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/spill-offset-calculation.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

297 lines
13 KiB
LLVM
Raw Normal View History

[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s
; Test that the VGPR spiller correctly switches to SGPR offsets when the
; instruction offset field would overflow, and that it accounts for memory
; swizzling.
; CHECK-LABEL: test_inst_offset_kernel
define amdgpu_kernel void @test_inst_offset_kernel() {
entry:
; Occupy 4092 bytes of scratch, so the offset of the spill of %a just fits in
; the instruction offset field.
%alloca = alloca i8, i32 4088, align 4, addrspace(5)
%buf = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%aptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4092 ; 4-byte Folded Spill
%a = load volatile i32, i32 addrspace(5)* %aptr
; Force %a to spill.
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
%outptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
store volatile i32 %a, i32 addrspace(5)* %outptr
ret void
}
; CHECK-LABEL: test_sgpr_offset_kernel
define amdgpu_kernel void @test_sgpr_offset_kernel() {
entry:
; Occupy 4096 bytes of scratch, so the offset of the spill of %a does not
; fit in the instruction, and has to live in the SGPR offset.
%alloca = alloca i8, i32 4092, align 4, addrspace(5)
%buf = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%aptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
; 0x40000 / 64 = 4096 (for wave64)
AMDGPU: Enable the scavenger for large frames Another test is needed for the case where the scavenge fail, but there's another issue with that which needs an additional fix. llvm-svn: 357093
2019-03-28 01:14:32 +08:00
; CHECK: s_add_u32 s6, s7, 0x40000
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s6 ; 4-byte Folded Spill
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%a = load volatile i32, i32 addrspace(5)* %aptr
; Force %a to spill
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
%outptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
store volatile i32 %a, i32 addrspace(5)* %outptr
ret void
}
AMDGPU: Add testcase I meant to merge into r357093 llvm-svn: 357097
2019-03-28 01:31:26 +08:00
; CHECK-LABEL: test_sgpr_offset_kernel_scavenge_fail
define amdgpu_kernel void @test_sgpr_offset_kernel_scavenge_fail() #1 {
entry:
; Occupy 4096 bytes of scratch, so the offset of the spill of %a does not
; fit in the instruction, and has to live in the SGPR offset.
%alloca = alloca i8, i32 4092, align 4, addrspace(5)
%buf = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%aptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
; 0x40000 / 64 = 4096 (for wave64)
%a = load volatile i32, i32 addrspace(5)* %aptr
%asm = call { i32, i32, i32, i32, i32, i32, i32, i32 } asm sideeffect "", "=s,=s,=s,=s,=s,=s,=s,=s"()
%asm0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 0
%asm1 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 1
%asm2 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 2
%asm3 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 3
%asm4 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 4
%asm5 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 5
%asm6 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 6
%asm7 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 7
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}"() #0
; CHECK: s_add_u32 s7, s7, 0x40000
; CHECK: buffer_load_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s7 ; 4-byte Folded Reload
; CHECK: s_sub_u32 s7, s7, 0x40000
; Force %a to spill with no free SGPRs
call void asm sideeffect "", "s,s,s,s,s,s,s,s,v"(i32 %asm0, i32 %asm1, i32 %asm2, i32 %asm3, i32 %asm4, i32 %asm5, i32 %asm6, i32 %asm7, i32 %a)
ret void
}
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
; CHECK-LABEL: test_sgpr_offset_function_scavenge_fail
define void @test_sgpr_offset_function_scavenge_fail() #2 {
entry:
; Occupy 4096 bytes of scratch, so the offset of the spill of %a does not
; fit in the instruction, and has to live in the SGPR offset.
%alloca = alloca i8, i32 4096, align 4, addrspace(5)
%buf = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%aptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
%asm.0 = call { i32, i32, i32, i32, i32, i32, i32, i32 } asm sideeffect "", "=s,=s,=s,=s,=s,=s,=s,=s"()
%asm0.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 0
%asm1.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 1
%asm2.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 2
%asm3.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 3
%asm4.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 4
%asm5.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 5
%asm6.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 6
%asm7.0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm.0, 7
; 0x40000 / 64 = 4096 (for wave64)
%a = load volatile i32, i32 addrspace(5)* %aptr
; CHECK: s_add_u32 s32, s32, 0x40000
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s32 ; 4-byte Folded Spill
; CHECK: s_sub_u32 s32, s32, 0x40000
call void asm sideeffect "", "s,s,s,s,s,s,s,s,v"(i32 %asm0.0, i32 %asm1.0, i32 %asm2.0, i32 %asm3.0, i32 %asm4.0, i32 %asm5.0, i32 %asm6.0, i32 %asm7.0, i32 %a)
%asm = call { i32, i32, i32, i32, i32, i32, i32, i32 } asm sideeffect "", "=s,=s,=s,=s,=s,=s,=s,=s"()
%asm0 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 0
%asm1 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 1
%asm2 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 2
%asm3 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 3
%asm4 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 4
%asm5 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 5
%asm6 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 6
%asm7 = extractvalue { i32, i32, i32, i32, i32, i32, i32, i32 } %asm, 7
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}"() #0
; CHECK: s_add_u32 s32, s32, 0x40000
; CHECK: buffer_load_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s32 ; 4-byte Folded Reload
; CHECK: s_sub_u32 s32, s32, 0x40000
; Force %a to spill with no free SGPRs
call void asm sideeffect "", "s,s,s,s,s,s,s,s,v"(i32 %asm0, i32 %asm1, i32 %asm2, i32 %asm3, i32 %asm4, i32 %asm5, i32 %asm6, i32 %asm7, i32 %a)
ret void
}
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
; CHECK-LABEL: test_sgpr_offset_subregs_kernel
define amdgpu_kernel void @test_sgpr_offset_subregs_kernel() {
entry:
; Occupy 4088 bytes of scratch, so that the spill of the last subreg of %a
; still fits below offset 4096 (4088 + 8 - 4 = 4092), and can be placed in
; the instruction offset field.
