[X86][BtVer2] Improved latency and throughput of float/vector loads and stores.

This patch introduces the following changes to the btver2 scheduling model:

- The number of micro opcodes for YMM loads and stores is now 2 (it was
  incorrectly set to 1 for both aligned and misaligned loads/stores).

- Increased the number of AGU resource cycles for YMM loads and stores
  to 2cy (instead of 1cy).

- Removed JFPU01 and JFPX from the list of resources consumed by pure
  float/vector loads (no MMX).

I verified with llvm-exegesis that pure XMM/YMM loads are no-pipe. Those
are dispatched to the FPU but not really issues on JFPU01.

Differential Revision: https://reviews.llvm.org/D68871

llvm-svn: 374765

diff --git a/llvm/lib/Target/X86/X86ScheduleBtVer2.td b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
index 3addf04..d0421d9 100644 (file)
--- a/llvm/lib/Target/X86/X86ScheduleBtVer2.td
+++ b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
@@ -501,14 +501,14 @@ defm : X86WriteRes<WriteFLD0,          [JFPU1, JSTC], 3, [1,1], 1>;
 defm : X86WriteRes<WriteFLD1,          [JFPU1, JSTC], 3, [1,1], 1>;
 defm : X86WriteRes<WriteFLDC,          [JFPU1, JSTC], 3, [1,1], 1>;
 defm : X86WriteRes<WriteFLoad,         [JLAGU, JFPU01, JFPX], 5, [1, 1, 1], 1>;
-defm : X86WriteRes<WriteFLoadX,        [JLAGU, JFPU01, JFPX], 5, [1, 1, 1], 1>;
-defm : X86WriteRes<WriteFLoadY,        [JLAGU, JFPU01, JFPX], 5, [1, 1, 1], 1>;
+defm : X86WriteRes<WriteFLoadX,        [JLAGU], 5, [1], 1>;
+defm : X86WriteRes<WriteFLoadY,        [JLAGU], 5, [2], 2>;
 defm : X86WriteRes<WriteFMaskedLoad,   [JLAGU, JFPU01, JFPX], 6, [1, 2, 2], 1>;
 defm : X86WriteRes<WriteFMaskedLoadY,  [JLAGU, JFPU01, JFPX], 6, [2, 4, 4], 2>;
 
 defm : X86WriteRes<WriteFStore,        [JSAGU, JFPU1,  JSTC], 2, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteFStoreX,       [JSAGU, JFPU1,  JSTC], 1, [1, 1, 1], 1>;
-defm : X86WriteRes<WriteFStoreY,       [JSAGU, JFPU1,  JSTC], 1, [1, 1, 1], 1>;
+defm : X86WriteRes<WriteFStoreY,       [JSAGU, JFPU1,  JSTC], 1, [2, 2, 2], 2>;
 defm : X86WriteRes<WriteFStoreNT,      [JSAGU, JFPU1,  JSTC], 3, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteFStoreNTX,     [JSAGU, JFPU1,  JSTC], 3, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteFStoreNTY,     [JSAGU, JFPU1,  JSTC], 3, [2, 2, 2], 1>;
@@ -657,8 +657,8 @@ defm : X86WriteResUnsupported<WriteCvtPS2PHZSt>;
 ////////////////////////////////////////////////////////////////////////////////
 
 defm : X86WriteRes<WriteVecLoad,          [JLAGU, JFPU01, JVALU], 5, [1, 1, 1], 1>;
-defm : X86WriteRes<WriteVecLoadX,         [JLAGU, JFPU01, JVALU], 5, [1, 1, 1], 1>;
-defm : X86WriteRes<WriteVecLoadY,         [JLAGU, JFPU01, JVALU], 5, [1, 1, 1], 1>;
+defm : X86WriteRes<WriteVecLoadX,         [JLAGU], 5, [1], 1>;
+defm : X86WriteRes<WriteVecLoadY,         [JLAGU], 5, [2], 2>;
 defm : X86WriteRes<WriteVecLoadNT,        [JLAGU, JFPU01, JVALU], 5, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteVecLoadNTY,       [JLAGU, JFPU01, JVALU], 5, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteVecMaskedLoad,    [JLAGU, JFPU01, JVALU], 6, [1, 2, 2], 1>;
@@ -666,7 +666,7 @@ defm : X86WriteRes<WriteVecMaskedLoadY,   [JLAGU, JFPU01, JVALU], 6, [2, 4, 4],
 
