Skip to content

[AMDGPU][True16][CodeGen] enable true16 for more codegen test patch 2 #131210

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
Merged
merged 1 commit into from
Mar 14, 2025
Merged

[AMDGPU][True16][CodeGen] enable true16 for more codegen test patch 2 #131210

merged 1 commit into from
Mar 14, 2025

Conversation

broxigarchen
Copy link
Contributor

This is a NFC patch.

Enable true16 mode for more CodeGen tests

@broxigarchen broxigarchen force-pushed the main-true16-codegen-more-test-2 branch from 006acc6 to 606d614 Compare March 13, 2025 20:38
@broxigarchen broxigarchen marked this pull request as ready for review March 13, 2025 20:47
@llvmbot
Copy link
Member

llvmbot commented Mar 13, 2025

@llvm/pr-subscribers-backend-amdgpu

Author: Brox Chen (broxigarchen)

Changes

This is a NFC patch.

Enable true16 mode for more CodeGen tests

Patch is 972.16 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/131210.diff

23 Files Affected:

  • (modified) llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll (+129-61)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll (+315-155)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.bf16.ll (+875-414)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll (+1405-356)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.maximum.f16.ll (+753-383)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.minimum.f16.ll (+753-383)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.minnum.f16.ll (+141-66)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll (+82-39)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll (+129-61)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.sqrt.f16.ll (+84-40)
  • (modified) llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll (+82-39)
  • (modified) llvm/test/CodeGen/AMDGPU/load-constant-i16.ll (+261-123)
  • (modified) llvm/test/CodeGen/AMDGPU/select-fabs-fneg-extract.f16.ll (+1567-396)
  • (modified) llvm/test/CodeGen/AMDGPU/select-fabs-fneg-extract.v2f16.ll (+2437-651)
  • (modified) llvm/test/CodeGen/AMDGPU/select.f16.ll (+1315-668)
  • (modified) llvm/test/CodeGen/AMDGPU/strict_fadd.f16.ll (+116-40)
  • (modified) llvm/test/CodeGen/AMDGPU/strict_fma.f16.ll (+94-43)
  • (modified) llvm/test/CodeGen/AMDGPU/strict_fmul.f16.ll (+156-42)
  • (modified) llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll (+334-29)
  • (modified) llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll (+233-123)
  • (modified) llvm/test/CodeGen/AMDGPU/sub.v2i16.ll (+49-23)
  • (modified) llvm/test/CodeGen/AMDGPU/v_madak_f16.ll (+126-61)
  • (modified) llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll (+524-261)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll
index 867025adca944..644c88457714b 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll
@@ -3,8 +3,10 @@
 ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX8 %s
 ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX9 %s
 ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX10 %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX11 %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX12 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GFX11,GFX11-TRUE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GFX11,GFX11-FAKE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1200 -mattr=+real-true16 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GFX12,GFX12-TRUE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1200 -mattr=-real-true16 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GFX12,GFX12-FAKE16 %s
 
 define amdgpu_kernel void @cos_f16(ptr addrspace(1) %r, ptr addrspace(1) %a) {
 ; GFX6-LABEL: cos_f16:
@@ -69,31 +71,57 @@ define amdgpu_kernel void @cos_f16(ptr addrspace(1) %r, ptr addrspace(1) %a) {
 ; GFX10-NEXT:    global_store_short v0, v1, s[0:1]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: cos_f16:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    global_load_u16 v1, v0, s[2:3]
-; GFX11-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_cos_f16_e32 v1, v1
-; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
-; GFX11-NEXT:    s_endpgm
+; GFX11-TRUE16-LABEL: cos_f16:
+; GFX11-TRUE16:       ; %bb.0:
+; GFX11-TRUE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT:    global_load_d16_b16 v0, v1, s[2:3]
+; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT:    v_mul_f16_e32 v0.l, 0.15915494, v0.l
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT:    v_cos_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT:    global_store_b16 v1, v0, s[0:1]
+; GFX11-TRUE16-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: cos_f16:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    global_load_u16 v1, v0, s[2:3]
-; GFX12-NEXT:    s_wait_loadcnt 0x0
-; GFX12-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT:    v_cos_f16_e32 v1, v1
-; GFX12-NEXT:    global_store_b16 v0, v1, s[0:1]
-; GFX12-NEXT:    s_endpgm
+; GFX11-FAKE16-LABEL: cos_f16:
+; GFX11-FAKE16:       ; %bb.0:
+; GFX11-FAKE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT:    global_load_u16 v1, v0, s[2:3]
+; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
+; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT:    v_cos_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT:    global_store_b16 v0, v1, s[0:1]
+; GFX11-FAKE16-NEXT:    s_endpgm
+;
+; GFX12-TRUE16-LABEL: cos_f16:
+; GFX12-TRUE16:       ; %bb.0:
+; GFX12-TRUE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX12-TRUE16-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-TRUE16-NEXT:    s_wait_kmcnt 0x0
+; GFX12-TRUE16-NEXT:    global_load_d16_b16 v0, v1, s[2:3]
+; GFX12-TRUE16-NEXT:    s_wait_loadcnt 0x0
+; GFX12-TRUE16-NEXT:    v_mul_f16_e32 v0.l, 0.15915494, v0.l
+; GFX12-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-TRUE16-NEXT:    v_cos_f16_e32 v0.l, v0.l
+; GFX12-TRUE16-NEXT:    global_store_b16 v1, v0, s[0:1]
+; GFX12-TRUE16-NEXT:    s_endpgm
+;
+; GFX12-FAKE16-LABEL: cos_f16:
+; GFX12-FAKE16:       ; %bb.0:
+; GFX12-FAKE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX12-FAKE16-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-FAKE16-NEXT:    s_wait_kmcnt 0x0
+; GFX12-FAKE16-NEXT:    global_load_u16 v1, v0, s[2:3]
+; GFX12-FAKE16-NEXT:    s_wait_loadcnt 0x0
+; GFX12-FAKE16-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
+; GFX12-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-FAKE16-NEXT:    v_cos_f16_e32 v1, v1
+; GFX12-FAKE16-NEXT:    global_store_b16 v0, v1, s[0:1]
+; GFX12-FAKE16-NEXT:    s_endpgm
   %a.val = load half, ptr addrspace(1) %a
   %r.val = call half @llvm.cos.f16(half %a.val)
   store half %r.val, ptr addrspace(1) %r
@@ -184,42 +212,79 @@ define amdgpu_kernel void @cos_v2f16(ptr addrspace(1) %r, ptr addrspace(1) %a) {
 ; GFX10-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: cos_v2f16:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
-; GFX11-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-NEXT:    v_lshrrev_b32_e32 v2, 16, v1
-; GFX11-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    v_mul_f16_e32 v2, 0.15915494, v2
-; GFX11-NEXT:    v_cos_f16_e32 v1, v1
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_cos_f16_e32 v2, v2
-; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_pack_b32_f16 v1, v1, v2
-; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
-; GFX11-NEXT:    s_endpgm
+; GFX11-TRUE16-LABEL: cos_v2f16:
+; GFX11-TRUE16:       ; %bb.0:
