From 5b652f77e02f4aaa4872daf134b56f5d7388f402 Mon Sep 17 00:00:00 2001
From: Jay Foad <jay.foad@amd.com>
Date: Thu, 8 Sep 2022 14:13:17 +0100
Subject: [PATCH] [AMDGPU] Add basic tests for emitting v_fma_f16 and friends

---
 llvm/test/CodeGen/AMDGPU/fma.f16.ll | 102 ++++++++++++++++++++++++++++++++++++
 1 file changed, 102 insertions(+)
 create mode 100644 llvm/test/CodeGen/AMDGPU/fma.f16.ll

diff --git a/llvm/test/CodeGen/AMDGPU/fma.f16.ll b/llvm/test/CodeGen/AMDGPU/fma.f16.ll
new file mode 100644
index 0000000..20d39ef
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/fma.f16.ll
@@ -0,0 +1,102 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx900 < %s | FileCheck %s -check-prefixes=GFX9,GFX9-SDAG
+; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx900 < %s | FileCheck %s -check-prefixes=GFX9,GFX9-GISEL
+; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx1010 < %s | FileCheck %s -check-prefixes=GFX10,GFX10-SDAG
+; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx1010 < %s | FileCheck %s -check-prefixes=GFX10,GFX10-GISEL
+
+declare half @llvm.fma.f16(half, half, half)
+declare half @llvm.maxnum.f16(half, half)
+
+define half @test_fma(half %x, half %y, half %z) {
+; GFX9-LABEL: test_fma:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_fma_f16 v0, v0, v1, v2
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_fma:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_fma_f16 v0, v0, v1, v2
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %r = call half @llvm.fma.f16(half %x, half %y, half %z)
+  ret half %r
+}
+
+; GFX10+ has v_fmac_f16.
+define half @test_fmac(half %x, half %y, half %z) {
+; GFX9-LABEL: test_fmac:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_fma_f16 v0, v1, v2, v0
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_fmac:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_fmac_f16_e32 v0, v1, v2
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %r = call half @llvm.fma.f16(half %y, half %z, half %x)
+  ret half %r
+}
+
+; GFX10+ has v_fmaak_f16.
+define half @test_fmaak(half %x, half %y, half %z) {
+; GFX9-SDAG-LABEL: test_fmaak:
+; GFX9-SDAG:       ; %bb.0:
+; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-SDAG-NEXT:    s_movk_i32 s4, 0x4200
+; GFX9-SDAG-NEXT:    v_fma_f16 v0, v0, v1, s4
+; GFX9-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-GISEL-LABEL: test_fmaak:
+; GFX9-GISEL:       ; %bb.0:
+; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, 0x4200
+; GFX9-GISEL-NEXT:    v_fma_f16 v0, v0, v1, v2
+; GFX9-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-SDAG-LABEL: test_fmaak:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-SDAG-NEXT:    v_fmaak_f16 v0, v0, v1, 0x4200
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-GISEL-LABEL: test_fmaak:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-GISEL-NEXT:    v_fma_f16 v0, v0, v1, 0x4200
+; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %r = call half @llvm.fma.f16(half %x, half %y, half 0xH4200)
+  ret half %r
+}
+
+; GFX10+ has v_fmamk_f16.
+define half @test_fmamk(half %x, half %y, half %z) {
+; GFX9-SDAG-LABEL: test_fmamk:
+; GFX9-SDAG:       ; %bb.0:
+; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-SDAG-NEXT:    s_movk_i32 s4, 0x4200
+; GFX9-SDAG-NEXT:    v_fma_f16 v0, v0, s4, v2
+; GFX9-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-GISEL-LABEL: test_fmamk:
+; GFX9-GISEL:       ; %bb.0:
+; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, 0x4200
+; GFX9-GISEL-NEXT:    v_fma_f16 v0, v0, v1, v2
+; GFX9-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_fmamk:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_fmamk_f16 v0, v0, 0x4200, v2
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %r = call half @llvm.fma.f16(half %x, half 0xH4200, half %z)
+  ret half %r
+}
-- 
2.7.4