%alloca = alloca i8, i32 4084, align 4, addrspace(5)
%bufv1 = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%bufv2 = bitcast i8 addrspace(5)* %alloca to <2 x i32> addrspace(5)*
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4088 ; 4-byte Folded Spill
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4092 ; 4-byte Folded Spill
%aptr = getelementptr <2 x i32>, <2 x i32> addrspace(5)* %bufv2, i32 1
%a = load volatile <2 x i32>, <2 x i32> addrspace(5)* %aptr
; Force %a to spill.
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
; Ensure the alloca sticks around.
%bptr = getelementptr i32, i32 addrspace(5)* %bufv1, i32 1
%b = load volatile i32, i32 addrspace(5)* %bptr
; Ensure the spill is of the full super-reg.
call void asm sideeffect "; $0", "r"(<2 x i32> %a)
ret void
}
; CHECK-LABEL: test_inst_offset_subregs_kernel
define amdgpu_kernel void @test_inst_offset_subregs_kernel() {
entry:
; Occupy 4092 bytes of scratch, so that the spill of the last subreg of %a
; does not fit below offset 4096 (4092 + 8 - 4 = 4096), and has to live
; in the SGPR offset.
%alloca = alloca i8, i32 4088, align 4, addrspace(5)
%bufv1 = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%bufv2 = bitcast i8 addrspace(5)* %alloca to <2 x i32> addrspace(5)*
; 0x3ff00 / 64 = 4092 (for wave64)
AMDGPU: Enable the scavenger for large frames Another test is needed for the case where the scavenge fail, but there's another issue with that which needs an additional fix. llvm-svn: 357093
2019-03-28 01:14:32 +08:00
; CHECK: s_add_u32 s6, s7, 0x3ff00
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s6 ; 4-byte Folded Spill
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s6 offset:4 ; 4-byte Folded Spill
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%aptr = getelementptr <2 x i32>, <2 x i32> addrspace(5)* %bufv2, i32 1
%a = load volatile <2 x i32>, <2 x i32> addrspace(5)* %aptr
; Force %a to spill.
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
; Ensure the alloca sticks around.
%bptr = getelementptr i32, i32 addrspace(5)* %bufv1, i32 1
%b = load volatile i32, i32 addrspace(5)* %bptr
; Ensure the spill is of the full super-reg.
call void asm sideeffect "; $0", "r"(<2 x i32> %a)
ret void
}
; CHECK-LABEL: test_inst_offset_function
define void @test_inst_offset_function() {
entry:
; Occupy 4092 bytes of scratch, so the offset of the spill of %a just fits in
; the instruction offset field.
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
%alloca = alloca i8, i32 4092, align 4, addrspace(5)
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%buf = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%aptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4092 ; 4-byte Folded Spill
%a = load volatile i32, i32 addrspace(5)* %aptr
; Force %a to spill.
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
%outptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
store volatile i32 %a, i32 addrspace(5)* %outptr
ret void
}
; CHECK-LABEL: test_sgpr_offset_function
define void @test_sgpr_offset_function() {
entry:
; Occupy 4096 bytes of scratch, so the offset of the spill of %a does not
; fit in the instruction, and has to live in the SGPR offset.
