Skip to content

GitLab

Commit 00810928 authored by OscarXu's avatar OscarXu
Browse files

replace fixed label with dynamic label

parent af02d656
Show whitespace changes
Inline Side-by-side
Showing with 5 additions and 5 deletions
include/ck_tile/ops/flatmm/block/uk/flatmm_sn_uk_gfx9_32x128x512_1x4x1_16x16x16_itl.inc
View file @ 00810928
...@@ -159,7 +159,7 @@ ...@@ -159,7 +159,7 @@
" ds_write_b64 %[v_sfl_sst], [%[c6],%[c7]] offset:23168 \n" " ds_write_b64 %[v_sfl_sst], [%[c6],%[c7]] offset:23168 \n"
" s_mov_b32 s80, 0 \n" " s_mov_b32 s80, 0 \n"
" s_waitcnt vmcnt(24) \n" " s_waitcnt vmcnt(24) \n"
"label_0AA6: \n" "coreloop_top_%=: \n"
" s_waitcnt vmcnt(30) & lgkmcnt(0) \n" " s_waitcnt vmcnt(30) & lgkmcnt(0) \n"
" s_barrier \n" _UK_MFMA_ " [%[c0], %[c1], %[c2], %[c3]], acc[0:1], v[128:129], 0 \n" " s_barrier \n" _UK_MFMA_ " [%[c0], %[c1], %[c2], %[c3]], acc[0:1], v[128:129], 0 \n"
" ds_read_b32 v10, %[v_sfl_sld] offset:16640 \n" " ds_read_b32 v10, %[v_sfl_sld] offset:16640 \n"
...@@ -396,7 +396,7 @@ ...@@ -396,7 +396,7 @@
"%[c15]", "%[c15]",
"%[c7]") " s_addk_i32 s80, 0x0080 \n" "%[c7]") " s_addk_i32 s80, 0x0080 \n"
" s_cmp_lt_i32 s80, %[s_loop_cnt] \n" " s_cmp_lt_i32 s80, %[s_loop_cnt] \n"
" s_cbranch_scc0 label_0EC1 \n" " s_cbranch_scc0 loop_atomic_%= \n"
" s_waitcnt vmcnt(30) & lgkmcnt(0) \n" " s_waitcnt vmcnt(30) & lgkmcnt(0) \n"
" s_barrier \n" _UK_MFMA_ " s_barrier \n" _UK_MFMA_
" [%[c16], %[c17], %[c18], %[c19]], acc[128:129], v[128:129], 0 \n" " [%[c16], %[c17], %[c18], %[c19]], acc[128:129], v[128:129], 0 \n"
...@@ -693,9 +693,9 @@ ...@@ -693,9 +693,9 @@
"%[c31]", "%[c31]",
"%[c23]") " s_addk_i32 s80, 0x0080 \n" "%[c23]") " s_addk_i32 s80, 0x0080 \n"
" s_cmp_lt_i32 s80, %[s_loop_cnt] \n" " s_cmp_lt_i32 s80, %[s_loop_cnt] \n"
" s_cbranch_scc0 label_0EC1 \n" " s_cbranch_scc0 loop_atomic_%= \n"
" s_branch label_0AA6 \n" " s_branch coreloop_top_%= \n"
" label_0EC1: \n" " loop_atomic_%=: \n"
" s_waitcnt lgkmcnt(0) \n" " s_waitcnt lgkmcnt(0) \n"
" s_barrier \n" " s_barrier \n"
" ds_read_b32 v10, %[v_sfl_sld] offset:16640 \n" " ds_read_b32 v10, %[v_sfl_sld] offset:16640 \n"
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment