[llvm] bb62af7 - [AMDGPU][True16][CodeGen] true16 codegen for valu op (#124797)
via llvm-commits
llvm-commits at lists.llvm.org
Wed Feb 26 07:50:52 PST 2025
Author: Brox Chen
Date: 2025-02-26T10:50:49-05:00
New Revision: bb62af7d14f7fe1301311234352f9652d45ba354
URL: https://github.com/llvm/llvm-project/commit/bb62af7d14f7fe1301311234352f9652d45ba354
DIFF: https://github.com/llvm/llvm-project/commit/bb62af7d14f7fe1301311234352f9652d45ba354.diff
LOG: [AMDGPU][True16][CodeGen] true16 codegen for valu op (#124797)
true16 selection for valu ops, enable `real-true16` attribute and update
the codegen test
Added:
Modified:
llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
llvm/test/CodeGen/AMDGPU/fmul.f16.ll
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rcp.f16.ll
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.f16.ll
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.f16.ll
llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll
llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index 2ef628a7c569c..d57074272487e 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -815,7 +815,7 @@ void SIInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
if (Fix16BitCopies) {
if (((Size == 16) != (SrcSize == 16))) {
// Non-VGPR Src and Dst will later be expanded back to 32 bits.
- assert(ST.hasTrue16BitInsts());
+ assert(ST.useRealTrue16Insts());
Register &RegToFix = (Size == 32) ? DestReg : SrcReg;
MCRegister SubReg = RI.getSubReg(RegToFix, AMDGPU::lo16);
RegToFix = SubReg;
@@ -989,7 +989,7 @@ void SIInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
return;
}
- if (ST.hasTrue16BitInsts()) {
+ if (ST.useRealTrue16Insts()) {
if (IsSGPRSrc) {
assert(SrcLow);
SrcReg = NewSrcReg;
@@ -5581,9 +5581,11 @@ unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) const {
return ST.useRealTrue16Insts() ? AMDGPU::V_FLOOR_F16_t16_e64
: AMDGPU::V_FLOOR_F16_fake16_e64;
case AMDGPU::S_TRUNC_F16:
- return AMDGPU::V_TRUNC_F16_fake16_e64;
+ return ST.useRealTrue16Insts() ? AMDGPU::V_TRUNC_F16_t16_e64
+ : AMDGPU::V_TRUNC_F16_fake16_e64;
case AMDGPU::S_RNDNE_F16:
- return AMDGPU::V_RNDNE_F16_fake16_e64;
+ return ST.useRealTrue16Insts() ? AMDGPU::V_RNDNE_F16_t16_e64
+ : AMDGPU::V_RNDNE_F16_fake16_e64;
case AMDGPU::S_ADD_F32: return AMDGPU::V_ADD_F32_e64;
case AMDGPU::S_SUB_F32: return AMDGPU::V_SUB_F32_e64;
case AMDGPU::S_MIN_F32: return AMDGPU::V_MIN_F32_e64;
@@ -5591,17 +5593,27 @@ unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) const {
case AMDGPU::S_MINIMUM_F32: return AMDGPU::V_MINIMUM_F32_e64;
case AMDGPU::S_MAXIMUM_F32: return AMDGPU::V_MAXIMUM_F32_e64;
case AMDGPU::S_MUL_F32: return AMDGPU::V_MUL_F32_e64;
- case AMDGPU::S_ADD_F16: return AMDGPU::V_ADD_F16_fake16_e64;
- case AMDGPU::S_SUB_F16: return AMDGPU::V_SUB_F16_fake16_e64;
- case AMDGPU::S_MIN_F16: return AMDGPU::V_MIN_F16_fake16_e64;
- case AMDGPU::S_MAX_F16: return AMDGPU::V_MAX_F16_fake16_e64;
+ case AMDGPU::S_ADD_F16:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_ADD_F16_t16_e64
+ : AMDGPU::V_ADD_F16_fake16_e64;
+ case AMDGPU::S_SUB_F16:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_SUB_F16_t16_e64
+ : AMDGPU::V_SUB_F16_fake16_e64;
+ case AMDGPU::S_MIN_F16:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_MIN_F16_t16_e64
+ : AMDGPU::V_MIN_F16_fake16_e64;
+ case AMDGPU::S_MAX_F16:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_MAX_F16_t16_e64
+ : AMDGPU::V_MAX_F16_fake16_e64;
case AMDGPU::S_MINIMUM_F16:
return ST.useRealTrue16Insts() ? AMDGPU::V_MINIMUM_F16_t16_e64
: AMDGPU::V_MINIMUM_F16_fake16_e64;
case AMDGPU::S_MAXIMUM_F16:
return ST.useRealTrue16Insts() ? AMDGPU::V_MAXIMUM_F16_t16_e64
: AMDGPU::V_MAXIMUM_F16_fake16_e64;
- case AMDGPU::S_MUL_F16: return AMDGPU::V_MUL_F16_fake16_e64;
+ case AMDGPU::S_MUL_F16:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_MUL_F16_t16_e64
+ : AMDGPU::V_MUL_F16_fake16_e64;
case AMDGPU::S_CVT_PK_RTZ_F16_F32: return AMDGPU::V_CVT_PKRTZ_F16_F32_e64;
case AMDGPU::S_FMAC_F32: return AMDGPU::V_FMAC_F32_e64;
case AMDGPU::S_FMAC_F16:
@@ -5666,15 +5678,25 @@ unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) const {
return ST.useRealTrue16Insts() ? AMDGPU::V_CMP_NLT_F16_t16_e64
: AMDGPU::V_CMP_NLT_F16_fake16_e64;
case AMDGPU::V_S_EXP_F32_e64: return AMDGPU::V_EXP_F32_e64;
- case AMDGPU::V_S_EXP_F16_e64: return AMDGPU::V_EXP_F16_fake16_e64;
+ case AMDGPU::V_S_EXP_F16_e64:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_EXP_F16_t16_e64
+ : AMDGPU::V_EXP_F16_fake16_e64;
case AMDGPU::V_S_LOG_F32_e64: return AMDGPU::V_LOG_F32_e64;
