lavu/riscv: require B or zba explicitly

Rémi Denis-Courmont committed 2 years ago

Commit 5f101731 parent e91a8cc2 files · +19 −19Browse files

Unified · Split

2 files

libavutil/riscv/fixed_dsp_rvv.S +7 −7
libavutil/riscv/float_dsp_rvv.S +12 −12

@@ -20,7 +20,7 @@ · 19 unmodified lines
20		20
21	#include "asm.S"	21	#include "asm.S"
22		22
23	func ff_vector_fmul_window_scaled_rvv, zve64x	23	func ff_vector_fmul_window_scaled_rvv, zve64x, zba
24	csrwi vxrm, 0	24	csrwi vxrm, 0
25	vsetvli t0, zero, e16, m1, ta, ma	25	vsetvli t0, zero, e16, m1, ta, ma
26	sh2add a2, a4, a2	26	sh2add a2, a4, a2
@@ -68,7 +68,7 @@ func ff_vector_fmul_window_scaled_rvv, zve64x · 41 unmodified lines
68	ret	68	ret
69	endfunc	69	endfunc
70		70
71	func ff_vector_fmul_window_fixed_rvv, zve64x	71	func ff_vector_fmul_window_fixed_rvv, zve64x, zba
72	csrwi vxrm, 0	72	csrwi vxrm, 0
73	vsetvli t0, zero, e16, m1, ta, ma	73	vsetvli t0, zero, e16, m1, ta, ma
74	sh2add a2, a4, a2	74	sh2add a2, a4, a2
@@ -112,7 +112,7 @@ func ff_vector_fmul_window_fixed_rvv, zve64x · 37 unmodified lines
112	ret	112	ret
113	endfunc	113	endfunc
114		114
115	func ff_vector_fmul_fixed_rvv, zve32x	115	func ff_vector_fmul_fixed_rvv, zve32x, zba
116	csrwi vxrm, 0	116	csrwi vxrm, 0
117	1:	117	1:
118	vsetvli t0, a3, e32, m4, ta, ma	118	vsetvli t0, a3, e32, m4, ta, ma
@@ -129,7 +129,7 @@ func ff_vector_fmul_fixed_rvv, zve32x · 10 unmodified lines
129	ret	129	ret
130	endfunc	130	endfunc
131		131
132	func ff_vector_fmul_reverse_fixed_rvv, zve32x	132	func ff_vector_fmul_reverse_fixed_rvv, zve32x, zba
133	csrwi vxrm, 0	133	csrwi vxrm, 0
134	// e16/m4 and e32/m8 are possible but slow the gathers down.	134	// e16/m4 and e32/m8 are possible but slow the gathers down.
135	vsetvli t0, zero, e16, m1, ta, ma	135	vsetvli t0, zero, e16, m1, ta, ma
@@ -155,7 +155,7 @@ func ff_vector_fmul_reverse_fixed_rvv, zve32x · 19 unmodified lines
155	ret	155	ret
156	endfunc	156	endfunc
157		157
158	func ff_vector_fmul_add_fixed_rvv, zve32x	158	func ff_vector_fmul_add_fixed_rvv, zve32x, zba
159	csrwi vxrm, 0	159	csrwi vxrm, 0
160	1:	160	1:
161	vsetvli t0, a4, e32, m8, ta, ma	161	vsetvli t0, a4, e32, m8, ta, ma
@@ -175,7 +175,7 @@ func ff_vector_fmul_add_fixed_rvv, zve32x · 13 unmodified lines
175	ret	175	ret
176	endfunc	176	endfunc
177		177
178	func ff_scalarproduct_fixed_rvv, zve64x	178	func ff_scalarproduct_fixed_rvv, zve64x, zba
179	li t1, 1 << 30	179	li t1, 1 << 30
180	vsetvli t0, zero, e64, m8, ta, ma	180	vsetvli t0, zero, e64, m8, ta, ma
181	vmv.v.x v8, zero	181	vmv.v.x v8, zero
@@ -198,7 +198,7 @@ func ff_scalarproduct_fixed_rvv, zve64x · 16 unmodified lines
198	endfunc	198	endfunc
199		199
200	// (a0) = (a0) + (a1), (a1) = (a0) - (a1) [0..a2-1]	200	// (a0) = (a0) + (a1), (a1) = (a0) - (a1) [0..a2-1]
201	func ff_butterflies_fixed_rvv, zve32x	201	func ff_butterflies_fixed_rvv, zve32x, zba
202	1:	202	1:
203	vsetvli t0, a2, e32, m4, ta, ma	203	vsetvli t0, a2, e32, m4, ta, ma
204	vle32.v v16, (a0)	204	vle32.v v16, (a0)

libavutil/riscv/float_dsp_rvv.S+12−12

@@ -21,7 +21,7 @@ · 20 unmodified lines
21	#include "asm.S"	21	#include "asm.S"
22		22
23	// (a0) = (a1) * (a2) [0..a3-1]	23	// (a0) = (a1) * (a2) [0..a3-1]
24	func ff_vector_fmul_rvv, zve32f	24	func ff_vector_fmul_rvv, zve32f, zba
25	1:	25	1:
26	vsetvli t0, a3, e32, m8, ta, ma	26	vsetvli t0, a3, e32, m8, ta, ma
27	vle32.v v16, (a1)	27	vle32.v v16, (a1)
@@ -38,7 +38,7 @@ func ff_vector_fmul_rvv, zve32f · 10 unmodified lines
38	endfunc	38	endfunc
39		39
40	// (a0) += (a1) * fa0 [0..a2-1]	40	// (a0) += (a1) * fa0 [0..a2-1]
41	func ff_vector_fmac_scalar_rvv, zve32f	41	func ff_vector_fmac_scalar_rvv, zve32f, zba
42	NOHWF fmv.w.x fa0, a2	42	NOHWF fmv.w.x fa0, a2
43	NOHWF mv a2, a3	43	NOHWF mv a2, a3
44	1:	44	1:
@@ -57,7 +57,7 @@ NOHWF mv a2, a3 · 12 unmodified lines
57	endfunc	57	endfunc
58		58
59	// (a0) = (a1) * fa0 [0..a2-1]	59	// (a0) = (a1) * fa0 [0..a2-1]
60	func ff_vector_fmul_scalar_rvv, zve32f	60	func ff_vector_fmul_scalar_rvv, zve32f, zba
61	NOHWF fmv.w.x fa0, a2	61	NOHWF fmv.w.x fa0, a2
62	NOHWF mv a2, a3	62	NOHWF mv a2, a3
63	1:	63	1:
@@ -73,7 +73,7 @@ NOHWF mv a2, a3 · 9 unmodified lines
73	ret	73	ret
74	endfunc	74	endfunc
75		75
76	func ff_vector_fmul_window_rvv, zve32f	76	func ff_vector_fmul_window_rvv, zve32f, zba
77	// a0: dst, a1: src0, a2: src1, a3: window, a4: length	77	// a0: dst, a1: src0, a2: src1, a3: window, a4: length
78	// e16/m2 and e32/m4 are possible but slower due to gather.	78	// e16/m2 and e32/m4 are possible but slower due to gather.
79	vsetvli t0, zero, e16, m1, ta, ma	79	vsetvli t0, zero, e16, m1, ta, ma
@@ -113,7 +113,7 @@ func ff_vector_fmul_window_rvv, zve32f · 33 unmodified lines
113	endfunc	113	endfunc
114		114
115	// (a0) = (a1) * (a2) + (a3) [0..a4-1]	115	// (a0) = (a1) * (a2) + (a3) [0..a4-1]
116	func ff_vector_fmul_add_rvv, zve32f	116	func ff_vector_fmul_add_rvv, zve32f, zba
117	1:	117	1:
118	vsetvli t0, a4, e32, m8, ta, ma	118	vsetvli t0, a4, e32, m8, ta, ma
119	vle32.v v8, (a1)	119	vle32.v v8, (a1)
@@ -133,7 +133,7 @@ endfunc · 13 unmodified lines
133		133
134	// TODO factor vrsub, separate last iteration?	134	// TODO factor vrsub, separate last iteration?
135	// (a0) = (a1) * reverse(a2) [0..a3-1]	135	// (a0) = (a1) * reverse(a2) [0..a3-1]
136	func ff_vector_fmul_reverse_rvv, zve32f	136	func ff_vector_fmul_reverse_rvv, zve32f, zba
137	// e16/m4 and e32/m8 are possible but slower due to gather.	137	// e16/m4 and e32/m8 are possible but slower due to gather.
138	vsetvli t0, zero, e16, m1, ta, ma	138	vsetvli t0, zero, e16, m1, ta, ma
