mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-17 18:31:04 +00:00
f303d037f2
Use ComputeNumSignBits instead of checking for i8 / i16 which only worked when AMDIL was lying about having legal i8 / i16. If an integer is known to fit in 24-bits, we can do division faster with float ops. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@213843 91177308-0d34-0410-b5e6-96231b3b80d8
121 lines
3.5 KiB
LLVM
121 lines
3.5 KiB
LLVM
; RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
|
|
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
|
|
|
|
; FUNC-LABEL: @sdiv24_i8
|
|
; SI: V_CVT_F32_I32
|
|
; SI: V_CVT_F32_I32
|
|
; SI: V_RCP_F32
|
|
; SI: V_CVT_I32_F32
|
|
|
|
; EG: INT_TO_FLT
|
|
; EG-DAG: INT_TO_FLT
|
|
; EG-DAG: RECIP_IEEE
|
|
; EG: FLT_TO_INT
|
|
define void @sdiv24_i8(i8 addrspace(1)* %out, i8 addrspace(1)* %in) {
|
|
%den_ptr = getelementptr i8 addrspace(1)* %in, i8 1
|
|
%num = load i8 addrspace(1) * %in
|
|
%den = load i8 addrspace(1) * %den_ptr
|
|
%result = sdiv i8 %num, %den
|
|
store i8 %result, i8 addrspace(1)* %out
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: @sdiv24_i16
|
|
; SI: V_CVT_F32_I32
|
|
; SI: V_CVT_F32_I32
|
|
; SI: V_RCP_F32
|
|
; SI: V_CVT_I32_F32
|
|
|
|
; EG: INT_TO_FLT
|
|
; EG-DAG: INT_TO_FLT
|
|
; EG-DAG: RECIP_IEEE
|
|
; EG: FLT_TO_INT
|
|
define void @sdiv24_i16(i16 addrspace(1)* %out, i16 addrspace(1)* %in) {
|
|
%den_ptr = getelementptr i16 addrspace(1)* %in, i16 1
|
|
%num = load i16 addrspace(1) * %in, align 2
|
|
%den = load i16 addrspace(1) * %den_ptr, align 2
|
|
%result = sdiv i16 %num, %den
|
|
store i16 %result, i16 addrspace(1)* %out, align 2
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: @sdiv24_i32
|
|
; SI: V_CVT_F32_I32
|
|
; SI: V_CVT_F32_I32
|
|
; SI: V_RCP_F32
|
|
; SI: V_CVT_I32_F32
|
|
|
|
; EG: INT_TO_FLT
|
|
; EG-DAG: INT_TO_FLT
|
|
; EG-DAG: RECIP_IEEE
|
|
; EG: FLT_TO_INT
|
|
define void @sdiv24_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
|
|
%den_ptr = getelementptr i32 addrspace(1)* %in, i32 1
|
|
%num = load i32 addrspace(1) * %in, align 4
|
|
%den = load i32 addrspace(1) * %den_ptr, align 4
|
|
%num.i24.0 = shl i32 %num, 8
|
|
%den.i24.0 = shl i32 %den, 8
|
|
%num.i24 = ashr i32 %num.i24.0, 8
|
|
%den.i24 = ashr i32 %den.i24.0, 8
|
|
%result = sdiv i32 %num.i24, %den.i24
|
|
store i32 %result, i32 addrspace(1)* %out, align 4
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: @sdiv25_i32
|
|
; SI-NOT: V_CVT_F32_I32
|
|
; SI-NOT: V_RCP_F32
|
|
|
|
; EG-NOT: INT_TO_FLT
|
|
; EG-NOT: RECIP_IEEE
|
|
define void @sdiv25_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
|
|
%den_ptr = getelementptr i32 addrspace(1)* %in, i32 1
|
|
%num = load i32 addrspace(1) * %in, align 4
|
|
%den = load i32 addrspace(1) * %den_ptr, align 4
|
|
%num.i24.0 = shl i32 %num, 7
|
|
%den.i24.0 = shl i32 %den, 7
|
|
%num.i24 = ashr i32 %num.i24.0, 7
|
|
%den.i24 = ashr i32 %den.i24.0, 7
|
|
%result = sdiv i32 %num.i24, %den.i24
|
|
store i32 %result, i32 addrspace(1)* %out, align 4
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: @test_no_sdiv24_i32_1
|
|
; SI-NOT: V_CVT_F32_I32
|
|
; SI-NOT: V_RCP_F32
|
|
|
|
; EG-NOT: INT_TO_FLT
|
|
; EG-NOT: RECIP_IEEE
|
|
define void @test_no_sdiv24_i32_1(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
|
|
%den_ptr = getelementptr i32 addrspace(1)* %in, i32 1
|
|
%num = load i32 addrspace(1) * %in, align 4
|
|
%den = load i32 addrspace(1) * %den_ptr, align 4
|
|
%num.i24.0 = shl i32 %num, 8
|
|
%den.i24.0 = shl i32 %den, 7
|
|
%num.i24 = ashr i32 %num.i24.0, 8
|
|
%den.i24 = ashr i32 %den.i24.0, 7
|
|
%result = sdiv i32 %num.i24, %den.i24
|
|
store i32 %result, i32 addrspace(1)* %out, align 4
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: @test_no_sdiv24_i32_2
|
|
; SI-NOT: V_CVT_F32_I32
|
|
; SI-NOT: V_RCP_F32
|
|
|
|
; EG-NOT: INT_TO_FLT
|
|
; EG-NOT: RECIP_IEEE
|
|
define void @test_no_sdiv24_i32_2(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
|
|
%den_ptr = getelementptr i32 addrspace(1)* %in, i32 1
|
|
%num = load i32 addrspace(1) * %in, align 4
|
|
%den = load i32 addrspace(1) * %den_ptr, align 4
|
|
%num.i24.0 = shl i32 %num, 7
|
|
%den.i24.0 = shl i32 %den, 8
|
|
%num.i24 = ashr i32 %num.i24.0, 7
|
|
%den.i24 = ashr i32 %den.i24.0, 8
|
|
%result = sdiv i32 %num.i24, %den.i24
|
|
store i32 %result, i32 addrspace(1)* %out, align 4
|
|
ret void
|
|
}
|