+; GFX11-TRUE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT:    global_load_b32 v0, v1, s[2:3]
+; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT:    v_mul_f16_e32 v0.l, 0.15915494, v0.l
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT:    v_mul_f16_e32 v0.h, 0.15915494, v2.l
+; GFX11-TRUE16-NEXT:    v_cos_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT:    v_cos_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT:    s_waitcnt_depctr 0xfff
+; GFX11-TRUE16-NEXT:    v_pack_b32_f16 v0, v0.l, v0.h
+; GFX11-TRUE16-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-TRUE16-NEXT:    s_endpgm
+;
+; GFX11-FAKE16-LABEL: cos_v2f16:
+; GFX11-FAKE16:       ; %bb.0:
+; GFX11-FAKE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT:    v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-FAKE16-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
+; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT:    v_mul_f16_e32 v2, 0.15915494, v2
+; GFX11-FAKE16-NEXT:    v_cos_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT:    v_cos_f16_e32 v2, v2
+; GFX11-FAKE16-NEXT:    s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT:    v_pack_b32_f16 v1, v1, v2
+; GFX11-FAKE16-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-FAKE16-NEXT:    s_endpgm
+;
+; GFX12-TRUE16-LABEL: cos_v2f16:
+; GFX12-TRUE16:       ; %bb.0:
+; GFX12-TRUE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX12-TRUE16-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-TRUE16-NEXT:    s_wait_kmcnt 0x0
+; GFX12-TRUE16-NEXT:    global_load_b32 v2, v1, s[2:3]
+; GFX12-TRUE16-NEXT:    s_wait_loadcnt 0x0
+; GFX12-TRUE16-NEXT:    v_mul_f16_e32 v0.l, 0.15915494, v2.l
+; GFX12-TRUE16-NEXT:    v_mul_f16_e32 v0.h, 0.15915494, v2.l
+; GFX12-TRUE16-NEXT:    ; kill: def $vgpr2 killed $vgpr2_lo16 killed $exec
+; GFX12-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT:    v_cos_f16_e32 v0.l, v0.l
+; GFX12-TRUE16-NEXT:    v_cos_f16_e32 v0.h, v0.h
+; GFX12-TRUE16-NEXT:    s_delay_alu instid0(TRANS32_DEP_1)
+; GFX12-TRUE16-NEXT:    v_pack_b32_f16 v0, v0.l, v0.h
+; GFX12-TRUE16-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-TRUE16-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: cos_v2f16:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    global_load_b32 v1, v0, s[2:3]
-; GFX12-NEXT:    s_wait_loadcnt 0x0
-; GFX12-NEXT:    v_lshrrev_b32_e32 v2, 16, v1
-; GFX12-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-NEXT:    v_mul_f16_e32 v2, 0.15915494, v2
-; GFX12-NEXT:    v_cos_f16_e32 v1, v1
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_1)
-; GFX12-NEXT:    v_cos_f16_e32 v2, v2
-; GFX12-NEXT:    v_pack_b32_f16 v1, v1, v2
-; GFX12-NEXT:    global_store_b32 v0, v1, s[0:1]
-; GFX12-NEXT:    s_endpgm
+; GFX12-FAKE16-LABEL: cos_v2f16:
+; GFX12-FAKE16:       ; %bb.0:
+; GFX12-FAKE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX12-FAKE16-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-FAKE16-NEXT:    s_wait_kmcnt 0x0
+; GFX12-FAKE16-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX12-FAKE16-NEXT:    s_wait_loadcnt 0x0
+; GFX12-FAKE16-NEXT:    v_lshrrev_b32_e32 v2, 16, v1
+; GFX12-FAKE16-NEXT:    v_mul_f16_e32 v1, 0.15915494, v1
+; GFX12-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-FAKE16-NEXT:    v_mul_f16_e32 v2, 0.15915494, v2
+; GFX12-FAKE16-NEXT:    v_cos_f16_e32 v1, v1
+; GFX12-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_1)
+; GFX12-FAKE16-NEXT:    v_cos_f16_e32 v2, v2
+; GFX12-FAKE16-NEXT:    v_pack_b32_f16 v1, v1, v2
+; GFX12-FAKE16-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-FAKE16-NEXT:    s_endpgm
   %a.val = load <2 x half>, ptr addrspace(1) %a
   %r.val = call <2 x half> @llvm.cos.v2f16(<2 x half> %a.val)
   store <2 x half> %r.val, ptr addrspace(1) %r
@@ -228,3 +293,6 @@ define amdgpu_kernel void @cos_v2f16(ptr addrspace(1) %r, ptr addrspace(1) %a) {
 
