[AMDGPU] s_get_shader_cycles_u64 gfx1250 instruction (#152390)

It is the same as reading SHADER_CYCLES_LO and SHADER_CYCLES_HI
but with a single instruction.

Author: rampitec

llvm/lib/Target/AMDGPU/AMDGPU.td

@@ -2612,6 +2612,10 @@ def HasPkMinMax3Insts :
   Predicate<"Subtarget->hasPkMinMax3Insts()">,
   AssemblerPredicate<(any_of FeatureGFX1250Insts)>;
 
+def HasSGetShaderCyclesInst :
+  Predicate<"Subtarget->hasSGetShaderCyclesInst()">,
+  AssemblerPredicate<(any_of FeatureGFX1250Insts)>;
+
 def HasImageInsts : Predicate<"Subtarget->hasImageInsts()">,
   AssemblerPredicate<(all_of FeatureImageInsts)>;
 

llvm/lib/Target/AMDGPU/GCNSubtarget.h

@@ -1562,6 +1562,9 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
   // \returns true if the target has V_PK_{MIN|MAX}3_{I|U}16 instructions.
   bool hasPkMinMax3Insts() const { return GFX1250Insts; }
 
+  // \returns ture if target has S_GET_SHADER_CYCLES_U64 instruction.
+  bool hasSGetShaderCyclesInst() const { return GFX1250Insts; }
+
   // \returns true if target has S_SETPRIO_INC_WG instruction.
   bool hasSetPrioIncWgInst() const { return HasSetPrioIncWgInst; }
 

llvm/lib/Target/AMDGPU/SOPInstructions.td

@@ -1653,6 +1653,12 @@ def S_SETPRIO_INC_WG : SOPP_Pseudo <"s_setprio_inc_wg", (ins i16imm:$simm16), "$
   let SubtargetPredicate = HasSetPrioIncWgInst;
 }
 
+def S_GET_SHADER_CYCLES_U64 : SOP1_64_0 <"s_get_shader_cycles_u64",
+  [(set i64:$sdst, (readcyclecounter))]> {
+  let SubtargetPredicate = HasSGetShaderCyclesInst;
+  let hasSideEffects = 1;
+}
+
 let Uses = [EXEC, M0] in {
 def S_SENDMSG : SOPP_Pseudo <"s_sendmsg" , (ins SendMsg:$simm16), "$simm16",
   [(int_amdgcn_s_sendmsg (i32 timm:$simm16), M0)]> {
@@ -2145,6 +2151,7 @@ defm S_ALLOC_VGPR                 : SOP1_Real_gfx12<0x053>;
 defm S_SLEEP_VAR                  : SOP1_IMM_Real_gfx12<0x058>;
 
 // GFX1250
+defm S_GET_SHADER_CYCLES_U64      : SOP1_Real_gfx12<0x06>;
 defm S_ADD_PC_I64                 : SOP1_Real_gfx12<0x04b>;
 
 //===----------------------------------------------------------------------===//

llvm/test/CodeGen/AMDGPU/readcyclecounter.ll

@@ -10,6 +10,8 @@
 ; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GETREG,GETREG-GISEL -check-prefix=GCN %s
 ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s
 ; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s
 
 declare i64 @llvm.readcyclecounter() #0
 
@@ -21,6 +23,7 @@ declare i64 @llvm.readcyclecounter() #0
 ; GFX12:       s_getreg_b32 [[HI2:s[0-9]+]], hwreg(HW_REG_SHADER_CYCLES_HI)
 ; GFX12:       s_cmp_eq_u32 [[HI1]], [[HI2]]
 ; GFX12:       s_cselect_b32 {{s[0-9]+}}, [[LO1]], 0
+; GFX1250:     s_get_shader_cycles_u64 s{{\[[0-9]+:[0-9]+\]}}
 ; GCN-DAG:     kmcnt
 ; MEMTIME:     store_dwordx2
 ; SIVI-NOT:    kmcnt
@@ -53,6 +56,7 @@ define amdgpu_kernel void @test_readcyclecounter(ptr addrspace(1) %out) #0 {
 ; GFX12:       s_getreg_b32 [[HI1:s[0-9]+]], hwreg(HW_REG_SHADER_CYCLES_HI)
 ; GFX12:       s_getreg_b32 [[LO1:s[0-9]+]], hwreg(HW_REG_SHADER_CYCLES_LO)
 ; GFX12:       s_getreg_b32 [[HI2:s[0-9]+]], hwreg(HW_REG_SHADER_CYCLES_HI)
+; GFX1250:     s_get_shader_cycles_u64 s{{\[[0-9]+:[0-9]+\]}}
 ; GCN-DAG:     s_load_{{dword|b32|b64}}
 ; GETREG-DAG:  s_getreg_b32 s{{[0-9]+}}, hwreg(HW_REG_SHADER_CYCLES, 0, 20)
 ; GFX12:       s_cmp_eq_u32 [[HI1]], [[HI2]]

llvm/test/MC/AMDGPU/gfx1250_asm_sop1.s

@@ -45,6 +45,10 @@ s_rfe_i64 s[2:3]
 s_rfe_b64 s[2:3]
 // GFX1250: s_rfe_i64 s[2:3]                        ; encoding: [0x02,0x4a,0x80,0xbe]
 
+s_get_shader_cycles_u64 s[2:3]
+// GFX1250: s_get_shader_cycles_u64 s[2:3]          ; encoding: [0x00,0x06,0x82,0xbe]
+// GFX12-ERR: :[[@LINE-2]]:{{[0-9]+}}: error: instruction not supported on this GPU
+
 s_barrier_signal -3
 // GFX1250: s_barrier_signal -3                     ; encoding: [0xc3,0x4e,0x80,0xbe]
 

llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_sop1.txt

@@ -12,6 +12,9 @@
 # GFX1250: s_add_pc_i64 s[2:3]                     ; encoding: [0x02,0x4b,0x80,0xbe]
 0x02,0x4b,0x80,0xbe
 
+# GFX1250: s_get_shader_cycles_u64 s[2:3]          ; encoding: [0x00,0x06,0x82,0xbe]
+0x00,0x06,0x82,0xbe
+
 # GFX1250: s_barrier_signal -3                     ; encoding: [0xc3,0x4e,0x80,0xbe]
 0xc3,0x4e,0x80,0xbe