From deecfb1082d35c2c14ef6ffdbf4cbf6c9b9c985c Mon Sep 17 00:00:00 2001
From: negge <negge@dgql.org>
Date: Thu, 27 Sep 2018 01:27:42 +0000
Subject: [PATCH] Revert "Embed the rust idct functions. (#613)"
This reverts commit ef97e0ec0768c3d32b9ad2295e72c1d5eee17861.
---
src/transform.rs | 268 ++++++++++++++++++++++++++---------------------
1 file changed, 146 insertions(+), 122 deletions(-)
diff --git a/src/transform.rs b/src/transform.rs
index cade42a3..c9c46705 100644
--- a/src/transform.rs
+++ b/src/transform.rs
@@ -39,14 +39,6 @@ fn half_btf(w0: i32, in0: i32, w1: i32, in1: i32, bit: usize) -> i32 {
round_shift(result, bit)
}
-// performs full butterfly rotation by PI/4
-#[inline]
-fn full_btf_sqrt2(in0: i32, in1: i32, bit: usize) -> (i32, i32) {
- let a = (in0 + in1) * COSPI_INV[32];
- let b = (in0 - in1) * COSPI_INV[32];
- (round_shift(a, bit), round_shift(b, bit))
-}
-
#[inline]
fn round_shift(value: i32, bit: usize) -> i32 {
if bit <= 0 {
@@ -72,11 +64,9 @@ fn av1_idct4(input: &[i32], output: &mut [i32], range: usize) {
let stg1 = [input[0], input[2], input[1], input[3]];
// stage 2
- // TODO: can we inline the return tuple
- let (a, b) = full_btf_sqrt2(stg1[0], stg1[1], cos_bit);
let stg2 = [
- a,
- b,
+ half_btf(COSPI_INV[32], stg1[0], COSPI_INV[32], stg1[1], cos_bit),
+ half_btf(COSPI_INV[32], stg1[0], -COSPI_INV[32], stg1[1], cos_bit),
half_btf(COSPI_INV[48], stg1[2], -COSPI_INV[16], stg1[3], cos_bit),
half_btf(COSPI_INV[16], stg1[2], COSPI_INV[48], stg1[3], cos_bit)
];
@@ -140,52 +130,61 @@ fn av1_iidentity4(input: &[i32], output: &mut [i32], _range: usize) {
}
fn av1_idct8(input: &[i32], output: &mut [i32], range: usize) {
- // call idct4
- let temp_in = [ input[0], input[2], input[4], input[6] ];
- let mut temp_out: [i32; 4] = [0; 4];
- av1_idct4(&temp_in, &mut temp_out, range);
-
+ // TODO: call idct4
let cos_bit = 12;
-
// stage 0
// stage 1
- let stg1 = [ input[1], input[5], input[3], input[7] ];
+ let stg1 = [
+ input[0], input[4], input[2], input[6], input[1], input[5], input[3],
+ input[7],
+ ];
// stage 2
let stg2 = [
- half_btf(COSPI_INV[56], stg1[0], -COSPI_INV[8], stg1[3], cos_bit),
- half_btf(COSPI_INV[24], stg1[1], -COSPI_INV[40], stg1[2], cos_bit),
- half_btf(COSPI_INV[40], stg1[1], COSPI_INV[24], stg1[2], cos_bit),
- half_btf(COSPI_INV[8], stg1[0], COSPI_INV[56], stg1[3], cos_bit)
+ stg1[0],
+ stg1[1],
+ stg1[2],
+ stg1[3],
+ half_btf(COSPI_INV[56], stg1[4], -COSPI_INV[8], stg1[7], cos_bit),
+ half_btf(COSPI_INV[24], stg1[5], -COSPI_INV[40], stg1[6], cos_bit),
+ half_btf(COSPI_INV[40], stg1[5], COSPI_INV[24], stg1[6], cos_bit),
+ half_btf(COSPI_INV[8], stg1[4], COSPI_INV[56], stg1[7], cos_bit)
];
// stage 3
let stg3 = [
- clamp_value(stg2[0] + stg2[1], range),
- clamp_value(stg2[0] - stg2[1], range),
- clamp_value(-stg2[2] + stg2[3], range),
- clamp_value(stg2[2] + stg2[3], range)
+ half_btf(COSPI_INV[32], stg2[0], COSPI_INV[32], stg2[1], cos_bit),
+ half_btf(COSPI_INV[32], stg2[0], -COSPI_INV[32], stg2[1], cos_bit),
+ half_btf(COSPI_INV[48], stg2[2], -COSPI_INV[16], stg2[3], cos_bit),
+ half_btf(COSPI_INV[16], stg2[2], COSPI_INV[48], stg2[3], cos_bit),
+ clamp_value(stg2[4] + stg2[5], range),
+ clamp_value(stg2[4] - stg2[5], range),
+ clamp_value(-stg2[6] + stg2[7], range),
+ clamp_value(stg2[6] + stg2[7], range)
];
// stage 4
- let (a, b) = full_btf_sqrt2(stg3[2], stg3[1], cos_bit);
let stg4 = [
- stg3[0],
- b,
- a,
- stg3[3]
+ clamp_value(stg3[0] + stg3[3], range),
+ clamp_value(stg3[1] + stg3[2], range),
+ clamp_value(stg3[1] - stg3[2], range),
+ clamp_value(stg3[0] - stg3[3], range),
+ stg3[4],
+ half_btf(-COSPI_INV[32], stg3[5], COSPI_INV[32], stg3[6], cos_bit),
+ half_btf(COSPI_INV[32], stg3[5], COSPI_INV[32], stg3[6], cos_bit),
+ stg3[7]
];
// stage 5
- output[0] = clamp_value(temp_out[0] + stg4[3], range);
- output[1] = clamp_value(temp_out[1] + stg4[2], range);
- output[2] = clamp_value(temp_out[2] + stg4[1], range);
- output[3] = clamp_value(temp_out[3] + stg4[0], range);
- output[4] = clamp_value(temp_out[3] - stg4[0], range);
- output[5] = clamp_value(temp_out[2] - stg4[1], range);
- output[6] = clamp_value(temp_out[1] - stg4[2], range);
- output[7] = clamp_value(temp_out[0] - stg4[3], range);
+ output[0] = clamp_value(stg4[0] + stg4[7], range);
+ output[1] = clamp_value(stg4[1] + stg4[6], range);
+ output[2] = clamp_value(stg4[2] + stg4[5], range);
+ output[3] = clamp_value(stg4[3] + stg4[4], range);
+ output[4] = clamp_value(stg4[3] - stg4[4], range);
+ output[5] = clamp_value(stg4[2] - stg4[5], range);
+ output[6] = clamp_value(stg4[1] - stg4[6], range);
+ output[7] = clamp_value(stg4[0] - stg4[7], range);
}
fn av1_iadst8(input: &[i32], output: &mut [i32], range: usize) {
@@ -247,17 +246,15 @@ fn av1_iadst8(input: &[i32], output: &mut [i32], range: usize) {
];
// stage 6
- let (a, b) = full_btf_sqrt2(stg5[2], stg5[3], cos_bit);
- let (c, d) = full_btf_sqrt2(stg5[6], stg5[7], cos_bit);
let stg6 = [
stg5[0],
stg5[1],
- a,
- b,
+ half_btf(COSPI_INV[32], stg5[2], COSPI_INV[32], stg5[3], cos_bit),
+ half_btf(COSPI_INV[32], stg5[2], -COSPI_INV[32], stg5[3], cos_bit),
stg5[4],
stg5[5],
- c,
- d,
+ half_btf(COSPI_INV[32], stg5[6], COSPI_INV[32], stg5[7], cos_bit),
+ half_btf(COSPI_INV[32], stg5[6], -COSPI_INV[32], stg5[7], cos_bit)
];
// stage 7
@@ -278,57 +275,74 @@ fn av1_iidentity8(input: &[i32], output: &mut [i32], _range: usize) {
}
fn av1_idct16(input: &[i32], output: &mut [i32], range: usize) {
- // call idct8
- let temp_in = [
- input[0], input[2], input[4], input[6], input[8], input[10], input[12],
- input[14]
- ];
- let mut temp_out: [i32; 8] = [0; 8];