- case AMDGPU::V_S_LOG_F16_e64: return AMDGPU::V_LOG_F16_fake16_e64;
+ case AMDGPU::V_S_LOG_F16_e64:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_LOG_F16_t16_e64
+ : AMDGPU::V_LOG_F16_fake16_e64;
case AMDGPU::V_S_RCP_F32_e64: return AMDGPU::V_RCP_F32_e64;
- case AMDGPU::V_S_RCP_F16_e64: return AMDGPU::V_RCP_F16_fake16_e64;
+ case AMDGPU::V_S_RCP_F16_e64:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_RCP_F16_t16_e64
+ : AMDGPU::V_RCP_F16_fake16_e64;
case AMDGPU::V_S_RSQ_F32_e64: return AMDGPU::V_RSQ_F32_e64;
- case AMDGPU::V_S_RSQ_F16_e64: return AMDGPU::V_RSQ_F16_fake16_e64;
+ case AMDGPU::V_S_RSQ_F16_e64:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_RSQ_F16_t16_e64
+ : AMDGPU::V_RSQ_F16_fake16_e64;
case AMDGPU::V_S_SQRT_F32_e64: return AMDGPU::V_SQRT_F32_e64;
- case AMDGPU::V_S_SQRT_F16_e64: return AMDGPU::V_SQRT_F16_fake16_e64;
+ case AMDGPU::V_S_SQRT_F16_e64:
+ return ST.useRealTrue16Insts() ? AMDGPU::V_SQRT_F16_t16_e64
+ : AMDGPU::V_SQRT_F16_fake16_e64;
}
llvm_unreachable(
"Unexpected scalar opcode without corresponding vector one!");
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
index d81faf91801b0..235ec22ba5c60 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
@@ -3,7 +3,8 @@
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=fiji < %s | FileCheck -check-prefix=GFX8 %s
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefix=GFX10 %s
-; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefix=GFX11 %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
define float @v_pow_f32(float %x, float %y) {
; GFX6-LABEL: v_pow_f32:
@@ -371,19 +372,33 @@ define half @v_pow_f16(half %x, half %y) {
; GFX10-NEXT: v_exp_f16_e32 v0, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-LABEL: v_pow_f16:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_log_f16_e32 v0, v0
-; GFX11-NEXT: v_cvt_f32_f16_e32 v1, v1
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_mul_dx9_zero_f32_e32 v0, v0, v1
-; GFX11-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_exp_f16_e32 v0, v0
-; GFX11-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-LABEL: v_pow_f16:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mul_dx9_zero_f32_e32 v0, v0, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-FAKE16-LABEL: v_pow_f16:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_mul_dx9_zero_f32_e32 v0, v0, v1
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%pow = call half @llvm.pow.f16(half %x, half %y)
ret half %pow
}
@@ -474,31 +489,54 @@ define <2 x half> @v_pow_v2f16(<2 x half> %x, <2 x half> %y) {
; GFX10-NEXT: v_lshl_or_b32 v0, v0, 16, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-LABEL: v_pow_v2f16:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_log_f16_e32 v2, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX11-NEXT: v_cvt_f32_f16_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_log_f16_e32 v0, v0
-; GFX11-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX11-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
-; GFX11-NEXT: v_cvt_f16_f32_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT: v_exp_f16_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_exp_f16_e32 v0, v0
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-NEXT: v_lshl_or_b32 v0, v0, 16, v1
-; GFX11-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-LABEL: v_pow_v2f16:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-FAKE16-LABEL: v_pow_v2f16:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v2, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v0, v0, 16, v1
+; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x, <2 x half> %y)
ret <2 x half> %pow
}
@@ -597,33 +635,57 @@ define <2 x half> @v_pow_v2f16_fneg_lhs(<2 x half> %x, <2 x half> %y) {
; GFX10-NEXT: v_lshl_or_b32 v0, v0, 16, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-LABEL: v_pow_v2f16_fneg_lhs:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX11-NEXT: v_cvt_f32_f16_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX11-NEXT: v_log_f16_e32 v2, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX11-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_log_f16_e32 v0, v0
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX11-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX11-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cvt_f16_f32_e32 v1, v1
-; GFX11-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_exp_f16_e32 v1, v1
-; GFX11-NEXT: v_exp_f16_e32 v0, v0
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_lshl_or_b32 v0, v0, 16, v1
