caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
1daf2994de49d1ecba4bee4e6842aa8a564cbc96
clang-p2996/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll
Jay Foad f2c164c815 [AMDGPU] Do not wait for vscnt on function entry and return 
SIInsertWaitcnts inserts waitcnt instructions to resolve data
dependencies. The GFX10+ vscnt (VMEM store count) counter is never used
in this way. It is only used to resolve memory dependencies, and that is
handled by SIMemoryLegalizer. Hence there is no need to conservatively
wait for vscnt to be 0 on function entry and before returns.

Differential Revision: https://reviews.llvm.org/D153537
2023-07-04 12:22:38 +01:00

585 lines
24 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -global-isel=0 -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefixes=GCN,GFX6,GFX6-SDAG %s
; RUN: llc -global-isel=0 -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,GFX8,GFX8-SDAG %s
; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9,GFX9-SDAG %s
; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GCN,GFX11,GFX11-SDAG %s
; RUN: llc -global-isel=1 -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefixes=GCN,GFX6,GFX6-GISEL %s
; RUN: llc -global-isel=1 -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,GFX8,GFX8-GISEL %s
; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9,GFX9-GISEL %s
; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GCN,GFX11,GFX11-GISEL %s
define float @test_ldexp_f32_i32(ptr addrspace(1) %out, float %a, i32 %b) {
; GFX6-LABEL: test_ldexp_f32_i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_ldexp_f32_e32 v0, v2, v3
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: test_ldexp_f32_i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_ldexp_f32 v0, v2, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: test_ldexp_f32_i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_ldexp_f32 v0, v2, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: test_ldexp_f32_i32:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_ldexp_f32 v0, v2, v3
; GFX11-NEXT: s_setpc_b64 s[30:31]
%result = call float @llvm.ldexp.f32.i32(float %a, i32 %b)
ret float %result
}
define <2 x float> @test_ldexp_v2f32_v2i32(ptr addrspace(1) %out, <2 x float> %a, <2 x i32> %b) {
; GFX6-LABEL: test_ldexp_v2f32_v2i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_ldexp_f32_e32 v0, v2, v4
; GFX6-NEXT: v_ldexp_f32_e32 v1, v3, v5
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: test_ldexp_v2f32_v2i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_ldexp_f32 v0, v2, v4
; GFX8-NEXT: v_ldexp_f32 v1, v3, v5
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: test_ldexp_v2f32_v2i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_ldexp_f32 v0, v2, v4
; GFX9-NEXT: v_ldexp_f32 v1, v3, v5
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: test_ldexp_v2f32_v2i32:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_ldexp_f32 v0, v2, v4
; GFX11-NEXT: v_ldexp_f32 v1, v3, v5
; GFX11-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float> %a, <2 x i32> %b)
ret <2 x float> %result
}
define <3 x float> @test_ldexp_v3f32_v3i32(ptr addrspace(1) %out, <3 x float> %a, <3 x i32> %b) {
; GFX6-LABEL: test_ldexp_v3f32_v3i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_ldexp_f32_e32 v0, v2, v5
; GFX6-NEXT: v_ldexp_f32_e32 v1, v3, v6
; GFX6-NEXT: v_ldexp_f32_e32 v2, v4, v7
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: test_ldexp_v3f32_v3i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_ldexp_f32 v0, v2, v5
; GFX8-NEXT: v_ldexp_f32 v1, v3, v6
; GFX8-NEXT: v_ldexp_f32 v2, v4, v7
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: test_ldexp_v3f32_v3i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_ldexp_f32 v0, v2, v5
; GFX9-NEXT: v_ldexp_f32 v1, v3, v6
; GFX9-NEXT: v_ldexp_f32 v2, v4, v7
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: test_ldexp_v3f32_v3i32:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_ldexp_f32 v0, v2, v5
; GFX11-NEXT: v_ldexp_f32 v1, v3, v6
; GFX11-NEXT: v_ldexp_f32 v2, v4, v7
; GFX11-NEXT: s_setpc_b64 s[30:31]
%result = call <3 x float> @llvm.ldexp.v3f32.v3i32(<3 x float> %a, <3 x i32> %b)
ret <3 x float> %result
}
define <4 x float> @test_ldexp_v4f32_v4i32(ptr addrspace(1) %out, <4 x float> %a, <4 x i32> %b) {