- av1_idct8(&temp_in, &mut temp_out, range);
-
let cos_bit = 12;
// stage 0
// stage 1
let stg1 = [
- input[1], input[9], input[5], input[13], input[3], input[11], input[7],
- input[15]
+ input[0], input[8], input[4], input[12], input[2], input[10], input[6],
+ input[14], input[1], input[9], input[5], input[13], input[3], input[11],
+ input[7], input[15],
];
// stage 2
let stg2 = [
- half_btf(COSPI_INV[60], stg1[0], -COSPI_INV[4], stg1[7], cos_bit),
- half_btf(COSPI_INV[28], stg1[1], -COSPI_INV[36], stg1[6], cos_bit),
- half_btf(COSPI_INV[44], stg1[2], -COSPI_INV[20], stg1[5], cos_bit),
- half_btf(COSPI_INV[12], stg1[3], -COSPI_INV[52], stg1[4], cos_bit),
- half_btf(COSPI_INV[52], stg1[3], COSPI_INV[12], stg1[4], cos_bit),
- half_btf(COSPI_INV[20], stg1[2], COSPI_INV[44], stg1[5], cos_bit),
- half_btf(COSPI_INV[36], stg1[1], COSPI_INV[28], stg1[6], cos_bit),
- half_btf(COSPI_INV[4], stg1[0], COSPI_INV[60], stg1[7], cos_bit)
+ stg1[0],
+ stg1[1],
+ stg1[2],
+ stg1[3],
+ stg1[4],
+ stg1[5],
+ stg1[6],
+ stg1[7],
+ half_btf(COSPI_INV[60], stg1[8], -COSPI_INV[4], stg1[15], cos_bit),
+ half_btf(COSPI_INV[28], stg1[9], -COSPI_INV[36], stg1[14], cos_bit),
+ half_btf(COSPI_INV[44], stg1[10], -COSPI_INV[20], stg1[13], cos_bit),
+ half_btf(COSPI_INV[12], stg1[11], -COSPI_INV[52], stg1[12], cos_bit),
+ half_btf(COSPI_INV[52], stg1[11], COSPI_INV[12], stg1[12], cos_bit),
+ half_btf(COSPI_INV[20], stg1[10], COSPI_INV[44], stg1[13], cos_bit),
+ half_btf(COSPI_INV[36], stg1[9], COSPI_INV[28], stg1[14], cos_bit),
+ half_btf(COSPI_INV[4], stg1[8], COSPI_INV[60], stg1[15], cos_bit)
];
// stage 3
let stg3 = [
- clamp_value(stg2[0] + stg2[1], range),
- clamp_value(stg2[0] - stg2[1], range),
- clamp_value(-stg2[2] + stg2[3], range),
- clamp_value(stg2[2] + stg2[3], range),
- clamp_value(stg2[4] + stg2[5], range),
- clamp_value(stg2[4] - stg2[5], range),
- clamp_value(-stg2[6] + stg2[7], range),
- clamp_value(stg2[6] + stg2[7], range)
+ stg2[0],
+ stg2[1],
+ stg2[2],
+ stg2[3],
+ half_btf(COSPI_INV[56], stg2[4], -COSPI_INV[8], stg2[7], cos_bit),
+ half_btf(COSPI_INV[24], stg2[5], -COSPI_INV[40], stg2[6], cos_bit),
+ half_btf(COSPI_INV[40], stg2[5], COSPI_INV[24], stg2[6], cos_bit),
+ half_btf(COSPI_INV[8], stg2[4], COSPI_INV[56], stg2[7], cos_bit),
+ clamp_value(stg2[8] + stg2[9], range),
+ clamp_value(stg2[8] - stg2[9], range),
+ clamp_value(-stg2[10] + stg2[11], range),
+ clamp_value(stg2[10] + stg2[11], range),
+ clamp_value(stg2[12] + stg2[13], range),
+ clamp_value(stg2[12] - stg2[13], range),
+ clamp_value(-stg2[14] + stg2[15], range),
+ clamp_value(stg2[14] + stg2[15], range)
];
// stage 4
let stg4 = [
- stg3[0],
- half_btf(-COSPI_INV[16], stg3[1], COSPI_INV[48], stg3[6], cos_bit),
- half_btf(-COSPI_INV[48], stg3[2], -COSPI_INV[16], stg3[5], cos_bit),
- stg3[3],
