Skip to content

[AMDGPU][NFC] Pre-commit test for redundant s_cmp_lg_* sX, 0 removal #162351

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
Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

[AMDGPU][NFC] Pre-commit test for redundant s_cmp_lg_* sX, 0 removal #162351

wants to merge 2 commits into from
+583 −0

Conversation

LU-JOHN
Copy link
Contributor

@LU-JOHN LU-JOHN commented Oct 7, 2025

Pre-commit test for redundant s_cmp_lg_* sX, 0 removal.

@llvmbot
Copy link
Member

llvmbot commented Oct 7, 2025

@llvm/pr-subscribers-backend-amdgpu

Author: None (LU-JOHN)

Changes

Pre-commit test for redundant s_cmp_lg_* sX, 0 removal.

Patch is 20.75 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/162351.diff

1 Files Affected:

  • (added) llvm/test/CodeGen/AMDGPU/s_cmp_0.ll (+585)
diff --git a/llvm/test/CodeGen/AMDGPU/s_cmp_0.ll b/llvm/test/CodeGen/AMDGPU/s_cmp_0.ll
new file mode 100644
index 0000000000000..8dc846c862200
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/s_cmp_0.ll
@@ -0,0 +1,585 @@
+; 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 %s
+
+declare i32 @llvm.ctpop.i32(i32)
+declare i64 @llvm.ctpop.i64(i64)
+declare i32 @llvm.amdgcn.s.quadmask.i32(i32)
+declare i64 @llvm.amdgcn.s.quadmask.i64(i64)
+
+define amdgpu_ps i32 @shl32(i32 inreg %val0, i32 inreg %val1) {
+; CHECK-LABEL: shl32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_lshl_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = shl i32 %val0, %val1
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @shl64(i64 inreg %val0, i64 inreg %val1) {
+; CHECK-LABEL: shl64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_lshl_b64 s[0:1], s[0:1], s2
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = shl i64 %val0, %val1
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @lshr32(i32 inreg %val0, i32 inreg %val1) {
+; CHECK-LABEL: lshr32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_lshr_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = lshr i32 %val0, %val1
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @lshr64(i64 inreg %val0, i64 inreg %val1) {
+; CHECK-LABEL: lshr64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_lshr_b64 s[0:1], s[0:1], s2
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = lshr i64 %val0, %val1
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @and32(i32 inreg %val0, i32 inreg %val1) {
+; CHECK-LABEL: and32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_and_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = and i32 %val0, %val1
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @and64(i64 inreg %val0, i64 inreg %val1) {
+; CHECK-LABEL: and64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_and_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = and i64 %val0, %val1
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @or32(i32 inreg %val0, i32 inreg %val1) {
+; CHECK-LABEL: or32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_or_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = or i32 %val0, %val1
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @or64(i64 inreg %val0, i64 inreg %val1) {
+; CHECK-LABEL: or64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_or_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = or i64 %val0, %val1
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @xor32(i32 inreg %val0, i32 inreg %val1) {
+; CHECK-LABEL: xor32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_xor_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = xor i32 %val0, %val1
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @xor64(i64 inreg %val0, i64 inreg %val1) {
+; CHECK-LABEL: xor64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_xor_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = xor i64 %val0, %val1
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @nand32(i32 inreg %val0, i32 inreg %val1, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: nand32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_nand_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dword v[0:1], v2, off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = and i32 %val0, %val1
+  %result2 = xor i32 %result, -1
+  store i32 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i32 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @nand64(i64 inreg %val0, i64 inreg %val1, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: nand64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_nand_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = and i64 %val0, %val1
+  %result2 = xor i64 %result, -1
+  store i64 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @nor32(i32 inreg %val0, i32 inreg %val1, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: nor32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_nor_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dword v[0:1], v2, off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = or i32 %val0, %val1
+  %result2 = xor i32 %result, -1
+  store i32 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i32 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @nor64(i64 inreg %val0, i64 inreg %val1, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: nor64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_nor_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = or i64 %val0, %val1
+  %result2 = xor i64 %result, -1
+  store i64 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @xnor32(i32 inreg %val0, i32 inreg %val1, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: xnor32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_xnor_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dword v[0:1], v2, off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = xor i32 %val0, %val1
+  %result2 = xor i32 %result, -1
+  store i32 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i32 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @xnor64(i64 inreg %val0, i64 inreg %val1, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: xnor64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_xnor_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = xor i64 %val0, %val1
+  %result2 = xor i64 %result, -1
+  store i64 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @andn232(i32 inreg %val0, i32 inreg %val1) {
+; CHECK-LABEL: andn232:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_andn2_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %nval1 = xor i32 %val1, -1
+  %result = and i32 %val0, %nval1
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @nandn264(i64 inreg %val0, i64 inreg %val1) {