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
%alloca = alloca i8, i32 4096, align 4, addrspace(5)
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%buf = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%aptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
; 0x40000 / 64 = 4096 (for wave64)
AMDGPU: Always use s33 for global scratch wave offset Every called function could possibly need this to calculate the absolute address of stack objectst, and this avoids inserting a copy around every call site in the kernel. It's also somewhat cleaner to keep this in a callee saved SGPR. llvm-svn: 363990
2019-06-21 05:58:24 +08:00
; CHECK: s_add_u32 s4, s32, 0x40000
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s4 ; 4-byte Folded Spill
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%a = load volatile i32, i32 addrspace(5)* %aptr
; Force %a to spill
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
%outptr = getelementptr i32, i32 addrspace(5)* %buf, i32 1
store volatile i32 %a, i32 addrspace(5)* %outptr
ret void
}
; CHECK-LABEL: test_sgpr_offset_subregs_function
define void @test_sgpr_offset_subregs_function() {
entry:
; Occupy 4088 bytes of scratch, so that the spill of the last subreg of %a
; still fits below offset 4096 (4088 + 8 - 4 = 4092), and can be placed in
; the instruction offset field.
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
%alloca = alloca i8, i32 4088, align 4, addrspace(5)
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%bufv1 = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%bufv2 = bitcast i8 addrspace(5)* %alloca to <2 x i32> addrspace(5)*
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4088 ; 4-byte Folded Spill
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4092 ; 4-byte Folded Spill
%aptr = getelementptr <2 x i32>, <2 x i32> addrspace(5)* %bufv2, i32 1
%a = load volatile <2 x i32>, <2 x i32> addrspace(5)* %aptr
; Force %a to spill.
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
; Ensure the alloca sticks around.
%bptr = getelementptr i32, i32 addrspace(5)* %bufv1, i32 1
%b = load volatile i32, i32 addrspace(5)* %bptr
; Ensure the spill is of the full super-reg.
call void asm sideeffect "; $0", "r"(<2 x i32> %a)
ret void
}
; CHECK-LABEL: test_inst_offset_subregs_function
define void @test_inst_offset_subregs_function() {
entry:
; Occupy 4092 bytes of scratch, so that the spill of the last subreg of %a
; does not fit below offset 4096 (4092 + 8 - 4 = 4096), and has to live
; in the SGPR offset.
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
%alloca = alloca i8, i32 4092, align 4, addrspace(5)
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%bufv1 = bitcast i8 addrspace(5)* %alloca to i32 addrspace(5)*
%bufv2 = bitcast i8 addrspace(5)* %alloca to <2 x i32> addrspace(5)*
; 0x3ff00 / 64 = 4092 (for wave64)
AMDGPU: Always use s33 for global scratch wave offset Every called function could possibly need this to calculate the absolute address of stack objectst, and this avoids inserting a copy around every call site in the kernel. It's also somewhat cleaner to keep this in a callee saved SGPR. llvm-svn: 363990
2019-06-21 05:58:24 +08:00
; CHECK: s_add_u32 s4, s32, 0x3ff00
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s4 ; 4-byte Folded Spill
; CHECK: buffer_store_dword v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s4 offset:4 ; 4-byte Folded Spill
[AMDGPU] Fix VGPR spills where offset doesn't fit in 12 bits Scale the offset of VGPR spills by the wave size when it cannot fit in the 12-bit offset immediate field and so is added to the soffset SGPR. This accounts for hardware swizzling of scratch memory. Differential Revision: https://reviews.llvm.org/D49448 llvm-svn: 338060
2018-07-27 03:47:51 +08:00
%aptr = getelementptr <2 x i32>, <2 x i32> addrspace(5)* %bufv2, i32 1
%a = load volatile <2 x i32>, <2 x i32> addrspace(5)* %aptr
; Force %a to spill.
call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}" ()
; Ensure the alloca sticks around.
%bptr = getelementptr i32, i32 addrspace(5)* %bufv1, i32 1
%b = load volatile i32, i32 addrspace(5)* %bptr
; Ensure the spill is of the full super-reg.
call void asm sideeffect "; $0", "r"(<2 x i32> %a)
ret void
}
AMDGPU: Add testcase I meant to merge into r357093 llvm-svn: 357097
2019-03-28 01:31:26 +08:00
attributes #0 = { nounwind }
AMDGPU: Make s34 the FP register Make the FP register callee saved. This is tricky because now the FP needs to be spilled in the prolog relative to the incoming SP register, rather than the frame register used throughout the rest of the function. I don't like how this bypassess the standard mechanism for CSR spills just to get the correct insert point. I may look for a better solution, since all CSR VGPRs may also need to have all lanes activated. Another option might be to make getFrameIndexReference change the base register if the frame index is a CSR, and then try to figure out the right insertion point in emitProlog. If there is a free VGPR lane available for SGPR spilling, try to use it for the FP. If that would require intrtoducing a new VGPR spill, try to use a free call clobbered SGPR. Only fallback to introducing a new VGPR spill as a last resort. This also doesn't attempt to handle SGPR spilling with scalar stores. llvm-svn: 365372
2019-07-09 03:03:38 +08:00
attributes #1 = { nounwind "amdgpu-num-sgpr"="17" "amdgpu-num-vgpr"="8" }
attributes #2 = { nounwind "amdgpu-num-sgpr"="14" "amdgpu-num-vgpr"="8" }