[AMDGPU] Use s_cmp instead of s_cmpk

Don't bother pre-shrinking "s_cmp_lg_u32 reg, 0" to s_cmpk_lg_u32
because 0 is already an inline constant so the s_cmpk form is no
smaller.

This is just for consistency with the surrounding code and to simplify a
downstream patch.

Differential Revision: https://reviews.llvm.org/D138993

diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index d55b730..0fe6ffc 100644 (file)
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -4202,7 +4202,7 @@ MachineBasicBlock *SITargetLowering::EmitInstrWithCustomInserter(
             .addImm(0);
       }
     } else {
-      BuildMI(*BB, MII, DL, TII->get(AMDGPU::S_CMPK_LG_U32))
+      BuildMI(*BB, MII, DL, TII->get(AMDGPU::S_CMP_LG_U32))
           .addReg(Src2.getReg())
           .addImm(0);
     }

diff --git a/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll b/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll
index 776de7d..e307239 100644 (file)
--- a/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll
+++ b/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll
@@ -50,7 +50,7 @@ define i32 @s_add_co_select_user() {
 ; GFX10-NEXT:    s_load_dword s4, s[4:5], 0x0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_add_co_u32 v0, s5, s4, s4
-; GFX10-NEXT:    s_cmpk_lg_u32 s5, 0x0
+; GFX10-NEXT:    s_cmp_lg_u32 s5, 0
 ; GFX10-NEXT:    s_addc_u32 s5, s4, 0
 ; GFX10-NEXT:    s_cselect_b32 s6, -1, 0
 ; GFX10-NEXT:    s_and_b32 s6, s6, exec_lo
@@ -69,7 +69,7 @@ define i32 @s_add_co_select_user() {
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_add_co_u32 v0, s1, s0, s0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT:    s_cmpk_lg_u32 s1, 0x0
+; GFX11-NEXT:    s_cmp_lg_u32 s1, 0
 ; GFX11-NEXT:    s_addc_u32 s1, s0, 0
 ; GFX11-NEXT:    s_cselect_b32 s2, -1, 0
 ; GFX11-NEXT:    s_and_b32 s2, s2, exec_lo
@@ -154,7 +154,7 @@ define amdgpu_kernel void @s_add_co_br_user(i32 %i) {
 ; GFX10-NEXT:    s_cmp_lt_u32 s1, s0
 ; GFX10-NEXT:    s_cselect_b32 s1, -1, 0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s1
-; GFX10-NEXT:    s_cmpk_lg_u32 s1, 0x0
+; GFX10-NEXT:    s_cmp_lg_u32 s1, 0
 ; GFX10-NEXT:    s_addc_u32 s0, s0, 0
 ; GFX10-NEXT:    v_cmp_ge_u32_e32 vcc_lo, s0, v0
 ; GFX10-NEXT:    s_cbranch_vccnz .LBB1_2
@@ -181,7 +181,7 @@ define amdgpu_kernel void @s_add_co_br_user(i32 %i) {
 ; GFX11-NEXT:    s_cmp_lt_u32 s1, s0
 ; GFX11-NEXT:    s_cselect_b32 s1, -1, 0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s1
-; GFX11-NEXT:    s_cmpk_lg_u32 s1, 0x0
+; GFX11-NEXT:    s_cmp_lg_u32 s1, 0
 ; GFX11-NEXT:    s_addc_u32 s0, s0, 0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_cmp_ge_u32_e32 vcc_lo, s0, v0

diff --git a/llvm/test/CodeGen/AMDGPU/udiv.ll b/llvm/test/CodeGen/AMDGPU/udiv.ll
index 889c3fc..3cbb8fa 100644 (file)
--- a/llvm/test/CodeGen/AMDGPU/udiv.ll
+++ b/llvm/test/CodeGen/AMDGPU/udiv.ll
@@ -2807,7 +2807,7 @@ define i64 @v_test_udiv64_mulhi_fold(i64 %arg) {
 ; GFX1030-NEXT:    s_add_u32 s4, 0x4237, s4
 ; GFX1030-NEXT:    s_addc_u32 s5, 0, 0
 ; GFX1030-NEXT:    v_add_co_u32 v2, s4, 0xa9000000, s4
-; GFX1030-NEXT:    s_cmpk_lg_u32 s4, 0x0
+; GFX1030-NEXT:    s_cmp_lg_u32 s4, 0
 ; GFX1030-NEXT:    s_addc_u32 s5, s5, 0xa7c5
 ; GFX1030-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX1030-NEXT:    s_mul_i32 s6, s5, 0xfffe7960
@@ -2830,7 +2830,7 @@ define i64 @v_test_udiv64_mulhi_fold(i64 %arg) {
 ; GFX1030-NEXT:    s_add_u32 s4, s4, s7
 ; GFX1030-NEXT:    s_addc_u32 s6, 0, s6
 ; GFX1030-NEXT:    v_add_co_u32 v4, s4, v2, s4
-; GFX1030-NEXT:    s_cmpk_lg_u32 s4, 0x0
+; GFX1030-NEXT:    s_cmp_lg_u32 s4, 0
 ; GFX1030-NEXT:    s_addc_u32 s4, s5, s6
 ; GFX1030-NEXT:    v_mul_hi_u32 v8, v0, v4
 ; GFX1030-NEXT:    v_mad_u64_u32 v[2:3], null, v0, s4, 0