f2c164c815
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
252 lines
11 KiB
LLVM
252 lines
11 KiB
LLVM
; 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_i16(ptr addrspace(1) %out, float %a, i16 %b) #0 {
|
|
; %result = call float @llvm.experimental.constrained.ldexp.f32.i16(float %a, i16 %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
; ret float %result
|
|
; }
|
|
|
|
define float @test_ldexp_f32_i32(ptr addrspace(1) %out, float %a, i32 %b) #0 {
|
|
; 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.experimental.constrained.ldexp.f32.i32(float %a, i32 %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
ret float %result
|
|
}
|
|
|
|
; define <2 x float> @test_ldexp_v2f32_v2i16(ptr addrspace(1) %out, <2 x float> %a, <2 x i16> %b) #0 {
|
|
; %result = call <2 x float> @llvm.experimental.constrained.ldexp.v2f32.v2i16(<2 x float> %a, <2 x i16> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
; ret <2 x float> %result
|
|
; }
|
|
|
|
define <2 x float> @test_ldexp_v2f32_v2i32(ptr addrspace(1) %out, <2 x float> %a, <2 x i32> %b) #0 {
|
|
; GFX6-SDAG-LABEL: test_ldexp_v2f32_v2i32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v1, v3, v5
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v2, v4
|
|
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX8-SDAG-LABEL: test_ldexp_v2f32_v2i32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v1, v3, v5
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v0, v2, v4
|
|
; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-SDAG-LABEL: test_ldexp_v2f32_v2i32:
|
|
; GFX9-SDAG: ; %bb.0:
|
|
; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v1, v3, v5
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v0, v2, v4
|
|
; GFX9-SDAG-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]
|
|
;
|
|
; GFX6-GISEL-LABEL: test_ldexp_v2f32_v2i32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v2, v4
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v1, v3, v5
|
|
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX8-GISEL-LABEL: test_ldexp_v2f32_v2i32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v0, v2, v4
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v1, v3, v5
|
|
; GFX8-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-GISEL-LABEL: test_ldexp_v2f32_v2i32:
|
|
; GFX9-GISEL: ; %bb.0:
|
|
; GFX9-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v0, v2, v4
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v1, v3, v5
|
|
; GFX9-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
%result = call <2 x float> @llvm.experimental.constrained.ldexp.v2f32.v2i32(<2 x float> %a, <2 x i32> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
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) #0 {
|
|
; GFX6-SDAG-LABEL: test_ldexp_v3f32_v3i32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v4, v4, v7
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v1, v3, v6
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v2, v5
|
|
; GFX6-SDAG-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX8-SDAG-LABEL: test_ldexp_v3f32_v3i32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v4, v4, v7
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v1, v3, v6
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v0, v2, v5
|
|
; GFX8-SDAG-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-SDAG-LABEL: test_ldexp_v3f32_v3i32:
|
|
; GFX9-SDAG: ; %bb.0:
|
|
; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v4, v4, v7
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v1, v3, v6
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v0, v2, v5
|
|
; GFX9-SDAG-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX9-SDAG-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]
|
|
;
|
|
; GFX6-GISEL-LABEL: test_ldexp_v3f32_v3i32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v2, v5
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v1, v3, v6
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v2, v4, v7
|
|
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX8-GISEL-LABEL: test_ldexp_v3f32_v3i32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v0, v2, v5
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v1, v3, v6
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v2, v4, v7
|
|
; GFX8-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-GISEL-LABEL: test_ldexp_v3f32_v3i32:
|
|
; GFX9-GISEL: ; %bb.0:
|
|
; GFX9-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v0, v2, v5
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v1, v3, v6
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v2, v4, v7
|
|
; GFX9-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
%result = call <3 x float> @llvm.experimental.constrained.ldexp.v3f32.v3i32(<3 x float> %a, <3 x i32> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
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) #0 {
|
|
; GFX6-SDAG-LABEL: test_ldexp_v4f32_v4i32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v5, v5, v9
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v4, v4, v8
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v1, v3, v7
|
|
; GFX6-SDAG-NEXT: v_ldexp_f32_e32 v0, v2, v6
|
|
; GFX6-SDAG-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX6-SDAG-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX6-SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX8-SDAG-LABEL: test_ldexp_v4f32_v4i32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v5, v5, v9
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v4, v4, v8
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v1, v3, v7
|
|
; GFX8-SDAG-NEXT: v_ldexp_f32 v0, v2, v6
|
|
; GFX8-SDAG-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX8-SDAG-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-SDAG-LABEL: test_ldexp_v4f32_v4i32:
|
|
; GFX9-SDAG: ; %bb.0:
|
|
; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v5, v5, v9
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v4, v4, v8
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v1, v3, v7
|
|
; GFX9-SDAG-NEXT: v_ldexp_f32 v0, v2, v6
|
|
; GFX9-SDAG-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX9-SDAG-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX9-SDAG-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]
|
|
;
|
|
; GFX6-GISEL-LABEL: test_ldexp_v4f32_v4i32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v0, v2, v6
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v1, v3, v7
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v2, v4, v8
|
|
; GFX6-GISEL-NEXT: v_ldexp_f32_e32 v3, v5, v9
|
|
; GFX6-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX8-GISEL-LABEL: test_ldexp_v4f32_v4i32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v0, v2, v6
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v1, v3, v7
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v2, v4, v8
|
|
; GFX8-GISEL-NEXT: v_ldexp_f32 v3, v5, v9
|
|
; GFX8-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-GISEL-LABEL: test_ldexp_v4f32_v4i32:
|
|
; GFX9-GISEL: ; %bb.0:
|
|
; GFX9-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v0, v2, v6
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v1, v3, v7
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v2, v4, v8
|
|
; GFX9-GISEL-NEXT: v_ldexp_f32 v3, v5, v9
|
|
; GFX9-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
%result = call <4 x float> @llvm.experimental.constrained.ldexp.v4f32.v4i32(<4 x float> %a, <4 x i32> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
ret <4 x float> %result
|
|
}
|
|
|
|
declare float @llvm.experimental.constrained.ldexp.f32.i16(float, i16, metadata, metadata) #1
|
|
declare float @llvm.experimental.constrained.ldexp.f32.i32(float, i32, metadata, metadata) #1
|
|
declare <2 x float> @llvm.experimental.constrained.ldexp.v2f32.v2i16(<2 x float>, <2 x i16>, metadata, metadata) #1
|
|
declare <2 x float> @llvm.experimental.constrained.ldexp.v2f32.v2i32(<2 x float>, <2 x i32>, metadata, metadata) #1
|
|
declare <3 x float> @llvm.experimental.constrained.ldexp.v3f32.v3i32(<3 x float>, <3 x i32>, metadata, metadata) #1
|
|
declare <4 x float> @llvm.experimental.constrained.ldexp.v4f32.v4i32(<4 x float>, <4 x i32>, metadata, metadata) #1
|
|
|
|
attributes #0 = { strictfp }
|
|
attributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite) }
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; GCN: {{.*}}
|
|
; GFX11-GISEL: {{.*}}
|
|
; GFX11-SDAG: {{.*}}
|