 defm : X86WriteRes<WriteVecStore,         [JSAGU, JFPU1,   JSTC], 2, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteVecStoreX,        [JSAGU, JFPU1,   JSTC], 1, [1, 1, 1], 1>;
-defm : X86WriteRes<WriteVecStoreY,        [JSAGU, JFPU1,   JSTC], 1, [1, 1, 1], 1>;
+defm : X86WriteRes<WriteVecStoreY,        [JSAGU, JFPU1,   JSTC], 1, [2, 2, 2], 2>;
 defm : X86WriteRes<WriteVecStoreNT,       [JSAGU, JFPU1,   JSTC], 2, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteVecStoreNTY,      [JSAGU, JFPU1,   JSTC], 2, [2, 2, 2], 1>;
 defm : X86WriteRes<WriteVecMaskedStore,   [JSAGU, JFPU01, JVALU], 6, [1, 1, 4], 1>;

diff --git a/llvm/test/tools/llvm-mca/X86/BtVer2/bottleneck-hints-3.s b/llvm/test/tools/llvm-mca/X86/BtVer2/bottleneck-hints-3.s
index d78e887..1f13a1f 100644 (file)
--- a/llvm/test/tools/llvm-mca/X86/BtVer2/bottleneck-hints-3.s
+++ b/llvm/test/tools/llvm-mca/X86/BtVer2/bottleneck-hints-3.s
@@ -83,17 +83,17 @@ vmovaps %xmm0, 48(%rdi)
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
-# CHECK-NEXT:  -      -      -     2.00   2.00   4.00   4.00   4.00    -     4.00   4.00    -      -      -
+# CHECK-NEXT:  -      -      -      -      -      -     4.00   4.00    -     4.00   4.00    -      -      -
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
-# CHECK-NEXT:  -      -      -      -     1.00   1.00    -     1.00    -      -      -      -      -      -     vmovaps        (%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        (%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, (%rdi)
-# CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     vmovaps        16(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        16(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 16(%rdi)
-# CHECK-NEXT:  -      -      -      -     1.00   1.00    -     1.00    -      -      -      -      -      -     vmovaps        32(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        32(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 32(%rdi)
-# CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     vmovaps        48(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        48(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 48(%rdi)
 
 # CHECK:      Timeline view:

diff --git a/llvm/test/tools/llvm-mca/X86/BtVer2/load-store-alias.s b/llvm/test/tools/llvm-mca/X86/BtVer2/load-store-alias.s
index b5f3204..e6294d6 100644 (file)
--- a/llvm/test/tools/llvm-mca/X86/BtVer2/load-store-alias.s
+++ b/llvm/test/tools/llvm-mca/X86/BtVer2/load-store-alias.s
@@ -56,17 +56,17 @@ vmovaps %xmm0, 48(%rdi)
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
-# CHECK-NEXT:  -      -      -     2.00   2.00   3.99   4.01   4.00    -     4.00   4.00    -      -      -
+# CHECK-NEXT:  -      -      -      -      -      -     4.00   4.00    -     4.00   4.00    -      -      -
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
-# CHECK-NEXT:  -      -      -      -     1.00   0.99   0.01   1.00    -      -      -      -      -      -     vmovaps        (%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        (%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, (%rdi)
-# CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     vmovaps        16(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        16(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 16(%rdi)
-# CHECK-NEXT:  -      -      -      -     1.00   1.00    -     1.00    -      -      -      -      -      -     vmovaps        32(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        32(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 32(%rdi)
-# CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     vmovaps        48(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        48(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 48(%rdi)
 
 # CHECK:      Timeline view:

diff --git a/llvm/test/tools/llvm-mca/X86/BtVer2/memcpy-like-test.s b/llvm/test/tools/llvm-mca/X86/BtVer2/memcpy-like-test.s
index 6d17586..4162d85 100644 (file)
--- a/llvm/test/tools/llvm-mca/X86/BtVer2/memcpy-like-test.s
+++ b/llvm/test/tools/llvm-mca/X86/BtVer2/memcpy-like-test.s
@@ -56,17 +56,17 @@ vmovaps %xmm0, 48(%rdi)
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
-# CHECK-NEXT:  -      -      -     2.00   2.00   3.97   4.03   4.00    -     4.00   4.00    -      -      -
+# CHECK-NEXT:  -      -      -      -      -      -     4.00   4.00    -     4.00   4.00    -      -      -
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
-# CHECK-NEXT:  -      -      -      -     1.00   0.98   0.02   1.00    -      -      -      -      -      -     vmovaps        (%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        (%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, (%rdi)
-# CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     vmovaps        16(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        16(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 16(%rdi)
-# CHECK-NEXT:  -      -      -      -     1.00   0.99   0.01   1.00    -      -      -      -      -      -     vmovaps        32(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        32(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 32(%rdi)
-# CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     vmovaps        48(%rsi), %xmm0
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        48(%rsi), %xmm0
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, 48(%rdi)
 