-; GFX11-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-LABEL: v_pow_v2f16_fneg_lhs:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-FAKE16-LABEL: v_pow_v2f16_fneg_lhs:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-FAKE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v2, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v0, v0, 16, v1
+; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%x.fneg = fneg <2 x half> %x
%pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x.fneg, <2 x half> %y)
ret <2 x half> %pow
@@ -723,32 +785,56 @@ define <2 x half> @v_pow_v2f16_fneg_rhs(<2 x half> %x, <2 x half> %y) {
; GFX10-NEXT: v_lshl_or_b32 v0, v0, 16, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-LABEL: v_pow_v2f16_fneg_rhs:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_log_f16_e32 v2, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX11-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_log_f16_e32 v0, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX11-NEXT: v_cvt_f32_f16_e32 v1, v1
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX11-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX11-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
-; GFX11-NEXT: v_cvt_f16_f32_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT: v_exp_f16_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_exp_f16_e32 v0, v0
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-NEXT: v_lshl_or_b32 v0, v0, 16, v1
-; GFX11-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-LABEL: v_pow_v2f16_fneg_rhs:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-FAKE16-LABEL: v_pow_v2f16_fneg_rhs:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v2, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v0, v0, 16, v1
+; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%y.fneg = fneg <2 x half> %y
%pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x, <2 x half> %y.fneg)
ret <2 x half> %pow
@@ -856,34 +942,60 @@ define <2 x half> @v_pow_v2f16_fneg_lhs_rhs(<2 x half> %x, <2 x half> %y) {
; GFX10-NEXT: v_lshl_or_b32 v0, v0, 16, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-LABEL: v_pow_v2f16_fneg_lhs_rhs:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX11-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_log_f16_e32 v2, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX11-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX11-NEXT: v_cvt_f32_f16_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_log_f16_e32 v0, v0
-; GFX11-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX11-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
-; GFX11-NEXT: v_cvt_f16_f32_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT: v_exp_f16_e32 v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_exp_f16_e32 v0, v0
-; GFX11-NEXT: s_waitcnt_depctr 0xfff
-; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-NEXT: v_lshl_or_b32 v0, v0, 16, v1
-; GFX11-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-LABEL: v_pow_v2f16_fneg_lhs_rhs:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
+; GFX11-TRUE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-FAKE16-LABEL: v_pow_v2f16_fneg_lhs_rhs:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-FAKE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
+; GFX11-FAKE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v2, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_log_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v0, v0
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_exp_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_waitcnt_depctr 0xfff
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v0, v0, 16, v1
+; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%x.fneg = fneg <2 x half> %x
%y.fneg = fneg <2 x half> %y
%pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x.fneg, <2 x half> %y.fneg)
diff --git a/llvm/test/CodeGen/AMDGPU/fmul.f16.ll b/llvm/test/CodeGen/AMDGPU/fmul.f16.ll
index 2cc5159c29f7f..e9877ae5144f5 100644
--- a/llvm/test/CodeGen/AMDGPU/fmul.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmul.f16.ll
@@ -2,7 +2,8 @@
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=SI %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX89,VI %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX89,GFX9 %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX11 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,+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=-flat-for-global,-real-true16 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,GFX11-FAKE16 %s
