llvm-project/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.load.a16.ll

129 lines
5.5 KiB
LLVM
Raw Normal View History

; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN %s
; GCN-LABEL: {{^}}load.f32.1d:
[AMDGPU] Add support for TFE/LWE in image intrinsics. 2nd try TFE and LWE support requires extra result registers that are written in the event of a failure in order to detect that failure case. The specific use-case that initiated these changes is sparse texture support. This means that if image intrinsics are used with either option turned on, the programmer must ensure that the return type can contain all of the expected results. This can result in redundant registers since the vector size must be a power-of-2. This change takes roughly 6 parts: 1. Modify the instruction defs in tablegen to add new instruction variants that can accomodate the extra return values. 2. Updates to lowerImage in SIISelLowering.cpp to accomodate setting TFE or LWE (where the bulk of the work for these instruction types is now done) 3. Extra verification code to catch cases where intrinsics have been used but insufficient return registers are used. 4. Modification to the adjustWritemask optimisation to account for TFE/LWE being enabled (requires extra registers to be maintained for error return value). 5. An extra pass to zero initialize the error value return - this is because if the error does not occur, the register is not written and thus must be zeroed before use. Also added a new (on by default) option to ensure ALL return values are zero-initialized that is required for sparse texture support. 6. Disable the inst_combine optimization in the presence of tfe/lwe (later TODO for this to re-enable and handle correctly). There's an additional fix now to avoid a dmask=0 For an image intrinsic with tfe where all result channels except tfe were unused, I was getting an image instruction with dmask=0 and only a single vgpr result for tfe. That is incorrect because the hardware assumes there is at least one vgpr result, plus the one for tfe. Fixed by forcing dmask to 1, which gives the desired two vgpr result with tfe in the second one. The TFE or LWE result is returned from the intrinsics using an aggregate type. Look in the test code provided to see how this works, but in essence IR code to invoke the intrinsic looks as follows: %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0) %v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.err = extractvalue {<4 x float>, i32} %v, 1 This re-submit of the change also includes a slight modification in SIISelLowering.cpp to work-around a compiler bug for the powerpc_le platform that caused a buildbot failure on a previous submission. Differential revision: https://reviews.llvm.org/D48826 Change-Id: If222bc03642e76cf98059a6bef5d5bffeda38dda Work around for ppcle compiler bug Change-Id: Ie284cf24b2271215be1b9dc95b485fd15000e32b llvm-svn: 351054
2019-01-14 19:55:24 +08:00
; GCN: image_load v0, v0, s[0:7] dmask:0x1 unorm a16
define amdgpu_ps <4 x float> @load.f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 1, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v2f32.1d:
[AMDGPU] Add support for TFE/LWE in image intrinsics. 2nd try TFE and LWE support requires extra result registers that are written in the event of a failure in order to detect that failure case. The specific use-case that initiated these changes is sparse texture support. This means that if image intrinsics are used with either option turned on, the programmer must ensure that the return type can contain all of the expected results. This can result in redundant registers since the vector size must be a power-of-2. This change takes roughly 6 parts: 1. Modify the instruction defs in tablegen to add new instruction variants that can accomodate the extra return values. 2. Updates to lowerImage in SIISelLowering.cpp to accomodate setting TFE or LWE (where the bulk of the work for these instruction types is now done) 3. Extra verification code to catch cases where intrinsics have been used but insufficient return registers are used. 4. Modification to the adjustWritemask optimisation to account for TFE/LWE being enabled (requires extra registers to be maintained for error return value). 5. An extra pass to zero initialize the error value return - this is because if the error does not occur, the register is not written and thus must be zeroed before use. Also added a new (on by default) option to ensure ALL return values are zero-initialized that is required for sparse texture support. 6. Disable the inst_combine optimization in the presence of tfe/lwe (later TODO for this to re-enable and handle correctly). There's an additional fix now to avoid a dmask=0 For an image intrinsic with tfe where all result channels except tfe were unused, I was getting an image instruction with dmask=0 and only a single vgpr result for tfe. That is incorrect because the hardware assumes there is at least one vgpr result, plus the one for tfe. Fixed by forcing dmask to 1, which gives the desired two vgpr result with tfe in the second one. The TFE or LWE result is returned from the intrinsics using an aggregate type. Look in the test code provided to see how this works, but in essence IR code to invoke the intrinsic looks as follows: %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0) %v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.err = extractvalue {<4 x float>, i32} %v, 1 This re-submit of the change also includes a slight modification in SIISelLowering.cpp to work-around a compiler bug for the powerpc_le platform that caused a buildbot failure on a previous submission. Differential revision: https://reviews.llvm.org/D48826 Change-Id: If222bc03642e76cf98059a6bef5d5bffeda38dda Work around for ppcle compiler bug Change-Id: Ie284cf24b2271215be1b9dc95b485fd15000e32b llvm-svn: 351054