 # CHECK:      Timeline view:

diff --git a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-avx1.s b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-avx1.s
index 1378a11..4e7a941 100644 (file)
--- a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-avx1.s
+++ b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-avx1.s
@@ -1217,7 +1217,7 @@ vzeroupper
 # CHECK-NEXT:  1      1     0.50                        vinsertps      $1, %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  1      6     1.00    *                   vinsertps      $1, (%rax), %xmm1, %xmm2
 # CHECK-NEXT:  1      5     1.00    *                   vlddqu (%rax), %xmm2
-# CHECK-NEXT:  1      5     1.00    *                   vlddqu (%rax), %ymm2
+# CHECK-NEXT:  2      5     2.00    *                   vlddqu (%rax), %ymm2
 # CHECK-NEXT:  1      3     1.00    *             U     vldmxcsr       (%rax)
 # CHECK-NEXT:  63     34    21.00   *      *      U     vmaskmovdqu    %xmm0, %xmm1
 # CHECK-NEXT:  1      6     1.00    *                   vmaskmovpd     (%rax), %xmm0, %xmm2
@@ -1256,14 +1256,14 @@ vzeroupper
 # CHECK-NEXT:  1      1     1.00           *            vmovapd        %xmm0, (%rax)
 # CHECK-NEXT:  1      5     1.00    *                   vmovapd        (%rax), %xmm2
 # CHECK-NEXT:  2      1     1.00                        vmovapd        %ymm0, %ymm2
-# CHECK-NEXT:  1      1     1.00           *            vmovapd        %ymm0, (%rax)
-# CHECK-NEXT:  1      5     1.00    *                   vmovapd        (%rax), %ymm2
+# CHECK-NEXT:  2      1     2.00           *            vmovapd        %ymm0, (%rax)
+# CHECK-NEXT:  2      5     2.00    *                   vmovapd        (%rax), %ymm2
 # CHECK-NEXT:  1      1     0.50                        vmovaps        %xmm0, %xmm2
 # CHECK-NEXT:  1      1     1.00           *            vmovaps        %xmm0, (%rax)
 # CHECK-NEXT:  1      5     1.00    *                   vmovaps        (%rax), %xmm2
 # CHECK-NEXT:  2      1     1.00                        vmovaps        %ymm0, %ymm2
-# CHECK-NEXT:  1      1     1.00           *            vmovaps        %ymm0, (%rax)
-# CHECK-NEXT:  1      5     1.00    *                   vmovaps        (%rax), %ymm2
+# CHECK-NEXT:  2      1     2.00           *            vmovaps        %ymm0, (%rax)
+# CHECK-NEXT:  2      5     2.00    *                   vmovaps        (%rax), %ymm2
 # CHECK-NEXT:  2      8     0.50                        vmovd  %eax, %xmm2
 # CHECK-NEXT:  1      5     1.00    *                   vmovd  (%rax), %xmm2
 # CHECK-NEXT:  1      4     1.00                        vmovd  %xmm0, %ecx
@@ -1276,14 +1276,14 @@ vzeroupper
 # CHECK-NEXT:  1      1     1.00           *            vmovdqa        %xmm0, (%rax)
 # CHECK-NEXT:  1      5     1.00    *                   vmovdqa        (%rax), %xmm2
 # CHECK-NEXT:  2      1     1.00                        vmovdqa        %ymm0, %ymm2
-# CHECK-NEXT:  1      1     1.00           *            vmovdqa        %ymm0, (%rax)
-# CHECK-NEXT:  1      5     1.00    *                   vmovdqa        (%rax), %ymm2
+# CHECK-NEXT:  2      1     2.00           *            vmovdqa        %ymm0, (%rax)
+# CHECK-NEXT:  2      5     2.00    *                   vmovdqa        (%rax), %ymm2
 # CHECK-NEXT:  1      1     0.50                        vmovdqu        %xmm0, %xmm2
 # CHECK-NEXT:  1      1     1.00           *            vmovdqu        %xmm0, (%rax)
 # CHECK-NEXT:  1      5     1.00    *                   vmovdqu        (%rax), %xmm2
 # CHECK-NEXT:  2      1     1.00                        vmovdqu        %ymm0, %ymm2
-# CHECK-NEXT:  1      1     1.00           *            vmovdqu        %ymm0, (%rax)
-# CHECK-NEXT:  1      5     1.00    *                   vmovdqu        (%rax), %ymm2
+# CHECK-NEXT:  2      1     2.00           *            vmovdqu        %ymm0, (%rax)
+# CHECK-NEXT:  2      5     2.00    *                   vmovdqu        (%rax), %ymm2
 # CHECK-NEXT:  1      1     0.50                        vmovhlps       %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  1      1     0.50                        vmovlhps       %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  1      2     1.00           *            vmovhpd        %xmm0, (%rax)
@@ -1329,14 +1329,14 @@ vzeroupper
 # CHECK-NEXT:  1      1     1.00           *            vmovupd        %xmm0, (%rax)
 # CHECK-NEXT:  1      5     1.00    *                   vmovupd        (%rax), %xmm2
 # CHECK-NEXT:  2      1     1.00                        vmovupd        %ymm0, %ymm2
-# CHECK-NEXT:  1      1     1.00           *            vmovupd        %ymm0, (%rax)