+; CHECK-LABEL: nandn264:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_andn2_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %nval1 = xor i64 %val1, -1
+  %result = and i64 %val0, %nval1
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @orn232(i32 inreg %val0, i32 inreg %val1) {
+; CHECK-LABEL: orn232:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_orn2_b32 s0, s0, s1
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %nval1 = xor i32 %val1, -1
+  %result = or i32 %val0, %nval1
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @orn264(i64 inreg %val0, i64 inreg %val1) {
+; CHECK-LABEL: orn264:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_orn2_b64 s[0:1], s[0:1], s[2:3]
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %nval1 = xor i64 %val1, -1
+  %result = or i64 %val0, %nval1
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bfe_i32(i32 inreg %val0) {
+; CHECK-LABEL: bfe_i32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_bfe_i32 s0, s0, 0x80010
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %shl = shl i32 %val0, 8
+  %result = ashr i32 %shl, 24
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bfe_i64(i64 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: bfe_i64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_bfe_i64 s[2:3], s[0:1], 0x80000
+; CHECK-NEXT:    s_and_b32 s0, s0, 0xff
+; CHECK-NEXT:    s_mov_b32 s1, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v3, s3
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %shl = shl i64 %val0, 56
+  %result = ashr i64 %shl, 56
+  store i64 %result, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bfe_u32(i32 inreg %val0) {
+; CHECK-LABEL: bfe_u32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_bfe_u32 s0, s0, 0x80010
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    ; return to shader part epilog
+  %shl = shl i32 %val0, 8
+  %result = lshr i32 %shl, 24
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bfe_u64(i64 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: bfe_u64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_and_b32 s0, s0, 0xff
+; CHECK-NEXT:    s_mov_b32 s1, 0
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %shl = shl i64 %val0, 56
+  %result = lshr i64 %shl, 56
+  store i64 %result, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bcnt032(i32 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: bcnt032:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_bcnt1_i32_b32 s0, s0
+; CHECK-NEXT:    s_sub_i32 s0, 32, s0
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dword v[0:1], v2, off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = call i32 @llvm.ctpop.i32(i32 %val0) nounwind readnone
+  %result2 = sub i32 32, %result
+  store i32 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i32 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bcnt064(i64 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: bcnt064:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_bcnt1_i32_b64 s0, s[0:1]
+; CHECK-NEXT:    s_sub_u32 s0, 64, s0
+; CHECK-NEXT:    s_subb_u32 s1, 0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = call i64 @llvm.ctpop.i64(i64 %val0) nounwind readnone
+  %result2 = sub i64 64, %result
+  store i64 %result2, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result2, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bcnt132(i32 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: bcnt132:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_bcnt1_i32_b32 s0, s0
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dword v[0:1], v2, off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = call i32 @llvm.ctpop.i32(i32 %val0) nounwind readnone
+  store i32 %result, ptr addrspace(1) %ptr
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @bcnt164(i64 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: bcnt164:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_bcnt1_i32_b64 s0, s[0:1]
+; CHECK-NEXT:    s_mov_b32 s1, 0
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = call i64 @llvm.ctpop.i64(i64 %val0) nounwind readnone
+  store i64 %result, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @quadmask32(i32 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: quadmask32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_quadmask_b32 s0, s0
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dword v[0:1], v2, off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = call i32 @llvm.amdgcn.s.quadmask.i32(i32 %val0) nounwind readnone
+  store i32 %result, ptr addrspace(1) %ptr
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @quadmask64(i64 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: quadmask64:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_quadmask_b64 s[0:1], s[0:1]
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    s_cmp_lg_u64 s[0:1], 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = call i64 @llvm.amdgcn.s.quadmask.i64(i64 %val0) nounwind readnone
+  store i64 %result, ptr addrspace(1) %ptr
+  %cmp = icmp ne i64 %result, 0
+  %zext = zext i1 %cmp to i32
+  ret i32 %zext
+}
+
+define amdgpu_ps i32 @not32(i32 inreg %val0, ptr addrspace(1) %ptr) {
+; CHECK-LABEL: not32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_not_b32 s0, s0
+; CHECK-NEXT:    s_cmp_lg_u32 s0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_cselect_b64 s[0:1], -1, 0
+; CHECK-NEXT:    global_store_dword v[0:1], v2, off
+; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
+; CHECK-NEXT:    v_readfirstlane_b32 s0, v0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ; return to shader part epilog
+  %result = xor i32 %val0, -1
+  store i32 %result, ptr addrspace(1) %ptr
+  %cmp = icmp ne i32 %result, 0
+  %zext = zext...
[truncated]

@LU-JOHN
Copy link
Contributor Author

LU-JOHN commented Oct 7, 2025

Deleting redundant s_cmp_lg* instruction done in #162352.

arsenm
arsenm reviewed Oct 8, 2025
View reviewed changes
llvm/test/CodeGen/AMDGPU/s_cmp_0.ll
; CHECK-NEXT: ; return to shader part epilog
%result = and i32 %val0, %val1
%result2 = xor i32 %result, -1
store i32 %result2, ptr addrspace(1) %ptr
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Using VALU stores for tests for scalar handling is suspicious, are these just for adding an extra use?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Using VALU stores for tests for scalar handling is suspicious, are these just for adding an extra use?

The stores are done to ensure the compiler cannot optimize the operation into another instruction.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@arsenm arsenm arsenm left review comments

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
@LU-JOHN @llvmbot @arsenm