aarch64: Manually tweak vertical alignment/indentation in tx_float_neon.S

Favour left aligned columns over right aligned columns.

In principle either style should be ok, but some of the cases
easily lead to incorrect indentation in the surrounding code (see
a couple of cases fixed up in the preceding patch), and show up in
automatic indentation correction attempts.

Signed-off-by: Martin Storsjö <martin@martin.st>
This commit is contained in:
Martin Storsjö 2023-10-17 14:04:41 +03:00
parent 7f905f3672
commit cada4597ca

View File

@ -733,12 +733,12 @@ FFT16_FN ns_float, 1
add x11, x1, x21, lsl #1
add x12, x1, x22
ldp q0, q1, [x1, #((0 + \part)*32 + \off)]
ldp q4, q5, [x1, #((2 + \part)*32 + \off)]
ldp q2, q3, [x10, #((0 + \part)*32 + \off)]
ldp q6, q7, [x10, #((2 + \part)*32 + \off)]
ldp q0, q1, [x1, #((0 + \part)*32 + \off)]
ldp q4, q5, [x1, #((2 + \part)*32 + \off)]
ldp q2, q3, [x10, #((0 + \part)*32 + \off)]
ldp q6, q7, [x10, #((2 + \part)*32 + \off)]
ldp q8, q9, [x11, #((0 + \part)*32 + \off)]
ldp q8, q9, [x11, #((0 + \part)*32 + \off)]
ldp q10, q11, [x11, #((2 + \part)*32 + \off)]
ldp q12, q13, [x12, #((0 + \part)*32 + \off)]
ldp q14, q15, [x12, #((2 + \part)*32 + \off)]
@ -747,12 +747,12 @@ FFT16_FN ns_float, 1
v8, v9, v10, v11, v12, v13, v14, v15, \
x7, x8, x9, 0
stp q0, q1, [x1, #((0 + \part)*32 + \off)]
stp q4, q5, [x1, #((2 + \part)*32 + \off)]
stp q2, q3, [x10, #((0 + \part)*32 + \off)]
stp q6, q7, [x10, #((2 + \part)*32 + \off)]
stp q0, q1, [x1, #((0 + \part)*32 + \off)]
stp q4, q5, [x1, #((2 + \part)*32 + \off)]
stp q2, q3, [x10, #((0 + \part)*32 + \off)]
stp q6, q7, [x10, #((2 + \part)*32 + \off)]
stp q8, q9, [x11, #((0 + \part)*32 + \off)]
stp q8, q9, [x11, #((0 + \part)*32 + \off)]
stp q12, q13, [x11, #((2 + \part)*32 + \off)]
stp q10, q11, [x12, #((0 + \part)*32 + \off)]
stp q14, q15, [x12, #((2 + \part)*32 + \off)]
@ -775,12 +775,12 @@ FFT16_FN ns_float, 1
add x12, x15, #((\part)*32 + \off)
add x13, x16, #((\part)*32 + \off)
ldp q0, q1, [x10]
ldp q4, q5, [x10, #(2*32)]
ldp q2, q3, [x11]
ldp q6, q7, [x11, #(2*32)]
ldp q0, q1, [x10]
ldp q4, q5, [x10, #(2*32)]
ldp q2, q3, [x11]
ldp q6, q7, [x11, #(2*32)]
ldp q8, q9, [x12]
ldp q8, q9, [x12]
ldp q10, q11, [x12, #(2*32)]
ldp q12, q13, [x13]
ldp q14, q15, [x13, #(2*32)]
@ -800,10 +800,10 @@ FFT16_FN ns_float, 1
zip1 v22.2d, v3.2d, v7.2d
zip2 v23.2d, v3.2d, v7.2d