- stg3[4],
- half_btf(-COSPI_INV[16], stg3[2], COSPI_INV[48], stg3[5], cos_bit),
- half_btf(COSPI_INV[48], stg3[1], COSPI_INV[16], stg3[6], cos_bit),
- stg3[7]
+ half_btf(COSPI_INV[32], stg3[0], COSPI_INV[32], stg3[1], cos_bit),
+ half_btf(COSPI_INV[32], stg3[0], -COSPI_INV[32], stg3[1], cos_bit),
+ half_btf(COSPI_INV[48], stg3[2], -COSPI_INV[16], stg3[3], cos_bit),
+ half_btf(COSPI_INV[16], stg3[2], COSPI_INV[48], stg3[3], cos_bit),
+ clamp_value(stg3[4] + stg3[5], range),
+ clamp_value(stg3[4] - stg3[5], range),
+ clamp_value(-stg3[6] + stg3[7], range),
+ clamp_value(stg3[6] + stg3[7], range),
+ stg3[8],
+ half_btf(-COSPI_INV[16], stg3[9], COSPI_INV[48], stg3[14], cos_bit),
+ half_btf(-COSPI_INV[48], stg3[10], -COSPI_INV[16], stg3[13], cos_bit),
+ stg3[11],
+ stg3[12],
+ half_btf(-COSPI_INV[16], stg3[10], COSPI_INV[48], stg3[13], cos_bit),
+ half_btf(COSPI_INV[48], stg3[9], COSPI_INV[16], stg3[14], cos_bit),
+ stg3[15]
];
// stage 5
@@ -337,43 +351,57 @@ fn av1_idct16(input: &[i32], output: &mut [i32], range: usize) {
clamp_value(stg4[1] + stg4[2], range),
clamp_value(stg4[1] - stg4[2], range),
clamp_value(stg4[0] - stg4[3], range),
- clamp_value(-stg4[4] + stg4[7], range),
- clamp_value(-stg4[5] + stg4[6], range),
- clamp_value(stg4[5] + stg4[6], range),
- clamp_value(stg4[4] + stg4[7], range)
+ stg4[4],
+ half_btf(-COSPI_INV[32], stg4[5], COSPI_INV[32], stg4[6], cos_bit),
+ half_btf(COSPI_INV[32], stg4[5], COSPI_INV[32], stg4[6], cos_bit),
+ stg4[7],
+ clamp_value(stg4[8] + stg4[11], range),
+ clamp_value(stg4[9] + stg4[10], range),
+ clamp_value(stg4[9] - stg4[10], range),
+ clamp_value(stg4[8] - stg4[11], range),
+ clamp_value(-stg4[12] + stg4[15], range),
+ clamp_value(-stg4[13] + stg4[14], range),
+ clamp_value(stg4[13] + stg4[14], range),
+ clamp_value(stg4[12] + stg4[15], range)
];
// stage 6
- let (a, b) = full_btf_sqrt2(stg5[5], stg5[2], cos_bit);
- let (c, d) = full_btf_sqrt2(stg5[4], stg5[3], cos_bit);
let stg6 = [
- stg5[0],
- stg5[1],
- b,
- d,
- c,
- a,
- stg5[6],
- stg5[7]
+ clamp_value(stg5[0] + stg5[7], range),
+ clamp_value(stg5[1] + stg5[6], range),
+ clamp_value(stg5[2] + stg5[5], range),
+ clamp_value(stg5[3] + stg5[4], range),
+ clamp_value(stg5[3] - stg5[4], range),
+ clamp_value(stg5[2] - stg5[5], range),
+ clamp_value(stg5[1] - stg5[6], range),
+ clamp_value(stg5[0] - stg5[7], range),
+ stg5[8],
+ stg5[9],
+ half_btf(-COSPI_INV[32], stg5[10], COSPI_INV[32], stg5[13], cos_bit),
+ half_btf(-COSPI_INV[32], stg5[11], COSPI_INV[32], stg5[12], cos_bit),
+ half_btf(COSPI_INV[32], stg5[11], COSPI_INV[32], stg5[12], cos_bit),
+ half_btf(COSPI_INV[32], stg5[10], COSPI_INV[32], stg5[13], cos_bit),
+ stg5[14],
+ stg5[15]
];
// stage 7
- output[0] = clamp_value(temp_out[0] + stg6[7], range);
- output[1] = clamp_value(temp_out[1] + stg6[6], range);
