diff options
author | NAKAMURA Takumi <geek4civic@gmail.com> | 2025-01-09 18:31:57 +0900 |
---|---|---|
committer | NAKAMURA Takumi <geek4civic@gmail.com> | 2025-01-09 18:33:27 +0900 |
commit | df025ebf872052c0761d44a3ef9b65e9675af8a8 (patch) | |
tree | 9b4e94583e2536546d6606270bcdf846c95e1ba2 /llvm/test/CodeGen/AMDGPU/minimumnum.ll | |
parent | 4428c9d0b1344179f85a72e183a44796976521e3 (diff) | |
parent | bdcf47e4bcb92889665825654bb80a8bbe30379e (diff) | |
download | llvm-users/chapuni/cov/single/loop.zip llvm-users/chapuni/cov/single/loop.tar.gz llvm-users/chapuni/cov/single/loop.tar.bz2 |
Merge branch 'users/chapuni/cov/single/base' into users/chapuni/cov/single/loopusers/chapuni/cov/single/loop
Conflicts:
clang/lib/CodeGen/CoverageMappingGen.cpp
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/minimumnum.ll')
-rw-r--r-- | llvm/test/CodeGen/AMDGPU/minimumnum.ll | 120 |
1 files changed, 60 insertions, 60 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/minimumnum.ll b/llvm/test/CodeGen/AMDGPU/minimumnum.ll index 9e0b7da..f5fb85d 100644 --- a/llvm/test/CodeGen/AMDGPU/minimumnum.ll +++ b/llvm/test/CodeGen/AMDGPU/minimumnum.ll @@ -1792,11 +1792,11 @@ define <3 x half> @v_minimumnum_v3f16(<3 x half> %x, <3 x half> %y) { ; GFX8-NEXT: v_max_f16_e32 v2, v2, v2 ; GFX8-NEXT: v_max_f16_e32 v0, v0, v0 ; GFX8-NEXT: v_min_f16_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD -; GFX8-NEXT: v_min_f16_e32 v0, v0, v2 -; GFX8-NEXT: v_max_f16_e32 v2, v3, v3 +; GFX8-NEXT: v_max_f16_e32 v3, v3, v3 ; GFX8-NEXT: v_max_f16_e32 v1, v1, v1 +; GFX8-NEXT: v_min_f16_e32 v0, v0, v2 +; GFX8-NEXT: v_min_f16_e32 v1, v1, v3 ; GFX8-NEXT: v_or_b32_e32 v0, v0, v4 -; GFX8-NEXT: v_min_f16_e32 v1, v1, v2 ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_minimumnum_v3f16: @@ -1858,8 +1858,8 @@ define <3 x half> @v_minimumnum_v3f16_nnan(<3 x half> %x, <3 x half> %y) { ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_min_f16_sdwa v4, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX8-NEXT: v_min_f16_e32 v0, v0, v2 -; GFX8-NEXT: v_or_b32_e32 v0, v0, v4 ; GFX8-NEXT: v_min_f16_e32 v1, v1, v3 +; GFX8-NEXT: v_or_b32_e32 v0, v0, v4 ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_minimumnum_v3f16_nnan: @@ -1901,20 +1901,20 @@ define <4 x half> @v_minimumnum_v4f16(<4 x half> %x, <4 x half> %y) { ; GFX8-LABEL: v_minimumnum_v4f16: ; GFX8: ; %bb.0: ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX8-NEXT: v_max_f16_sdwa v4, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v5, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_e32 v2, v2, v2 -; GFX8-NEXT: v_max_f16_e32 v0, v0, v0 +; GFX8-NEXT: v_max_f16_sdwa v4, v3, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v5, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX8-NEXT: v_min_f16_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD -; GFX8-NEXT: v_min_f16_e32 v0, v0, v2 -; GFX8-NEXT: v_or_b32_e32 v0, v0, v4 -; GFX8-NEXT: v_max_f16_sdwa v2, v3, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v4, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v5, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v6, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX8-NEXT: v_max_f16_e32 v3, v3, v3 ; GFX8-NEXT: v_max_f16_e32 v1, v1, v1 -; GFX8-NEXT: v_min_f16_sdwa