139 lines · plain
1# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py2# RUN: llvm-mca -mtriple=x86_64-unknown-unknown -mcpu=bdver2 -iterations=100 -resource-pressure=false -timeline -timeline-max-iterations=2 < %s | FileCheck %s3 4# In this test, the VDIVPS takes 38 cycles to write to register YMM3. The first5# VADDPS does not depend on the VDIVPS (the WAW dependency is eliminated at6# register renaming stage). So the first VADDPS can be executed in parallel to7# the VDIVPS. That VADDPS also writes to register XMM3, and the upper half of8# YMM3 is implicitly cleared. As a consequence, the definition of YMM3 from the9# VDIVPS is killed, and the subsequent VADDPS instructions don't need to wait10# for the VDIVPS to complete.11# The block reciprocal throughput is limited by the VDIVPS reciprocal throughput12# (which is 38 cycles). The sequence of VADDPS can be executed in parallel on13# the FPA unit; their latency is "hidden" by the long latency of the VDIVPS.14 15vdivps %ymm0, %ymm1, %ymm316vaddps %xmm0, %xmm1, %xmm317vaddps %ymm3, %ymm1, %ymm418vaddps %ymm3, %ymm1, %ymm419vaddps %ymm3, %ymm1, %ymm420vaddps %ymm3, %ymm1, %ymm421vaddps %ymm3, %ymm1, %ymm422vaddps %ymm3, %ymm1, %ymm423vaddps %ymm3, %ymm1, %ymm424vaddps %ymm3, %ymm1, %ymm425vaddps %ymm3, %ymm1, %ymm426vaddps %ymm3, %ymm1, %ymm427vaddps %ymm3, %ymm1, %ymm428vaddps %ymm3, %ymm1, %ymm429vaddps %ymm3, %ymm1, %ymm430vaddps %ymm3, %ymm1, %ymm431vaddps %ymm3, %ymm1, %ymm432vandps %xmm4, %xmm1, %xmm033 34# CHECK: Iterations: 10035# CHECK-NEXT: Instructions: 180036# CHECK-NEXT: Total Cycles: 320337# CHECK-NEXT: Total uOps: 340038 39# CHECK: Dispatch Width: 440# CHECK-NEXT: uOps Per Cycle: 1.0641# CHECK-NEXT: IPC: 0.5642# CHECK-NEXT: Block RThroughput: 24.543 44# CHECK: Instruction Info:45# CHECK-NEXT: [1]: #uOps46# CHECK-NEXT: [2]: Latency47# CHECK-NEXT: [3]: RThroughput48# CHECK-NEXT: [4]: MayLoad49# CHECK-NEXT: [5]: MayStore50# CHECK-NEXT: [6]: HasSideEffects (U)51 52# CHECK: [1] [2] [3] [4] [5] [6] Instructions:53# CHECK-NEXT: 2 9 9.00 vdivps %ymm0, %ymm1, %ymm354# CHECK-NEXT: 1 5 1.00 vaddps %xmm0, %xmm1, %xmm355# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm456# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm457# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm458# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm459# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm460# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm461# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm462# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm463# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm464# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm465# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm466# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm467# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm468# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm469# CHECK-NEXT: 2 5 1.00 vaddps %ymm3, %ymm1, %ymm470# CHECK-NEXT: 1 2 0.50 vandps %xmm4, %xmm1, %xmm071 72# CHECK: Timeline view:73# CHECK-NEXT: 0123456789 0123456789 012345678974# CHECK-NEXT: Index 0123456789 0123456789 0123456789 012345675 76# CHECK: [0,0] DeeeeeeeeeER . . . . . . . . . . .. vdivps %ymm0, %ymm1, %ymm377# CHECK-NEXT: [0,1] DeeeeeE----R . . . . . . . . . . .. vaddps %xmm0, %xmm1, %xmm378# CHECK-NEXT: [0,2] .D====eeeeeER . . . . . . . . . . .. vaddps %ymm3, %ymm1, %ymm479# CHECK-NEXT: [0,3] .D======eeeeeER. . . . . . . . . . .. vaddps %ymm3, %ymm1, %ymm480# CHECK-NEXT: [0,4] . D=======eeeeeER . . . . . . . . . .. vaddps %ymm3, %ymm1, %ymm481# CHECK-NEXT: [0,5] . D=========eeeeeER . . . . . . . . . .. vaddps %ymm3, %ymm1, %ymm482# CHECK-NEXT: [0,6] . D==========eeeeeER . . . . . . . . .. vaddps %ymm3, %ymm1, %ymm483# CHECK-NEXT: [0,7] . D============eeeeeER . . . . . . . . .. vaddps %ymm3, %ymm1, %ymm484# CHECK-NEXT: [0,8] . D=============eeeeeER. . . . . . . . .. vaddps %ymm3, %ymm1, %ymm485# CHECK-NEXT: [0,9] . D==============eeeeeER . . . . . . . .. vaddps %ymm3, %ymm1, %ymm486# CHECK-NEXT: [0,10] . D==============eeeeeER . . . . . . . .. vaddps %ymm3, %ymm1, %ymm487# CHECK-NEXT: [0,11] . D===============eeeeeER . . . . . . . .. vaddps %ymm3, %ymm1, %ymm488# CHECK-NEXT: [0,12] . .D===============eeeeeER . . . . . . . .. vaddps %ymm3, %ymm1, %ymm489# CHECK-NEXT: [0,13] . .D================eeeeeER. . . . . . . .. vaddps %ymm3, %ymm1, %ymm490# CHECK-NEXT: [0,14] . . D================eeeeeER . . . . . . .. vaddps %ymm3, %ymm1, %ymm491# CHECK-NEXT: [0,15] . . D=================eeeeeER . . . . . . .. vaddps %ymm3, %ymm1, %ymm492# CHECK-NEXT: [0,16] . . D=================eeeeeER . . . . . . .. vaddps %ymm3, %ymm1, %ymm493# CHECK-NEXT: [0,17] . . D======================eeER. . . . . . .. vandps %xmm4, %xmm1, %xmm094# CHECK-NEXT: [1,0] . . D=======================eeeeeeeeeER . . . . .. vdivps %ymm0, %ymm1, %ymm395# CHECK-NEXT: [1,1] . . D=======================eeeeeE----R . . . . .. vaddps %xmm0, %xmm1, %xmm396# CHECK-NEXT: [1,2] . . D===========================eeeeeER. . . . .. vaddps %ymm3, %ymm1, %ymm497# CHECK-NEXT: [1,3] . . D=============================eeeeeER . . . .. vaddps %ymm3, %ymm1, %ymm498# CHECK-NEXT: [1,4] . . .D==============================eeeeeER . . . .. vaddps %ymm3, %ymm1, %ymm499# CHECK-NEXT: [1,5] . . .D================================eeeeeER . . .. vaddps %ymm3, %ymm1, %ymm4100# CHECK-NEXT: [1,6] . . . D=================================eeeeeER . . .. vaddps %ymm3, %ymm1, %ymm4101# CHECK-NEXT: [1,7] . . . D===================================eeeeeER. . .. vaddps %ymm3, %ymm1, %ymm4102# CHECK-NEXT: [1,8] . . . D====================================eeeeeER . .. vaddps %ymm3, %ymm1, %ymm4103# CHECK-NEXT: [1,9] . . . D=====================================eeeeeER . .. vaddps %ymm3, %ymm1, %ymm4104# CHECK-NEXT: [1,10] . . . D=====================================eeeeeER . .. vaddps %ymm3, %ymm1, %ymm4105# CHECK-NEXT: [1,11] . . . D======================================eeeeeER. .. vaddps %ymm3, %ymm1, %ymm4106# CHECK-NEXT: [1,12] . . . D======================================eeeeeER .. vaddps %ymm3, %ymm1, %ymm4107# CHECK-NEXT: [1,13] . . . D=======================================eeeeeER .. vaddps %ymm3, %ymm1, %ymm4108# CHECK-NEXT: [1,14] . . . .D=======================================eeeeeER .. vaddps %ymm3, %ymm1, %ymm4109# CHECK-NEXT: [1,15] . . . .D========================================eeeeeER .. vaddps %ymm3, %ymm1, %ymm4110# CHECK-NEXT: [1,16] . . . . D========================================eeeeeER.. vaddps %ymm3, %ymm1, %ymm4111# CHECK-NEXT: [1,17] . . . . D=============================================eeER vandps %xmm4, %xmm1, %xmm0112 113# CHECK: Average Wait times (based on the timeline view):114# CHECK-NEXT: [0]: Executions115# CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue116# CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready117# CHECK-NEXT: [3]: Average time elapsed from WB until retire stage118 119# CHECK: [0] [1] [2] [3]120# CHECK-NEXT: 0. 2 12.5 0.5 0.0 vdivps %ymm0, %ymm1, %ymm3121# CHECK-NEXT: 1. 2 12.5 0.5 4.0 vaddps %xmm0, %xmm1, %xmm3122# CHECK-NEXT: 2. 2 16.5 0.0 0.0 vaddps %ymm3, %ymm1, %ymm4123# CHECK-NEXT: 3. 2 18.5 2.0 0.0 vaddps %ymm3, %ymm1, %ymm4124# CHECK-NEXT: 4. 2 19.5 4.0 0.0 vaddps %ymm3, %ymm1, %ymm4125# CHECK-NEXT: 5. 2 21.5 6.0 0.0 vaddps %ymm3, %ymm1, %ymm4126# CHECK-NEXT: 6. 2 22.5 8.0 0.0 vaddps %ymm3, %ymm1, %ymm4127# CHECK-NEXT: 7. 2 24.5 10.0 0.0 vaddps %ymm3, %ymm1, %ymm4128# CHECK-NEXT: 8. 2 25.5 12.0 0.0 vaddps %ymm3, %ymm1, %ymm4129# CHECK-NEXT: 9. 2 26.5 13.0 0.0 vaddps %ymm3, %ymm1, %ymm4130# CHECK-NEXT: 10. 2 26.5 14.0 0.0 vaddps %ymm3, %ymm1, %ymm4131# CHECK-NEXT: 11. 2 27.5 15.0 0.0 vaddps %ymm3, %ymm1, %ymm4132# CHECK-NEXT: 12. 2 27.5 16.0 0.0 vaddps %ymm3, %ymm1, %ymm4133# CHECK-NEXT: 13. 2 28.5 17.0 0.0 vaddps %ymm3, %ymm1, %ymm4134# CHECK-NEXT: 14. 2 28.5 17.5 0.0 vaddps %ymm3, %ymm1, %ymm4135# CHECK-NEXT: 15. 2 29.5 18.5 0.0 vaddps %ymm3, %ymm1, %ymm4136# CHECK-NEXT: 16. 2 29.5 19.0 0.0 vaddps %ymm3, %ymm1, %ymm4137# CHECK-NEXT: 17. 2 34.5 0.0 0.0 vandps %xmm4, %xmm1, %xmm0138# CHECK-NEXT: 2 24.0 9.6 0.2 <total>139