AMDGPU: Add baseline test for #161651

[arsenm]

No description provided.

[arsenm]

• AMDGPU: Add baseline test for #161651 #165921 👈 (View in Graphite)
• main

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

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/165921.diff

1 Files Affected:

• (added) llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll (+236)

diff --git a/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll b/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll
new file mode 100644
index 0000000000000..22e4a24435f12
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll
@@ -0,0 +1,236 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 < %s | FileCheck -check-prefix=GFX11 %s
+
+define i16 @v_underflow_compare_fold_i16(i16 %a, i16 %b) #0 {
+; GFX9-LABEL: v_underflow_compare_fold_i16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_sub_u16_e32 v1, v0, v1
+; GFX9-NEXT:    v_min_u16_e32 v0, v1, v0
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_underflow_compare_fold_i16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_sub_nc_u16 v0.h, v0.l, v1.l
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_min_u16 v0.l, v0.h, v0.l
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %sub = sub i16 %a, %b
+  %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a)
+  ret i16 %cond
+}
+
+define i32 @v_underflow_compare_fold_i32(i32 %a, i32 %b) #0 {
+; GFX9-LABEL: v_underflow_compare_fold_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_sub_u32_e32 v1, v0, v1
+; GFX9-NEXT:    v_min_u32_e32 v0, v1, v0
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_underflow_compare_fold_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_sub_nc_u32_e32 v1, v0, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_min_u32_e32 v0, v1, v0
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %sub = sub i32 %a, %b
+  %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a)
+  ret i32 %cond
+}
+
+define i32 @v_underflow_compare_fold_i32_commute(i32 %a, i32 %b) #0 {
+; GFX9-LABEL: v_underflow_compare_fold_i32_commute:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_sub_u32_e32 v1, v0, v1
+; GFX9-NEXT:    v_min_u32_e32 v0, v0, v1
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_underflow_compare_fold_i32_commute:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_sub_nc_u32_e32 v1, v0, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_min_u32_e32 v0, v0, v1
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %sub = sub i32 %a, %b
+  %cond = call i32 @llvm.umin.i32(i32 %a, i32 %sub)
+  ret i32 %cond
+}
+
+define i32 @v_underflow_compare_fold_i32_multi_use(i32 %a, i32 %b, ptr addrspace(1) %ptr) #0 {
+; GFX9-LABEL: v_underflow_compare_fold_i32_multi_use:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_sub_u32_e32 v1, v0, v1
+; GFX9-NEXT:    v_min_u32_e32 v0, v1, v0
+; GFX9-NEXT:    global_store_dword v[2:3], v1, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_underflow_compare_fold_i32_multi_use:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_sub_nc_u32_e32 v1, v0, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_min_u32_e32 v0, v1, v0
+; GFX11-NEXT:    global_store_b32 v[2:3], v1, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %sub = sub i32 %a, %b
+  store i32 %sub, ptr addrspace(1) %ptr
+  %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a)
+  ret i32 %cond
+}
+
+define i64 @v_underflow_compare_fold_i64(i64 %a, i64 %b) #0 {
+; GFX9-LABEL: v_underflow_compare_fold_i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_sub_co_u32_e32 v2, vcc, v0, v2
+; GFX9-NEXT:    v_subb_co_u32_e32 v3, vcc, v1, v3, vcc
+; GFX9-NEXT:    v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1]
+; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX9-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_underflow_compare_fold_i64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_sub_co_u32 v2, vcc_lo, v0, v2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo
+; GFX11-NEXT:    v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1]
+; GFX11-NEXT:    v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %sub = sub i64 %a, %b
+  %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a)
+  ret i64 %cond
+}
+
+define i64 @v_underflow_compare_fold_i64_commute(i64 %a, i64 %b) #0 {
