Skip to content

Commit 6ec4c9c

Browse files
committed
[MCA][X86] Add scatter instruction test coverage for #105675
1 parent eb549da commit 6ec4c9c

File tree

8 files changed

+320
-8
lines changed

8 files changed

+320
-8
lines changed

llvm/test/tools/llvm-mca/X86/Generic/resources-avx512.s

+27-1
Original file line numberDiff line numberDiff line change
@@ -814,6 +814,11 @@ vpermq %zmm16, %zmm17, %zmm19 {z}{k1}
814814
vpermq (%rax), %zmm17, %zmm19 {z}{k1}
815815
vpermq (%rax){1to8}, %zmm17, %zmm19 {z}{k1}
816816

817+
vpscatterdd %zmm1, (%rdx,%zmm0,4) {%k1}
818+
vpscatterdq %zmm1, (%rdx,%ymm0,4) {%k1}
819+
vpscatterqd %ymm1, (%rdx,%zmm0,4) {%k1}
820+
vpscatterqq %zmm1, (%rdx,%zmm0,4) {%k1}
821+
817822
vpshufd $0, %zmm16, %zmm19
818823
vpshufd $0, (%rax), %zmm19
819824
vpshufd $0, (%rax){1to16}, %zmm19
@@ -884,6 +889,11 @@ vpunpcklqdq %zmm16, %zmm17, %zmm19 {z}{k1}
884889
vpunpcklqdq (%rax), %zmm17, %zmm19 {z}{k1}
885890
vpunpcklqdq (%rax){1to8}, %zmm17, %zmm19 {z}{k1}
886891

892+
vscatterdps %zmm1, (%rdx,%zmm0,4) {%k1}
893+
vscatterdpd %zmm1, (%rdx,%ymm0,4) {%k1}
894+
vscatterqps %ymm1, (%rdx,%zmm0,4) {%k1}
895+
vscatterqpd %zmm1, (%rdx,%zmm0,4) {%k1}
896+
887897
vshuff32x4 $0, %zmm16, %zmm17, %zmm19
888898
vshuff32x4 $0, (%rax), %zmm17, %zmm19
889899
vshuff32x4 $0, (%rax){1to16}, %zmm17, %zmm19
@@ -1792,6 +1802,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1}
17921802
# CHECK-NEXT: 1 1 1.00 vpermq %zmm16, %zmm17, %zmm19 {%k1} {z}
17931803
# CHECK-NEXT: 2 8 1.00 * vpermq (%rax), %zmm17, %zmm19 {%k1} {z}
17941804
# CHECK-NEXT: 2 8 1.00 * vpermq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z}
1805+
# CHECK-NEXT: 1 1 1.00 * vpscatterdd %zmm1, (%rdx,%zmm0,4) {%k1}
1806+
# CHECK-NEXT: 1 1 1.00 * vpscatterdq %zmm1, (%rdx,%ymm0,4) {%k1}
1807+
# CHECK-NEXT: 1 1 1.00 * vpscatterqd %ymm1, (%rdx,%zmm0,4) {%k1}
1808+
# CHECK-NEXT: 1 1 1.00 * vpscatterqq %zmm1, (%rdx,%zmm0,4) {%k1}
17951809
# CHECK-NEXT: 1 1 1.00 vpshufd $0, %zmm16, %zmm19
17961810
# CHECK-NEXT: 2 8 1.00 * vpshufd $0, (%rax), %zmm19
17971811
# CHECK-NEXT: 2 8 1.00 * vpshufd $0, (%rax){1to16}, %zmm19
@@ -1855,6 +1869,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1}
18551869
# CHECK-NEXT: 1 1 1.00 vpunpcklqdq %zmm16, %zmm17, %zmm19 {%k1} {z}
18561870
# CHECK-NEXT: 2 8 1.00 * vpunpcklqdq (%rax), %zmm17, %zmm19 {%k1} {z}
18571871
# CHECK-NEXT: 2 8 1.00 * vpunpcklqdq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z}
1872+
# CHECK-NEXT: 1 1 1.00 * vscatterdps %zmm1, (%rdx,%zmm0,4) {%k1}
1873+
# CHECK-NEXT: 1 1 1.00 * vscatterdpd %zmm1, (%rdx,%ymm0,4) {%k1}
1874+
# CHECK-NEXT: 1 1 1.00 * vscatterqps %ymm1, (%rdx,%zmm0,4) {%k1}
1875+
# CHECK-NEXT: 1 1 1.00 * vscatterqpd %zmm1, (%rdx,%zmm0,4) {%k1}
18581876
# CHECK-NEXT: 1 1 1.00 vshuff32x4 $0, %zmm16, %zmm17, %zmm19
18591877
# CHECK-NEXT: 2 8 1.00 * vshuff32x4 $0, (%rax), %zmm17, %zmm19
18601878
# CHECK-NEXT: 2 8 1.00 * vshuff32x4 $0, (%rax){1to16}, %zmm17, %zmm19
@@ -2032,7 +2050,7 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1}
20322050

