 # CHECK-NEXT:  1      1     0.33                        vxorpd  %xmm16, %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  2      7     0.50    *                   vxorpd  (%rax), %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  2      7     0.50    *                   vxorpd  (%rax){1to2}, %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  1      1     0.33                        vxorpd  %xmm16, %xmm17, %xmm19 {%k1} {z}
 # CHECK-NEXT:  2      7     0.50    *                   vxorpd  (%rax), %xmm17, %xmm19 {%k1} {z}

 # CHECK-NEXT: [5]   - SKXPort3
 # CHECK-NEXT: [6]   - SKXPort4
 # CHECK-NEXT: [7]   - SKXPort5
 # CHECK-NEXT: [8]   - SKXPort6
 # CHECK-NEXT: [9]   - SKXPort7
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]
-# CHECK-NEXT:  -      -     64.00  64.00  59.00  59.00   -     76.00   -      -
+# CHECK-NEXT:  -      -     67.00  67.00  59.00  59.00   -     70.00   -      -
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    Instructions:
 # CHECK-NEXT:  -      -     0.33   0.33    -      -      -     0.33    -      -     vandnpd %xmm16, %xmm17, %xmm19
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vandnpd (%rax), %xmm17, %xmm19
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vandnpd (%rax){1to2}, %xmm17, %xmm19
 # CHECK-NEXT:  -      -     0.33   0.33    -      -      -     0.33    -      -     vandnpd %xmm16, %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vandnpd (%rax), %xmm17, %xmm19 {%k1}

 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclassps  $171, (%rax){1to4}, %k1 {%k2}
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -     vfpclassps  $171, %ymm16, %k1
 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclasspsy $171, (%rax), %k1
 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclassps  $171, (%rax){1to8}, %k1
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -     vfpclassps  $171, %ymm16, %k1 {%k2}
 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclasspsy $171, (%rax), %k1 {%k2}
 # CHECK-NEXT:  -      -      -      -     0.50   0.50    -     1.00    -      -     vfpclassps  $171, (%rax){1to8}, %k1 {%k2}
 # CHECK-NEXT:  -      -     1.50   1.50    -      -      -      -      -      -     vpmullq %xmm16, %xmm17, %xmm19
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %xmm17, %xmm19
+# CHECK-NEXT:  -      -     1.50   1.50   0.50   0.50    -      -      -      -     vpmullq (%rax), %xmm17, %xmm19
 # CHECK-NEXT:  -      -     1.50   1.50    -      -      -      -      -      -     vpmullq %xmm16, %xmm17, %xmm19 {%k1}
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %xmm17, %xmm19 {%k1}
+# CHECK-NEXT:  -      -     1.50   1.50   0.50   0.50    -      -      -      -     vpmullq (%rax), %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  -      -     1.50   1.50    -      -      -      -      -      -     vpmullq %xmm16, %xmm17, %xmm19 {%k1} {z}
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %xmm17, %xmm19 {%k1} {z}
+# CHECK-NEXT:  -      -     1.50   1.50   0.50   0.50    -      -      -      -     vpmullq (%rax), %xmm17, %xmm19 {%k1} {z}
 # CHECK-NEXT:  -      -     1.50   1.50    -      -      -      -      -      -     vpmullq %ymm16, %ymm17, %ymm19
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %ymm17, %ymm19
+# CHECK-NEXT:  -      -     1.50   1.50   0.50   0.50    -      -      -      -     vpmullq (%rax), %ymm17, %ymm19
 # CHECK-NEXT:  -      -     1.50   1.50    -      -      -      -      -      -     vpmullq %ymm16, %ymm17, %ymm19 {%k1}
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %ymm17, %ymm19 {%k1}
+# CHECK-NEXT:  -      -     1.50   1.50   0.50   0.50    -      -      -      -     vpmullq (%rax), %ymm17, %ymm19 {%k1}
 # CHECK-NEXT:  -      -     1.50   1.50    -      -      -      -      -      -     vpmullq %ymm16, %ymm17, %ymm19 {%k1} {z}
-# CHECK-NEXT:  -      -     1.00   1.00   0.50   0.50    -     1.00    -      -     vpmullq (%rax), %ymm17, %ymm19 {%k1} {z}
+# CHECK-NEXT:  -      -     1.50   1.50   0.50   0.50    -      -      -      -     vpmullq (%rax), %ymm17, %ymm19 {%k1} {z}
 # CHECK-NEXT:  -      -     0.33   0.33    -      -      -     0.33    -      -     vxorpd  %xmm16, %xmm17, %xmm19
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vxorpd  (%rax), %xmm17, %xmm19
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vxorpd  (%rax){1to2}, %xmm17, %xmm19
 # CHECK-NEXT:  -      -     0.33   0.33    -      -      -     0.33    -      -     vxorpd  %xmm16, %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vxorpd  (%rax), %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vxorpd  (%rax){1to2}, %xmm17, %xmm19 {%k1}
 # CHECK-NEXT:  -      -     0.33   0.33    -      -      -     0.33    -      -     vxorpd  %xmm16, %xmm17, %xmm19 {%k1} {z}
 # CHECK-NEXT:  -      -     0.33   0.33   0.50   0.50    -     0.33    -      -     vxorpd  (%rax), %xmm17, %xmm19 {%k1} {z}