ldp q0, q1, [x10, #(1*32)]
ldp q4, q5, [x10, #(3*32)]
ldp q2, q3, [x11, #(1*32)]
ldp q6, q7, [x11, #(3*32)]
ldp q0, q1, [x10, #(1*32)]
ldp q4, q5, [x10, #(3*32)]
ldp q2, q3, [x11, #(1*32)]
ldp q6, q7, [x11, #(3*32)]
st1 { v16.4s, v17.4s, v18.4s, v19.4s }, [x10], #64
st1 { v20.4s, v21.4s, v22.4s, v23.4s }, [x11], #64
@ -817,7 +817,7 @@ FFT16_FN ns_float, 1
zip1 v26.2d, v11.2d, v15.2d
zip2 v27.2d, v11.2d, v15.2d
ldp q8, q9, [x12, #(1*32)]
ldp q8, q9, [x12, #(1*32)]
ldp q10, q11, [x12, #(3*32)]
ldp q12, q13, [x13, #(1*32)]
ldp q14, q15, [x13, #(3*32)]
@ -875,9 +875,9 @@ function ff_tx_fft32_\name\()_neon, export=1
SETUP_SR_RECOMB 32, x7, x8, x9
SETUP_LUT \no_perm
LOAD_INPUT 0, 1, 2, 3, x2, \no_perm
LOAD_INPUT 4, 5, 6, 7, x2, \no_perm
LOAD_INPUT 8, 9, 10, 11, x2, \no_perm
LOAD_INPUT 0, 1, 2, 3, x2, \no_perm
LOAD_INPUT 4, 5, 6, 7, x2, \no_perm
LOAD_INPUT 8, 9, 10, 11, x2, \no_perm
LOAD_INPUT 12, 13, 14, 15, x2, \no_perm
FFT8_X2 v8, v9, v10, v11, v12, v13, v14, v15
@ -982,37 +982,37 @@ function ff_tx_fft_sr_\name\()_neon, export=1
32:
SETUP_SR_RECOMB 32, x7, x8, x9
LOAD_INPUT 0, 1, 2, 3, x2, \no_perm
LOAD_INPUT 4, 6, 5, 7, x2, \no_perm, 1
LOAD_INPUT 8, 9, 10, 11, x2, \no_perm
LOAD_INPUT 0, 1, 2, 3, x2, \no_perm
LOAD_INPUT 4, 6, 5, 7, x2, \no_perm, 1
LOAD_INPUT 8, 9, 10, 11, x2, \no_perm
LOAD_INPUT 12, 13, 14, 15, x2, \no_perm
FFT8_X2 v8, v9, v10, v11, v12, v13, v14, v15
FFT16 v0, v1, v2, v3, v4, v6, v5, v7
SR_COMBINE v0, v1, v2, v3, v4, v6, v5, v7, \
v8, v9, v10, v11, v12, v13, v14, v15, \
x7, x8, x9, 0
SR_COMBINE v0, v1, v2, v3, v4, v6, v5, v7, \
v8, v9, v10, v11, v12, v13, v14, v15, \
x7, x8, x9, 0
stp q2, q3, [x1, #32*1]
stp q6, q7, [x1, #32*3]
stp q2, q3, [x1, #32*1]
stp q6, q7, [x1, #32*3]
stp q10, q11, [x1, #32*5]
stp q14, q15, [x1, #32*7]
cmp w20, #32
b.gt 64f
stp q0, q1, [x1, #32*0]
stp q4, q5, [x1, #32*2]
stp q8, q9, [x1, #32*4]
stp q0, q1, [x1, #32*0]
stp q4, q5, [x1, #32*2]
stp q8, q9, [x1, #32*4]
stp q12, q13, [x1, #32*6]
ret
64:
SETUP_SR_RECOMB 64, x7, x8, x9
LOAD_INPUT 2, 3, 10, 11, x2, \no_perm, 1
LOAD_INPUT 6, 14, 7, 15, x2, \no_perm, 1
LOAD_INPUT 2, 3, 10, 11, x2, \no_perm, 1
LOAD_INPUT 6, 14, 7, 15, x2, \no_perm, 1
FFT16 v2, v3, v10, v11, v6, v14, v7, v15
@ -1033,38 +1033,38 @@ function ff_tx_fft_sr_\name\()_neon, export=1
// TODO: investigate doing the 2 combines like in deinterleave
