aarch64 JIT: optimized soft AES

26 fewer instructions per call
SChernykh · Sep 28, 2023 · c29bf53 · c29bf53
1 parent a99d803
commit c29bf53
Showing 1 changed file with 59 additions and 90 deletions.
diff --git a/src/jit_compiler_a64_static.S b/src/jit_compiler_a64_static.S
@@ -539,83 +539,54 @@ DECL(randomx_program_aarch64_v2_FE_mix_soft_aes):
 	stp	x16, x30, [sp, 128]
 	stp	q0, q1, [sp, 144]
 
-	# f0 = aesenc(f0, e0), f1 = aesdec(f1, e0), f2 = aesenc(f2, e0), f3 = aesdec(f3, e0)
-
-	mov	v1.16b, v20.16b
+	adr	x18, randomx_aes_lut_enc
+	adr	x19, randomx_aes_lut_dec
 
+	# f0 = aesenc(f0, e0), f0 = aesenc(f0, e1), f0 = aesenc(f0, e2), f0 = aesenc(f0, e3)
 	mov	v0.16b, v16.16b
+	mov	v1.16b, v20.16b
 	bl	randomx_soft_aesenc
-	mov	v16.16b, v0.16b
-
-	mov	v0.16b, v17.16b
-	bl	randomx_soft_aesdec
-	mov	v17.16b, v0.16b
-
-	mov	v0.16b, v18.16b
-	bl	randomx_soft_aesenc
-	mov	v18.16b, v0.16b
-
-	mov	v0.16b, v19.16b
-	bl	randomx_soft_aesdec
-	mov	v19.16b, v0.16b
-
-	# f0 = aesenc(f0, e1), f1 = aesdec(f1, e1), f2 = aesenc(f2, e1), f3 = aesdec(f3, e1)
-
 	mov	v1.16b, v21.16b
-
-	mov	v0.16b, v16.16b
-	bl	randomx_soft_aesenc
-	mov	v16.16b, v0.16b
-
-	mov	v0.16b, v17.16b
-	bl	randomx_soft_aesdec
-	mov	v17.16b, v0.16b
-
-	mov	v0.16b, v18.16b
 	bl	randomx_soft_aesenc
-	mov	v18.16b, v0.16b
-
-	mov	v0.16b, v19.16b
-	bl	randomx_soft_aesdec
-	mov	v19.16b, v0.16b
-
-	# f0 = aesenc(f0, e2), f1 = aesdec(f1, e2), f2 = aesenc(f2, e2), f3 = aesdec(f3, e2)
-
 	mov	v1.16b, v22.16b
-
-	mov	v0.16b, v16.16b
+	bl	randomx_soft_aesenc
+	mov	v1.16b, v23.16b
 	bl	randomx_soft_aesenc
 	mov	v16.16b, v0.16b
 
+	# f1 = aesdec(f1, e0), f1 = aesdec(f1, e1), f1 = aesdec(f1, e2), f1 = aesdec(f1, e3)
 	mov	v0.16b, v17.16b
+	mov	v1.16b, v20.16b
 	bl	randomx_soft_aesdec
-	mov	v17.16b, v0.16b
-
-	mov	v0.16b, v18.16b
-	bl	randomx_soft_aesenc
-	mov	v18.16b, v0.16b
-
-	mov	v0.16b, v19.16b
+	mov	v1.16b, v21.16b
+	bl	randomx_soft_aesdec
+	mov	v1.16b, v22.16b
 	bl	randomx_soft_aesdec
-	mov	v19.16b, v0.16b
-
-	# f0 = aesenc(f0, e3), f1 = aesdec(f1, e3), f2 = aesenc(f2, e3), f3 = aesdec(f3, e3)
-
 	mov	v1.16b, v23.16b
-
-	mov	v0.16b, v16.16b
-	bl	randomx_soft_aesenc
-	mov	v16.16b, v0.16b
-
-	mov	v0.16b, v17.16b
 	bl	randomx_soft_aesdec
 	mov	v17.16b, v0.16b
 