20332051
# CHECK: Resource pressure per iteration:
20342052
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1]
2035-
# CHECK-NEXT: - 1506.00 198.00 335.00 17.00 523.00 300.00 300.00
2053+
# CHECK-NEXT: - 1506.00 198.00 335.00 25.00 523.00 304.00 304.00
20362054

20372055
# CHECK: Resource pressure by instruction:
20382056
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1] Instructions:
@@ -2750,6 +2768,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1}
27502768
# CHECK-NEXT: - - - - - 1.00 - - vpermq %zmm16, %zmm17, %zmm19 {%k1} {z}
27512769
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpermq (%rax), %zmm17, %zmm19 {%k1} {z}
27522770
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpermq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z}
2771+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterdd %zmm1, (%rdx,%zmm0,4) {%k1}
2772+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterdq %zmm1, (%rdx,%ymm0,4) {%k1}
2773+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterqd %ymm1, (%rdx,%zmm0,4) {%k1}
2774+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterqq %zmm1, (%rdx,%zmm0,4) {%k1}
27532775
# CHECK-NEXT: - - - - - 1.00 - - vpshufd $0, %zmm16, %zmm19
27542776
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpshufd $0, (%rax), %zmm19
27552777
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpshufd $0, (%rax){1to16}, %zmm19
@@ -2813,6 +2835,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1}
28132835
# CHECK-NEXT: - - - - - 1.00 - - vpunpcklqdq %zmm16, %zmm17, %zmm19 {%k1} {z}
28142836
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpunpcklqdq (%rax), %zmm17, %zmm19 {%k1} {z}
28152837
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpunpcklqdq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z}
2838+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterdps %zmm1, (%rdx,%zmm0,4) {%k1}
2839+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterdpd %zmm1, (%rdx,%ymm0,4) {%k1}
2840+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterqps %ymm1, (%rdx,%zmm0,4) {%k1}
2841+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterqpd %zmm1, (%rdx,%zmm0,4) {%k1}
28162842
# CHECK-NEXT: - - - - - 1.00 - - vshuff32x4 $0, %zmm16, %zmm17, %zmm19
28172843
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vshuff32x4 $0, (%rax), %zmm17, %zmm19
28182844
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vshuff32x4 $0, (%rax){1to16}, %zmm17, %zmm19

llvm/test/tools/llvm-mca/X86/Generic/resources-avx512vl.s

+53-1
Original file line numberDiff line numberDiff line change
@@ -1344,6 +1344,16 @@ vpmulld %ymm16, %ymm17, %ymm19 {z}{k1}
13441344
vpmulld (%rax), %ymm17, %ymm19 {z}{k1}
13451345
vpmulld (%rax){1to8}, %ymm17, %ymm19 {z}{k1}
13461346

1347+
vpscatterdd %xmm1, (%rdx,%xmm0,4) {%k1}
1348+
vpscatterdq %xmm1, (%rdx,%xmm0,4) {%k1}
1349+
vpscatterqd %xmm1, (%rdx,%xmm0,4) {%k1}
1350+
vpscatterqq %xmm1, (%rdx,%xmm0,4) {%k1}
1351+
1352+
vpscatterdd %ymm1, (%rdx,%ymm0,4) {%k1}
1353+
vpscatterdq %ymm1, (%rdx,%xmm0,4) {%k1}
1354+
vpscatterqd %xmm1, (%rdx,%ymm0,4) {%k1}
1355+
vpscatterqq %ymm1, (%rdx,%ymm0,4) {%k1}
1356+
13471357
vpshufd $0, %xmm16, %xmm19
13481358
vpshufd $0, (%rax), %xmm19
13491359
vpshufd $0, (%rax){1to4}, %xmm19
@@ -1500,6 +1510,16 @@ vpunpckldq %ymm16, %ymm17, %ymm19 {z}{k1}
15001510
vpunpckldq (%rax), %ymm17, %ymm19 {z}{k1}
15011511
vpunpckldq (%rax){1to8}, %ymm17, %ymm19 {z}{k1}
15021512

