From 01dd9602edeced7fc9972d6cabbdc3861e1b592b Mon Sep 17 00:00:00 2001 From: Sebastian Neubauer Date: Thu, 8 Apr 2021 17:22:32 +0200 Subject: [PATCH] [AMDGPU] Add TransVALU to gfx10 Instructions on the transcendental unit are executed in parallel to the normal VALU, so add this as an extra resource. This doesn't seem to have any effect, but it should be more correct. Differential Revision: https://reviews.llvm.org/D100123 --- lib/Target/AMDGPU/SISchedule.td | 7 +- .../llvm-mca/AMDGPU/gfx10-add-sequence.s | 17 ++-- test/tools/llvm-mca/AMDGPU/gfx10-double.s | 67 +++++++------- test/tools/llvm-mca/AMDGPU/gfx10-trans.s | 88 +++++++++++++++++++ 4 files changed, 136 insertions(+), 43 deletions(-) create mode 100644 test/tools/llvm-mca/AMDGPU/gfx10-trans.s diff --git a/lib/Target/AMDGPU/SISchedule.td b/lib/Target/AMDGPU/SISchedule.td index adb79dd93d2..b24c061af7a 100644 --- a/lib/Target/AMDGPU/SISchedule.td +++ b/lib/Target/AMDGPU/SISchedule.td @@ -107,6 +107,9 @@ def HWVMEM : ProcResource<1> { def HWVALU : ProcResource<1> { let BufferSize = 1; } +def HWTransVALU : ProcResource<1> { // Transcendental VALU + let BufferSize = 1; +} def HWRC : ProcResource<1> { // Register destination cache let BufferSize = 1; } @@ -240,14 +243,14 @@ let SchedModel = GFX10SpeedModel in { def : HWWriteRes; def : HWWriteRes; def : HWWriteRes; -def : HWWriteRes; +def : HWWriteRes; def : HWWriteRes; def : HWWriteRes; def : HWWriteRes; def : HWWriteRes; def : HWWriteRes; def : HWWriteRes; -def : HWWriteRes; +def : HWWriteRes; def : HWWriteRes; def : HWWriteRes; diff --git a/test/tools/llvm-mca/AMDGPU/gfx10-add-sequence.s b/test/tools/llvm-mca/AMDGPU/gfx10-add-sequence.s index be57731389b..ec902e5b0fd 100644 --- a/test/tools/llvm-mca/AMDGPU/gfx10-add-sequence.s +++ b/test/tools/llvm-mca/AMDGPU/gfx10-add-sequence.s @@ -34,18 +34,19 @@ v_add_f32 v2, v1, v0 # CHECK-NEXT: [2] - HWLGKM # CHECK-NEXT: [3] - HWRC # CHECK-NEXT: [4] - HWSALU -# CHECK-NEXT: [5] - HWVALU -# CHECK-NEXT: [6] - HWVMEM +# CHECK-NEXT: [5] - HWTransVALU +# CHECK-NEXT: [6] - HWVALU +# CHECK-NEXT: [7] - HWVMEM # CHECK: Resource pressure per iteration: -# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] -# CHECK-NEXT: - - - 3.00 - 3.00 - +# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] +# CHECK-NEXT: - - - 3.00 - - 3.00 - # CHECK: Resource pressure by instruction: -# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] Instructions: -# CHECK-NEXT: - - - 1.00 - 1.00 - v_add_f32_e32 v0, v0, v0 -# CHECK-NEXT: - - - 1.00 - 1.00 - v_add_f32_e32 v1, v1, v1 -# CHECK-NEXT: - - - 1.00 - 1.00 - v_add_f32_e32 v2, v1, v0 +# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] Instructions: +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_add_f32_e32 v0, v0, v0 +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_add_f32_e32 v1, v1, v1 +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_add_f32_e32 v2, v1, v0 # CHECK: Timeline view: # CHECK-NEXT: 01 diff --git a/test/tools/llvm-mca/AMDGPU/gfx10-double.s b/test/tools/llvm-mca/AMDGPU/gfx10-double.s index 3a0991d875f..de19418fac5 100644 --- a/test/tools/llvm-mca/AMDGPU/gfx10-double.s +++ b/test/tools/llvm-mca/AMDGPU/gfx10-double.s @@ -93,43 +93,44 @@ v_sqrt_f64 v[4:5], v[4:5] # CHECK-NEXT: [2] - HWLGKM # CHECK-NEXT: [3] - HWRC # CHECK-NEXT: [4] - HWSALU -# CHECK-NEXT: [5] - HWVALU -# CHECK-NEXT: [6] - HWVMEM +# CHECK-NEXT: [5] - HWTransVALU +# CHECK-NEXT: [6] - HWVALU +# CHECK-NEXT: [7] - HWVMEM # CHECK: Resource pressure per iteration: -# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] -# CHECK-NEXT: - - - 29.00 1.00 28.00 - +# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] +# CHECK-NEXT: - - - 29.00 1.00 3.00 28.00 - # CHECK: Resource pressure by instruction: -# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] Instructions: -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cvt_i32_f64_e32 v0, v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cvt_f64_i32_e32 v[2:3], v2 -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cvt_f32_f64_e32 