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
90 lines
4.8 KiB
LLVM
90 lines
4.8 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 double @test_ldexp_f64_i16(ptr addrspace(1) %out, double %a, i16 %b) #0 {
|
|
; %result = call double @llvm.experimental.constrained.ldexp.f64.i16(double %a, i16 %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
; ret double %result
|
|
; }
|
|
|
|
define double @test_ldexp_f64_i32(ptr addrspace(1) %out, double %a, i32 %b) #0 {
|
|
; 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[2:3], v4
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
%result = call double @llvm.experimental.constrained.ldexp.f64.i32(double %a, i32 %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
ret double %result
|
|
}
|
|
|
|
; define <2 x double> @test_ldexp_v2f64_v2i16(ptr addrspace(1) %out, <2 x double> %a, <2 x i16> %b) #0 {
|
|
; %result = call <2 x double> @llvm.experimental.constrained.ldexp.v2f64.v2i16(<2 x double> %a, <2 x i16> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
; ret <2 x double> %result
|
|
; }
|
|
|
|
define <2 x double> @test_ldexp_v2f64_v2i32(ptr addrspace(1) %out, <2 x double> %a, <2 x i32> %b) #0 {
|
|
; 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[2:3], v6
|
|
; GCN-NEXT: v_ldexp_f64 v[2:3], v[4:5], v7
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
%result = call <2 x double> @llvm.experimental.constrained.ldexp.v2f64.v2i32(<2 x double> %a, <2 x i32> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
ret <2 x double> %result
|
|
}
|
|
|
|
define <3 x double> @test_ldexp_v3f64_v3i32(ptr addrspace(1) %out, <3 x double> %a, <3 x i32> %b) #0 {
|
|
; GCN-LABEL: test_ldexp_v3f64_v3i32:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: v_ldexp_f64 v[0:1], v[2:3], v8
|
|
; GCN-NEXT: v_ldexp_f64 v[2:3], v[4:5], v9
|
|
; GCN-NEXT: v_ldexp_f64 v[4:5], v[6:7], v10
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
%result = call <3 x double> @llvm.experimental.constrained.ldexp.v3f64.v3i32(<3 x double> %a, <3 x i32> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
ret <3 x double> %result
|
|
}
|
|
|
|
define <4 x double> @test_ldexp_v4f64_v4i32(ptr addrspace(1) %out, <4 x double> %a, <4 x i32> %b) #0 {
|
|
; GCN-LABEL: test_ldexp_v4f64_v4i32:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: v_ldexp_f64 v[0:1], v[2:3], v10
|
|
; GCN-NEXT: v_ldexp_f64 v[2:3], v[4:5], v11
|
|
; GCN-NEXT: v_ldexp_f64 v[4:5], v[6:7], v12
|
|
; GCN-NEXT: v_ldexp_f64 v[6:7], v[8:9], v13
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
%result = call <4 x double> @llvm.experimental.constrained.ldexp.v4f64.v4i32(<4 x double> %a, <4 x i32> %b, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
ret <4 x double> %result
|
|
}
|
|
|
|
declare double @llvm.experimental.constrained.ldexp.f64.i16(double, i16, metadata, metadata) #1
|
|
declare double @llvm.experimental.constrained.ldexp.f64.i32(double, i32, metadata, metadata) #1
|
|
declare <2 x double> @llvm.experimental.constrained.ldexp.v2f64.v2i16(<2 x double>, <2 x i16>, metadata, metadata) #1
|
|
declare <2 x double> @llvm.experimental.constrained.ldexp.v2f64.v2i32(<2 x double>, <2 x i32>, metadata, metadata) #1
|
|
declare <3 x double> @llvm.experimental.constrained.ldexp.v3f64.v3i32(<3 x double>, <3 x i32>, metadata, metadata) #1
|
|
declare <4 x double> @llvm.experimental.constrained.ldexp.v4f64.v4i32(<4 x double>, <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:
|
|
; GFX11: {{.*}}
|
|
; GFX11-GISEL: {{.*}}
|
|
; GFX11-SDAG: {{.*}}
|
|
; GFX6: {{.*}}
|
|
; GFX6-GISEL: {{.*}}
|
|
; GFX6-SDAG: {{.*}}
|
|
; GFX8: {{.*}}
|
|
; GFX8-GISEL: {{.*}}
|
|
; GFX8-SDAG: {{.*}}
|
|
; GFX9: {{.*}}
|
|
; GFX9-GISEL: {{.*}}
|
|
; GFX9-SDAG: {{.*}}
|