1513+
vscatterdps %xmm1, (%rdx,%xmm0,4) {%k1}
1514+
vscatterdpd %xmm1, (%rdx,%xmm0,4) {%k1}
1515+
vscatterqps %xmm1, (%rdx,%xmm0,4) {%k1}
1516+
vscatterqpd %xmm1, (%rdx,%xmm0,4) {%k1}
1517+
1518+
vscatterdps %ymm1, (%rdx,%ymm0,4) {%k1}
1519+
vscatterdpd %ymm1, (%rdx,%xmm0,4) {%k1}
1520+
vscatterqps %xmm1, (%rdx,%ymm0,4) {%k1}
1521+
vscatterqpd %ymm1, (%rdx,%ymm0,4) {%k1}
1522+
15031523
vshuff32x4 $0, %ymm16, %ymm17, %ymm19
15041524
vshuff32x4 $0, (%rax), %ymm17, %ymm19
15051525
vshuff32x4 $0, (%rax){1to8}, %ymm17, %ymm19
@@ -2897,6 +2917,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1}
28972917
# CHECK-NEXT: 1 5 1.00 vpmulld %ymm16, %ymm17, %ymm19 {%k1} {z}
28982918
# CHECK-NEXT: 2 12 1.00 * vpmulld (%rax), %ymm17, %ymm19 {%k1} {z}
28992919
# CHECK-NEXT: 2 12 1.00 * vpmulld (%rax){1to8}, %ymm17, %ymm19 {%k1} {z}
2920+
# CHECK-NEXT: 1 1 1.00 * vpscatterdd %xmm1, (%rdx,%xmm0,4) {%k1}
2921+
# CHECK-NEXT: 1 1 1.00 * vpscatterdq %xmm1, (%rdx,%xmm0,4) {%k1}
2922+
# CHECK-NEXT: 1 1 1.00 * vpscatterqd %xmm1, (%rdx,%xmm0,4) {%k1}
2923+
# CHECK-NEXT: 1 1 1.00 * vpscatterqq %xmm1, (%rdx,%xmm0,4) {%k1}
2924+
# CHECK-NEXT: 1 1 1.00 * vpscatterdd %ymm1, (%rdx,%ymm0,4) {%k1}
2925+
# CHECK-NEXT: 1 1 1.00 * vpscatterdq %ymm1, (%rdx,%xmm0,4) {%k1}
2926+
# CHECK-NEXT: 1 1 1.00 * vpscatterqd %xmm1, (%rdx,%ymm0,4) {%k1}
2927+
# CHECK-NEXT: 1 1 1.00 * vpscatterqq %ymm1, (%rdx,%ymm0,4) {%k1}
29002928
# CHECK-NEXT: 1 1 0.50 vpshufd $0, %xmm16, %xmm19
29012929
# CHECK-NEXT: 2 7 0.50 * vpshufd $0, (%rax), %xmm19
29022930
# CHECK-NEXT: 2 7 0.50 * vpshufd $0, (%rax){1to4}, %xmm19
@@ -3035,6 +3063,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1}
30353063
# CHECK-NEXT: 1 1 1.00 vpunpckldq %ymm16, %ymm17, %ymm19 {%k1} {z}
30363064
# CHECK-NEXT: 2 8 1.00 * vpunpckldq (%rax), %ymm17, %ymm19 {%k1} {z}
30373065
# CHECK-NEXT: 2 8 1.00 * vpunpckldq (%rax){1to8}, %ymm17, %ymm19 {%k1} {z}
3066+
# CHECK-NEXT: 1 1 1.00 * vscatterdps %xmm1, (%rdx,%xmm0,4) {%k1}
3067+
# CHECK-NEXT: 1 1 1.00 * vscatterdpd %xmm1, (%rdx,%xmm0,4) {%k1}
3068+
# CHECK-NEXT: 1 1 1.00 * vscatterqps %xmm1, (%rdx,%xmm0,4) {%k1}
3069+
# CHECK-NEXT: 1 1 1.00 * vscatterqpd %xmm1, (%rdx,%xmm0,4) {%k1}
3070+
# CHECK-NEXT: 1 1 1.00 * vscatterdps %ymm1, (%rdx,%ymm0,4) {%k1}
3071+
# CHECK-NEXT: 1 1 1.00 * vscatterdpd %ymm1, (%rdx,%xmm0,4) {%k1}
3072+
# CHECK-NEXT: 1 1 1.00 * vscatterqps %xmm1, (%rdx,%ymm0,4) {%k1}
3073+
# CHECK-NEXT: 1 1 1.00 * vscatterqpd %ymm1, (%rdx,%ymm0,4) {%k1}
30383074
# CHECK-NEXT: 1 1 1.00 vshuff32x4 $0, %ymm16, %ymm17, %ymm19
30393075
# CHECK-NEXT: 2 8 1.00 * vshuff32x4 $0, (%rax), %ymm17, %ymm19
30403076
# CHECK-NEXT: 2 8 1.00 * vshuff32x4 $0, (%rax){1to8}, %ymm17, %ymm19
@@ -3228,7 +3264,7 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1}
32283264

