2017-04-06 23:03:28 +08:00
|
|
|
; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole=0 -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=NOSDWA -check-prefix=GCN %s
|
|
|
|
|
; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=SDWA -check-prefix=GCN %s
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}add_shr_i32:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_add_i32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v[[DST]]
|
|
|
|
|
; NOSDWA-NOT: v_add_i32_sdwa
|
|
|
|
|
|
|
|
|
|
; SDWA: v_add_i32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @add_shr_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
%a = load i32, i32 addrspace(1)* %in, align 4
|
|
|
|
|
%shr = lshr i32 %a, 16
|
|
|
|
|
%add = add i32 %a, %shr
|
|
|
|
|
store i32 %add, i32 addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}sub_shr_i32:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_subrev_i32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v[[DST]]
|
|
|
|
|
; NOSDWA-NOT: v_subrev_i32_sdwa
|
|
|
|
|
|
|
|
|
|
; SDWA: v_subrev_i32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @sub_shr_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
%a = load i32, i32 addrspace(1)* %in, align 4
|
|
|
|
|
%shr = lshr i32 %a, 16
|
|
|
|
|
%sub = sub i32 %shr, %a
|
|
|
|
|
store i32 %sub, i32 addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_shr_i32:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v[[DST1]], v[[DST0]]
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
|
|
|
|
; SDWA: v_mul_u32_u24_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_shr_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in1, i32 addrspace(1)* %in2) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
%a = load i32, i32 addrspace(1)* %in1, align 4
|
|
|
|
|
%b = load i32, i32 addrspace(1)* %in2, align 4
|
|
|
|
|
%shra = lshr i32 %a, 16
|
|
|
|
|
%shrb = lshr i32 %b, 16
|
|
|
|
|
%mul = mul i32 %shra, %shrb
|
|
|
|
|
store i32 %mul, i32 addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_i16:
|
|
|
|
|
; NOSDWA: v_mul_lo_i32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA: v_mul_lo_i32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_i16(i16 addrspace(1)* %out, i16 addrspace(1)* %ina, i16 addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load i16, i16 addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load i16, i16 addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul i16 %a, %b
|
|
|
|
|
store i16 %mul, i16 addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v2i16:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v[[DST_MUL:[0-9]+]], v[[DST1]], v[[DST0]]
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MUL]]
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v[[DST_SHL]], v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL_LO:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL_HI:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
2017-03-31 19:42:43 +08:00
|
|
|
; SDWA: v_or_b32_sdwa v{{[0-9]+}}, v[[DST_MUL_HI]], v[[DST_MUL_LO]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %ina, <2 x i16> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <2 x i16>, <2 x i16> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <2 x i16>, <2 x i16> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul <2 x i16> %a, %b
|
|
|
|
|
store <2 x i16> %mul, <2 x i16> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v4i16:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL0:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL1:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL2:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL3:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_or_b32_sdwa v{{[0-9]+}}, v[[DST_MUL3]], v[[DST_MUL2]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_or_b32_sdwa v{{[0-9]+}}, v[[DST_MUL1]], v[[DST_MUL0]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v4i16(<4 x i16> addrspace(1)* %out, <4 x i16> addrspace(1)* %ina, <4 x i16> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <4 x i16>, <4 x i16> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <4 x i16>, <4 x i16> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul <4 x i16> %a, %b
|
|
|
|
|
store <4 x i16> %mul, <4 x i16> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v8i16:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL0:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL1:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL2:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL3:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL4:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL5:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL6:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa v[[DST_MUL7:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_or_b32_sdwa v{{[0-9]+}}, v[[DST_MUL7]], v[[DST_MUL6]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_or_b32_sdwa v{{[0-9]+}}, v[[DST_MUL5]], v[[DST_MUL4]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_or_b32_sdwa v{{[0-9]+}}, v[[DST_MUL3]], v[[DST_MUL2]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_or_b32_sdwa v{{[0-9]+}}, v[[DST_MUL1]], v[[DST_MUL0]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v8i16(<8 x i16> addrspace(1)* %out, <8 x i16> addrspace(1)* %ina, <8 x i16> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <8 x i16>, <8 x i16> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <8 x i16>, <8 x i16> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul <8 x i16> %a, %b