 declare half @llvm.cos.f16(half %a)
 declare <2 x half> @llvm.cos.v2f16(<2 x half> %a)
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX11: {{.*}}
+; GFX12: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll
index 814f44477f528..61991c8b409dd 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll
@@ -5,8 +5,10 @@
 ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=fiji -denormal-fp-math=ieee -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=VI-DENORM %s
 ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1010 -denormal-fp-math=preserve-sign -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX10-FLUSH %s
 ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1010 -denormal-fp-math=ieee -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX10-DENORM %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -denormal-fp-math=preserve-sign -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-FLUSH %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -denormal-fp-math=ieee -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-DENORM %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -denormal-fp-math=preserve-sign -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-FLUSH,GFX11-FLUSH-TRUE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -denormal-fp-math=preserve-sign -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-FLUSH,GFX11-FLUSH-FAKE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -denormal-fp-math=ieee -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-DENORM,GFX11-DENORM-TRUE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -denormal-fp-math=ieee -denormal-fp-math-f32=preserve-sign -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-DENORM,GFX11-DENORM-FAKE16 %s
 
 declare half @llvm.fmuladd.f16(half %a, half %b, half %c)
 declare <2 x half> @llvm.fmuladd.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c)
@@ -160,64 +162,122 @@ define amdgpu_kernel void @fmuladd_f16(
 ; GFX10-DENORM-NEXT:    buffer_store_short v2, off, s[0:3], 0
 ; GFX10-DENORM-NEXT:    s_endpgm
 ;
-; GFX11-FLUSH-LABEL: fmuladd_f16:
-; GFX11-FLUSH:       ; %bb.0:
-; GFX11-FLUSH-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
-; GFX11-FLUSH-NEXT:    s_mov_b32 s10, -1
-; GFX11-FLUSH-NEXT:    s_mov_b32 s11, 0x31016000
-; GFX11-FLUSH-NEXT:    s_mov_b32 s14, s10
-; GFX11-FLUSH-NEXT:    s_mov_b32 s15, s11
-; GFX11-FLUSH-NEXT:    s_mov_b32 s18, s10
-; GFX11-FLUSH-NEXT:    s_mov_b32 s19, s11
-; GFX11-FLUSH-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-FLUSH-NEXT:    s_mov_b32 s12, s2
-; GFX11-FLUSH-NEXT:    s_mov_b32 s13, s3
-; GFX11-FLUSH-NEXT:    s_mov_b32 s16, s4
-; GFX11-FLUSH-NEXT:    s_mov_b32 s17, s5
-; GFX11-FLUSH-NEXT:    buffer_load_u16 v0, off, s[12:15], 0
-; GFX11-FLUSH-NEXT:    buffer_load_u16 v1, off, s[16:19], 0
-; GFX11-FLUSH-NEXT:    s_mov_b32 s4, s6
-; GFX11-FLUSH-NEXT:    s_mov_b32 s5, s7
-; GFX11-FLUSH-NEXT:    s_mov_b32 s6, s10
-; GFX11-FLUSH-NEXT:    s_mov_b32 s7, s11
-; GFX11-FLUSH-NEXT:    s_mov_b32 s8, s0
-; GFX11-FLUSH-NEXT:    buffer_load_u16 v2, off, s[4:7], 0
-; GFX11-FLUSH-NEXT:    s_mov_b32 s9, s1
-; GFX11-FLUSH-NEXT:    s_waitcnt vmcnt(1)
-; GFX11-FLUSH-NEXT:    v_mul_f16_e32 v0, v0, v1
-; GFX11-FLUSH-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-FLUSH-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FLUSH-NEXT:    v_add_f16_e32 v0, v0, v2
-; GFX11-FLUSH-NEXT:    buffer_store_b16 v0, off, s[8:11], 0
-; GFX11-FLUSH-NEXT:    s_endpgm
+; GFX11-FLUSH-TRUE16-LABEL: fmuladd_f16:
+; GFX11-FLUSH-TRUE16:       ; %bb.0:
+; GFX11-FLUSH-TRUE16-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s10, -1
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s11, 0x31016000
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s14, s10
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s15, s11
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s18, s10
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s19, s11
+; GFX11-FLUSH-TRUE16-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s12, s2
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s13, s3
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s16, s4
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s17, s5
+; GFX11-FLUSH-TRUE16-NEXT:    buffer_load_u16 v0, off, s[12:15], 0
+; GFX11-FLUSH-TRUE16-NEXT:    buffer_load_u16 v1, off, s[16:19], 0
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s12, s6
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s13, s7
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s8, s0
+; GFX11-FLUSH-TRUE16-NEXT:    buffer_load_u16 v2, off, s[12:15], 0
