AMDGPU: Handle gfx950 XDL-write-VGPR-VALU-Mem-Exp wait state change (llvm#126727)

(cherry picked from commit a2263eb)

Author: VigneshwarJ

llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp

@@ -2593,12 +2593,14 @@ static int GFX940_XDL_N_PassWriteVgprVALUWawWaitStates(int NumPasses) {
   return NumPasses + 3;
 }
 
-static int GFX940_XDL_N_PassWriteVgprVALUMemExpReadWaitStates(int NumPasses) {
-  // 2 pass -> 5
-  // 4 pass -> 7
-  // 8 pass -> 11
-  // 16 pass -> 19
-  return NumPasses + 3;
+static int GFX940_XDL_N_PassWriteVgprVALUMemExpReadWaitStates(int NumPasses,
+                                                              bool IsGFX950) {
+  // xdl def cycles | gfx940 | gfx950
+  // 2 pass         |  5        5
+  // 4 pass         |  7        8
+  // 8 pass         |  11       12
+  // 16 pass        |  19       20
+  return NumPasses + 3 + (NumPasses != 2 && IsGFX950);
 }
 
 static int GFX940_SMFMA_N_PassWriteVgprVALUMemExpReadWaitStates(int NumPasses) {
@@ -2749,7 +2751,8 @@ int GCNHazardRecognizer::checkMAIVALUHazards(MachineInstr *MI) {
       } else if (ST.hasGFX940Insts()) {
         NeedWaitStates =
             isXDL(ST, *MFMA)
-                ? GFX940_XDL_N_PassWriteVgprVALUMemExpReadWaitStates(NumPasses)
+                ? GFX940_XDL_N_PassWriteVgprVALUMemExpReadWaitStates(
+                      NumPasses, ST.hasGFX950Insts())
                 : GFX940_SMFMA_N_PassWriteVgprVALUMemExpReadWaitStates(
                       NumPasses);
       } else {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll

@@ -49,7 +49,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16(<8 x bfloat> %arg0, <8 x
 ; GCN-NEXT:    v_mov_b32_e32 v9, s13
 ; GCN-NEXT:    v_mov_b32_e32 v10, s14
 ; GCN-NEXT:    v_mov_b32_e32 v11, s15
-; GCN-NEXT:    s_nop 3
+; GCN-NEXT:    s_nop 4
 ; GCN-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
@@ -122,7 +122,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__flags(<8 x bfloat> %arg0
 ; GCN-NEXT:    v_mov_b32_e32 v9, s13
 ; GCN-NEXT:    v_mov_b32_e32 v10, s14
 ; GCN-NEXT:    v_mov_b32_e32 v11, s15
-; GCN-NEXT:    s_nop 3
+; GCN-NEXT:    s_nop 4
 ; GCN-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
@@ -179,7 +179,7 @@ define <16 x float> @test_mfma_f32_32x32x16_bf16__mac(<8 x bfloat> %arg0, <8 x b
 ; GCN-NEXT:    s_nop 1
 ; GCN-NEXT:    v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[0:15]
 ; GCN-NEXT:    s_nop 7
-; GCN-NEXT:    s_nop 2
+; GCN-NEXT:    s_nop 3
 ; GCN-NEXT:    v_accvgpr_read_b32 v0, a0
 ; GCN-NEXT:    v_accvgpr_read_b32 v1, a1
 ; GCN-NEXT:    v_accvgpr_read_b32 v2, a2
@@ -224,7 +224,7 @@ define <16 x float> @test_mfma_f32_32x32x16_bf16__mac__flags(<8 x bfloat> %arg0,
 ; GCN-NEXT:    s_nop 1
 ; GCN-NEXT:    v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:1 abid:1 blgp:1
 ; GCN-NEXT:    s_nop 7
-; GCN-NEXT:    s_nop 2
+; GCN-NEXT:    s_nop 3
 ; GCN-NEXT:    v_accvgpr_read_b32 v0, a0
 ; GCN-NEXT:    v_accvgpr_read_b32 v1, a1
 ; GCN-NEXT:    v_accvgpr_read_b32 v2, a2
@@ -396,7 +396,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd_mac(<8 x bfloat>
 ; GCN-NEXT:    v_mov_b32_e32 v16, 0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_nop 7
-; GCN-NEXT:    s_nop 0
+; GCN-NEXT:    s_nop 1
 ; GCN-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
 ; GCN-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
 ; GCN-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
@@ -431,7 +431,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd_mac_flags(<8 x bf
 ; GCN-NEXT:    v_mov_b32_e32 v16, 0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_nop 7
-; GCN-NEXT:    s_nop 0
+; GCN-NEXT:    s_nop 1
 ; GCN-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
 ; GCN-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
 ; GCN-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16