// TODO: experiment with spilling to gprs and converting to HALF or full
SR_COMBINE_LITE v0, v1, v8, v9, \
v2, v3, v16, v17, \
SR_COMBINE_LITE v0, v1, v8, v9, \
v2, v3, v16, v17, \
v24, v25, v26, v27, \
v28, v29, v30, 0
stp q0, q1, [x1, #32* 0]
stp q8, q9, [x1, #32* 4]
stp q2, q3, [x1, #32* 8]
stp q0, q1, [x1, #32* 0]
stp q8, q9, [x1, #32* 4]
stp q2, q3, [x1, #32* 8]
stp q16, q17, [x1, #32*12]
SR_COMBINE_HALF v4, v5, v12, v13, \
v6, v7, v20, v21, \
SR_COMBINE_HALF v4, v5, v12, v13, \
v6, v7, v20, v21, \
v24, v25, v26, v27, \
v28, v29, v30, v0, v1, v8, 1
stp q4, q20, [x1, #32* 2]
stp q4, q20, [x1, #32* 2]
stp q12, q21, [x1, #32* 6]
stp q6, q5, [x1, #32*10]
stp q7, q13, [x1, #32*14]
stp q6, q5, [x1, #32*10]
stp q7, q13, [x1, #32*14]
ldp q2, q3, [x1, #32*1]
ldp q6, q7, [x1, #32*3]
ldp q2, q3, [x1, #32*1]
ldp q6, q7, [x1, #32*3]
ldp q12, q13, [x1, #32*5]
ldp q16, q17, [x1, #32*7]
SR_COMBINE v2, v3, v12, v13, v6, v16, v7, v17, \
SR_COMBINE v2, v3, v12, v13, v6, v16, v7, v17, \
v10, v11, v14, v15, v18, v19, v22, v23, \
x7, x8, x9, 0, \
x7, x8, x9, 0, \
v24, v25, v26, v27, v28, v29, v30, v8, v0, v1, v4, v5
stp q2, q3, [x1, #32* 1]
stp q6, q7, [x1, #32* 3]
stp q2, q3, [x1, #32* 1]
stp q6, q7, [x1, #32* 3]
stp q12, q13, [x1, #32* 5]
stp q16, q17, [x1, #32* 7]
@ -1198,13 +1198,13 @@ SR_TRANSFORM_DEF 131072
mov x10, v23.d[0]
mov x11, v23.d[1]
SR_COMBINE_LITE v0, v1, v8, v9, \
v2, v3, v16, v17, \
SR_COMBINE_LITE v0, v1, v8, v9, \
v2, v3, v16, v17, \
v24, v25, v26, v27, \
v28, v29, v30, 0
SR_COMBINE_HALF v4, v5, v12, v13, \
v6, v7, v20, v21, \
SR_COMBINE_HALF v4, v5, v12, v13, \
v6, v7, v20, v21, \
v24, v25, v26, v27, \
v28, v29, v30, v23, v24, v26, 1
@ -1236,7 +1236,7 @@ SR_TRANSFORM_DEF 131072
zip2 v3.2d, v17.2d, v13.2d
// stp is faster by a little on A53, but this is faster on M1s (theory)
ldp q8, q9, [x1, #32*1]
ldp q8, q9, [x1, #32*1]
ldp q12, q13, [x1, #32*5]
st1 { v23.4s, v24.4s, v25.4s, v26.4s }, [x12], #64 // 32* 0...1
@ -1247,12 +1247,12 @@ SR_TRANSFORM_DEF 131072
mov v23.d[0], x10
mov v23.d[1], x11
ldp q6, q7, [x1, #32*3]
ldp q6, q7, [x1, #32*3]
ldp q16, q17, [x1, #32*7]
SR_COMBINE v8, v9, v12, v13, v6, v16, v7, v17, \
SR_COMBINE v8, v9, v12, v13, v6, v16, v7, v17, \
v10, v11, v14, v15, v18, v19, v22, v23, \
x7, x8, x9, 0, \
x7, x8, x9, 0, \
v24, v25, v26, v27, v28, v29, v30, v4, v0, v1, v5, v20
zip1 v0.2d, v8.2d, v6.2d