[llvm] f407f2d - AMDGPU: Test a few more cycle counts for mfma ops
Matt Arsenault via llvm-commits
llvm-commits at lists.llvm.org
Mon Mar 4 07:00:13 PST 2024
Author: Matt Arsenault
Date: 2024-03-04T20:30:01+05:30
New Revision: f407f2df2ced195264d1a81df84143dc87c0cd24
URL: https://github.com/llvm/llvm-project/commit/f407f2df2ced195264d1a81df84143dc87c0cd24
DIFF: https://github.com/llvm/llvm-project/commit/f407f2df2ced195264d1a81df84143dc87c0cd24.diff
LOG: AMDGPU: Test a few more cycle counts for mfma ops
Added:
Modified:
llvm/test/tools/llvm-mca/AMDGPU/gfx940-mfma.s
Removed:
################################################################################
diff --git a/llvm/test/tools/llvm-mca/AMDGPU/gfx940-mfma.s b/llvm/test/tools/llvm-mca/AMDGPU/gfx940-mfma.s
index 323acbccfe5499..e7ddeee97ea924 100644
--- a/llvm/test/tools/llvm-mca/AMDGPU/gfx940-mfma.s
+++ b/llvm/test/tools/llvm-mca/AMDGPU/gfx940-mfma.s
@@ -1,18 +1,9 @@
# RUN: llvm-mca -mtriple=amdgcn -mcpu=gfx940 --timeline --iterations=1 --timeline-max-cycles=0 < %s | FileCheck %s
# CHECK: Iterations: 1
-# CHECK: Instructions: 58
-# CHECK: Total Cycles: 543
-# CHECK: Total uOps: 58
-
-# CHECK: Resources:
-# CHECK: [0] - HWBranch
-# CHECK: [1] - HWExport
-# CHECK: [2] - HWLGKM
-# CHECK: [3] - HWSALU
-# CHECK: [4] - HWVALU
-# CHECK: [5] - HWVMEM
-# CHECK: [6] - HWXDL
+# CHECK: Instructions: 78
+# CHECK: Total Cycles: 699
+# CHECK: Total uOps: 78
v_mfma_f32_16x16x4_f32 a[0:3], v0, v1, a[2:5]
v_mfma_f32_16x16x4_f32 v[0:3], v0, v1, v[2:5]
@@ -101,6 +92,57 @@ v_smfmac_i32_16x16x64_i8 a[10:13], v[2:3], a[4:7], v9
v_smfmac_i32_32x32x32_i8 v[10:25], a[2:3], v[4:7], v10 cbsz:3 abid:1
v_smfmac_i32_32x32x32_i8 a[10:25], v[2:3], a[4:7], v11
+v_mfma_f32_16x16x32_bf8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
+v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
+
+v_mfma_f32_16x16x32_bf8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
+v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
+
+v_mfma_f32_16x16x32_fp8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
+v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
+
+v_mfma_f32_16x16x32_fp8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
+v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
+
+v_mfma_f32_32x32x16_bf8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
+v_mfma_f32_32x32x16_fp8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
+v_mfma_f32_32x32x16_bf8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
+v_mfma_f32_32x32x16_fp8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
+
+v_smfmac_f32_16x16x64_bf8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+v_smfmac_f32_16x16x64_bf8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+v_smfmac_f32_16x16x64_fp8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+v_smfmac_f32_16x16x64_fp8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+
+v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
+v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
+v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
+v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
+
+# CHECK: Instruction Info:
+# CHECK-NEXT:[1]: #uOps
+# CHECK-NEXT:[2]: Latency
+# CHECK-NEXT:[3]: RThroughput
+# CHECK-NEXT:[4]: MayLoad
+# CHECK-NEXT:[5]: MayStore
+# CHECK-NEXT:[6]: HasSideEffects (U)
+
+# CHECK: [1] [2] [3] [4] [5] [6] Instructions:
+
+# CHECK: 1 8 1.00 U v_mfma_f64_4x4x4_4b_f64 a[0:1], v[0:1], a[2:3], a[2:3]
+# CHECK-NEXT: 1 8 1.00 U v_mfma_f64_4x4x4_4b_f64 v[0:1], v[0:1], v[2:3], v[2:3]
+# CHECK-NEXT: 1 20 1.00 U v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7]
+# CHECK-NEXT: 1 20 1.00 U v_mfma_f64_16x16x4_f64 v[0:7], v[0:1], v[2:3], v[0:7]
+
+# CHECK: Resources:
+# CHECK: [0] - HWBranch
+# CHECK: [1] - HWExport
+# CHECK: [2] - HWLGKM
+# CHECK: [3] - HWSALU
+# CHECK: [4] - HWVALU
+# CHECK: [5] - HWVMEM
+# CHECK: [6] - HWXDL
+
# CHECK: [0] [1] [2] [3] [4] [5] [6] Instructions:
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_16x16x4_f32 a[0:3], v0, v1, a[2:5]
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_16x16x4_f32 v[0:3], v0, v1, v[2:5]
@@ -160,3 +202,23 @@ v_smfmac_i32_32x32x32_i8 a[10:25], v[2:3], a[4:7], v11
# CHECK-NEXT: - - - - - - 4.00 v_smfmac_i32_16x16x64_i8 a[10:13], v[2:3], a[4:7], v9
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_i32_32x32x32_i8 v[10:25], a[2:3], v[4:7], v10 cbsz:3 abid:1
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_i32_32x32x32_i8 a[10:25], v[2:3], a[4:7], v11
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
+# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
+# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_bf8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
+# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_fp8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
+# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_bf8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
+# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_fp8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
+# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_bf8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_bf8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_fp8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_fp8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
+# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
+# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
+# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
+# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
More information about the llvm-commits
mailing list