Skip to content

AMDGPU: Add a few mfma test with immediate splat src2 #150627

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Jul 26, 2025
Merged

AMDGPU: Add a few mfma test with immediate splat src2 #150627

merged 1 commit into from
Jul 26, 2025
+347 −0

Conversation

arsenm
Copy link
Contributor

@arsenm arsenm commented Jul 25, 2025

No description provided.

@arsenm arsenm mentioned this pull request Jul 25, 2025
Merged
@arsenm Graphite App
Copy link
Contributor Author

arsenm commented Jul 25, 2025 •
edited
Loading

This stack of pull requests is managed by Graphite. Learn more about stacking.

@arsenm arsenm mentioned this pull request Jul 25, 2025
Merged
@llvmbot
Copy link
Member

llvmbot commented Jul 25, 2025

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/150627.diff

1 Files Affected:

  • (modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll (+347)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
index f77844558460f..d7257a49b00bb 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
@@ -3121,6 +3121,159 @@ bb:
   ret void
 }
 
+define amdgpu_kernel void @test_mfma_i32_16x16x4i8_splatimm_src2_64(ptr addrspace(1) %arg) #0 {
+; NOLIT-SRCC-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; NOLIT-SRCC:       ; %bb.0: ; %bb
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a4, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a5, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a6, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a7, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a8, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a9, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a10, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a11, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a12, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a13, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a14, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a15, 64
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; NOLIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v16, 0
+; NOLIT-SRCC-NEXT:    v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; NOLIT-SRCC-NEXT:    s_nop 7
+; NOLIT-SRCC-NEXT:    s_nop 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v15, a15
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v14, a14
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v13, a13
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v12, a12
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v7, a7
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v6, a6
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v5, a5
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v4, a4
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v11, a11
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v10, a10
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v9, a9
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v8, a8
+; NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; NOLIT-SRCC-NEXT:    s_endpgm
+;
+; LIT-SRCC-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; LIT-SRCC:       ; %bb.0: ; %bb
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; LIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v16, 0
+; LIT-SRCC-NEXT:    v_mfma_i32_16x16x4i8 a[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3
+; LIT-SRCC-NEXT:    s_nop 7
+; LIT-SRCC-NEXT:    s_nop 1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v15, a15
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v14, a14
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v13, a13
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v12, a12
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v7, a7
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v6, a6
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v5, a5
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v4, a4
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v11, a11
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v10, a10
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v9, a9
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v8, a8
+; LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; LIT-SRCC-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 2
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mfma_i32_16x16x4i8 a[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    v_mov_b32_e32 v0, 1
+; GFX942-NEXT:    v_mov_b32_e32 v1, 2
+; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    v_mfma_i32_16x16x4_4b_i8 a[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 64
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, 2
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    v_mfma_i32_16x16x4_4b_i8 v[0:15], v17, v18, v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
+bb:
+  %in.1 = load <16 x i32>, ptr addrspace(1) %arg
+  %mai.1 = tail call <16 x i32> @llvm.amdgcn.mfma.i32.16x16x4i8(i32 1, i32 2, <16 x i32> splat (i32 64), i32 1, i32 2, i32 3)
+  store <16 x i32> %mai.1, ptr addrspace(1) %arg
+  ret void
+}
+
 define amdgpu_kernel void @test_mfma_i32_4x4x4i8(ptr addrspace(1) %arg) #0 {
 ; NOLIT-SRCC-LABEL: test_mfma_i32_4x4x4i8:
 ; NOLIT-SRCC:       ; %bb.0: ; %bb
@@ -3239,6 +3392,200 @@ bb:
   ret void
 }
 