-# CHECK-NEXT:  1      5     1.00    *                   vmovupd        (%rax), %ymm2
+# CHECK-NEXT:  2      1     2.00           *            vmovupd        %ymm0, (%rax)
+# CHECK-NEXT:  2      5     2.00    *                   vmovupd        (%rax), %ymm2
 # CHECK-NEXT:  1      1     0.50                        vmovups        %xmm0, %xmm2
 # CHECK-NEXT:  1      1     1.00           *            vmovups        %xmm0, (%rax)
 # CHECK-NEXT:  1      5     1.00    *                   vmovups        (%rax), %xmm2
 # CHECK-NEXT:  2      1     1.00                        vmovups        %ymm0, %ymm2
-# CHECK-NEXT:  1      1     1.00           *            vmovups        %ymm0, (%rax)
-# CHECK-NEXT:  1      5     1.00    *                   vmovups        (%rax), %ymm2
+# CHECK-NEXT:  2      1     2.00           *            vmovups        %ymm0, (%rax)
+# CHECK-NEXT:  2      5     2.00    *                   vmovups        (%rax), %ymm2
 # CHECK-NEXT:  3      3     2.00                        vmpsadbw       $1, %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  3      8     2.00    *                   vmpsadbw       $1, (%rax), %xmm1, %xmm2
 # CHECK-NEXT:  1      4     2.00                        vmulpd %xmm0, %xmm1, %xmm2
@@ -1740,7 +1740,7 @@ vzeroupper
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
-# CHECK-NEXT: 86.00  30.00   -     362.00 907.00 449.50 480.50 414.00  -     78.00  154.00 135.50 159.50 38.00
+# CHECK-NEXT: 86.00  30.00   -     358.00 903.00 442.50 479.50 421.00  -     84.00  160.00 132.50 156.50 38.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
@@ -1930,8 +1930,8 @@ vzeroupper
 # CHECK-NEXT:  -      -      -     1.00   1.00   1.00   1.00   1.00    -      -      -      -      -      -     vinsertf128    $1, (%rax), %ymm1, %ymm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vinsertps      $1, %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vinsertps      $1, (%rax), %xmm1, %xmm2
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     vlddqu (%rax), %xmm2
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     vlddqu (%rax), %ymm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vlddqu (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     2.00    -      -      -      -      -      -     vlddqu (%rax), %ymm2
 # CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vldmxcsr       (%rax)
 # CHECK-NEXT: 21.00  21.00   -     1.00    -     1.00   2.00   2.00    -     16.00  2.00    -      -      -     vmaskmovdqu    %xmm0, %xmm1
 # CHECK-NEXT:  -      -      -     1.00   1.00   1.00   1.00   1.00    -      -      -      -      -      -     vmaskmovpd     (%rax), %xmm0, %xmm2
@@ -1968,16 +1968,16 @@ vzeroupper
 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     vminss (%rax), %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vmovapd        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovapd        %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovapd        (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovapd        (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00   1.00   1.00   1.00    -      -      -      -      -      -      -     vmovapd        %ymm0, %ymm2
-# CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovapd        %ymm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovapd        (%rax), %ymm2
+# CHECK-NEXT:  -      -      -      -      -      -     2.00    -      -     2.00   2.00    -      -      -     vmovapd        %ymm0, (%rax)
+# CHECK-NEXT:  -      -      -      -      -      -      -     2.00    -      -      -      -      -      -     vmovapd        (%rax), %ymm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vmovaps        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovaps        (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovaps        (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00   1.00   1.00   1.00    -      -      -      -      -      -      -     vmovaps        %ymm0, %ymm2
-# CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovaps        %ymm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovaps        (%rax), %ymm2
+# CHECK-NEXT:  -      -      -      -      -      -     2.00    -      -     2.00   2.00    -      -      -     vmovaps        %ymm0, (%rax)
+# CHECK-NEXT:  -      -      -      -      -      -      -     2.00    -      -      -      -      -      -     vmovaps        (%rax), %ymm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vmovd  %eax, %xmm2
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     vmovd  (%rax), %xmm2