2019-01-14 19:55:24 +08:00
; GCN: image_load v[0:1], v0, s[0:7] dmask:0x3 unorm a16
define amdgpu_ps <4 x float> @load.v2f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 3, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v3f32.1d:
; GCN: image_load v[0:2], v0, s[0:7] dmask:0x7 unorm a16
define amdgpu_ps <4 x float> @load.v3f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 7, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v4f32.1d:
; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm a16
define amdgpu_ps <4 x float> @load.v4f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 15, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.f32.2d:
[AMDGPU] Add support for TFE/LWE in image intrinsics. 2nd try TFE and LWE support requires extra result registers that are written in the event of a failure in order to detect that failure case. The specific use-case that initiated these changes is sparse texture support. This means that if image intrinsics are used with either option turned on, the programmer must ensure that the return type can contain all of the expected results. This can result in redundant registers since the vector size must be a power-of-2. This change takes roughly 6 parts: 1. Modify the instruction defs in tablegen to add new instruction variants that can accomodate the extra return values. 2. Updates to lowerImage in SIISelLowering.cpp to accomodate setting TFE or LWE (where the bulk of the work for these instruction types is now done) 3. Extra verification code to catch cases where intrinsics have been used but insufficient return registers are used. 4. Modification to the adjustWritemask optimisation to account for TFE/LWE being enabled (requires extra registers to be maintained for error return value). 5. An extra pass to zero initialize the error value return - this is because if the error does not occur, the register is not written and thus must be zeroed before use. Also added a new (on by default) option to ensure ALL return values are zero-initialized that is required for sparse texture support. 6. Disable the inst_combine optimization in the presence of tfe/lwe (later TODO for this to re-enable and handle correctly). There's an additional fix now to avoid a dmask=0 For an image intrinsic with tfe where all result channels except tfe were unused, I was getting an image instruction with dmask=0 and only a single vgpr result for tfe. That is incorrect because the hardware assumes there is at least one vgpr result, plus the one for tfe. Fixed by forcing dmask to 1, which gives the desired two vgpr result with tfe in the second one. The TFE or LWE result is returned from the intrinsics using an aggregate type. Look in the test code provided to see how this works, but in essence IR code to invoke the intrinsic looks as follows: %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0) %v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.err = extractvalue {<4 x float>, i32} %v, 1 This re-submit of the change also includes a slight modification in SIISelLowering.cpp to work-around a compiler bug for the powerpc_le platform that caused a buildbot failure on a previous submission. Differential revision: https://reviews.llvm.org/D48826 Change-Id: If222bc03642e76cf98059a6bef5d5bffeda38dda Work around for ppcle compiler bug Change-Id: Ie284cf24b2271215be1b9dc95b485fd15000e32b llvm-svn: 351054
2019-01-14 19:55:24 +08:00
; GCN: image_load v0, v0, s[0:7] dmask:0x1 unorm a16
define amdgpu_ps <4 x float> @load.f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%y = extractelement <2 x i16> %coords, i32 1
%v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 1, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v2f32.2d:
[AMDGPU] Add support for TFE/LWE in image intrinsics. 2nd try TFE and LWE support requires extra result registers that are written in the event of a failure in order to detect that failure case. The specific use-case that initiated these changes is sparse texture support. This means that if image intrinsics are used with either option turned on, the programmer must ensure that the return type can contain all of the expected results. This can result in redundant registers since the vector size must be a power-of-2. This change takes roughly 6 parts: 1. Modify the instruction defs in tablegen to add new instruction variants that can accomodate the extra return values. 