+	# f2 = aesenc(f2, e0), f2 = aesenc(f2, e1), f2 = aesenc(f2, e2), f2 = aesenc(f2, e3)
 	mov	v0.16b, v18.16b
+	mov	v1.16b, v20.16b
+	bl	randomx_soft_aesenc
+	mov	v1.16b, v21.16b
+	bl	randomx_soft_aesenc
+	mov	v1.16b, v22.16b
+	bl	randomx_soft_aesenc
+	mov	v1.16b, v23.16b
 	bl	randomx_soft_aesenc
 	mov	v18.16b, v0.16b
 
+	# f3 = aesdec(f3, e0), f3 = aesdec(f3, e1), f3 = aesdec(f3, e2), f3 = aesdec(f3, e3)
 	mov	v0.16b, v19.16b
+	mov	v1.16b, v20.16b
+	bl	randomx_soft_aesdec
+	mov	v1.16b, v21.16b
+	bl	randomx_soft_aesdec
+	mov	v1.16b, v22.16b
+	bl	randomx_soft_aesdec
+	mov	v1.16b, v23.16b
 	bl	randomx_soft_aesdec
 	mov	v19.16b, v0.16b
 
@@ -644,7 +615,6 @@ randomx_soft_aesenc:
 	umov	w11, v0.b[3]
 	umov	w5, v0.b[0]
 	umov	w16, v0.b[4]
-	adr	x0, randomx_aes_lut_enc
 	add	x4, x4, 256
 	add	x1, x1, 512
 	add	x12, x12, 768
@@ -654,20 +624,20 @@ randomx_soft_aesenc:
 	add	x9, x9, 256
 	add	x2, x2, 512
 	add	x11, x11, 768
-	ldr	w10, [x0, x4, lsl 2]
-	ldr	w15, [x0, x5, lsl 2]
+	ldr	w10, [x18, x4, lsl 2]
+	ldr	w15, [x18, x5, lsl 2]
 	umov	w13, v0.b[8]
-	ldr	w14, [x0, x12, lsl 2]
+	ldr	w14, [x18, x12, lsl 2]
 	umov	w6, v0.b[1]
-	ldr	w1, [x0, x1, lsl 2]
+	ldr	w1, [x18, x1, lsl 2]
 	eor	w10, w10, w15
-	ldr	w2, [x0, x2, lsl 2]
+	ldr	w2, [x18, x2, lsl 2]
 	umov	w5, v0.b[6]
-	ldr	w9, [x0, x9, lsl 2]
+	ldr	w9, [x18, x9, lsl 2]
 	umov	w4, v0.b[11]
-	ldr	w12, [x0, x16, lsl 2]
+	ldr	w12, [x18, x16, lsl 2]
 	eor	w1, w1, w14
-	ldr	w11, [x0, x11, lsl 2]
+	ldr	w11, [x18, x11, lsl 2]
 	eor	w1, w1, w10
 	add	x8, x8, 512
 	add	x3, x3, 256
@@ -677,21 +647,21 @@ randomx_soft_aesenc:
 	eor	w1, w2, w11
 	umov	w10, v0.b[12]
 	eor	w1, w1, w9
-	ldr	w3, [x0, x3, lsl 2]
+	ldr	w3, [x18, x3, lsl 2]
 	add	x6, x6, 256
-	ldr	w9, [x0, x13, lsl 2]
+	ldr	w9, [x18, x13, lsl 2]
 	ins	v28.s[1], w1
-	ldr	w2, [x0, x8, lsl 2]
+	ldr	w2, [x18, x8, lsl 2]
 	add	x5, x5, 512
-	ldr	w7, [x0, x7, lsl 2]
+	ldr	w7, [x18, x7, lsl 2]
 	add	x4, x4, 768