 # CHECK-NEXT: 1.00    -      -     1.00    -     1.00    -      -      -      -      -      -      -      -     vmovd  %xmm0, %ecx
@@ -1988,16 +1988,16 @@ vzeroupper
 # CHECK-NEXT:  -      -      -     1.00   1.00   1.00   1.00   2.00    -      -      -      -      -      -     vmovddup       (%rax), %ymm2
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -     0.50   0.50    -     vmovdqa        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovdqa        %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     vmovdqa        (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovdqa        (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -      -     1.00   1.00    -      -      -      -     1.00   1.00    -     vmovdqa        %ymm0, %ymm2
-# CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovdqa        %ymm0, (%rax)
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     vmovdqa        (%rax), %ymm2
+# CHECK-NEXT:  -      -      -      -      -      -     2.00    -      -     2.00   2.00    -      -      -     vmovdqa        %ymm0, (%rax)
+# CHECK-NEXT:  -      -      -      -      -      -      -     2.00    -      -      -      -      -      -     vmovdqa        (%rax), %ymm2
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -     0.50   0.50    -     vmovdqu        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovdqu        %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     vmovdqu        (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovdqu        (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -      -     1.00   1.00    -      -      -      -     1.00   1.00    -     vmovdqu        %ymm0, %ymm2
-# CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovdqu        %ymm0, (%rax)
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     vmovdqu        (%rax), %ymm2
+# CHECK-NEXT:  -      -      -      -      -      -     2.00    -      -     2.00   2.00    -      -      -     vmovdqu        %ymm0, (%rax)
+# CHECK-NEXT:  -      -      -      -      -      -      -     2.00    -      -      -      -      -      -     vmovdqu        (%rax), %ymm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vmovhlps       %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vmovlhps       %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovhpd        %xmm0, (%rax)
@@ -2041,16 +2041,16 @@ vzeroupper
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovss (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vmovupd        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovupd        %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovupd        (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovupd        (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00   1.00   1.00   1.00    -      -      -      -      -      -      -     vmovupd        %ymm0, %ymm2
-# CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovupd        %ymm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovupd        (%rax), %ymm2
+# CHECK-NEXT:  -      -      -      -      -      -     2.00    -      -     2.00   2.00    -      -      -     vmovupd        %ymm0, (%rax)
+# CHECK-NEXT:  -      -      -      -      -      -      -     2.00    -      -      -      -      -      -     vmovupd        (%rax), %ymm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     vmovups        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovups        %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovups        (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     vmovups        (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00   1.00   1.00   1.00    -      -      -      -      -      -      -     vmovups        %ymm0, %ymm2
-# CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     vmovups        %ymm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     vmovups        (%rax), %ymm2
+# CHECK-NEXT:  -      -      -      -      -      -     2.00    -      -     2.00   2.00    -      -      -     vmovups        %ymm0, (%rax)
+# CHECK-NEXT:  -      -      -      -      -      -      -     2.00    -      -      -      -      -      -     vmovups        (%rax), %ymm2
 # CHECK-NEXT:  -      -      -      -      -     1.00    -      -      -      -      -      -      -     2.00   vmpsadbw       $1, %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -      -      -     1.00    -     1.00    -      -      -      -      -     2.00   vmpsadbw       $1, (%rax), %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -      -     2.00    -     1.00    -      -      -      -      -      -      -     vmulpd %xmm0, %xmm1, %xmm2

