maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ps.live.ll

63 lines
2.2 KiB
LLVM
Raw Normal View History

[llvm] Remove redundant check-prefix=CHECK from tests. NFC. Reviewers: t.p.northover, oren_ben_simhon, niravd, mcrosier Reviewed By: oren_ben_simhon, mcrosier Subscribers: nhaehnle, javed.absar, llvm-commits Differential Revision: https://reviews.llvm.org/D35466 llvm-svn: 308193
2017-07-18 01:32:45 +08:00
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck %s
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
; CHECK-LABEL: {{^}}test1:
AMDGPU: VALU carry-in and v_cndmask condition cannot be EXEC The hardware will only forward EXEC_LO; the high 32 bits will be zero. Additionally, inline constants do not work. At least, v_addc_u32_e64 v0, vcc, v0, v1, -1 which could conceivably be used to combine (v0 + v1 + 1) into a single instruction, acts as if all carry-in bits are zero. The llvm.amdgcn.ps.live test is adjusted; it would be nice to combine s_mov_b64 s[0:1], exec v_cndmask_b32_e64 v0, v1, v2, s[0:1] into v_mov_b32 v0, v3 but it's not particularly high priority. Fixes dEQP-GLES31.functional.shaders.helper_invocation.value.* llvm-svn: 314522
2017-09-29 23:37:31 +08:00
; CHECK: s_mov_b64 s[0:1], exec
; CHECK: v_cndmask_b32_e64 v0, 0, 1, s[0:1]
;
; Note: The hardware doesn't implement EXEC as src2 for v_cndmask.
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
;
; Note: We could generate better code here if we recognized earlier that
; there is no WQM use and therefore llvm.amdgcn.ps.live is constant. However,
; the expectation is that the intrinsic will be used in non-trivial shaders,
; so such an optimization doesn't seem worth the effort.
AMDGPU: Convert image intrinsic uses in tests llvm-svn: 298386
2017-03-22 00:24:12 +08:00
define amdgpu_ps float @test1() #0 {
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
%live = call i1 @llvm.amdgcn.ps.live()
%live.32 = zext i1 %live to i32
%r = bitcast i32 %live.32 to float
ret float %r
}
; CHECK-LABEL: {{^}}test2:
; CHECK: s_mov_b64 [[LIVE:s\[[0-9]+:[0-9]+\]]], exec
; CHECK-DAG: s_wqm_b64 exec, exec
; CHECK-DAG: v_cndmask_b32_e64 [[VAR:v[0-9]+]], 0, 1, [[LIVE]]
; CHECK: image_sample v0, [[VAR]],
AMDGPU: Convert image intrinsic uses in tests llvm-svn: 298386
2017-03-22 00:24:12 +08:00
define amdgpu_ps float @test2() #0 {
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
%live = call i1 @llvm.amdgcn.ps.live()
%live.32 = zext i1 %live to i32
AMDGPU: Convert image intrinsic uses in tests llvm-svn: 298386
2017-03-22 00:24:12 +08:00
%live.32.bc = bitcast i32 %live.32 to float
AMDGPU: Convert test cases to the dimension-aware intrinsics Summary: Also explicitly port over some tests in llvm.amdgcn.image.* that were missing. Some tests are removed because they no longer apply (i.e. explicitly testing building an address vector via insertelement). This is in preparation for the eventual removal of the old-style intrinsics. Some additional notes: - constant-address-space-32bit.ll: change some GCN-NEXT to GCN because the instruction schedule was subtly altered - insert_vector_elt.ll: the old test didn't actually test anything, because %tmp1 was not used; remove the load, because it doesn't work (Because of the amdgpu_ps calling convention? In any case, it's orthogonal to what the test claims to be testing.) Change-Id: Idfa99b6512ad139e755e82b8b89548ab08f0afcf Reviewers: arsenm, rampitec Subscribers: MatzeB, qcolombet, kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, javed.absar, llvm-commits Differential Revision: https://reviews.llvm.org/D48018 llvm-svn: 335229
2018-06-21 21:37:19 +08:00
%t = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %live.32.bc, <8 x i32> undef, <4 x i32> undef, i1 0, i32 0, i32 0)
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
%r = extractelement <4 x float> %t, i32 0
ret float %r
}
; CHECK-LABEL: {{^}}test3:
; CHECK: s_mov_b64 [[LIVE:s\[[0-9]+:[0-9]+\]]], exec
; CHECK-DAG: s_wqm_b64 exec, exec
; CHECK-DAG: s_xor_b64 [[HELPER:s\[[0-9]+:[0-9]+\]]], [[LIVE]], -1
