[llvm] c1b2920 - [AMDGPU] Autogenerate llvm.amdgcn.fcmp.ll
Pierre van Houtryve via llvm-commits
llvm-commits at lists.llvm.org
Wed Oct 26 00:00:39 PDT 2022
Author: Pierre van Houtryve
Date: 2022-10-26T07:00:34Z
New Revision: c1b2920c6ef1ef23d7d6805b9aefe3c325708e15
URL: https://github.com/llvm/llvm-project/commit/c1b2920c6ef1ef23d7d6805b9aefe3c325708e15
DIFF: https://github.com/llvm/llvm-project/commit/c1b2920c6ef1ef23d7d6805b9aefe3c325708e15.diff
LOG: [AMDGPU] Autogenerate llvm.amdgcn.fcmp.ll
Prep commit for adding GISel run lines to that test.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D136591
Added:
Modified:
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll
index 6ff66b94fdfed..85d19dbf51eab 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll
@@ -1,5 +1,7 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr="+wavefrontsize64" -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX11 %s
+; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9 %s
declare i64 @llvm.amdgcn.fcmp.f32(float, float, i32) #0
declare i64 @llvm.amdgcn.fcmp.f64(double, double, i32) #0
@@ -8,18 +10,88 @@ declare float @llvm.fabs.f32(float) #0
declare i64 @llvm.amdgcn.fcmp.f16(half, half, i32) #0
declare half @llvm.fabs.f16(half) #0
-; GCN-LABEL: {{^}}v_fcmp_f32_oeq_with_fabs:
-; GCN: v_cmp_eq_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}}, |{{v[0-9]+}}|
define amdgpu_kernel void @v_fcmp_f32_oeq_with_fabs(i64 addrspace(1)* %out, float %src, float %a) {
+; GFX11-LABEL: v_fcmp_f32_oeq_with_fabs:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_eq_f32_e64 s[2:3], s2, |s3|
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oeq_with_fabs:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s3
+; VI-NEXT: v_cmp_eq_f32_e64 s[2:3], s2, |v0|
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_oeq_with_fabs:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s3
+; GFX9-NEXT: v_cmp_eq_f32_e64 s[2:3], s2, |v0|
+; GFX9-NEXT: v_mov_b32_e32 v0, s2
+; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT: s_endpgm
%temp = call float @llvm.fabs.f32(float %a)
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float %temp, i32 1)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_oeq_both_operands_with_fabs:
-; GCN: v_cmp_eq_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, |{{s[0-9]+}}|, |{{v[0-9]+}}|
define amdgpu_kernel void @v_fcmp_f32_oeq_both_operands_with_fabs(i64 addrspace(1)* %out, float %src, float %a) {
+; GFX11-LABEL: v_fcmp_f32_oeq_both_operands_with_fabs:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_eq_f32_e64 s[2:3], |s2|, |s3|
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oeq_both_operands_with_fabs:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s3
+; VI-NEXT: v_cmp_eq_f32_e64 s[2:3], |s2|, |v0|
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_oeq_both_operands_with_fabs:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s3
+; GFX9-NEXT: v_cmp_eq_f32_e64 s[2:3], |s2|, |v0|
+; GFX9-NEXT: v_mov_b32_e32 v0, s2
+; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT: s_endpgm
%temp = call float @llvm.fabs.f32(float %a)
%src_input = call float @llvm.fabs.f32(float %src)
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src_input, float %temp, i32 1)
@@ -27,227 +99,1232 @@ define amdgpu_kernel void @v_fcmp_f32_oeq_both_operands_with_fabs(i64 addrspace(
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32:
-; GCN-NOT: v_cmp_eq_f32_e64
define amdgpu_kernel void @v_fcmp_f32(i64 addrspace(1)* %out, float %src) {
+; GCN-LABEL: v_fcmp_f32:
+; GCN: ; %bb.0:
+; GCN-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 -1)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_oeq:
-; GCN: v_cmp_eq_f32_e64
define amdgpu_kernel void @v_fcmp_f32_oeq(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_oeq:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_eq_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oeq:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_eq_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_oeq:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_eq_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 1)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_one:
-; GCN: v_cmp_neq_f32_e64
define amdgpu_kernel void @v_fcmp_f32_one(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_one:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_neq_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_one:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_neq_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_one:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_neq_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 6)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_ogt:
-; GCN: v_cmp_gt_f32_e64
define amdgpu_kernel void @v_fcmp_f32_ogt(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_ogt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_lt_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ogt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_gt_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_ogt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_gt_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 2)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_oge:
-; GCN: v_cmp_ge_f32_e64
define amdgpu_kernel void @v_fcmp_f32_oge(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_oge:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_le_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oge:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_ge_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_oge:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_ge_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 3)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_olt:
-; GCN: v_cmp_lt_f32_e64
define amdgpu_kernel void @v_fcmp_f32_olt(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_olt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_gt_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_olt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_lt_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_olt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_lt_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 4)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_ole:
-; GCN: v_cmp_le_f32_e64
define amdgpu_kernel void @v_fcmp_f32_ole(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_ole:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_ge_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ole:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_le_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_ole:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_le_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 5)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_ueq:
-; GCN: v_cmp_nlg_f32_e64
define amdgpu_kernel void @v_fcmp_f32_ueq(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_ueq:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nlg_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ueq:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nlg_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_ueq:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nlg_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 9)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_une:
-; GCN: v_cmp_neq_f32_e64
define amdgpu_kernel void @v_fcmp_f32_une(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_une:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_neq_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_une:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_neq_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_une:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_neq_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 14)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_ugt:
-; GCN: v_cmp_nle_f32_e64
define amdgpu_kernel void @v_fcmp_f32_ugt(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_ugt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nge_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ugt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nle_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_ugt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nle_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 10)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_uge:
-; GCN: v_cmp_nlt_f32_e64
define amdgpu_kernel void @v_fcmp_f32_uge(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_uge:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_ngt_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_uge:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nlt_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_uge:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nlt_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 11)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_ult:
-; GCN: v_cmp_nge_f32_e64