define amdgpu_kernel void @fmul_f16(
; SI-LABEL: fmul_f16:
@@ -54,29 +55,55 @@ define amdgpu_kernel void @fmul_f16(
; GFX89-NEXT: buffer_store_short v0, off, s[4:7], 0
; GFX89-NEXT: s_endpgm
;
-; GFX11-LABEL: fmul_f16:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_clause 0x1
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34
-; GFX11-NEXT: s_mov_b32 s10, -1
-; GFX11-NEXT: s_mov_b32 s11, 0x31016000
-; GFX11-NEXT: s_mov_b32 s14, s10
-; GFX11-NEXT: s_mov_b32 s15, s11
-; GFX11-NEXT: s_mov_b32 s6, s10
-; GFX11-NEXT: s_mov_b32 s7, s11
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s12, s2
-; GFX11-NEXT: s_mov_b32 s13, s3
-; GFX11-NEXT: buffer_load_u16 v0, off, s[12:15], 0 glc dlc
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: buffer_load_u16 v1, off, s[4:7], 0 glc dlc
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s0
-; GFX11-NEXT: s_mov_b32 s9, s1
-; GFX11-NEXT: v_mul_f16_e32 v0, v0, v1
-; GFX11-NEXT: buffer_store_b16 v0, off, s[8:11], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: fmul_f16:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_clause 0x1
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_load_b64 s[4:5], s[4:5], 0x34
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s14, s10
+; GFX11-TRUE16-NEXT: s_mov_b32 s15, s11
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s10
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, s11
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s12, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s13, s3
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[12:15], 0 glc dlc
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: buffer_load_u16 v1, off, s[4:7], 0 glc dlc
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s0
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s1
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v0.h
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: fmul_f16:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_clause 0x1
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_load_b64 s[4:5], s[4:5], 0x34
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s14, s10
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, s11
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, s10
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, s11
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s12, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s13, s3
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[12:15], 0 glc dlc
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: buffer_load_u16 v1, off, s[4:7], 0 glc dlc
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s0
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s1
+; GFX11-FAKE16-NEXT: v_mul_f16_e32 v0, v0, v1
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_endpgm
ptr addrspace(1) %r,
ptr addrspace(1) %a,
ptr addrspace(1) %b) {
@@ -127,23 +154,41 @@ define amdgpu_kernel void @fmul_f16_imm_a(
; GFX89-NEXT: buffer_store_short v0, off, s[4:7], 0
; GFX89-NEXT: s_endpgm
;
-; GFX11-LABEL: fmul_f16_imm_a:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_u16 v0, off, s[8:11], 0 glc dlc
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: v_mul_f16_e32 v0, 0x4200, v0
-; GFX11-NEXT: buffer_store_b16 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: fmul_f16_imm_a:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0 glc dlc
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x4200, v0.l
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: fmul_f16_imm_a:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0 glc dlc
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: v_mul_f16_e32 v0, 0x4200, v0
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
ptr addrspace(1) %r,
ptr addrspace(1) %b) {
entry:
@@ -192,23 +237,41 @@ define amdgpu_kernel void @fmul_f16_imm_b(
; GFX89-NEXT: buffer_store_short v0, off, s[4:7], 0
; GFX89-NEXT: s_endpgm
;
-; GFX11-LABEL: fmul_f16_imm_b:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_u16 v0, off, s[8:11], 0 glc dlc
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: v_mul_f16_e32 v0, 4.0, v0
-; GFX11-NEXT: buffer_store_b16 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: fmul_f16_imm_b:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0 glc dlc
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: v_mul_f16_e32 v0.l, 4.0, v0.l
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: fmul_f16_imm_b:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0 glc dlc
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: v_mul_f16_e32 v0, 4.0, v0