 	eor	w1, w3, w9
-	ldr	w3, [x0, x6, lsl 2]
+	ldr	w3, [x18, x6, lsl 2]
 	eor	w2, w2, w7
-	ldr	w6, [x0, x10, lsl 2]
+	ldr	w6, [x18, x10, lsl 2]
 	eor	w2, w2, w1
-	ldr	w1, [x0, x5, lsl 2]
-	ldr	w0, [x0, x4, lsl 2]
+	ldr	w1, [x18, x5, lsl 2]
+	ldr	w0, [x18, x4, lsl 2]
 	eor	w3, w3, w6
 	ins	v28.s[2], w2
 	eor	w0, w1, w0
@@ -709,7 +679,6 @@ randomx_soft_aesdec:
 	umov	w11, v0.b[1]
 	umov	w4, v0.b[0]
 	umov	w16, v0.b[4]
-	adr	x0, randomx_aes_lut_dec
 	add	x3, x3, 768
 	add	x1, x1, 512
 	add	x12, x12, 256
@@ -719,20 +688,20 @@ randomx_soft_aesdec:
 	add	x9, x9, 768
 	add	x2, x2, 512
 	add	x11, x11, 256
-	ldr	w15, [x0, x3, lsl 2]
-	ldr	w10, [x0, x4, lsl 2]
+	ldr	w15, [x19, x3, lsl 2]
+	ldr	w10, [x19, x4, lsl 2]
 	umov	w13, v0.b[8]
-	ldr	w14, [x0, x12, lsl 2]
+	ldr	w14, [x19, x12, lsl 2]
 	umov	w5, v0.b[9]
-	ldr	w1, [x0, x1, lsl 2]
+	ldr	w1, [x19, x1, lsl 2]
 	umov	w3, v0.b[6]
-	ldr	w12, [x0, x9, lsl 2]
+	ldr	w12, [x19, x9, lsl 2]
 	umov	w4, v0.b[3]
-	ldr	w9, [x0, x16, lsl 2]
+	ldr	w9, [x19, x16, lsl 2]
 	eor	w1, w1, w15
-	ldr	w2, [x0, x2, lsl 2]
+	ldr	w2, [x19, x2, lsl 2]
 	eor	w10, w10, w14
-	ldr	w11, [x0, x11, lsl 2]
+	ldr	w11, [x19, x11, lsl 2]
 	eor	w1, w1, w10
 	add	x8, x8, 256
 	add	x6, x6, 512
@@ -742,21 +711,21 @@ randomx_soft_aesdec:
 	eor	w1, w9, w11
 	eor	w1, w2, w1
 	umov	w9, v0.b[12]
-	ldr	w2, [x0, x13, lsl 2]
+	ldr	w2, [x19, x13, lsl 2]
 	add	x5, x5, 256
-	ldr	w8, [x0, x8, lsl 2]
+	ldr	w8, [x19, x8, lsl 2]
 	ins	v28.s[1], w1
-	ldr	w6, [x0, x6, lsl 2]
+	ldr	w6, [x19, x6, lsl 2]
 	add	x3, x3, 512
-	ldr	w7, [x0, x7, lsl 2]
+	ldr	w7, [x19, x7, lsl 2]
 	add	x4, x4, 768
 	eor	w2, w2, w8
-	ldr	w1, [x0, x9, lsl 2]
+	ldr	w1, [x19, x9, lsl 2]
 	eor	w6, w6, w7
-	ldr	w3, [x0, x3, lsl 2]
+	ldr	w3, [x19, x3, lsl 2]
 	eor	w2, w2, w6
-	ldr	w4, [x0, x4, lsl 2]
-	ldr	w5, [x0, x5, lsl 2]
+	ldr	w4, [x19, x4, lsl 2]
+	ldr	w5, [x19, x5, lsl 2]
 	ins	v28.s[2], w2
 	eor	w0, w1, w5
 	eor	w1, w3, w4