[PATCH] D105742: [AMDGPU] Make V_CVT_I32_F64/V_CVT_F64_I32 rematerializable.

Stanislav Mekhanoshin via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Jul 9 16:10:45 PDT 2021


rampitec marked an inline comment as done.
rampitec added inline comments.


================
Comment at: llvm/test/tools/llvm-mca/AMDGPU/gfx10-double.s:63
 # CHECK-NEXT:  1      22    1.00                  U     v_cvt_f64_i32_e32 v[2:3], v2
-# CHECK-NEXT:  1      22    1.00                  U     v_cvt_f32_f64_e32 v4, v[4:5]
-# CHECK-NEXT:  1      22    1.00                  U     v_cvt_f64_f32_e32 v[6:7], v6
-# CHECK-NEXT:  1      22    1.00                  U     v_cvt_u32_f64_e32 v8, v[8:9]
-# CHECK-NEXT:  1      22    1.00                  U     v_cvt_f64_u32_e32 v[10:11], v10
+# CHECK-NEXT:  1      5     1.00                  U     v_cvt_f32_f64_e32 v4, v[4:5]
+# CHECK-NEXT:  1      5     1.00                  U     v_cvt_f64_f32_e32 v[6:7], v6
----------------
Something is wrong here, latency should not been changed...


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D105742/new/

https://reviews.llvm.org/D105742



More information about the llvm-commits mailing list