+; GFX11-FLUSH-TRUE16-NEXT:    s_mov_b32 s9, s1
+; GFX11-FLUSH-TRUE16-NEXT:    s_waitcnt vmcnt(1)
+; GFX11-FLUSH-TRUE16-NEXT:    v_mul_f16_e32 v0.l, v0.l, v1.l
+; GFX11-FLUSH-TRUE16-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-FLUSH-TRUE16-NEXT:    v_mov_b16_e32 v1.l, v2.l
+; GFX11-FLUSH-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FLUSH-TRUE16-NEXT:    v_add_f16_e32 v0.l, v0.l, v1.l
+; GFX11-FLUSH-TRUE16-NEXT:    buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-FLUSH-TRUE16-NEXT:    s_endpgm
 ;
-; GFX11-DENORM-LABEL: fmuladd_f16:
-; GFX11-DENORM:       ; %bb.0:
-; GFX11-DENORM-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
-; GFX11-DENORM-NEXT:    s_mov_b32 s10, -1
-; GFX11-DENORM-NEXT:    s_mov_b32 s11, 0x31016000
-; GFX11-DENORM-NEXT:    s_mov_b32 s14, s10
-; GFX11-DENORM-NEXT:    s_mov_b32 s15, s11
-; GFX11-DENORM-NEXT:    s_mov_b32 s18, s10
-; GFX11-DENORM-NEXT:    s_mov_b32 s19, s11
-; GFX11-DENORM-NEXT:    s_mov_b32 s22, s10
-; GFX11-DENORM-NEXT:    s_mov_b32 s23, s11
-; GFX11-DENORM-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-DENORM-NEXT:    s_mov_b32 s12, s2
-; GFX11-DENORM-NEXT:    s_mov_b32 s13, s3
-; GFX11-DENORM-NEXT:    s_mov_b32 s16, s4
-; GFX11-DENORM-NEXT:    s_mov_b32 s17, s5
-; GFX11-DENORM-NEXT:    s_mov_b32 s20, s6
-; GFX11-DENORM-NEXT:    s_mov_b32 s21, s7
-; GFX11-DENORM-NEXT:    buffer_load_u16 v0, off, s[12:15], 0
-; GFX11-DENORM-NEXT:    buffer_load_u16 v1, off, s[16:19], 0
-; GFX11-DENORM-NEXT:    buffer_load_u16 v2, off, s[20:23], 0
-; GFX11-DENORM-NEXT:    s_mov_b32 s8, s0
-; GFX11-DENORM-NEXT:    s_mov_b32 s9, s1
-; GFX11-DENORM-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-DENORM-NEXT:    v_fmac_f16_e32 v2, v0, v1
-; GFX11-DENORM-NEXT:    buffer_store_b16 v2, off, s[8:11], 0
-; GFX11-DENORM-NEXT:    s_endpgm
+; GFX11-FLUSH-FAKE16-LABEL: fmuladd_f16:
+; GFX11-FLUSH-FAKE16:       ; %bb.0:
+; GFX11-FLUSH-FAKE16-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s10, -1
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s11, 0x31016000
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s14, s10
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s15, s11
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s18, s10
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s19, s11
+; GFX11-FLUSH-FAKE16-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s12, s2
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s13, s3
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s16, s4
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s17, s5
+; GFX11-FLUSH-FAKE16-NEXT:    buffer_load_u16 v0, off, s[12:15], 0
+; GFX11-FLUSH-FAKE16-NEXT:    buffer_load_u16 v1, off, s[16:19], 0
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s4, s6
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s5, s7
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s6, s10
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s7, s11
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s8, s0
+; GFX11-FLUSH-FAKE16-NEXT:    buffer_load_u16 v2, off, s[4:7], 0
+; GFX11-FLUSH-FAKE16-NEXT:    s_mov_b32 s9, s1
+; GFX11-FLUSH-FAKE16-NEXT:    s_waitcnt vmcnt(1)
+; GFX11-FLUSH-FAKE16-NEXT:    v_mul_f16_e32 v0, v0, v1
+; GFX11-FLUSH-FAKE16-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-FLUSH-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FLUSH-FAKE16-NEXT:    v_add_f16_e32 v0, v0, v2
+; GFX11-FLUSH-FAKE16-NEXT:    buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-FLUSH-FAKE16-NEXT:    s_endpgm
+;
+; GFX11-DENORM-TRUE16-LABEL: fmuladd_f16:
+; GFX11-DENORM-TRUE16:       ; %bb.0:
+; GFX11-DENORM-TRUE16-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
+; GFX11-DENORM-TRUE16-NEXT:    s_mov_b32 s10, -1
+; GFX11-DENORM-TRUE16-NEXT:    s_mov_b32 s11, 0x31016000
+; GFX11-DENORM-TRUE16-NEXT:    s_mov_b32 s14, s10
+; GFX11-DENORM-TRUE16-NEXT:    s_mov_b32 s15, s11
+; GFX11-DENORM-TRUE16-NEXT:    s_mov_b32 s18, s10
+; GFX11-DENORM-TRUE16-NEXT:    s...
[truncated]

@broxigarchen broxigarchen requested review from shiltian, kosarev, arsenm and Sisyph and removed request for kosarev March 14, 2025 02:39
@broxigarchen broxigarchen merged commit b1fe7da into llvm:main Mar 14, 2025
12 of 14 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@arsenm arsenm arsenm approved these changes

@shiltian shiltian Awaiting requested review from shiltian

@kosarev kosarev Awaiting requested review from kosarev

@Sisyph Sisyph Awaiting requested review from Sisyph

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
@broxigarchen @llvmbot @arsenm