v4, v[4:5] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cvt_f64_f32_e32 v[6:7], v6 -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cvt_u32_f64_e32 v8, v[8:9] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cvt_f64_u32_e32 v[10:11], v10 -# CHECK-NEXT: - - - 1.00 - 1.00 - v_frexp_exp_i32_f64_e32 v0, v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_frexp_mant_f64_e32 v[2:3], v[2:3] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_fract_f64_e32 v[4:5], v[4:5] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_trunc_f64_e32 v[0:1], v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_ceil_f64_e32 v[2:3], v[2:3] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_rndne_f64_e32 v[4:5], v[4:5] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_floor_f64_e32 v[6:7], v[6:7] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_fma_f64 v[0:1], v[0:1], v[0:1], v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_add_f64 v[2:3], v[2:3], v[2:3] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_mul_f64 v[4:5], v[4:5], v[4:5] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_min_f64 v[6:7], v[6:7], v[6:7] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_max_f64 v[8:9], v[8:9], v[8:9] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_div_fmas_f64 v[0:1], v[0:1], v[0:1], v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_div_fixup_f64 v[0:1], v[0:1], v[0:1], v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_ldexp_f64 v[2:3], v[2:3], v0 -# CHECK-NEXT: - - - 2.00 1.00 1.00 - v_div_scale_f64 v[0:1], vcc_lo, v[0:1], v[0:1], v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_trig_preop_f64 v[2:3], v[2:3], v0 -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cmp_eq_f64_e32 vcc_lo, v[0:1], v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_cmp_class_f64_e64 vcc_lo, v[2:3], s0 -# CHECK-NEXT: - - - 1.00 - 1.00 - v_rcp_f64_e32 v[0:1], v[0:1] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_rsq_f64_e32 v[2:3], v[2:3] -# CHECK-NEXT: - - - 1.00 - 1.00 - v_sqrt_f64_e32 v[4:5], v[4:5] +# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] Instructions: +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cvt_i32_f64_e32 v0, v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cvt_f64_i32_e32 v[2:3], v2 +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cvt_f32_f64_e32 v4, v[4:5] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cvt_f64_f32_e32 v[6:7], v6 +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cvt_u32_f64_e32 v8, v[8:9] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cvt_f64_u32_e32 v[10:11], v10 +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_frexp_exp_i32_f64_e32 v0, v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_frexp_mant_f64_e32 v[2:3], v[2:3] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_fract_f64_e32 v[4:5], v[4:5] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_trunc_f64_e32 v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_ceil_f64_e32 v[2:3], v[2:3] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_rndne_f64_e32 v[4:5], v[4:5] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_floor_f64_e32 v[6:7], v[6:7] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_fma_f64 v[0:1], v[0:1], v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_add_f64 v[2:3], v[2:3], v[2:3] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_mul_f64 v[4:5], v[4:5], v[4:5] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_min_f64 v[6:7], v[6:7], v[6:7] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_max_f64 v[8:9], v[8:9], v[8:9] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_div_fmas_f64 v[0:1], v[0:1], v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_div_fixup_f64 v[0:1], v[0:1], v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_ldexp_f64 v[2:3], v[2:3], v0 +# CHECK-NEXT: - - - 2.00 1.00 - 1.00 - v_div_scale_f64 v[0:1], vcc_lo, v[0:1], v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_trig_preop_f64 