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
ptr addrspace(1) %r,
ptr addrspace(1) %a) {
entry:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rcp.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rcp.f16.ll
index b78ea1033baac..c6ea12dd61651 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rcp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rcp.f16.ll
@@ -1,10 +1,14 @@
; RUN: llc -mtriple=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,+real-true16 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX11-TRUE16 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,-real-true16 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX11-FAKE16 %s
declare half @llvm.amdgcn.rcp.f16(half %a)
; GCN-LABEL: {{^}}rcp_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; VI: v_rcp_f16_e32 v[[R_F16:[0-9]+]], v[[A_F16]]
+; GFX11-TRUE16: v_rcp_f16_e32 v[[A_F16:[0-9]+]].l, v[[A_F16]].l
+; GFX11-FAKE16: v_rcp_f16_e32 v[[A_F16:[0-9]+]], v[[A_F16]]
; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm
define amdgpu_kernel void @rcp_f16(
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.f16.ll
index 8c003c96dddfe..0924e9a5c2314 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.f16.ll
@@ -1,10 +1,14 @@
; RUN: llc -mtriple=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,+real-true16 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX11-TRUE16 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,-real-true16 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX11-FAKE16 %s
declare half @llvm.amdgcn.rsq.f16(half %a)
; GCN-LABEL: {{^}}rsq_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; VI: v_rsq_f16_e32 v[[R_F16:[0-9]+]], v[[A_F16]]
+; GFX11-TRUE16: v_rsq_f16_e32 v[[A_F16:[0-9]+]].l, v[[A_F16]].l
+; GFX11-FAKE16: v_rsq_f16_e32 v[[A_F16:[0-9]+]], v[[A_F16]]
; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm
define amdgpu_kernel void @rsq_f16(
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.f16.ll
index bbfb88a4b22a3..aa994558162db 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.f16.ll
@@ -1,6 +1,10 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji < %s | FileCheck -check-prefix=GCN %s
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdhsa -mcpu=fiji < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck -check-prefixes=GFX11-SDAG-TRUE16 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck -check-prefixes=GFX11-SDAG-FAKE16 %s
+; RUN: llc -global-isel -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck -check-prefixes=GFX11-GISEL-TRUE16 %s
+; RUN: llc -global-isel -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck -check-prefixes=GFX11-GISEL-FAKE16 %s
define half @v_sqrt_f16(half %src) {
; GCN-LABEL: v_sqrt_f16:
@@ -8,6 +12,30 @@ define half @v_sqrt_f16(half %src) {
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_sqrt_f16_e32 v0, v0
; GCN-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-TRUE16-LABEL: v_sqrt_f16:
+; GFX11-SDAG-TRUE16: ; %bb.0:
+; GFX11-SDAG-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-TRUE16-NEXT: v_sqrt_f16_e32 v0.l, v0.l
+; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-FAKE16-LABEL: v_sqrt_f16:
+; GFX11-SDAG-FAKE16: ; %bb.0:
+; GFX11-SDAG-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-FAKE16-NEXT: v_sqrt_f16_e32 v0, v0
+; GFX11-SDAG-FAKE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-TRUE16-LABEL: v_sqrt_f16:
+; GFX11-GISEL-TRUE16: ; %bb.0:
+; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_sqrt_f16_e32 v0.l, v0.l
+; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-FAKE16-LABEL: v_sqrt_f16:
+; GFX11-GISEL-FAKE16: ; %bb.0:
+; GFX11-GISEL-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-FAKE16-NEXT: v_sqrt_f16_e32 v0, v0
+; GFX11-GISEL-FAKE16-NEXT: s_setpc_b64 s[30:31]
%sqrt = call half @llvm.amdgcn.sqrt.f16(half %src)
ret half %sqrt
}
@@ -18,6 +46,30 @@ define half @v_fabs_sqrt_f16(half %src) {
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_sqrt_f16_e64 v0, |v0|
; GCN-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-TRUE16-LABEL: v_fabs_sqrt_f16:
+; GFX11-SDAG-TRUE16: ; %bb.0:
+; GFX11-SDAG-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-TRUE16-NEXT: v_sqrt_f16_e64 v0.l, |v0.l|
+; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-FAKE16-LABEL: v_fabs_sqrt_f16:
+; GFX11-SDAG-FAKE16: ; %bb.0:
+; GFX11-SDAG-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-FAKE16-NEXT: v_sqrt_f16_e64 v0, |v0|
+; GFX11-SDAG-FAKE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-TRUE16-LABEL: v_fabs_sqrt_f16:
+; GFX11-GISEL-TRUE16: ; %bb.0:
+; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_sqrt_f16_e64 v0.l, |v0.l|
+; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-FAKE16-LABEL: v_fabs_sqrt_f16:
+; GFX11-GISEL-FAKE16: ; %bb.0:
+; GFX11-GISEL-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-FAKE16-NEXT: v_sqrt_f16_e64 v0, |v0|
+; GFX11-GISEL-FAKE16-NEXT: s_setpc_b64 s[30:31]
%fabs.src = call half @llvm.fabs.f16(half %src)
%sqrt = call half @llvm.amdgcn.sqrt.f16(half %fabs.src)
ret half %sqrt
@@ -29,6 +81,30 @@ define half @v_fneg_fabs_sqrt_f16(half %src) {
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_sqrt_f16_e64 v0, -|v0|
; GCN-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-TRUE16-LABEL: v_fneg_fabs_sqrt_f16:
+; GFX11-SDAG-TRUE16: ; %bb.0:
+; GFX11-SDAG-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-TRUE16-NEXT: v_sqrt_f16_e64 v0.l, -|v0.l|
+; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-FAKE16-LABEL: v_fneg_fabs_sqrt_f16:
+; GFX11-SDAG-FAKE16: ; %bb.0:
+; GFX11-SDAG-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-FAKE16-NEXT: v_sqrt_f16_e64 v0, -|v0|
+; GFX11-SDAG-FAKE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-TRUE16-LABEL: v_fneg_fabs_sqrt_f16:
+; GFX11-GISEL-TRUE16: ; %bb.0:
+; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_sqrt_f16_e64 v0.l, -|v0.l|
+; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-FAKE16-LABEL: v_fneg_fabs_sqrt_f16:
+; GFX11-GISEL-FAKE16: ; %bb.0:
+; GFX11-GISEL-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-FAKE16-NEXT: v_sqrt_f16_e64 v0, -|v0|
+; GFX11-GISEL-FAKE16-NEXT: s_setpc_b64 s[30:31]
%fabs.src = call half @llvm.fabs.f16(half %src)
%neg.fabs.src = fneg half %fabs.src
%sqrt = call half @llvm.amdgcn.sqrt.f16(half %neg.fabs.src)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
index 0517e41e3d651..3faf84e5d58c8 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
@@ -3,7 +3,8 @@
; RUN: llc -mtriple=amdgcn-- -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=VI %s
; RUN: llc -mtriple=amdgcn-- -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX9 %s
; RUN: llc -mtriple=amdgcn-- -mcpu=gfx1010 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX10 %s
-; RUN: llc -mtriple=amdgcn-- -mcpu=gfx1100 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefix=GFX11 %s
+; RUN: llc -mtriple=amdgcn-- -mcpu=gfx1100 -mattr=-flat-for-global,+real-true16 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GFX11,GFX11-TRUE16 %s
+; RUN: llc -mtriple=amdgcn-- -mcpu=gfx1100 -mattr=-flat-for-global,-real-true16 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GFX11,GFX11-FAKE16 %s
declare half @llvm.maxnum.f16(half %a, half %b)
declare <2 x half> @llvm.maxnum.v2f16(<2 x half> %a, <2 x half> %b)
@@ -113,32 +114,60 @@ define amdgpu_kernel void @maxnum_f16(
; GFX10-NEXT: buffer_store_short v0, off, s[4:7], 0
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: maxnum_f16:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_clause 0x1
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34
-; GFX11-NEXT: s_mov_b32 s10, -1
-; GFX11-NEXT: s_mov_b32 s11, 0x31016000
-; GFX11-NEXT: s_mov_b32 s14, s10
-; GFX11-NEXT: s_mov_b32 s15, s11
-; GFX11-NEXT: s_mov_b32 s6, s10
-; GFX11-NEXT: s_mov_b32 s7, s11
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s12, s2
-; GFX11-NEXT: s_mov_b32 s13, s3
-; GFX11-NEXT: buffer_load_u16 v0, off, s[12:15], 0 glc dlc
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: buffer_load_u16 v1, off, s[4:7], 0 glc dlc
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s0
-; GFX11-NEXT: s_mov_b32 s9, s1
-; GFX11-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX11-NEXT: v_max_f16_e32 v1, v1, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_max_f16_e32 v0, v0, v1
-; GFX11-NEXT: buffer_store_b16 v0, off, s[8:11], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: maxnum_f16:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_clause 0x1
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_load_b64 s[4:5], s[4:5], 0x34
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s14, s10
+; GFX11-TRUE16-NEXT: s_mov_b32 s15, s11
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s10
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, s11
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s12, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s13, s3
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[12:15], 0 glc dlc
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: buffer_load_u16 v1, off, s[4:7], 0 glc dlc