+define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_imm_src2_1(ptr addrspace(1) %arg) #0 {
+; NOLIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; NOLIT-SRCC:       ; %bb.0: ; %bb
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, 1
+; NOLIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v0, v1, a[0:3] cbsz:1 abid:2 blgp:3
+; NOLIT-SRCC-NEXT:    s_nop 3
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; NOLIT-SRCC-NEXT:    s_endpgm
+;
+; LIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; LIT-SRCC:       ; %bb.0: ; %bb
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; LIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v0, v1, 1 cbsz:1 abid:2 blgp:3
+; LIT-SRCC-NEXT:    s_nop 3
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; LIT-SRCC-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v0, v2, 1 cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_nop 3
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    v_mov_b32_e32 v0, 1
+; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    v_mfma_i32_4x4x4_16b_i8 a[0:3], v0, v2, 1 cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_nop 3
+; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v0, v5, v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 3
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
+bb:
+  %in.1 = load <4 x i32>, ptr addrspace(1) %arg
+  %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 1, i32 2, <4 x i32> splat (i32 1), i32 1, i32 2, i32 3)
+  store <4 x i32> %mai.1, ptr addrspace(1) %arg
+  ret void
+}
+
+define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_k_src2_1(ptr addrspace(1) %arg) #0 {
+; NOLIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; NOLIT-SRCC:       ; %bb.0:
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 0x41
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, 1
+; NOLIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, v0
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, v0
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, v0
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, v0
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 2
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v1, v0, a[0:3] cbsz:1 abid:2 blgp:3
+; NOLIT-SRCC-NEXT:    s_nop 3
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; NOLIT-SRCC-NEXT:    s_endpgm
+;
+; LIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; LIT-SRCC:       ; %bb.0:
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 0x41
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, 1
+; LIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, v0
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, v0
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, v0
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, v0
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 2
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v1, v0, a[0:3] cbsz:1 abid:2 blgp:3
+; LIT-SRCC-NEXT:    s_nop 3
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; LIT-SRCC-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0x41
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 1
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v1, v2, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_nop 3
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; GFX942:       ; %bb.0:
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0x41
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, v1
+; GFX942-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    v_mfma_i32_4x4x4_16b_i8 a[0:3], v1, v2, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_nop 3
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; GFX942-VGPR:       ; %bb.0:
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 1
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0x41
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v5, v6, v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 3
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
+  %in.1 = load <4 x i32>, ptr addrspace(1) %arg
+  %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 1, i32 2, <4 x i32> splat (i32 65), i32 1, i32 2, i32 3)
+  store <4 x i32> %mai.1, ptr addrspace(1) %arg
+  ret void
+}
+
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_forward_acc(ptr addrspace(1) %arg) #0 {
 ; NOLIT-SRCC-LABEL: test_mfma_f32_32x32x1f32_forward_acc:
 ; NOLIT-SRCC:       ; %bb.0: ; %bb

@arsenm arsenm marked this pull request as ready for review July 25, 2025 14:53
@arsenm Graphite App
Copy link
Contributor Author

arsenm commented Jul 25, 2025 •
edited
Loading

Merge activity

  • Jul 25, 4:01 PM UTC: A user started a stack merge that includes this pull request via Graphite.
  • Jul 25, 9:43 PM UTC: Graphite couldn't merge this pull request because a downstack PR AMDGPU: Add tests with forced selection of mfmas to VGPR form #150626 failed to merge.
  • Jul 26, 1:20 AM UTC: Graphite rebased this pull request as part of a merge.
  • Jul 26, 1:23 AM UTC: Graphite rebased this pull request as part of a merge.
  • Jul 26, 1:40 AM UTC: Graphite rebased this pull request as part of a merge.
  • Jul 26, 1:50 AM UTC: Graphite rebased this pull request as part of a merge.
  • Jul 26, 1:58 AM UTC: Graphite rebased this pull request as part of a merge.
  • Jul 26, 2:23 AM UTC: Graphite rebased this pull request as part of a merge.
  • Jul 26, 2:26 AM UTC: Graphite rebased this pull request as part of a merge.
  • Jul 26, 2:58 AM UTC: @arsenm merged this pull request with Graphite.

@arsenm arsenm force-pushed the users/arsenm/amdgpu/add-tests-mfma-forced-vgpr-selection branch 18 times, most recently from 4328b06 to 206ae8b Compare July 25, 2025 17:45
@arsenm arsenm force-pushed the users/arsenm/amdgpu/add-tests-mfma-forced-vgpr-selection branch 16 times, most recently from b81d061 to 86c6961 Compare July 25, 2025 23:48
@arsenm arsenm force-pushed the users/arsenm/amdgpu/add-tests-mfma-src2-imm branch from ad766da to 4ee7be5 Compare July 25, 2025 23:48
@arsenm arsenm force-pushed the users/arsenm/amdgpu/add-tests-mfma-forced-vgpr-selection branch from 86c6961 to 5810f83 Compare July 26, 2025 00:18
Base automatically changed from users/arsenm/amdgpu/add-tests-mfma-forced-vgpr-selection to main July 26, 2025 00:46
@arsenm arsenm force-pushed the users/arsenm/amdgpu/add-tests-mfma-src2-imm branch 7 times, most recently from a658fbc to 54837d8 Compare July 26, 2025 02:23
@arsenm arsenm force-pushed the users/arsenm/amdgpu/add-tests-mfma-src2-imm branch from 54837d8 to 4d8c839 Compare July 26, 2025 02:25
@arsenm arsenm merged commit 3f8a058 into main Jul 26, 2025
9 checks passed
@arsenm arsenm deleted the users/arsenm/amdgpu/add-tests-mfma-src2-imm branch July 26, 2025 02:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@shiltian shiltian shiltian approved these changes

@jrbyrnes jrbyrnes Awaiting requested review from jrbyrnes

@rampitec rampitec Awaiting requested review from rampitec

@srpande srpande Awaiting requested review from srpande

Assignees
No one assigned
Labels
backend:AMDGPU
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@arsenm @llvmbot @shiltian