From 180f1832e69d5253ec4db82aa99b86e246d3c989 Mon Sep 17 00:00:00 2001 From: Simon Pilgrim Date: Thu, 22 Aug 2024 18:07:44 +0100 Subject: [PATCH] [MCA][X86] Add scatter instruction test coverage for #105675 Missed IceLakeServer when I updated the other CPUs in 6ec4c9c3eb4a556f848dac37a2d6f0d46ecc6f02 --- .../X86/IceLakeServer/resources-avx512.s | 28 +++++++++- .../X86/IceLakeServer/resources-avx512vl.s | 54 ++++++++++++++++++- 2 files changed, 80 insertions(+), 2 deletions(-) diff --git a/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512.s b/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512.s index 1ff8eccf290a6ff..c4df992f3aebcad 100644 --- a/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512.s +++ b/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512.s @@ -814,6 +814,11 @@ vpermq %zmm16, %zmm17, %zmm19 {z}{k1} vpermq (%rax), %zmm17, %zmm19 {z}{k1} vpermq (%rax){1to8}, %zmm17, %zmm19 {z}{k1} +vpscatterdd %zmm1, (%rdx,%zmm0,4) {%k1} +vpscatterdq %zmm1, (%rdx,%ymm0,4) {%k1} +vpscatterqd %ymm1, (%rdx,%zmm0,4) {%k1} +vpscatterqq %zmm1, (%rdx,%zmm0,4) {%k1} + vpshufd $0, %zmm16, %zmm19 vpshufd $0, (%rax), %zmm19 vpshufd $0, (%rax){1to16}, %zmm19 @@ -884,6 +889,11 @@ vpunpcklqdq %zmm16, %zmm17, %zmm19 {z}{k1} vpunpcklqdq (%rax), %zmm17, %zmm19 {z}{k1} vpunpcklqdq (%rax){1to8}, %zmm17, %zmm19 {z}{k1} +vscatterdps %zmm1, (%rdx,%zmm0,4) {%k1} +vscatterdpd %zmm1, (%rdx,%ymm0,4) {%k1} +vscatterqps %ymm1, (%rdx,%zmm0,4) {%k1} +vscatterqpd %zmm1, (%rdx,%zmm0,4) {%k1} + vshuff32x4 $0, %zmm16, %zmm17, %zmm19 vshuff32x4 $0, (%rax), %zmm17, %zmm19 vshuff32x4 $0, (%rax){1to16}, %zmm17, %zmm19 @@ -1792,6 +1802,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1} # CHECK-NEXT: 1 3 1.00 vpermq %zmm16, %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: 2 10 1.00 * vpermq (%rax), %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: 2 10 1.00 * vpermq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z} +# CHECK-NEXT: 36 8 8.00 * vpscatterdd %zmm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: 19 7 4.00 * vpscatterdq %zmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: 1 1 0.50 * vpscatterqd %ymm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: 19 7 4.00 * vpscatterqq %zmm1, (%rdx,%zmm0,4) {%k1} # CHECK-NEXT: 1 1 1.00 vpshufd $0, %zmm16, %zmm19 # CHECK-NEXT: 2 8 1.00 * vpshufd $0, (%rax), %zmm19 # CHECK-NEXT: 2 8 1.00 * vpshufd $0, (%rax){1to16}, %zmm19 @@ -1855,6 +1869,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1} # CHECK-NEXT: 1 1 1.00 vpunpcklqdq %zmm16, %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: 2 8 1.00 * vpunpcklqdq (%rax), %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: 2 8 1.00 * vpunpcklqdq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z} +# CHECK-NEXT: 36 7 8.00 * vscatterdps %zmm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: 19 7 4.00 * vscatterdpd %zmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: 1 1 0.50 * vscatterqps %ymm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: 19 7 4.00 * vscatterqpd %zmm1, (%rdx,%zmm0,4) {%k1} # CHECK-NEXT: 1 3 1.00 vshuff32x4 $0, %zmm16, %zmm17, %zmm19 # CHECK-NEXT: 2 10 1.00 * vshuff32x4 $0, (%rax), %zmm17, %zmm19 # CHECK-NEXT: 2 10 1.00 * vshuff32x4 $0, (%rax){1to16}, %zmm17, %zmm19 @@ -2036,7 +2054,7 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1} # CHECK: Resource pressure per iteration: # CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] -# CHECK-NEXT: - 612.00 399.17 99.67 327.50 327.50 8.50 587.17 2.00 8.50 8.50 8.50 +# CHECK-NEXT: - 612.00 408.17 102.67 327.50 327.50 41.50 592.17 5.00 41.50 41.50 41.50 # CHECK: Resource pressure by instruction: # CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] Instructions: @@ -2754,6 +2772,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1} # CHECK-NEXT: - - - - - - - 1.00 - - - - vpermq %zmm16, %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vpermq (%rax), %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vpermq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z} +# CHECK-NEXT: - - 1.50 0.50 - - 8.00 1.50 0.50 8.00 8.00 8.00 vpscatterdd %zmm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 4.00 0.50 0.50 4.00 4.00 4.00 vpscatterdq %zmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: - - - - - - 0.50 - - 0.50 0.50 0.50 vpscatterqd %ymm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 4.00 0.50 0.50 4.00 4.00 4.00 vpscatterqq %zmm1, (%rdx,%zmm0,4) {%k1} # CHECK-NEXT: - - - - - - - 1.00 - - - - vpshufd $0, %zmm16, %zmm19 # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vpshufd $0, (%rax), %zmm19 # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vpshufd $0, (%rax){1to16}, %zmm19 @@ -2817,6 +2839,10 @@ vunpcklps (%rax){1to16}, %zmm17, %zmm19 {z}{k1} # CHECK-NEXT: - - - - - - - 1.00 - - - - vpunpcklqdq %zmm16, %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vpunpcklqdq (%rax), %zmm17, %zmm19 {%k1} {z} # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vpunpcklqdq (%rax){1to8}, %zmm17, %zmm19 {%k1} {z} +# CHECK-NEXT: - - 1.50 0.50 - - 8.00 1.50 0.50 8.00 8.00 8.00 vscatterdps %zmm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 4.00 0.50 0.50 4.00 4.00 4.00 vscatterdpd %zmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: - - - - - - 0.50 - - 0.50 0.50 0.50 vscatterqps %ymm1, (%rdx,%zmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 4.00 0.50 0.50 4.00 4.00 4.00 vscatterqpd %zmm1, (%rdx,%zmm0,4) {%k1} # CHECK-NEXT: - - - - - - - 1.00 - - - - vshuff32x4 $0, %zmm16, %zmm17, %zmm19 # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vshuff32x4 $0, (%rax), %zmm17, %zmm19 # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vshuff32x4 $0, (%rax){1to16}, %zmm17, %zmm19 diff --git a/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512vl.s b/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512vl.s index 375087ae0cfe4e7..00e5c3b03f6f529 100644 --- a/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512vl.s +++ b/llvm/test/tools/llvm-mca/X86/IceLakeServer/resources-avx512vl.s @@ -1344,6 +1344,16 @@ vpmulld %ymm16, %ymm17, %ymm19 {z}{k1} vpmulld (%rax), %ymm17, %ymm19 {z}{k1} vpmulld (%rax){1to8}, %ymm17, %ymm19 {z}{k1} +vpscatterdd %xmm1, (%rdx,%xmm0,4) {%k1} +vpscatterdq %xmm1, (%rdx,%xmm0,4) {%k1} +vpscatterqd %xmm1, (%rdx,%xmm0,4) {%k1} +vpscatterqq %xmm1, (%rdx,%xmm0,4) {%k1} + +vpscatterdd %ymm1, (%rdx,%ymm0,4) {%k1} +vpscatterdq %ymm1, (%rdx,%xmm0,4) {%k1} +vpscatterqd %xmm1, (%rdx,%ymm0,4) {%k1} +vpscatterqq %ymm1, (%rdx,%ymm0,4) {%k1} + vpshufd $0, %xmm16, %xmm19 vpshufd $0, (%rax), %xmm19 vpshufd $0, (%rax){1to4}, %xmm19 @@ -1500,6 +1510,16 @@ vpunpckldq %ymm16, %ymm17, %ymm19 {z}{k1} vpunpckldq (%rax), %ymm17, %ymm19 {z}{k1} vpunpckldq (%rax){1to8}, %ymm17, %ymm19 {z}{k1} +vscatterdps %xmm1, (%rdx,%xmm0,4) {%k1} +vscatterdpd %xmm1, (%rdx,%xmm0,4) {%k1} +vscatterqps %xmm1, (%rdx,%xmm0,4) {%k1} +vscatterqpd %xmm1, (%rdx,%xmm0,4) {%k1} + +vscatterdps %ymm1, (%rdx,%ymm0,4) {%k1} +vscatterdpd %ymm1, (%rdx,%xmm0,4) {%k1} +vscatterqps %xmm1, (%rdx,%ymm0,4) {%k1} +vscatterqpd %ymm1, (%rdx,%ymm0,4) {%k1} + vshuff32x4 $0, %ymm16, %ymm17, %ymm19 vshuff32x4 $0, (%rax), %ymm17, %ymm19 vshuff32x4 $0, (%rax){1to8}, %ymm17, %ymm19 @@ -2897,6 +2917,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1} # CHECK-NEXT: 2 10 1.00 vpmulld %ymm16, %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: 3 17 1.00 * vpmulld (%rax), %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: 3 17 1.00 * vpmulld (%rax){1to8}, %ymm17, %ymm19 {%k1} {z} +# CHECK-NEXT: 12 8 2.00 * vpscatterdd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 7 7 1.00 * vpscatterdq %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 8 8 1.00 * vpscatterqd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 7 7 1.00 * vpscatterqq %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 20 8 4.00 * vpscatterdd %ymm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: 11 7 2.00 * vpscatterdq %ymm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 8 8 1.00 * vpscatterqd %xmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: 11 7 2.00 * vpscatterqq %ymm1, (%rdx,%ymm0,4) {%k1} # CHECK-NEXT: 1 1 0.50 vpshufd $0, %xmm16, %xmm19 # CHECK-NEXT: 2 7 0.50 * vpshufd $0, (%rax), %xmm19 # CHECK-NEXT: 2 7 0.50 * vpshufd $0, (%rax){1to4}, %xmm19 @@ -3035,6 +3063,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1} # CHECK-NEXT: 1 1 0.50 vpunpckldq %ymm16, %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: 2 8 0.50 * vpunpckldq (%rax), %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: 2 8 0.50 * vpunpckldq (%rax){1to8}, %ymm17, %ymm19 {%k1} {z} +# CHECK-NEXT: 12 8 2.00 * vscatterdps %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 7 7 1.00 * vscatterdpd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 8 8 1.00 * vscatterqps %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 7 7 1.00 * vscatterqpd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 20 8 4.00 * vscatterdps %ymm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: 11 7 2.00 * vscatterdpd %ymm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: 8 8 1.00 * vscatterqps %xmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: 11 7 2.00 * vscatterqpd %ymm1, (%rdx,%ymm0,4) {%k1} # CHECK-NEXT: 1 3 1.00 vshuff32x4 $0, %ymm16, %ymm17, %ymm19 # CHECK-NEXT: 2 10 1.00 * vshuff32x4 $0, (%rax), %ymm17, %ymm19 # CHECK-NEXT: 2 10 1.00 * vshuff32x4 $0, (%rax){1to8}, %ymm17, %ymm19 @@ -3232,7 +3268,7 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1} # CHECK: Resource pressure