; GFX6-LABEL: test_ldexp_v4f32_v4i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_ldexp_f32_e32 v0, v2, v6
; GFX6-NEXT: v_ldexp_f32_e32 v1, v3, v7
; GFX6-NEXT: v_ldexp_f32_e32 v2, v4, v8
; GFX6-NEXT: v_ldexp_f32_e32 v3, v5, v9
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: test_ldexp_v4f32_v4i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_ldexp_f32 v0, v2, v6
; GFX8-NEXT: v_ldexp_f32 v1, v3, v7
; GFX8-NEXT: v_ldexp_f32 v2, v4, v8
; GFX8-NEXT: v_ldexp_f32 v3, v5, v9
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: test_ldexp_v4f32_v4i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_ldexp_f32 v0, v2, v6
; GFX9-NEXT: v_ldexp_f32 v1, v3, v7
; GFX9-NEXT: v_ldexp_f32 v2, v4, v8
; GFX9-NEXT: v_ldexp_f32 v3, v5, v9
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: test_ldexp_v4f32_v4i32:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_ldexp_f32 v0, v2, v6
; GFX11-NEXT: v_ldexp_f32 v1, v3, v7
; GFX11-NEXT: v_ldexp_f32 v2, v4, v8
; GFX11-NEXT: v_ldexp_f32 v3, v5, v9
; GFX11-NEXT: s_setpc_b64 s[30:31]
%result = call <4 x float> @llvm.ldexp.v4f32.v4i32(<4 x float> %a, <4 x i32> %b)
ret <4 x float> %result
}
define double @test_ldexp_f64_i32(double %a, i32 %b) {
; GCN-LABEL: test_ldexp_f64_i32:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_ldexp_f64 v[0:1], v[0:1], v2
; GCN-NEXT: s_setpc_b64 s[30:31]
%result = call double @llvm.ldexp.f64.i32(double %a, i32 %b)
ret double %result
}
define <2 x double> @test_ldexp_v2f64_v2i32(<2 x double> %a, <2 x i32> %b) {
; GCN-LABEL: test_ldexp_v2f64_v2i32:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_ldexp_f64 v[0:1], v[0:1], v4
; GCN-NEXT: v_ldexp_f64 v[2:3], v[2:3], v5
; GCN-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x double> @llvm.ldexp.v2f64.v2i32(<2 x double> %a, <2 x i32> %b)
ret <2 x double> %result
}
; Broken for DAG
; define float @test_ldexp_f32_i16(float %a, i16 %b) {
; %result = call float @llvm.ldexp.f32.i16(float %a, i16 %b)
; ret float %result
; }
; FIXME: Should be able to truncate to i32
; define float @test_ldexp_f32_i64(float %a, i64 %b) {
; %result = call float @llvm.ldexp.f32.i64(float %a, i64 %b)
; ret float %result
; }
; define <2 x float> @test_ldexp_v2f32_v2i16(<2 x float> %a, <2 x i16> %b) {
; %result = call <2 x float> @llvm.ldexp.v2f32.v2i16(<2 x float> %a, <2 x i16> %b)
; ret <2 x float> %result
; }
; FIXME: Should be able to truncate to i32
; define <2 x float> @test_ldexp_v2f32_v2i64(<2 x float> %a, <2 x i64> %b) {
; %result = call <2 x float> @llvm.ldexp.v2f32.v2i64(<2 x float> %a, <2 x i64> %b)
; ret <2 x float> %result
; }
define half @test_ldexp_f16_i8(half %a, i8 %b) {
; GFX6-SDAG-LABEL: test_ldexp_f16_i8:
; GFX6-SDAG: ; %bb.0:
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-SDAG-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-SDAG-NEXT: v_bfe_i32 v1, v1, 0, 8
; GFX6-SDAG-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v0, v1
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-SDAG-LABEL: test_ldexp_f16_i8:
; GFX8-SDAG: ; %bb.0:
; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-SDAG-NEXT: v_ldexp_f16_sdwa v0, v0, sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-SDAG-LABEL: test_ldexp_f16_i8:
; GFX9-SDAG: ; %bb.0:
; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-SDAG-NEXT: v_ldexp_f16_sdwa v0, v0, sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: test_ldexp_f16_i8:
; GFX11-SDAG: ; %bb.0:
; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT: v_bfe_i32 v1, v1, 0, 8
; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-GISEL-LABEL: test_ldexp_f16_i8:
; GFX6-GISEL: ; %bb.0:
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-GISEL-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-GISEL-NEXT: v_bfe_i32 v1, v1, 0, 8
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v0, v1
; GFX6-GISEL-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-GISEL-LABEL: test_ldexp_f16_i8:
; GFX8-GISEL: ; %bb.0:
; GFX8-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-GISEL-NEXT: v_bfe_i32 v1, v1, 0, 8
; GFX8-GISEL-NEXT: v_mov_b32_e32 v2, 0xffff8000
; GFX8-GISEL-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX8-GISEL-NEXT: v_med3_i32 v1, v1, v2, v3
; GFX8-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX8-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-GISEL-LABEL: test_ldexp_f16_i8:
; GFX9-GISEL: ; %bb.0:
; GFX9-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-GISEL-NEXT: v_bfe_i32 v1, v1, 0, 8
; GFX9-GISEL-NEXT: v_mov_b32_e32 v2, 0xffff8000
; GFX9-GISEL-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX9-GISEL-NEXT: v_med3_i32 v1, v1, v2, v3
; GFX9-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX9-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: test_ldexp_f16_i8:
; GFX11-GISEL: ; %bb.0:
; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT: v_bfe_i32 v1, v1, 0, 8
; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0x7fff
; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-GISEL-NEXT: v_med3_i32 v1, 0xffff8000, v1, v2
; GFX11-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31]
%result = call half @llvm.ldexp.f16.i8(half %a, i8 %b)
ret half %result
}
define half @test_ldexp_f16_i16(half %a, i16 %b) {
; GFX6-SDAG-LABEL: test_ldexp_f16_i16:
; GFX6-SDAG: ; %bb.0:
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-SDAG-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-SDAG-NEXT: v_bfe_i32 v1, v1, 0, 16
; GFX6-SDAG-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v0, v1
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: test_ldexp_f16_i16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: test_ldexp_f16_i16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: test_ldexp_f16_i16:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-GISEL-LABEL: test_ldexp_f16_i16:
; GFX6-GISEL: ; %bb.0:
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-GISEL-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-GISEL-NEXT: v_bfe_i32 v1, v1, 0, 16
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v0, v1
; GFX6-GISEL-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
%result = call half @llvm.ldexp.f16.i16(half %a, i16 %b)
ret half %result
}
define half @test_ldexp_f16_i32(half %a, i32 %b) {
; GFX6-SDAG-LABEL: test_ldexp_f16_i32:
; GFX6-SDAG: ; %bb.0:
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-SDAG-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-SDAG-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v0, v1
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-SDAG-LABEL: test_ldexp_f16_i32:
; GFX8-SDAG: ; %bb.0:
; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-SDAG-NEXT: s_movk_i32 s4, 0x8000
; GFX8-SDAG-NEXT: v_mov_b32_e32 v2, 0x7fff
; GFX8-SDAG-NEXT: v_med3_i32 v1, v1, s4, v2
; GFX8-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-SDAG-LABEL: test_ldexp_f16_i32:
; GFX9-SDAG: ; %bb.0:
; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-SDAG-NEXT: s_movk_i32 s4, 0x8000
; GFX9-SDAG-NEXT: v_mov_b32_e32 v2, 0x7fff
; GFX9-SDAG-NEXT: v_med3_i32 v1, v1, s4, v2
; GFX9-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: test_ldexp_f16_i32:
; GFX11-SDAG: ; %bb.0:
; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT: s_movk_i32 s0, 0x8000
; GFX11-SDAG-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-SDAG-NEXT: v_med3_i32 v1, v1, s0, 0x7fff
; GFX11-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-GISEL-LABEL: test_ldexp_f16_i32:
; GFX6-GISEL: ; %bb.0:
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-GISEL-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v0, v1
; GFX6-GISEL-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-GISEL-LABEL: test_ldexp_f16_i32:
; GFX8-GISEL: ; %bb.0:
; GFX8-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-GISEL-NEXT: v_mov_b32_e32 v2, 0xffff8000
; GFX8-GISEL-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX8-GISEL-NEXT: v_med3_i32 v1, v1, v2, v3
; GFX8-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX8-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-GISEL-LABEL: test_ldexp_f16_i32:
; GFX9-GISEL: ; %bb.0:
; GFX9-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-GISEL-NEXT: v_mov_b32_e32 v2, 0xffff8000
; GFX9-GISEL-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX9-GISEL-NEXT: v_med3_i32 v1, v1, v2, v3
; GFX9-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX9-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: test_ldexp_f16_i32:
; GFX11-GISEL: ; %bb.0:
; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0x7fff
; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-GISEL-NEXT: v_med3_i32 v1, 0xffff8000, v1, v2
; GFX11-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31]