diff --git a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse1.s b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse1.s
index a22c9d1..07efac5 100644 (file)
--- a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse1.s
+++ b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse1.s
@@ -337,7 +337,7 @@ xorps       (%rax), %xmm2
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
-# CHECK-NEXT: 15.00   -      -     47.00  184.00 58.50  66.50  58.00   -     9.00   25.00  9.50   9.50   2.00
+# CHECK-NEXT: 15.00   -      -     46.00  183.00 57.50  65.50  58.00   -     9.00   25.00  9.50   9.50   2.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
@@ -389,7 +389,7 @@ xorps       (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     minss  (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movaps %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movaps %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movaps (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     movaps (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movhlps        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movlhps        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movhps %xmm0, (%rax)
@@ -404,7 +404,7 @@ xorps       (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movss  (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movups %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movups %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movups (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     movups (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -     1.00    -     1.00    -      -      -      -      -      -      -     mulps  %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -     1.00    -     1.00   1.00    -      -      -      -      -      -     mulps  (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -     1.00    -     1.00    -      -      -      -      -      -      -     mulss  %xmm0, %xmm2

diff --git a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse2.s b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse2.s
index 924066a..63a29f9 100644 (file)
--- a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse2.s
+++ b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse2.s
@@ -693,7 +693,7 @@ xorpd       (%rax), %xmm2
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
-# CHECK-NEXT: 38.00  21.00   -     50.00  204.00 129.50 142.50 120.00  -     31.00  55.00  67.50  67.50  12.00
+# CHECK-NEXT: 38.00  21.00   -     49.00  203.00 127.50 140.50 120.00  -     31.00  55.00  66.50  66.50  12.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
@@ -767,17 +767,17 @@ xorpd       (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     minsd  (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movapd %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movapd %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movapd (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     movapd (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movd   %eax, %xmm2
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     movd   (%rax), %xmm2
 # CHECK-NEXT: 1.00    -      -     1.00    -     1.00    -      -      -      -      -      -      -      -     movd   %xmm0, %ecx
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movd   %xmm0, (%rax)
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -     0.50   0.50    -     movdqa %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movdqa %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     movdqa (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     movdqa (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -     0.50   0.50    -     movdqu %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movdqu %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     movdqu (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     movdqu (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -     0.50   0.50    -     movdq2q        %xmm0, %mm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movhpd %xmm0, (%rax)
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movhpd (%rax), %xmm2
@@ -799,7 +799,7 @@ xorpd       (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movsd  (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movupd %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -     1.00   1.00    -      -      -     movupd %xmm0, (%rax)
-# CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movupd (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     movupd (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -     2.00    -     1.00    -      -      -      -      -      -      -     mulpd  %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -     2.00    -     1.00   1.00    -      -      -      -      -      -     mulpd  (%rax), %xmm2
 # CHECK-NEXT:  -      -      -      -     2.00    -     1.00    -      -      -      -      -      -      -     mulsd  %xmm0, %xmm2

diff --git a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse3.s b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse3.s
index 581af93..77660b5 100644 (file)
--- a/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse3.s
+++ b/llvm/test/tools/llvm-mca/X86/BtVer2/resources-sse3.s
@@ -83,7 +83,7 @@ mwait
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
-# CHECK-NEXT: 1.00   1.00    -     15.00  3.00   15.50  3.50   10.00   -      -      -     0.50   0.50    -
+# CHECK-NEXT: 1.00   1.00    -     15.00  3.00   15.00  3.00   10.00   -      -      -      -      -      -
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
@@ -99,7 +99,7 @@ mwait
 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     hsubpd (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -      -      -      -      -      -      -      -     hsubps %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -     1.00    -      -      -      -      -      -     hsubps (%rax), %xmm2
-# CHECK-NEXT:  -      -      -      -      -     0.50   0.50   1.00    -      -      -     0.50   0.50    -     lddqu  (%rax), %xmm2
+# CHECK-NEXT:  -      -      -      -      -      -      -     1.00    -      -      -      -      -      -     lddqu  (%rax), %xmm2
 # CHECK-NEXT: 0.50   0.50    -      -      -      -      -      -      -      -      -      -      -      -     monitor
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50    -      -      -      -      -      -      -     movddup        %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50   0.50   0.50   1.00    -      -      -      -      -      -     movddup        (%rax), %xmm2