per iteration: # CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] -# CHECK-NEXT: - 423.00 438.33 413.33 492.50 492.50 16.00 722.33 4.00 16.00 16.00 16.00 +# CHECK-NEXT: - 423.00 462.33 421.33 492.50 492.50 44.00 738.33 12.00 44.00 44.00 44.00 # CHECK: Resource pressure by instruction: # CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] Instructions: @@ -4424,6 +4460,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1} # CHECK-NEXT: - - 1.00 1.00 - - - - - - - - vpmulld %ymm16, %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: - - 1.00 1.00 0.50 0.50 - - - - - - vpmulld (%rax), %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: - - 1.00 1.00 0.50 0.50 - - - - - - vpmulld (%rax){1to8}, %ymm17, %ymm19 {%k1} {z} +# CHECK-NEXT: - - 1.50 0.50 - - 2.00 1.50 0.50 2.00 2.00 2.00 vpscatterdd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 0.50 0.50 1.00 1.00 1.00 vpscatterdq %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 1.50 0.50 1.00 1.00 1.00 vpscatterqd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 0.50 0.50 1.00 1.00 1.00 vpscatterqq %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 4.00 1.50 0.50 4.00 4.00 4.00 vpscatterdd %ymm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 2.00 0.50 0.50 2.00 2.00 2.00 vpscatterdq %ymm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 1.50 0.50 1.00 1.00 1.00 vpscatterqd %xmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 2.00 0.50 0.50 2.00 2.00 2.00 vpscatterqq %ymm1, (%rdx,%ymm0,4) {%k1} # CHECK-NEXT: - - - 0.50 - - - 0.50 - - - - vpshufd $0, %xmm16, %xmm19 # CHECK-NEXT: - - - 0.50 0.50 0.50 - 0.50 - - - - vpshufd $0, (%rax), %xmm19 # CHECK-NEXT: - - - 0.50 0.50 0.50 - 0.50 - - - - vpshufd $0, (%rax){1to4}, %xmm19 @@ -4562,6 +4606,14 @@ vunpcklps (%rax){1to8}, %ymm17, %ymm19 {z}{k1} # CHECK-NEXT: - - - 0.50 - - - 0.50 - - - - vpunpckldq %ymm16, %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: - - - 0.50 0.50 0.50 - 0.50 - - - - vpunpckldq (%rax), %ymm17, %ymm19 {%k1} {z} # CHECK-NEXT: - - - 0.50 0.50 0.50 - 0.50 - - - - vpunpckldq (%rax){1to8}, %ymm17, %ymm19 {%k1} {z} +# CHECK-NEXT: - - 1.50 0.50 - - 2.00 1.50 0.50 2.00 2.00 2.00 vscatterdps %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 0.50 0.50 1.00 1.00 1.00 vscatterdpd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 1.50 0.50 1.00 1.00 1.00 vscatterqps %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 0.50 0.50 1.00 1.00 1.00 vscatterqpd %xmm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 4.00 1.50 0.50 4.00 4.00 4.00 vscatterdps %ymm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 2.00 0.50 0.50 2.00 2.00 2.00 vscatterdpd %ymm1, (%rdx,%xmm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 1.00 1.50 0.50 1.00 1.00 1.00 vscatterqps %xmm1, (%rdx,%ymm0,4) {%k1} +# CHECK-NEXT: - - 1.50 0.50 - - 2.00 0.50 0.50 2.00 2.00 2.00 vscatterqpd %ymm1, (%rdx,%ymm0,4) {%k1} # CHECK-NEXT: - - - - - - - 1.00 - - - - vshuff32x4 $0, %ymm16, %ymm17, %ymm19 # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vshuff32x4 $0, (%rax), %ymm17, %ymm19 # CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - - - vshuff32x4 $0, (%rax){1to8}, %ymm17, %ymm19