139	sh2add a2, a3, a2	139	sh2add a2, a3, a2
@@ -159,7 +159,7 @@ func ff_vector_fmul_reverse_rvv, zve32f · 19 unmodified lines
159	endfunc	159	endfunc
160		160
161	// (a0) = (a0) + (a1), (a1) = (a0) - (a1) [0..a2-1]	161	// (a0) = (a0) + (a1), (a1) = (a0) - (a1) [0..a2-1]
162	func ff_butterflies_float_rvv, zve32f	162	func ff_butterflies_float_rvv, zve32f, zba
163	1:	163	1:
164	vsetvli t0, a2, e32, m8, ta, ma	164	vsetvli t0, a2, e32, m8, ta, ma
165	vle32.v v16, (a0)	165	vle32.v v16, (a0)
@@ -177,7 +177,7 @@ func ff_butterflies_float_rvv, zve32f · 11 unmodified lines
177	endfunc	177	endfunc
178		178
179	// a0 = (a0).(a1) [0..a2-1]	179	// a0 = (a0).(a1) [0..a2-1]
180	func ff_scalarproduct_float_rvv, zve32f	180	func ff_scalarproduct_float_rvv, zve32f, zba
181	vsetvli t0, zero, e32, m8, ta, ma	181	vsetvli t0, zero, e32, m8, ta, ma
182	vmv.v.x v8, zero	182	vmv.v.x v8, zero
183	vmv.s.x v0, zero	183	vmv.s.x v0, zero
@@ -199,7 +199,7 @@ NOHWF fmv.x.w a0, fa0 · 15 unmodified lines
199	endfunc	199	endfunc
200		200
201	// (a0) = (a1) * (a2) [0..a3-1]	201	// (a0) = (a1) * (a2) [0..a3-1]
202	func ff_vector_dmul_rvv, zve64d	202	func ff_vector_dmul_rvv, zve64d, zba
203	1:	203	1:
204	vsetvli t0, a3, e64, m8, ta, ma	204	vsetvli t0, a3, e64, m8, ta, ma
205	vle64.v v16, (a1)	205	vle64.v v16, (a1)
@@ -216,7 +216,7 @@ func ff_vector_dmul_rvv, zve64d · 10 unmodified lines
216	endfunc	216	endfunc
217		217
218	// (a0) += (a1) * fa0 [0..a2-1]	218	// (a0) += (a1) * fa0 [0..a2-1]
219	func ff_vector_dmac_scalar_rvv, zve64d	219	func ff_vector_dmac_scalar_rvv, zve64d, zba
220	NOHWD fmv.d.x fa0, a2	220	NOHWD fmv.d.x fa0, a2
221	NOHWD mv a2, a3	221	NOHWD mv a2, a3
222	1:	222	1:
@@ -234,7 +234,7 @@ NOHWD mv a2, a3 · 11 unmodified lines
234	endfunc	234	endfunc
235		235
236	// (a0) = (a1) * fa0 [0..a2-1]	236	// (a0) = (a1) * fa0 [0..a2-1]
237	func ff_vector_dmul_scalar_rvv, zve64d	237	func ff_vector_dmul_scalar_rvv, zve64d, zba
238	NOHWD fmv.d.x fa0, a2	238	NOHWD fmv.d.x fa0, a2
239	NOHWD mv a2, a3	239	NOHWD mv a2, a3
240	1:	240	1:
@@ -250,7 +250,7 @@ NOHWD mv a2, a3 · 9 unmodified lines
250	ret	250	ret
251	endfunc	251	endfunc
252		252
253	func ff_scalarproduct_double_rvv, zve64f	253	func ff_scalarproduct_double_rvv, zve64f, zba
254	vsetvli t0, zero, e64, m8, ta, ma	254	vsetvli t0, zero, e64, m8, ta, ma
255	vmv.v.x v8, zero	255	vmv.v.x v8, zero
256	vmv.s.x v0, zero	256	vmv.s.x v0, zero

lavu/riscv: require B or zba explicitly

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code