%result = call half @llvm.ldexp.f16.i32(half %a, i32 %b)
ret half %result
}
define <2 x half> @test_ldexp_v2f16_v2i32(<2 x half> %a, <2 x i32> %b) {
; GFX6-SDAG-LABEL: test_ldexp_v2f16_v2i32:
; GFX6-SDAG: ; %bb.0:
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-SDAG-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-SDAG-NEXT: v_cvt_f16_f32_e32 v1, v1
; GFX6-SDAG-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-SDAG-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v0, v2
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v1, v1, v3
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-SDAG-LABEL: test_ldexp_v2f16_v2i32:
; GFX8-SDAG: ; %bb.0:
; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-SDAG-NEXT: s_movk_i32 s4, 0x8000
; GFX8-SDAG-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX8-SDAG-NEXT: v_med3_i32 v2, v2, s4, v3
; GFX8-SDAG-NEXT: v_med3_i32 v1, v1, s4, v3
; GFX8-SDAG-NEXT: v_ldexp_f16_sdwa v2, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX8-SDAG-NEXT: v_or_b32_e32 v0, v0, v2
; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-SDAG-LABEL: test_ldexp_v2f16_v2i32:
; GFX9-SDAG: ; %bb.0:
; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-SDAG-NEXT: s_movk_i32 s4, 0x8000
; GFX9-SDAG-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX9-SDAG-NEXT: v_med3_i32 v2, v2, s4, v3
; GFX9-SDAG-NEXT: v_med3_i32 v1, v1, s4, v3
; GFX9-SDAG-NEXT: v_ldexp_f16_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX9-SDAG-NEXT: v_pack_b32_f16 v0, v0, v2
; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: test_ldexp_v2f16_v2i32:
; GFX11-SDAG: ; %bb.0:
; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT: s_movk_i32 s0, 0x8000
; GFX11-SDAG-NEXT: v_lshrrev_b32_e32 v3, 16, v0
; GFX11-SDAG-NEXT: v_med3_i32 v2, v2, s0, 0x7fff
; GFX11-SDAG-NEXT: v_med3_i32 v1, v1, s0, 0x7fff
; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-SDAG-NEXT: v_ldexp_f16_e32 v2, v3, v2
; GFX11-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-SDAG-NEXT: v_pack_b32_f16 v0, v0, v2
; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-GISEL-LABEL: test_ldexp_v2f16_v2i32:
; GFX6-GISEL: ; %bb.0:
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-GISEL-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-GISEL-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v0, v2
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v1, v1, v3
; GFX6-GISEL-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-GISEL-NEXT: v_cvt_f16_f32_e32 v1, v1
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-GISEL-LABEL: test_ldexp_v2f16_v2i32:
; GFX8-GISEL: ; %bb.0:
; GFX8-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-GISEL-NEXT: v_mov_b32_e32 v3, 0xffff8000
; GFX8-GISEL-NEXT: v_mov_b32_e32 v4, 0x7fff
; GFX8-GISEL-NEXT: v_med3_i32 v1, v1, v3, v4
; GFX8-GISEL-NEXT: v_med3_i32 v2, v2, v3, v4
; GFX8-GISEL-NEXT: v_ldexp_f16_e32 v1, v0, v1
; GFX8-GISEL-NEXT: v_ldexp_f16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-GISEL-NEXT: v_or_b32_e32 v0, v1, v0
; GFX8-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-GISEL-LABEL: test_ldexp_v2f16_v2i32:
; GFX9-GISEL: ; %bb.0:
; GFX9-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-GISEL-NEXT: v_mov_b32_e32 v3, 0xffff8000
; GFX9-GISEL-NEXT: v_mov_b32_e32 v4, 0x7fff
; GFX9-GISEL-NEXT: v_med3_i32 v1, v1, v3, v4
; GFX9-GISEL-NEXT: v_med3_i32 v2, v2, v3, v4
; GFX9-GISEL-NEXT: v_ldexp_f16_e32 v1, v0, v1
; GFX9-GISEL-NEXT: v_ldexp_f16_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-GISEL-NEXT: v_lshl_or_b32 v0, v0, 16, v1
; GFX9-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: test_ldexp_v2f16_v2i32:
; GFX11-GISEL: ; %bb.0:
; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX11-GISEL-NEXT: v_lshrrev_b32_e32 v4, 16, v0
; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-GISEL-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
; GFX11-GISEL-NEXT: v_med3_i32 v2, 0xffff8000, v2, v3
; GFX11-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-GISEL-NEXT: v_ldexp_f16_e32 v1, v4, v2
; GFX11-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x half> @llvm.ldexp.v2f16.v2i32(<2 x half> %a, <2 x i32> %b)
ret <2 x half> %result
}
define <2 x half> @test_ldexp_v2f16_v2i16(<2 x half> %a, <2 x i16> %b) {
; GFX6-SDAG-LABEL: test_ldexp_v2f16_v2i16:
; GFX6-SDAG: ; %bb.0:
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-SDAG-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-SDAG-NEXT: v_cvt_f16_f32_e32 v1, v1
; GFX6-SDAG-NEXT: v_bfe_i32 v3, v3, 0, 16
; GFX6-SDAG-NEXT: v_bfe_i32 v2, v2, 0, 16
; GFX6-SDAG-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-SDAG-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v0, v2
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v1, v1, v3
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-SDAG-LABEL: test_ldexp_v2f16_v2i16:
; GFX8-SDAG: ; %bb.0:
; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-SDAG-NEXT: v_ldexp_f16_sdwa v2, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; GFX8-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX8-SDAG-NEXT: v_or_b32_e32 v0, v0, v2
; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-SDAG-LABEL: test_ldexp_v2f16_v2i16:
; GFX9-SDAG: ; %bb.0:
; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-SDAG-NEXT: v_ldexp_f16_sdwa v2, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; GFX9-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX9-SDAG-NEXT: v_pack_b32_f16 v0, v0, v2
; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: test_ldexp_v2f16_v2i16:
; GFX11-SDAG: ; %bb.0:
; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT: v_lshrrev_b32_e32 v2, 16, v1
; GFX11-SDAG-NEXT: v_lshrrev_b32_e32 v3, 16, v0
; GFX11-SDAG-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-SDAG-NEXT: v_ldexp_f16_e32 v2, v3, v2
; GFX11-SDAG-NEXT: v_pack_b32_f16 v0, v0, v2
; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-GISEL-LABEL: test_ldexp_v2f16_v2i16:
; GFX6-GISEL: ; %bb.0:
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-GISEL-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-GISEL-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX6-GISEL-NEXT: v_bfe_i32 v2, v2, 0, 16
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v0, v2
; GFX6-GISEL-NEXT: v_bfe_i32 v2, v3, 0, 16
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v1, v1, v2
; GFX6-GISEL-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-GISEL-NEXT: v_cvt_f16_f32_e32 v1, v1
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-GISEL-LABEL: test_ldexp_v2f16_v2i16:
; GFX8-GISEL: ; %bb.0:
; GFX8-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-GISEL-NEXT: v_ldexp_f16_e32 v2, v0, v1
; GFX8-GISEL-NEXT: v_ldexp_f16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; GFX8-GISEL-NEXT: v_or_b32_e32 v0, v2, v0
; GFX8-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-GISEL-LABEL: test_ldexp_v2f16_v2i16:
; GFX9-GISEL: ; %bb.0:
; GFX9-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-GISEL-NEXT: v_ldexp_f16_e32 v2, v0, v1
; GFX9-GISEL-NEXT: v_ldexp_f16_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; GFX9-GISEL-NEXT: v_lshl_or_b32 v0, v0, 16, v2
; GFX9-GISEL-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: test_ldexp_v2f16_v2i16:
; GFX11-GISEL: ; %bb.0:
; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-GISEL-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-NEXT: v_ldexp_f16_e32 v0, v0, v1
; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-GISEL-NEXT: v_ldexp_f16_e32 v1, v2, v3
; GFX11-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x half> @llvm.ldexp.v2f16.v2i16(<2 x half> %a, <2 x i16> %b)
ret <2 x half> %result
}
declare float @llvm.ldexp.f32.i32(float, i32) #0
declare float @llvm.ldexp.f32.i16(float, i16) #0
declare float @llvm.ldexp.f32.i64(float, i64) #0
declare half @llvm.ldexp.f16.i8(half, i8) #0
declare half @llvm.ldexp.f16.i16(half, i16) #0
declare half @llvm.ldexp.f16.i32(half, i32) #0
declare <2 x half> @llvm.ldexp.v2f16.v2i16(<2 x half>, <2 x i16>) #0
declare <2 x half> @llvm.ldexp.v2f16.v2i32(<2 x half>, <2 x i32>) #0
declare <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float>, <2 x i32>) #0
declare <2 x float> @llvm.ldexp.v2f32.v2i16(<2 x float>, <2 x i16>) #0
declare <2 x float> @llvm.ldexp.v2f32.v2i64(<2 x float>, <2 x i64>) #0
declare <3 x float> @llvm.ldexp.v3f32.v3i32(<3 x float>, <3 x i32>) #0
declare <4 x float> @llvm.ldexp.v4f32.v4i32(<4 x float>, <4 x i32>) #0
declare double @llvm.ldexp.f64.i32(double, i32) #0
declare <2 x double> @llvm.ldexp.v2f64.v2i32(<2 x double>, <2 x i32>) #0
attributes #0 = { nounwind readnone }
Reference in New Issue View Git Blame Copy Permalink