; CHECK_DAG: s_and_saveexec_b64 [[SAVED:s\[[0-9]+:[0-9]+\]]], [[HELPER]]
; CHECK: ; %dead
AMDGPU: Convert image intrinsic uses in tests llvm-svn: 298386
2017-03-22 00:24:12 +08:00
define amdgpu_ps float @test3(i32 %in) #0 {
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
entry:
%live = call i1 @llvm.amdgcn.ps.live()
br i1 %live, label %end, label %dead
dead:
%tc.dead = mul i32 %in, 2
br label %end
end:
%tc = phi i32 [ %in, %entry ], [ %tc.dead, %dead ]
AMDGPU: Convert image intrinsic uses in tests llvm-svn: 298386
2017-03-22 00:24:12 +08:00
%tc.bc = bitcast i32 %tc to float
AMDGPU: Convert test cases to the dimension-aware intrinsics Summary: Also explicitly port over some tests in llvm.amdgcn.image.* that were missing. Some tests are removed because they no longer apply (i.e. explicitly testing building an address vector via insertelement). This is in preparation for the eventual removal of the old-style intrinsics. Some additional notes: - constant-address-space-32bit.ll: change some GCN-NEXT to GCN because the instruction schedule was subtly altered - insert_vector_elt.ll: the old test didn't actually test anything, because %tmp1 was not used; remove the load, because it doesn't work (Because of the amdgpu_ps calling convention? In any case, it's orthogonal to what the test claims to be testing.) Change-Id: Idfa99b6512ad139e755e82b8b89548ab08f0afcf Reviewers: arsenm, rampitec Subscribers: MatzeB, qcolombet, kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, javed.absar, llvm-commits Differential Revision: https://reviews.llvm.org/D48018 llvm-svn: 335229
2018-06-21 21:37:19 +08:00
%t = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %tc.bc, <8 x i32> undef, <4 x i32> undef, i1 0, i32 0, i32 0) #0
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
%r = extractelement <4 x float> %t, i32 0
ret float %r
}
AMDGPU: Convert image intrinsic uses in tests llvm-svn: 298386
2017-03-22 00:24:12 +08:00
declare i1 @llvm.amdgcn.ps.live() #1
AMDGPU: Convert test cases to the dimension-aware intrinsics Summary: Also explicitly port over some tests in llvm.amdgcn.image.* that were missing. Some tests are removed because they no longer apply (i.e. explicitly testing building an address vector via insertelement). This is in preparation for the eventual removal of the old-style intrinsics. Some additional notes: - constant-address-space-32bit.ll: change some GCN-NEXT to GCN because the instruction schedule was subtly altered - insert_vector_elt.ll: the old test didn't actually test anything, because %tmp1 was not used; remove the load, because it doesn't work (Because of the amdgpu_ps calling convention? In any case, it's orthogonal to what the test claims to be testing.) Change-Id: Idfa99b6512ad139e755e82b8b89548ab08f0afcf Reviewers: arsenm, rampitec Subscribers: MatzeB, qcolombet, kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, javed.absar, llvm-commits Differential Revision: https://reviews.llvm.org/D48018 llvm-svn: 335229
2018-06-21 21:37:19 +08:00
declare <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32, float, <8 x i32>, <4 x i32>, i1, i32, i32) #2
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic Summary: This intrinsic returns true if the current thread belongs to a live pixel and false if it belongs to a pixel that we are executing only for derivative computation. It will be used by Mesa to implement gl_HelperInvocation. Note that for pixels that are killed during the shader, this implementation also returns true, but it doesn't matter because those pixels are always disabled in the EXEC mask. This unearthed a corner case in the instruction verifier, which complained about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but correct code, so make the verifier accept it as such. Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D19191 llvm-svn: 267102
2016-04-22 12:04:08 +08:00
AMDGPU: Convert image intrinsic uses in tests llvm-svn: 298386
2017-03-22 00:24:12 +08:00
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }
attributes #2 = { nounwind readonly }