+; GFX9-LABEL: v_underflow_compare_fold_i64_commute:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_sub_co_u32_e32 v2, vcc, v0, v2
+; GFX9-NEXT:    v_subb_co_u32_e32 v3, vcc, v1, v3, vcc
+; GFX9-NEXT:    v_cmp_lt_u64_e32 vcc, v[0:1], v[2:3]
+; GFX9-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; GFX9-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_underflow_compare_fold_i64_commute:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_sub_co_u32 v2, vcc_lo, v0, v2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo
+; GFX11-NEXT:    v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[2:3]
+; GFX11-NEXT:    v_dual_cndmask_b32 v0, v2, v0 :: v_dual_cndmask_b32 v1, v3, v1
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %sub = sub i64 %a, %b
+  %cond = call i64 @llvm.umin.i64(i64 %a, i64 %sub)
+  ret i64 %cond
+}
+
+define i64 @v_underflow_compare_fold_i64_multi_use(i64 %a, i64 %b, ptr addrspace(1) %ptr) #0 {
+; GFX9-LABEL: v_underflow_compare_fold_i64_multi_use:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_sub_co_u32_e32 v2, vcc, v0, v2
+; GFX9-NEXT:    v_subb_co_u32_e32 v3, vcc, v1, v3, vcc
+; GFX9-NEXT:    v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1]
+; GFX9-NEXT:    global_store_dwordx2 v[4:5], v[2:3], off
+; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX9-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_underflow_compare_fold_i64_multi_use:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_sub_co_u32 v2, vcc_lo, v0, v2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo
+; GFX11-NEXT:    v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1]
+; GFX11-NEXT:    global_store_b64 v[4:5], v[2:3], off
+; GFX11-NEXT:    v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %sub = sub i64 %a, %b
+  store i64 %sub, ptr addrspace(1) %ptr
+  %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a)
+  ret i64 %cond
+}
+
+define amdgpu_ps i16 @s_underflow_compare_fold_i16(i16 inreg %a, i16 inreg %b) #0 {
+; GFX9-LABEL: s_underflow_compare_fold_i16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_sub_i32 s1, s0, s1
+; GFX9-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX9-NEXT:    s_and_b32 s1, s1, 0xffff
+; GFX9-NEXT:    s_min_u32 s0, s1, s0
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: s_underflow_compare_fold_i16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_sub_i32 s1, s0, s1
+; GFX11-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX11-NEXT:    s_and_b32 s1, s1, 0xffff
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    s_min_u32 s0, s1, s0
+; GFX11-NEXT:    ; return to shader part epilog
+  %sub = sub i16 %a, %b
+  %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a)
+  ret i16 %cond
+}
+
+define amdgpu_ps i32 @s_underflow_compare_fold_i32(i32 inreg %a, i32 inreg %b) #0 {
+; GFX9-LABEL: s_underflow_compare_fold_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_sub_i32 s1, s0, s1
+; GFX9-NEXT:    s_min_u32 s0, s1, s0
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: s_underflow_compare_fold_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_sub_i32 s1, s0, s1
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    s_min_u32 s0, s1, s0
+; GFX11-NEXT:    ; return to shader part epilog
+  %sub = sub i32 %a, %b
+  %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a)
+  ret i32 %cond
+}
+
+define amdgpu_ps i64 @s_underflow_compare_fold_i64(i64 inreg %a, i64 inreg %b) #0 {
+; GFX9-LABEL: s_underflow_compare_fold_i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_sub_u32 s2, s0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    s_subb_u32 s3, s1, s3
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1]
+; GFX9-NEXT:    s_and_b64 s[4:5], vcc, exec
+; GFX9-NEXT:    s_cselect_b32 s1, s3, s1
+; GFX9-NEXT:    s_cselect_b32 s0, s2, s0
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: s_underflow_compare_fold_i64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_sub_u32 s2, s0, s2
+; GFX11-NEXT:    s_subb_u32 s3, s1, s3
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    v_cmp_lt_u64_e64 s4, s[2:3], s[0:1]
+; GFX11-NEXT:    s_and_b32 s4, s4, exec_lo
+; GFX11-NEXT:    s_cselect_b32 s0, s2, s0
+; GFX11-NEXT:    s_cselect_b32 s1, s3, s1
+; GFX11-NEXT:    ; return to shader part epilog
+  %sub = sub i64 %a, %b
+  %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a)
+  ret i64 %cond
+}
+
+attributes #0 = { nounwind }