32293265
# CHECK: Resource pressure per iteration:
32303266
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1]
3231-
# CHECK-NEXT: - 1935.00 278.00 579.50 32.00 738.50 486.50 486.50
3267+
# CHECK-NEXT: - 1935.00 278.00 579.50 48.00 738.50 494.50 494.50
32323268

32333269
# CHECK: Resource pressure by instruction:
32343270
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1] Instructions:
@@ -4420,6 +4456,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1}
44204456
# CHECK-NEXT: - - 1.00 - - - - - vpmulld %ymm16, %ymm17, %ymm19 {%k1} {z}
44214457
# CHECK-NEXT: - - 1.00 - - - 0.50 0.50 vpmulld (%rax), %ymm17, %ymm19 {%k1} {z}
44224458
# CHECK-NEXT: - - 1.00 - - - 0.50 0.50 vpmulld (%rax){1to8}, %ymm17, %ymm19 {%k1} {z}
4459+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterdd %xmm1, (%rdx,%xmm0,4) {%k1}
4460+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterdq %xmm1, (%rdx,%xmm0,4) {%k1}
4461+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterqd %xmm1, (%rdx,%xmm0,4) {%k1}
4462+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterqq %xmm1, (%rdx,%xmm0,4) {%k1}
4463+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterdd %ymm1, (%rdx,%ymm0,4) {%k1}
4464+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterdq %ymm1, (%rdx,%xmm0,4) {%k1}
4465+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterqd %xmm1, (%rdx,%ymm0,4) {%k1}
4466+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vpscatterqq %ymm1, (%rdx,%ymm0,4) {%k1}
44234467
# CHECK-NEXT: - - - 0.50 - 0.50 - - vpshufd $0, %xmm16, %xmm19
44244468
# CHECK-NEXT: - - - 0.50 - 0.50 0.50 0.50 vpshufd $0, (%rax), %xmm19
44254469
# CHECK-NEXT: - - - 0.50 - 0.50 0.50 0.50 vpshufd $0, (%rax){1to4}, %xmm19
@@ -4558,6 +4602,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1}
45584602
# CHECK-NEXT: - - - - - 1.00 - - vpunpckldq %ymm16, %ymm17, %ymm19 {%k1} {z}
45594603
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpunpckldq (%rax), %ymm17, %ymm19 {%k1} {z}
45604604
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vpunpckldq (%rax){1to8}, %ymm17, %ymm19 {%k1} {z}
4605+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterdps %xmm1, (%rdx,%xmm0,4) {%k1}
4606+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterdpd %xmm1, (%rdx,%xmm0,4) {%k1}
4607+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterqps %xmm1, (%rdx,%xmm0,4) {%k1}
4608+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterqpd %xmm1, (%rdx,%xmm0,4) {%k1}
4609+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterdps %ymm1, (%rdx,%ymm0,4) {%k1}
4610+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterdpd %ymm1, (%rdx,%xmm0,4) {%k1}
4611+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterqps %xmm1, (%rdx,%ymm0,4) {%k1}
4612+
# CHECK-NEXT: - - - - 1.00 - 0.50 0.50 vscatterqpd %ymm1, (%rdx,%ymm0,4) {%k1}
45614613
# CHECK-NEXT: - - - - - 1.00 - - vshuff32x4 $0, %ymm16, %ymm17, %ymm19
45624614
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vshuff32x4 $0, (%rax), %ymm17, %ymm19
45634615
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vshuff32x4 $0, (%rax){1to8}, %ymm17, %ymm19

0 commit comments

Comments
 (0)