|
|
|
|
|
store <8 x i16> %mul, <8 x i16> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_half:
|
|
|
|
|
; NOSDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_f16_sdwa
|
|
|
|
|
; SDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA-NOT: v_mul_f16_sdwa
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_half(half addrspace(1)* %out, half addrspace(1)* %ina, half addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load half, half addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load half, half addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = fmul half %a, %b
|
|
|
|
|
store half %mul, half addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v2half:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_f16_e32 v[[DST_MUL:[0-9]+]], v[[DST1]], v[[DST0]]
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MUL]]
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v[[DST_SHL]], v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_f16_sdwa
|
|
|
|
|
|
2017-04-01 03:53:03 +08:00
|
|
|
; SDWA-DAG: v_mul_f16_sdwa v[[DST_MUL_HI:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_f16_e32 v[[DST_MUL_LO:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL_HI]], v[[DST_MUL_LO]]
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v2half(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %ina, <2 x half> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <2 x half>, <2 x half> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <2 x half>, <2 x half> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = fmul <2 x half> %a, %b
|
|
|
|
|
store <2 x half> %mul, <2 x half> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v4half:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_f16_sdwa
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; SDWA-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v4half(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %ina, <4 x half> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <4 x half>, <4 x half> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <4 x half>, <4 x half> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = fmul <4 x half> %a, %b
|
|
|
|
|
store <4 x half> %mul, <4 x half> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v8half:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_f16_sdwa
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; SDWA-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v8half(<8 x half> addrspace(1)* %out, <8 x half> addrspace(1)* %ina, <8 x half> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <8 x half>, <8 x half> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <8 x half>, <8 x half> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = fmul <8 x half> %a, %b
|
|
|
|
|
store <8 x half> %mul, <8 x half> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_i8:
|
|
|
|
|
; NOSDWA: v_mul_lo_i32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA: v_mul_lo_i32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; SDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_i8(i8 addrspace(1)* %out, i8 addrspace(1)* %ina, i8 addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load i8, i8 addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load i8, i8 addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul i8 %a, %b
|
|
|
|
|
store i8 %mul, i8 addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v2i8:
|
|
|
|
|
; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
|
|
|
|
; SDWA: v_mul_u32_u24_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v2i8(<2 x i8> addrspace(1)* %out, <2 x i8> addrspace(1)* %ina, <2 x i8> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <2 x i8>, <2 x i8> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <2 x i8>, <2 x i8> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul <2 x i8> %a, %b
|
|
|
|
|
store <2 x i8> %mul, <2 x i8> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v4i8:
|
|
|
|
|
; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %ina, <4 x i8> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <4 x i8>, <4 x i8> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <4 x i8>, <4 x i8> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul <4 x i8> %a, %b
|
|
|
|
|
store <4 x i8> %mul, <4 x i8> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mul_v8i8:
|
|
|
|
|
; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-DAG: v_mul_u32_u24_sdwa
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mul_v8i8(<8 x i8> addrspace(1)* %out, <8 x i8> addrspace(1)* %ina, <8 x i8> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <8 x i8>, <8 x i8> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <8 x i8>, <8 x i8> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul <8 x i8> %a, %b
|
|
|
|
|
store <8 x i8> %mul, <8 x i8> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
2017-04-06 23:03:28 +08:00
|
|
|
; GCN-LABEL: {{^}}sitofp_v2i16_to_v2f16:
|
|
|
|
|
; NOSDWA-DAG: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 16
|
|
|
|
|
; NOSDWA-DAG: v_ashrrev_i32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-DAG: v_cvt_f32_i32_e32 v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-DAG: v_cvt_f32_i32_e32 v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_cvt_f32_i32_sdwa
|
|
|
|
|
|
|
|
|
|
; SDWA-DAG: v_cvt_f32_i32_sdwa v{{[0-9]+}}, sext(v{{[0-9]+}}) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0