v[2:3], v[2:3], v0 +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cmp_eq_f64_e32 vcc_lo, v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - - 1.00 - v_cmp_class_f64_e64 vcc_lo, v[2:3], s0 +# CHECK-NEXT: - - - 1.00 - 1.00 1.00 - v_rcp_f64_e32 v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - 1.00 1.00 - v_rsq_f64_e32 v[2:3], v[2:3] +# CHECK-NEXT: - - - 1.00 - 1.00 1.00 - v_sqrt_f64_e32 v[4:5], v[4:5] # CHECK: Timeline view: # CHECK-NEXT: 0123456789 0123456789 0123456789 0 diff --git a/test/tools/llvm-mca/AMDGPU/gfx10-trans.s b/test/tools/llvm-mca/AMDGPU/gfx10-trans.s new file mode 100644 index 00000000000..6f84465ecef --- /dev/null +++ b/test/tools/llvm-mca/AMDGPU/gfx10-trans.s @@ -0,0 +1,88 @@ +# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py +# RUN: llvm-mca -mtriple=amdgcn -mcpu=gfx1010 --timeline --iterations=1 < %s | FileCheck %s + +v_log_f32 v0, v0 +v_rcp_f32 v0, v0 +v_rsq_f32 v1, v1 +v_sqrt_f32 v2, v0 +v_rcp_f64 v[0:1], v[0:1] +v_rsq_f64 v[1:2], v[1:2] +v_sqrt_f64 v[2:3], v[0:1] + +# CHECK: Iterations: 1 +# CHECK-NEXT: Instructions: 7 +# CHECK-NEXT: Total Cycles: 94 +# CHECK-NEXT: Total uOps: 7 + +# CHECK: Dispatch Width: 1 +# CHECK-NEXT: uOps Per Cycle: 0.07 +# CHECK-NEXT: IPC: 0.07 +# CHECK-NEXT: Block RThroughput: 7.0 + +# 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-NEXT: 1 10 1.00 U v_log_f32_e32 v0, v0 +# CHECK-NEXT: 1 10 1.00 U v_rcp_f32_e32 v0, v0 +# CHECK-NEXT: 1 10 1.00 U v_rsq_f32_e32 v1, v1 +# CHECK-NEXT: 1 10 1.00 U v_sqrt_f32_e32 v2, v0 +# CHECK-NEXT: 1 24 1.00 U v_rcp_f64_e32 v[0:1], v[0:1] +# CHECK-NEXT: 1 24 1.00 U v_rsq_f64_e32 v[1:2], v[1:2] +# CHECK-NEXT: 1 24 1.00 U v_sqrt_f64_e32 v[2:3], v[0:1] + +# CHECK: Resources: +# CHECK-NEXT: [0] - HWBranch +# CHECK-NEXT: [1] - HWExport +# CHECK-NEXT: [2] - HWLGKM +# CHECK-NEXT: [3] - HWRC +# CHECK-NEXT: [4] - HWSALU +# CHECK-NEXT: [5] - HWTransVALU +# CHECK-NEXT: [6] - HWVALU +# CHECK-NEXT: [7] - HWVMEM + +# CHECK: Resource pressure per iteration: +# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] +# CHECK-NEXT: - - - 7.00 - 7.00 3.00 - + +# CHECK: Resource pressure by instruction: +# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] Instructions: +# CHECK-NEXT: - - - 1.00 - 1.00 - - v_log_f32_e32 v0, v0 +# CHECK-NEXT: - - - 1.00 - 1.00 - - v_rcp_f32_e32 v0, v0 +# CHECK-NEXT: - - - 1.00 - 1.00 - - v_rsq_f32_e32 v1, v1 +# CHECK-NEXT: - - - 1.00 - 1.00 - - v_sqrt_f32_e32 v2, v0 +# CHECK-NEXT: - - - 1.00 - 1.00 1.00 - v_rcp_f64_e32 v[0:1], v[0:1] +# CHECK-NEXT: - - - 1.00 - 1.00 1.00 - v_rsq_f64_e32 v[1:2], v[1:2] +# CHECK-NEXT: - - - 1.00 - 1.00 1.00 - v_sqrt_f64_e32 v[2:3], v[0:1] + +# CHECK: Timeline view: +# CHECK-NEXT: 0123456789 0123456789 0123456789 +# CHECK-NEXT: Index 0123456789 0123456789 0123456789 0123456789 + +# CHECK: [0,0] DeeeeeeeeeE . . . . . . . . . . . . v_log_f32_e32 v0, v0 +# CHECK-NEXT: [0,1] . . DeeeeeeeeeE . . . . . . . . . . v_rcp_f32_e32 v0, v0 +# CHECK-NEXT: [0,2] . . .DeeeeeeeeeE . . . . . . . . . . v_rsq_f32_e32 v1, v1 +# CHECK-NEXT: [0,3] . . . . DeeeeeeeeeE . . . . . . . . v_sqrt_f32_e32 v2, v0 +# CHECK-NEXT: [0,4] . . . . .DeeeeeeeeeeeeeeeeeeeeeeeE . . . . . v_rcp_f64_e32 v[0:1], v[0:1] +# CHECK-NEXT: [0,5] . . . . . . . . . DeeeeeeeeeeeeeeeeeeeeeeeE v_rsq_f64_e32 v[1:2], v[1:2] + +# CHECK: Average Wait times (based on the timeline view): +# CHECK-NEXT: [0]: Executions +# CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue +# CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready +# CHECK-NEXT: [3]: Average time elapsed from WB until retire stage + +# CHECK: [0] [1] [2] [3] +# CHECK-NEXT: 0. 1 0.0 0.0 0.0 v_log_f32_e32 v0, v0 +# CHECK-NEXT: 1. 1 0.0 0.0 0.0 v_rcp_f32_e32 v0, v0 +# CHECK-NEXT: 2. 1 0.0 0.0 0.0 v_rsq_f32_e32 v1, v1 +# CHECK-NEXT: 3. 1 0.0 0.0 0.0 v_sqrt_f32_e32 v2, v0 +# CHECK-NEXT: 4. 1 0.0 0.0 0.0 v_rcp_f64_e32 v[0:1], v[0:1] +# CHECK-NEXT: 5. 1 0.0 0.0 0.0 v_rsq_f64_e32 v[1:2], v[1:2] +# CHECK-NEXT: 6. 1 0.0 0.0 0.0 v_sqrt_f64_e32 v[2:3], v[0:1] +# CHECK-NEXT: 1 0.0 0.0 0.0