2. Updates to lowerImage in SIISelLowering.cpp to accomodate setting TFE or LWE (where the bulk of the work for these instruction types is now done) 3. Extra verification code to catch cases where intrinsics have been used but insufficient return registers are used. 4. Modification to the adjustWritemask optimisation to account for TFE/LWE being enabled (requires extra registers to be maintained for error return value). 5. An extra pass to zero initialize the error value return - this is because if the error does not occur, the register is not written and thus must be zeroed before use. Also added a new (on by default) option to ensure ALL return values are zero-initialized that is required for sparse texture support. 6. Disable the inst_combine optimization in the presence of tfe/lwe (later TODO for this to re-enable and handle correctly). There's an additional fix now to avoid a dmask=0 For an image intrinsic with tfe where all result channels except tfe were unused, I was getting an image instruction with dmask=0 and only a single vgpr result for tfe. That is incorrect because the hardware assumes there is at least one vgpr result, plus the one for tfe. Fixed by forcing dmask to 1, which gives the desired two vgpr result with tfe in the second one. The TFE or LWE result is returned from the intrinsics using an aggregate type. Look in the test code provided to see how this works, but in essence IR code to invoke the intrinsic looks as follows: %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0) %v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.err = extractvalue {<4 x float>, i32} %v, 1 This re-submit of the change also includes a slight modification in SIISelLowering.cpp to work-around a compiler bug for the powerpc_le platform that caused a buildbot failure on a previous submission. Differential revision: https://reviews.llvm.org/D48826 Change-Id: If222bc03642e76cf98059a6bef5d5bffeda38dda Work around for ppcle compiler bug Change-Id: Ie284cf24b2271215be1b9dc95b485fd15000e32b llvm-svn: 351054
2019-01-14 19:55:24 +08:00
; GCN: image_load v[0:1], v0, s[0:7] dmask:0x3 unorm a16
define amdgpu_ps <4 x float> @load.v2f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%y = extractelement <2 x i16> %coords, i32 1
%v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 3, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v3f32.2d:
; GCN: image_load v[0:2], v0, s[0:7] dmask:0x7 unorm a16
define amdgpu_ps <4 x float> @load.v3f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%y = extractelement <2 x i16> %coords, i32 1
%v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 7, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v4f32.2d:
; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm a16
define amdgpu_ps <4 x float> @load.v4f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
main_body:
%x = extractelement <2 x i16> %coords, i32 0
%y = extractelement <2 x i16> %coords, i32 1
%v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 15, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.f32.3d:
[AMDGPU] Add support for TFE/LWE in image intrinsics. 2nd try TFE and LWE support requires extra result registers that are written in the event of a failure in order to detect that failure case. The specific use-case that initiated these changes is sparse texture support. This means that if image intrinsics are used with either option turned on, the programmer must ensure that the return type can contain all of the expected results. This can result in redundant registers since the vector size must be a power-of-2. This change takes roughly 6 parts: 1. Modify the instruction defs in tablegen to add new instruction variants that can accomodate the extra return values. 2. Updates to lowerImage in SIISelLowering.cpp to accomodate setting TFE or LWE (where the bulk of the work for these instruction types is now done) 3. Extra verification code to catch cases where intrinsics have been used but insufficient return registers are used. 4. Modification to the adjustWritemask optimisation to account for TFE/LWE being enabled (requires extra registers to be maintained for error return value). 5. An extra pass to zero initialize the error value return - this is because if the error does not occur, the register is not written and thus must be zeroed before use. Also added a new (on by default) option to ensure ALL return values are zero-initialized that is required for sparse texture support. 6. Disable the inst_combine optimization in the presence of tfe/lwe (later TODO for this to re-enable and handle correctly). There's an additional fix now to avoid a dmask=0 For an image intrinsic with tfe where all result channels except tfe were unused, I was getting an image instruction with dmask=0 and only a single vgpr result for tfe. That is incorrect because the hardware assumes there is at least one vgpr result, plus the one for tfe. Fixed by forcing dmask to 1, which gives the desired two vgpr result with tfe in the second one. The TFE or LWE result is returned from the intrinsics using an aggregate type. Look in the test code provided to see how this works, but in essence IR code to invoke the intrinsic looks as follows: %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0) %v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.err = extractvalue {<4 x float>, i32} %v, 1 This re-submit of the change also includes a slight modification in SIISelLowering.cpp to work-around a compiler bug for the powerpc_le platform that caused a buildbot failure on a previous submission. Differential revision: https://reviews.llvm.org/D48826 Change-Id: If222bc03642e76cf98059a6bef5d5bffeda38dda Work around for ppcle compiler bug Change-Id: Ie284cf24b2271215be1b9dc95b485fd15000e32b llvm-svn: 351054