define amdgpu_kernel void @v_fcmp_f32_ult(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_ult:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nle_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ult:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nge_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_ult:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nge_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 12)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f32_ule:
-; GCN: v_cmp_ngt_f32_e64
define amdgpu_kernel void @v_fcmp_f32_ule(i64 addrspace(1)* %out, float %src) {
+; GFX11-LABEL: v_fcmp_f32_ule:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nlt_f32_e64 s[2:3], 0x42c80000, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ule:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_ngt_f32_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f32_ule:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x42c80000
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_ngt_f32_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 13)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_oeq:
-; GCN: v_cmp_eq_f64_e64
define amdgpu_kernel void @v_fcmp_f64_oeq(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_oeq:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_eq_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_oeq:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_eq_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_oeq:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_eq_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 1)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_one:
-; GCN: v_cmp_neq_f64_e64
define amdgpu_kernel void @v_fcmp_f64_one(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_one:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_neq_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_one:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_one:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 6)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_ogt:
-; GCN: v_cmp_gt_f64_e64
define amdgpu_kernel void @v_fcmp_f64_ogt(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_ogt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_gt_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ogt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_gt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_ogt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_gt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 2)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_oge:
-; GCN: v_cmp_ge_f64_e64
define amdgpu_kernel void @v_fcmp_f64_oge(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_oge:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_ge_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_oge:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_ge_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_oge:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_ge_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 3)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_olt:
-; GCN: v_cmp_lt_f64_e64
define amdgpu_kernel void @v_fcmp_f64_olt(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_olt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_lt_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_olt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_lt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_olt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_lt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 4)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_ole:
-; GCN: v_cmp_le_f64_e64
define amdgpu_kernel void @v_fcmp_f64_ole(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_ole:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_le_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ole:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_le_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_ole:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_le_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 5)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_ueq:
-; GCN: v_cmp_nlg_f64_e64
define amdgpu_kernel void @v_fcmp_f64_ueq(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_ueq:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nlg_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ueq:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nlg_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_ueq:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nlg_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 9)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_une:
-; GCN: v_cmp_neq_f64_e64
define amdgpu_kernel void @v_fcmp_f64_une(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_une:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_neq_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_une:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_une:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 14)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_ugt:
-; GCN: v_cmp_nle_f64_e64
define amdgpu_kernel void @v_fcmp_f64_ugt(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_ugt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nle_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ugt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nle_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_ugt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nle_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 10)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_uge:
-; GCN: v_cmp_nlt_f64_e64
define amdgpu_kernel void @v_fcmp_f64_uge(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_uge:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nlt_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_uge:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nlt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_uge:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nlt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 11)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_ult:
-; GCN: v_cmp_nge_f64_e64
define amdgpu_kernel void @v_fcmp_f64_ult(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_ult:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nge_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ult:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nge_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_ult:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nge_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 12)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f64_ule:
-; GCN: v_cmp_ngt_f64_e64
define amdgpu_kernel void @v_fcmp_f64_ule(i64 addrspace(1)* %out, double %src) {
+; GFX11-LABEL: v_fcmp_f64_ule:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_mov_b32 s4, 0
+; GFX11-NEXT: s_mov_b32 s5, 0x40590000
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_ngt_f64_e64 s[2:3], s[2:3], s[4:5]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ule:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_ngt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f64_ule:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0x40590000
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_ngt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 13)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_oeq_with_fabs:
-; VI: v_cmp_eq_f16_e64 {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}}, |{{v[0-9]+}}|
-; SI: v_cvt_f32_f16_e32 [[CVT0:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cvt_f32_f16_e64 [[CVT1:v[0-9]+]], |s{{[0-9]+}}|
-; SI: v_cmp_eq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[CVT0]], [[CVT1]]
define amdgpu_kernel void @v_fcmp_f16_oeq_with_fabs(i64 addrspace(1)* %out, half %src, half %a) {
+; GFX11-LABEL: v_fcmp_f16_oeq_with_fabs:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_lshr_b32 s3, s2, 16
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_cmp_eq_f16_e64 s[2:3], s2, |s3|
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oeq_with_fabs:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s3, s2, 16
+; VI-NEXT: v_mov_b32_e32 v0, s3
+; VI-NEXT: v_cmp_eq_f16_e64 s[2:3], s2, |v0|
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_oeq_with_fabs:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_lshr_b32 s0, s4, 16
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_cmp_eq_f16_e64 s[0:1], s4, |v0|
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%temp = call half @llvm.fabs.f16(half %a)
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half %temp, i32 1)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_oeq_both_operands_with_fabs:
-; VI: v_cmp_eq_f16_e64 {{s\[[0-9]+:[0-9]+\]}}, |{{s[0-9]+}}|, |{{v[0-9]+}}|
-; SI: v_cvt_f32_f16_e64 [[CVT0:v[0-9]+]], |s{{[0-9]+}}|
-; SI: v_cvt_f32_f16_e64 [[CVT1:v[0-9]+]], |s{{[0-9]+}}|
-; SI: v_cmp_eq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[CVT0]], [[CVT1]]