|
|
|
|
|
; SDWA-DAG: v_cvt_f32_i32_sdwa v{{[0-9]+}}, sext(v{{[0-9]+}}) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
|
|
|
|
|
|
|
|
|
|
define amdgpu_kernel void @sitofp_v2i16_to_v2f16(
|
|
|
|
|
<2 x half> addrspace(1)* %r,
|
|
|
|
|
<2 x i16> addrspace(1)* %a) {
|
|
|
|
|
entry:
|
|
|
|
|
%a.val = load <2 x i16>, <2 x i16> addrspace(1)* %a
|
|
|
|
|
%r.val = sitofp <2 x i16> %a.val to <2 x half>
|
|
|
|
|
store <2 x half> %r.val, <2 x half> addrspace(1)* %r
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}mac_v2half:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mac_f16_e32 v[[DST_MAC:[0-9]+]], v[[DST1]], v[[DST0]]
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MAC]]
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v[[DST_SHL]], v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mac_f16_sdwa
|
|
|
|
|
|
|
|
|
|
; SDWA: v_mac_f16_sdwa v[[DST_MAC:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
; SDWA: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MAC]]
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mac_v2half(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %ina, <2 x half> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <2 x half>, <2 x half> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <2 x half>, <2 x half> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = fmul <2 x half> %a, %b
|
|
|
|
|
%mac = fadd <2 x half> %mul, %b
|
|
|
|
|
store <2 x half> %mac, <2 x half> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; GCN-LABEL: {{^}}immediate_mul_v2i16:
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
; SDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @immediate_mul_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <2 x i16>, <2 x i16> addrspace(1)* %in, align 4
|
|
|
|
|
%mul = mul <2 x i16> %a, <i16 123, i16 321>
|
|
|
|
|
store <2 x i16> %mul, <2 x i16> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Double use of same src - should not convert it
|
|
|
|
|
; GCN-LABEL: {{^}}mulmul_v2i16:
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
|
|
|
|
|
; NOSDWA-NOT: v_mul_u32_u24_sdwa
|
|
|
|
|
|
|
|
|
|
; SDWA: v_mul_u32_u24_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
|
|
|
|
2017-03-22 05:39:51 +08:00
|
|
|
define amdgpu_kernel void @mulmul_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %ina, <2 x i16> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
|
|
|
|
%a = load <2 x i16>, <2 x i16> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <2 x i16>, <2 x i16> addrspace(1)* %inb, align 4
|
|
|
|
|
%mul = mul <2 x i16> %a, %b
|
|
|
|
|
%mul2 = mul <2 x i16> %mul, %b
|
|
|
|
|
store <2 x i16> %mul2, <2 x i16> addrspace(1)* %out, align 4
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
2017-04-12 17:36:05 +08:00
|
|
|
; GCN-LABEL: {{^}}add_bb_v2i16:
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
; NOSDWA-NOT: v_add_i32_sdwa
|
|
|
|
|
|
2017-04-12 17:36:05 +08:00
|
|
|
; SDWA: v_add_i32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
|
|
|
|
|
|
|
|
|
|
define amdgpu_kernel void @add_bb_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %ina, <2 x i16> addrspace(1)* %inb) {
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
entry:
|
2017-04-12 17:36:05 +08:00
|
|
|
%a = load <2 x i16>, <2 x i16> addrspace(1)* %ina, align 4
|
|
|
|
|
%b = load <2 x i16>, <2 x i16> addrspace(1)* %inb, align 4
|
|
|
|
|
br label %add_label
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
add_label:
|
2017-04-12 17:36:05 +08:00
|
|
|
%add = add <2 x i16> %a, %b
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
br label %store_label
|
|
|
|
|
store_label:
|
2017-04-12 17:36:05 +08:00
|
|
|
store <2 x i16> %add, <2 x i16> addrspace(1)* %out, align 4
|
[ADMGPU] SDWA peephole optimization pass.
Summary:
First iteration of SDWA peephole.
This pass tries to combine several instruction into one SDWA instruction. E.g. it converts:
'''
V_LSHRREV_B32_e32 %vreg0, 16, %vreg1
V_ADD_I32_e32 %vreg2, %vreg0, %vreg3
V_LSHLREV_B32_e32 %vreg4, 16, %vreg2
'''
Into:
'''
V_ADD_I32_sdwa %vreg4, %vreg1, %vreg3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
'''
Pass structure:
1. Iterate over machine instruction in basic block and try to apply "SDWA patterns" to each of them. SDWA patterns match machine instruction into either source or destination SDWA operand. E.g. ''' V_LSHRREV_B32_e32 %vreg0, 16, %vreg1''' is matched to source SDWA operand '''%vreg1 src_sel:WORD_1'''.
2. Iterate over found SDWA operands and find instruction that could be potentially coverted into SDWA. E.g. for source SDWA operand potential instruction are all instruction in this basic block that uses '''%vreg0'''
3. Iterate over all potential instructions and check if they can be converted into SDWA.
4. Convert instructions to SDWA.
This review contains basic implementation of SDWA peephole pass. This pass requires additional testing fot both correctness and performance (no performance testing done).
There are several ways this pass can be improved:
1. Make this pass work on whole function not only basic block. As I can see this can be done right now without changes to pass.
2. Introduce more SDWA patterns
3. Introduce mnemonics to limit when SDWA patterns should apply
Reviewers: vpykhtin, alex-t, arsenm, rampitec
Subscribers: wdng, nhaehnle, mgorny
Differential Revision: https://reviews.llvm.org/D30038
llvm-svn: 298365
2017-03-21 20:51:34 +08:00
|
|
|
ret void
|
2017-04-12 17:36:05 +08:00
|
|
|
}
|