v2, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_max_f16_e32 v2, v2, v2 +; GFX8-NEXT: v_max_f16_e32 v0, v0, v0 +; GFX8-NEXT: v_min_f16_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD ; GFX8-NEXT: v_min_f16_e32 v1, v1, v3 -; GFX8-NEXT: v_or_b32_e32 v1, v1, v2 +; GFX8-NEXT: v_min_f16_e32 v0, v0, v2 +; GFX8-NEXT: v_or_b32_e32 v0, v0, v5 +; GFX8-NEXT: v_or_b32_e32 v1, v1, v4 ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_minimumnum_v4f16: @@ -1974,12 +1974,12 @@ define <4 x half> @v_minimumnum_v4f16_nnan(<4 x half> %x, <4 x half> %y) { ; GFX8-LABEL: v_minimumnum_v4f16_nnan: ; GFX8: ; %bb.0: ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX8-NEXT: v_min_f16_sdwa v4, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_min_f16_e32 v0, v0, v2 -; GFX8-NEXT: v_min_f16_sdwa v2, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_min_f16_sdwa v4, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_min_f16_sdwa v5, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX8-NEXT: v_min_f16_e32 v1, v1, v3 -; GFX8-NEXT: v_or_b32_e32 v0, v0, v4 -; GFX8-NEXT: v_or_b32_e32 v1, v1, v2 +; GFX8-NEXT: v_min_f16_e32 v0, v0, v2 +; GFX8-NEXT: v_or_b32_e32 v0, v0, v5 +; GFX8-NEXT: v_or_b32_e32 v1, v1, v4 ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_minimumnum_v4f16_nnan: @@ -2021,27 +2021,27 @@ define <6 x half> @v_minimumnum_v6f16(<6 x half> %x, <6 x half> %y) { ; GFX8-LABEL: v_minimumnum_v6f16: ; GFX8: ; %bb.0: ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX8-NEXT: v_max_f16_sdwa v6, v3, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v7, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_e32 v3, v3, v3 -; GFX8-NEXT: v_max_f16_e32 v0, v0, v0 +; GFX8-NEXT: v_max_f16_sdwa v6, v5, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v7, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX8-NEXT: v_min_f16_sdwa v6, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD -; GFX8-NEXT: v_min_f16_e32 v0, v0, v3 -; GFX8-NEXT: v_or_b32_e32 v0, v0, v6 -; GFX8-NEXT: v_max_f16_sdwa v3, v4, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v6, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v7, v4, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v8, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_min_f16_sdwa v7, v8, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_max_f16_sdwa v8, v3, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v9, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_e32 v5, v5, v5 +; GFX8-NEXT: v_max_f16_e32 v2, v2, v2 ; GFX8-NEXT: v_max_f16_e32 v4, v4, v4 ; GFX8-NEXT: v_max_f16_e32 v1, v1, v1 -; GFX8-NEXT: v_min_f16_sdwa v3, v6, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_max_f16_e32 v3, v3, v3 +; GFX8-NEXT: v_max_f16_e32 v0, v0, v0 +; GFX8-NEXT: v_min_f16_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_min_f16_e32 v2, v2, v5 ; GFX8-NEXT: v_min_f16_e32 v1, v1, v4 -; GFX8-NEXT: v_or_b32_e32 v1, v1, v3 -; GFX8-NEXT: v_max_f16_sdwa v3, v5, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v4, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_min_f16_sdwa v3, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD -; GFX8-NEXT: v_max_f16_e32 v4, v5, v5 -; GFX8-NEXT: v_max_f16_e32 v2, v2, v2 -; GFX8-NEXT: v_min_f16_e32 v2, v2, v4 -; GFX8-NEXT: v_or_b32_e32 v2, v2, v3 +; GFX8-NEXT: v_min_f16_e32 v0, v0, v3 +; GFX8-NEXT: v_or_b32_e32 v0, v0, v8 +; GFX8-NEXT: v_or_b32_e32 v1, v1, v7 +; GFX8-NEXT: v_or_b32_e32 v2, v2, v6 ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_minimumnum_v6f16: @@ -2113,34 +2113,34 @@ define <8 x half> @v_minimumnum_v8f16(<8 x half> %x, <8 x half> %y) { ; GFX8-LABEL: v_minimumnum_v8f16: ; GFX8: ; %bb.0: ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX8-NEXT: v_max_f16_sdwa v8, v4, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v9, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_e32 v4, v4, v4 -; GFX8-NEXT: v_max_f16_e32 v0, v0, v0 +; GFX8-NEXT: v_max_f16_sdwa v8, v7, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v9, v3, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX8-NEXT: v_min_f16_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD -; GFX8-NEXT: v_min_f16_e32 v0, v0, v4 -; GFX8-NEXT: v_or_b32_e32 v0, v0, v8 -; GFX8-NEXT: v_max_f16_sdwa v4, v5, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v8, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v9, v6, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v10, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_min_f16_sdwa v9, v10, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_max_f16_sdwa v10, v5, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v11, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_min_f16_sdwa v10, v11, v10 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_max_f16_sdwa v11, v4, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_sdwa v12, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_max_f16_e32 v7, v7, v7 +; GFX8-NEXT: v_max_f16_e32 v3, v3, v3 +; GFX8-NEXT: v_max_f16_e32 v6, v6, v6 +; GFX8-NEXT: v_max_f16_e32 v2, v2, v2 ; GFX8-NEXT: v_max_f16_e32 v5, v5, v5 ; GFX8-NEXT: v_max_f16_e32 v1, v1, v1 -; GFX8-NEXT: v_min_f16_sdwa v4, v8, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_max_f16_e32 v4, v4, v4 +; GFX8-NEXT: v_max_f16_e32 v0, v0, v0 +; GFX8-NEXT: v_min_f16_sdwa v11, v12, v11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD +; GFX8-NEXT: v_min_f16_e32 v3, v3, v7 +; GFX8-NEXT: v_min_f16_e32 v2, v2, v6 ; GFX8-NEXT: v_min_f16_e32 v1, v1, v5 -; GFX8-NEXT: v_or_b32_e32 v1, v1, v4 -; GFX8-NEXT: v_max_f16_sdwa v4, v6, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v5, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_min_f16_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD -; GFX8-NEXT: v_max_f16_e32 v5, v6, v6 -; GFX8-NEXT: v_max_f16_e32 v2, v2, v2 -; GFX8-NEXT: v_min_f16_e32 v2, v2, v5 -; GFX8-NEXT: v_or_b32_e32 v2, v2, v4 -; GFX8-NEXT: v_max_f16_sdwa v4, v7, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_max_f16_sdwa v5, v3, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 -; GFX8-NEXT: v_min_f16_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD -; GFX8-NEXT: v_max_f16_e32 v5, v7, v7 -; GFX8-NEXT: v_max_f16_e32 v3, v3, v3 -; GFX8-NEXT: v_min_f16_e32 v3, v3, v5 -; GFX8-NEXT: v_or_b32_e32 v3, v3, v4 +; GFX8-NEXT: v_min_f16_e32 v0, v0, v4 +; GFX8-NEXT: v_or_b32_e32 v0, v0, v11 +; GFX8-NEXT: v_or_b32_e32 v1, v1, v10 +; GFX8-NEXT: v_or_b32_e32 v2, v2, v9 +; GFX8-NEXT: v_or_b32_e32 v3, v3, v8 ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_minimumnum_v8f16: |