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s0
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s1
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: maxnum_f16:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_clause 0x1
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_load_b64 s[4:5], s[4:5], 0x34
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s14, s10
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, s11
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, s10
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, s11
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s12, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s13, s3
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[12:15], 0 glc dlc
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: buffer_load_u16 v1, off, s[4:7], 0 glc dlc
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s0
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s1
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v1
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_endpgm
ptr addrspace(1) %r,
ptr addrspace(1) %a,
ptr addrspace(1) %b) #0 {
@@ -228,25 +257,45 @@ define amdgpu_kernel void @maxnum_f16_imm_a(
; GFX10-NEXT: buffer_store_short v0, off, s[4:7], 0
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: maxnum_f16_imm_a:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_u16 v0, off, s[8:11], 0
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_max_f16_e32 v0, 0x4200, v0
-; GFX11-NEXT: buffer_store_b16 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: maxnum_f16_imm_a:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, 0x4200, v0.l
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: maxnum_f16_imm_a:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, 0x4200, v0
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
ptr addrspace(1) %r,
ptr addrspace(1) %b) #0 {
entry:
@@ -334,25 +383,45 @@ define amdgpu_kernel void @maxnum_f16_imm_b(
; GFX10-NEXT: buffer_store_short v0, off, s[4:7], 0
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: maxnum_f16_imm_b:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_u16 v0, off, s[8:11], 0
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_max_f16_e32 v0, 4.0, v0
-; GFX11-NEXT: buffer_store_b16 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: maxnum_f16_imm_b:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, 4.0, v0.l
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: maxnum_f16_imm_b:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, 4.0, v0
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
ptr addrspace(1) %r,
ptr addrspace(1) %a) #0 {
entry:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll
index 795ed6d542a13..3a2bf9d009460 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll
@@ -2,7 +2,8 @@
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck --check-prefixes=SI %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX89,VI %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX89,GFX9 %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,+real-true16 -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-TRUE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,-real-true16 -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11-FAKE16 %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1200 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX12 %s
declare half @llvm.rint.f16(half %a)
@@ -47,23 +48,41 @@ define amdgpu_kernel void @rint_f16(
; GFX89-NEXT: buffer_store_short v0, off, s[4:7], 0
; GFX89-NEXT: s_endpgm
;
-; GFX11-LABEL: rint_f16:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_u16 v0, off, s[8:11], 0
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_rndne_f16_e32 v0, v0
-; GFX11-NEXT: buffer_store_b16 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: rint_f16:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: rint_f16:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_rndne_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
;
; GFX12-LABEL: rint_f16:
; GFX12: ; %bb.0: ; %entry
@@ -166,27 +185,49 @@ define amdgpu_kernel void @rint_v2f16(
; GFX9-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GFX9-NEXT: s_endpgm
;
-; GFX11-LABEL: rint_v2f16:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_b32 v0, off, s[8:11], 0
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX11-NEXT: v_rndne_f16_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_rndne_f16_e32 v1, v1
-; GFX11-NEXT: v_pack_b32_f16 v0, v0, v1