- output[2] = clamp_value(temp_out[2] + stg6[5], range);
- output[3] = clamp_value(temp_out[3] + stg6[4], range);
- output[4] = clamp_value(temp_out[4] + stg6[3], range);
- output[5] = clamp_value(temp_out[5] + stg6[2], range);
- output[6] = clamp_value(temp_out[6] + stg6[1], range);
- output[7] = clamp_value(temp_out[7] + stg6[0], range);
- output[8] = clamp_value(temp_out[7] - stg6[0], range);
- output[9] = clamp_value(temp_out[6] - stg6[1], range);
- output[10] = clamp_value(temp_out[5] - stg6[2], range);
- output[11] = clamp_value(temp_out[4] - stg6[3], range);
- output[12] = clamp_value(temp_out[3] - stg6[4], range);
- output[13] = clamp_value(temp_out[2] - stg6[5], range);
- output[14] = clamp_value(temp_out[1] - stg6[6], range);
- output[15] = clamp_value(temp_out[0] - stg6[7], range);
+ output[0] = clamp_value(stg6[0] + stg6[15], range);
+ output[1] = clamp_value(stg6[1] + stg6[14], range);
+ output[2] = clamp_value(stg6[2] + stg6[13], range);
+ output[3] = clamp_value(stg6[3] + stg6[12], range);
+ output[4] = clamp_value(stg6[4] + stg6[11], range);
+ output[5] = clamp_value(stg6[5] + stg6[10], range);
+ output[6] = clamp_value(stg6[6] + stg6[9], range);
+ output[7] = clamp_value(stg6[7] + stg6[8], range);
+ output[8] = clamp_value(stg6[7] - stg6[8], range);
+ output[9] = clamp_value(stg6[6] - stg6[9], range);
+ output[10] = clamp_value(stg6[5] - stg6[10], range);
+ output[11] = clamp_value(stg6[4] - stg6[11], range);
+ output[12] = clamp_value(stg6[3] - stg6[12], range);
+ output[13] = clamp_value(stg6[2] - stg6[13], range);
+ output[14] = clamp_value(stg6[1] - stg6[14], range);
+ output[15] = clamp_value(stg6[0] - stg6[15], range);
}
fn av1_iadst16(input: &[i32], output: &mut [i32], range: usize) {
@@ -508,27 +536,23 @@ fn av1_iadst16(input: &[i32], output: &mut [i32], range: usize) {
];
// stage 8
- let (a, b) = full_btf_sqrt2(stg7[2], stg7[3], cos_bit);
- let (c, d) = full_btf_sqrt2(stg7[6], stg7[7], cos_bit);
- let (e, f) = full_btf_sqrt2(stg7[10], stg7[11], cos_bit);
- let (g, h) = full_btf_sqrt2(stg7[14], stg7[15], cos_bit);
let stg8 = [
stg7[0],
stg7[1],
- a,
- b,
+ half_btf(COSPI_INV[32], stg7[2], COSPI_INV[32], stg7[3], cos_bit),
+ half_btf(COSPI_INV[32], stg7[2], -COSPI_INV[32], stg7[3], cos_bit),
stg7[4],
stg7[5],
- c,
- d,
+ half_btf(COSPI_INV[32], stg7[6], COSPI_INV[32], stg7[7], cos_bit),
+ half_btf(COSPI_INV[32], stg7[6], -COSPI_INV[32], stg7[7], cos_bit),
stg7[8],
stg7[9],
- e,
- f,
+ half_btf(COSPI_INV[32], stg7[10], COSPI_INV[32], stg7[11], cos_bit),
+ half_btf(COSPI_INV[32], stg7[10], -COSPI_INV[32], stg7[11], cos_bit),
stg7[12],
stg7[13],
- g,
- h,
+ half_btf(COSPI_INV[32], stg7[14], COSPI_INV[32], stg7[15], cos_bit),
+ half_btf(COSPI_INV[32], stg7[14], -COSPI_INV[32], stg7[15], cos_bit)
];
// stage 9
--
GitLab