2019-01-14 19:55:24 +08:00
; GCN: image_load v0, v[0:1], s[0:7] dmask:0x1 unorm a16
define amdgpu_ps <4 x float> @load.f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
main_body:
%x = extractelement <2 x i16> %coords_lo, i32 0
%y = extractelement <2 x i16> %coords_lo, i32 1
%z = extractelement <2 x i16> %coords_hi, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 1, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v2f32.3d:
[AMDGPU] Add support for TFE/LWE in image intrinsics. 2nd try TFE and LWE support requires extra result registers that are written in the event of a failure in order to detect that failure case. The specific use-case that initiated these changes is sparse texture support. This means that if image intrinsics are used with either option turned on, the programmer must ensure that the return type can contain all of the expected results. This can result in redundant registers since the vector size must be a power-of-2. This change takes roughly 6 parts: 1. Modify the instruction defs in tablegen to add new instruction variants that can accomodate the extra return values. 2. Updates to lowerImage in SIISelLowering.cpp to accomodate setting TFE or LWE (where the bulk of the work for these instruction types is now done) 3. Extra verification code to catch cases where intrinsics have been used but insufficient return registers are used. 4. Modification to the adjustWritemask optimisation to account for TFE/LWE being enabled (requires extra registers to be maintained for error return value). 5. An extra pass to zero initialize the error value return - this is because if the error does not occur, the register is not written and thus must be zeroed before use. Also added a new (on by default) option to ensure ALL return values are zero-initialized that is required for sparse texture support. 6. Disable the inst_combine optimization in the presence of tfe/lwe (later TODO for this to re-enable and handle correctly). There's an additional fix now to avoid a dmask=0 For an image intrinsic with tfe where all result channels except tfe were unused, I was getting an image instruction with dmask=0 and only a single vgpr result for tfe. That is incorrect because the hardware assumes there is at least one vgpr result, plus the one for tfe. Fixed by forcing dmask to 1, which gives the desired two vgpr result with tfe in the second one. The TFE or LWE result is returned from the intrinsics using an aggregate type. Look in the test code provided to see how this works, but in essence IR code to invoke the intrinsic looks as follows: %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0) %v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.err = extractvalue {<4 x float>, i32} %v, 1 This re-submit of the change also includes a slight modification in SIISelLowering.cpp to work-around a compiler bug for the powerpc_le platform that caused a buildbot failure on a previous submission. Differential revision: https://reviews.llvm.org/D48826 Change-Id: If222bc03642e76cf98059a6bef5d5bffeda38dda Work around for ppcle compiler bug Change-Id: Ie284cf24b2271215be1b9dc95b485fd15000e32b llvm-svn: 351054
2019-01-14 19:55:24 +08:00
; GCN: image_load v[0:1], v[0:1], s[0:7] dmask:0x3 unorm a16
define amdgpu_ps <4 x float> @load.v2f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
main_body:
%x = extractelement <2 x i16> %coords_lo, i32 0
%y = extractelement <2 x i16> %coords_lo, i32 1
%z = extractelement <2 x i16> %coords_hi, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 3, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v3f32.3d:
; GCN: image_load v[0:2], v[0:1], s[0:7] dmask:0x7 unorm a16
define amdgpu_ps <4 x float> @load.v3f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
main_body:
%x = extractelement <2 x i16> %coords_lo, i32 0
%y = extractelement <2 x i16> %coords_lo, i32 1
%z = extractelement <2 x i16> %coords_hi, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 7, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
; GCN-LABEL: {{^}}load.v4f32.3d:
; GCN: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm a16
define amdgpu_ps <4 x float> @load.v4f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
main_body:
%x = extractelement <2 x i16> %coords_lo, i32 0
%y = extractelement <2 x i16> %coords_lo, i32 1
%z = extractelement <2 x i16> %coords_hi, i32 0
%v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 15, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
ret <4 x float> %v
}
declare <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32, i16, <8 x i32>, i32, i32) #2
declare <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32, i16, i16, <8 x i32>, i32, i32) #2
declare <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32, i16, i16, i16, <8 x i32>, i32, i32) #2
attributes #0 = { nounwind }
attributes #1 = { nounwind readonly }