Skip to content

Commit

Permalink
[AMDGPU][True16][CodeGen] saddsat/ssubsat sdag for true16 format (#11…
Browse files Browse the repository at this point in the history 
…8245)

saddsat and ssubsat SDAG codeGen pattern for True16 format
  • Loading branch information
@broxigarchen
broxigarchen authored Dec 2, 2024
1 parent de6d0d2 commit 94316dd
Show file tree
Hide file tree
Showing 3 changed files with 96 additions and 28 deletions.
4 changes: 4 additions & 0 deletions llvm/lib/Target/AMDGPU/VOP3Instructions.td
View file
Original file line number Diff line number Diff line change
Expand Up @@ -763,6 +763,10 @@ let SubtargetPredicate = isGFX9Plus, True16Predicate = NotHasTrue16BitInsts in {
def : OpSelBinOpClampPat<saddsat, V_ADD_I16_e64>;
def : OpSelBinOpClampPat<ssubsat, V_SUB_I16_e64>;
} // End SubtargetPredicate = isGFX9Plus, True16Predicate = NotHasTrue16BitInsts
let True16Predicate = UseRealTrue16Insts in {
def : OpSelBinOpClampPat<saddsat, V_ADD_I16_t16_e64>;
def : OpSelBinOpClampPat<ssubsat, V_SUB_I16_t16_e64>;
} // End True16Predicate = UseRealTrue16Insts
let True16Predicate = UseFakeTrue16Insts in {
def : OpSelBinOpClampPat<saddsat, V_ADD_I16_fake16_e64>;
def : OpSelBinOpClampPat<ssubsat, V_SUB_I16_fake16_e64>;
Expand Down
60 changes: 46 additions & 14 deletions llvm/test/CodeGen/AMDGPU/saddsat.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,8 @@
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=fiji < %s | FileCheck --check-prefix=GFX8 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck --check-prefix=GFX9 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX10 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX11 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -mattr=+real-true16 -amdgpu-enable-delay-alu=0 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX11,GFX11-TRUE16 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -mattr=-real-true16 -amdgpu-enable-delay-alu=0 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX11,GFX11-FAKE16 %s

define i8 @v_saddsat_i8(i8 %lhs, i8 %rhs) {
; GFX6-LABEL: v_saddsat_i8:
Expand Down Expand Up @@ -34,14 +35,32 @@ define i8 @v_saddsat_i8(i8 %lhs, i8 %rhs) {
; GFX9-NEXT: v_ashrrev_i16_e32 v0, 8, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_saddsat_i8:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_lshlrev_b16 v1, 8, v1
; GFX10PLUS-NEXT: v_lshlrev_b16 v0, 8, v0
; GFX10PLUS-NEXT: v_add_nc_i16 v0, v0, v1 clamp
; GFX10PLUS-NEXT: v_ashrrev_i16 v0, 8, v0
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
; GFX10-LABEL: v_saddsat_i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
; GFX10-NEXT: v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT: v_add_nc_i16 v0, v0, v1 clamp
; GFX10-NEXT: v_ashrrev_i16 v0, 8, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: v_saddsat_i8:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v1.l
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GFX11-TRUE16-NEXT: v_add_nc_i16 v0.l, v0.l, v0.h clamp
; GFX11-TRUE16-NEXT: v_ashrrev_i16 v0.l, 8, v0.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_saddsat_i8:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v1, 8, v1
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v0, 8, v0
; GFX11-FAKE16-NEXT: v_add_nc_i16 v0, v0, v1 clamp
; GFX11-FAKE16-NEXT: v_ashrrev_i16 v0, 8, v0
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%result = call i8 @llvm.sadd.sat.i8(i8 %lhs, i8 %rhs)
ret i8 %result
}
Expand Down Expand Up @@ -76,11 +95,24 @@ define i16 @v_saddsat_i16(i16 %lhs, i16 %rhs) {
; GFX9-NEXT: v_add_i16 v0, v0, v1 clamp
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_saddsat_i16:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_add_nc_i16 v0, v0, v1 clamp
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
; GFX10-LABEL: v_saddsat_i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_add_nc_i16 v0, v0, v1 clamp
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: v_saddsat_i16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.l
; GFX11-TRUE16-NEXT: v_add_nc_i16 v0.l, v0.l, v0.h clamp
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_saddsat_i16:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: v_add_nc_i16 v0, v0, v1 clamp
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%result = call i16 @llvm.sadd.sat.i16(i16 %lhs, i16 %rhs)
ret i16 %result
}
Expand Down
60 changes: 46 additions & 14 deletions llvm/test/CodeGen/AMDGPU/ssubsat.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,8 @@
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=fiji < %s | FileCheck --check-prefix=GFX8 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck --check-prefix=GFX9 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX10 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX11 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -mattr=+real-true16 -amdgpu-enable-delay-alu=0 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX11,GFX11-TRUE16 %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -mattr=-real-true16 -amdgpu-enable-delay-alu=0 < %s | FileCheck --check-prefixes=GFX10PLUS,GFX11,GFX11-FAKE16 %s

define i8 @v_ssubsat_i8(i8 %lhs, i8 %rhs) {
; GFX6-LABEL: v_ssubsat_i8:
Expand Down Expand Up @@ -34,14 +35,32 @@ define i8 @v_ssubsat_i8(i8 %lhs, i8 %rhs) {
; GFX9-NEXT: v_ashrrev_i16_e32 v0, 8, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_ssubsat_i8:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_lshlrev_b16 v1, 8, v1
; GFX10PLUS-NEXT: v_lshlrev_b16 v0, 8, v0
; GFX10PLUS-NEXT: v_sub_nc_i16 v0, v0, v1 clamp
; GFX10PLUS-NEXT: v_ashrrev_i16 v0, 8, v0
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
; GFX10-LABEL: v_ssubsat_i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
; GFX10-NEXT: v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT: v_sub_nc_i16 v0, v0, v1 clamp
; GFX10-NEXT: v_ashrrev_i16 v0, 8, v0
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: v_ssubsat_i8:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v1.l
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GFX11-TRUE16-NEXT: v_sub_nc_i16 v0.l, v0.l, v0.h clamp
; GFX11-TRUE16-NEXT: v_ashrrev_i16 v0.l, 8, v0.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_ssubsat_i8:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v1, 8, v1
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v0, 8, v0
; GFX11-FAKE16-NEXT: v_sub_nc_i16 v0, v0, v1 clamp
; GFX11-FAKE16-NEXT: v_ashrrev_i16 v0, 8, v0
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%result = call i8 @llvm.ssub.sat.i8(i8 %lhs, i8 %rhs)
ret i8 %result
}
Expand Down Expand Up @@ -76,11 +95,24 @@ define i16 @v_ssubsat_i16(i16 %lhs, i16 %rhs) {
; GFX9-NEXT: v_sub_i16 v0, v0, v1 clamp
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_ssubsat_i16:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_sub_nc_i16 v0, v0, v1 clamp
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
; GFX10-LABEL: v_ssubsat_i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_sub_nc_i16 v0, v0, v1 clamp
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: v_ssubsat_i16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.l
; GFX11-TRUE16-NEXT: v_sub_nc_i16 v0.l, v0.l, v0.h clamp
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_ssubsat_i16:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: v_sub_nc_i16 v0, v0, v1 clamp
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%result = call i16 @llvm.ssub.sat.i16(i16 %lhs, i16 %rhs)
ret i16 %result
}
Expand Down

0 comments on commit 94316dd

Please sign in to comment.