-; GFX11-NEXT: buffer_store_b32 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: rint_v2f16:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v1.l
+; GFX11-TRUE16-NEXT: v_pack_b32_f16 v0, v0.l, v0.h
+; GFX11-TRUE16-NEXT: buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: rint_v2f16:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-FAKE16-NEXT: v_rndne_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_rndne_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: v_pack_b32_f16 v0, v0, v1
+; GFX11-FAKE16-NEXT: buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
;
; GFX12-LABEL: rint_v2f16:
; GFX12: ; %bb.0: ; %entry
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll
index 0d58afd1812de..c1ba985d37453 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll
@@ -1,7 +1,8 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefixes=SI %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=VI %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,+real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11-TRUE16 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global,-real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11-FAKE16 %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1200 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12 %s
declare half @llvm.trunc.f16(half %a)
@@ -46,23 +47,41 @@ define amdgpu_kernel void @trunc_f16(
; VI-NEXT: buffer_store_short v0, off, s[4:7], 0
; VI-NEXT: s_endpgm
;
-; GFX11-LABEL: trunc_f16:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_u16 v0, off, s[8:11], 0
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_trunc_f16_e32 v0, v0
-; GFX11-NEXT: buffer_store_b16 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: trunc_f16:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_trunc_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: trunc_f16:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_trunc_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
;
; GFX12-LABEL: trunc_f16:
; GFX12: ; %bb.0: ; %entry
@@ -145,27 +164,49 @@ define amdgpu_kernel void @trunc_v2f16(
; VI-NEXT: buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT: s_endpgm
;
-; GFX11-LABEL: trunc_v2f16:
-; GFX11: ; %bb.0: ; %entry
-; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
-; GFX11-NEXT: s_mov_b32 s6, -1
-; GFX11-NEXT: s_mov_b32 s7, 0x31016000
-; GFX11-NEXT: s_mov_b32 s10, s6
-; GFX11-NEXT: s_mov_b32 s11, s7
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s8, s2
-; GFX11-NEXT: s_mov_b32 s9, s3
-; GFX11-NEXT: s_mov_b32 s4, s0
-; GFX11-NEXT: buffer_load_b32 v0, off, s[8:11], 0
-; GFX11-NEXT: s_mov_b32 s5, s1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX11-NEXT: v_trunc_f16_e32 v0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_trunc_f16_e32 v1, v1
-; GFX11-NEXT: v_pack_b32_f16 v0, v0, v1
-; GFX11-NEXT: buffer_store_b32 v0, off, s[4:7], 0
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: trunc_v2f16:
+; GFX11-TRUE16: ; %bb.0: ; %entry
+; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, -1
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-TRUE16-NEXT: s_mov_b32 s10, s6
+; GFX11-TRUE16-NEXT: s_mov_b32 s11, s7
+; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-TRUE16-NEXT: s_mov_b32 s8, s2
+; GFX11-TRUE16-NEXT: s_mov_b32 s9, s3
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s0
+; GFX11-TRUE16-NEXT: buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-TRUE16-NEXT: v_trunc_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_trunc_f16_e32 v0.h, v1.l
+; GFX11-TRUE16-NEXT: v_pack_b32_f16 v0, v0.l, v0.h
+; GFX11-TRUE16-NEXT: buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: trunc_v2f16:
+; GFX11-FAKE16: ; %bb.0: ; %entry
+; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
+; GFX11-FAKE16-NEXT: s_mov_b32 s6, -1
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
+; GFX11-FAKE16-NEXT: s_mov_b32 s10, s6
+; GFX11-FAKE16-NEXT: s_mov_b32 s11, s7
+; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT: s_mov_b32 s8, s2
+; GFX11-FAKE16-NEXT: s_mov_b32 s9, s3
+; GFX11-FAKE16-NEXT: s_mov_b32 s4, s0
+; GFX11-FAKE16-NEXT: buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s5, s1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-FAKE16-NEXT: v_trunc_f16_e32 v0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_trunc_f16_e32 v1, v1
+; GFX11-FAKE16-NEXT: v_pack_b32_f16 v0, v0, v1
+; GFX11-FAKE16-NEXT: buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-FAKE16-NEXT: s_endpgm
;
; GFX12-LABEL: trunc_v2f16:
; GFX12: ; %bb.0: ; %entry
More information about the llvm-commits
mailing list