1 ; RUN: opt -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mattr=+half-rate-64-ops < %s | FileCheck -check-prefix=FASTF64 -check-prefix=ALL %s 2 ; RUN: opt -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mattr=-half-rate-64-ops < %s | FileCheck -check-prefix=SLOWF64 -check-prefix=ALL %s 3 4 ; ALL: 'fsub_f32' 5 ; ALL: estimated cost of 1 for {{.*}} fsub float 6 define void @fsub_f32(float addrspace(1)* %out, float addrspace(1)* %vaddr, float %b) #0 { 7 %vec = load float, float addrspace(1)* %vaddr 8 %add = fsub float %vec, %b 9 store float %add, float addrspace(1)* %out 10 ret void 11 } 12 13 ; ALL: 'fsub_v2f32' 14 ; ALL: estimated cost of 2 for {{.*}} fsub <2 x float> 15 define void @fsub_v2f32(<2 x float> addrspace(1)* %out, <2 x float> addrspace(1)* %vaddr, <2 x float> %b) #0 { 16 %vec = load <2 x float>, <2 x float> addrspace(1)* %vaddr 17 %add = fsub <2 x float> %vec, %b 18 store <2 x float> %add, <2 x float> addrspace(1)* %out 19 ret void 20 } 21 22 ; ALL: 'fsub_v3f32' 23 ; ALL: estimated cost of 3 for {{.*}} fsub <3 x float> 24 define void @fsub_v3f32(<3 x float> addrspace(1)* %out, <3 x float> addrspace(1)* %vaddr, <3 x float> %b) #0 { 25 %vec = load <3 x float>, <3 x float> addrspace(1)* %vaddr 26 %add = fsub <3 x float> %vec, %b 27 store <3 x float> %add, <3 x float> addrspace(1)* %out 28 ret void 29 } 30 31 ; ALL: 'fsub_f64' 32 ; FASTF64: estimated cost of 2 for {{.*}} fsub double 33 ; SLOWF64: estimated cost of 3 for {{.*}} fsub double 34 define void @fsub_f64(double addrspace(1)* %out, double addrspace(1)* %vaddr, double %b) #0 { 35 %vec = load double, double addrspace(1)* %vaddr 36 %add = fsub double %vec, %b 37 store double %add, double addrspace(1)* %out 38 ret void 39 } 40 41 ; ALL: 'fsub_v2f64' 42 ; FASTF64: estimated cost of 4 for {{.*}} fsub <2 x double> 43 ; SLOWF64: estimated cost of 6 for {{.*}} fsub <2 x double> 44 define void @fsub_v2f64(<2 x double> addrspace(1)* %out, <2 x double> addrspace(1)* %vaddr, <2 x double> %b) #0 { 45 %vec = load <2 x double>, <2 x double> addrspace(1)* %vaddr 46 %add = fsub <2 x double> %vec, %b 47 store <2 x double> %add, <2 x double> addrspace(1)* %out 48 ret void 49 } 50 51 ; ALL: 'fsub_v3f64' 52 ; FASTF64: estimated cost of 6 for {{.*}} fsub <3 x double> 53 ; SLOWF64: estimated cost of 9 for {{.*}} fsub <3 x double> 54 define void @fsub_v3f64(<3 x double> addrspace(1)* %out, <3 x double> addrspace(1)* %vaddr, <3 x double> %b) #0 { 55 %vec = load <3 x double>, <3 x double> addrspace(1)* %vaddr 56 %add = fsub <3 x double> %vec, %b 57 store <3 x double> %add, <3 x double> addrspace(1)* %out 58 ret void 59 } 60 61 ; ALL: 'fsub_f16' 62 ; ALL: estimated cost of 1 for {{.*}} fsub half 63 define void @fsub_f16(half addrspace(1)* %out, half addrspace(1)* %vaddr, half %b) #0 { 64 %vec = load half, half addrspace(1)* %vaddr 65 %add = fsub half %vec, %b 66 store half %add, half addrspace(1)* %out 67 ret void 68 } 69 70 ; ALL: 'fsub_v2f16' 71 ; ALL: estimated cost of 2 for {{.*}} fsub <2 x half> 72 define void @fsub_v2f16(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %vaddr, <2 x half> %b) #0 { 73 %vec = load <2 x half>, <2 x half> addrspace(1)* %vaddr 74 %add = fsub <2 x half> %vec, %b 75 store <2 x half> %add, <2 x half> addrspace(1)* %out 76 ret void 77 } 78 79 ; ALL: 'fsub_v4f16' 80 ; ALL: estimated cost of 4 for {{.*}} fsub <4 x half> 81 define void @fsub_v4f16(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %vaddr, <4 x half> %b) #0 { 82 %vec = load <4 x half>, <4 x half> addrspace(1)* %vaddr 83 %add = fsub <4 x half> %vec, %b 84 store <4 x half> %add, <4 x half> addrspace(1)* %out 85 ret void 86 } 87