define amdgpu_kernel void @v_fcmp_f16_oeq_both_operands_with_fabs(i64 addrspace(1)* %out, half %src, half %a) {
+; GFX11-LABEL: v_fcmp_f16_oeq_both_operands_with_fabs:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_lshr_b32 s3, s2, 16
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_cmp_eq_f16_e64 s[2:3], |s2|, |s3|
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oeq_both_operands_with_fabs:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s3, s2, 16
+; VI-NEXT: v_mov_b32_e32 v0, s3
+; VI-NEXT: v_cmp_eq_f16_e64 s[2:3], |s2|, |v0|
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_oeq_both_operands_with_fabs:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_lshr_b32 s0, s4, 16
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_cmp_eq_f16_e64 s[0:1], |s4|, |v0|
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%temp = call half @llvm.fabs.f16(half %a)
%src_input = call half @llvm.fabs.f16(half %src)
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src_input, half %temp, i32 1)
@@ -255,153 +1332,586 @@ define amdgpu_kernel void @v_fcmp_f16_oeq_both_operands_with_fabs(i64 addrspace(
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16:
-; GCN-NOT: v_cmp_eq_
define amdgpu_kernel void @v_fcmp_f16(i64 addrspace(1)* %out, half %src) {
+; GCN-LABEL: v_fcmp_f16:
+; GCN: ; %bb.0:
+; GCN-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 -1)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_oeq:
-; VI: v_cmp_eq_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_eq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_oeq(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_oeq:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_eq_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oeq:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_eq_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_oeq:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_eq_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 1)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_one:
-; VI: v_cmp_neq_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_neq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_one(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_one:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_neq_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_one:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_neq_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_one:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_neq_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 6)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_ogt:
-; VI: v_cmp_gt_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_lt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_ogt(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_ogt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_lt_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ogt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_gt_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_ogt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_gt_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 2)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_oge:
-; VI: v_cmp_ge_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_le_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_oge(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_oge:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_le_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oge:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_ge_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_oge:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_ge_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 3)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_olt:
-; VI: v_cmp_lt_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_gt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_olt(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_olt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_gt_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_olt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_lt_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_olt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_lt_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 4)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_ole:
-; VI: v_cmp_le_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_ge_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_ole(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_ole:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_ge_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ole:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_le_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_ole:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_le_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 5)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_ueq:
-; VI: v_cmp_nlg_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nlg_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_ueq(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_ueq:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nlg_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ueq:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nlg_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_ueq:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nlg_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 9)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_une:
-; VI: v_cmp_neq_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_neq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_une(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_une:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_neq_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_une:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_neq_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_une:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_neq_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 14)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_ugt:
-; VI: v_cmp_nle_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nge_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_ugt(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_ugt:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nge_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ugt:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nle_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_ugt:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nle_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 10)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_uge:
-; VI: v_cmp_nlt_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_ngt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_uge(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_uge:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_ngt_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_uge:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nlt_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_uge:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nlt_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 11)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_ult:
-; VI: v_cmp_nge_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nle_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_ult(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_ult:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nle_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ult:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_nge_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_ult:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_nge_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 12)
store i64 %result, i64 addrspace(1)* %out
ret void
}
-; GCN-LABEL: {{^}}v_fcmp_f16_ule:
-; VI: v_cmp_ngt_f16_e64
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nlt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
define amdgpu_kernel void @v_fcmp_f16_ule(i64 addrspace(1)* %out, half %src) {
+; GFX11-LABEL: v_fcmp_f16_ule:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_nlt_f16_e64 s[2:3], 0x5640, s2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ule:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v0, 0x5640
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_cmp_ngt_f16_e64 s[2:3], s2, v0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_fcmp_f16_ule:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
+; GFX9-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX9-NEXT: v_mov_b32_e32 v0, 0x5640
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_cmp_ngt_f16_e64 s[0:1], s4, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX9-NEXT: s_endpgm
%result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 13)
store i64 %result, i64 addrspace(1)* %out
ret void
More information about the llvm-commits
mailing list