Use new style method to call ASM functions for cn/2 & added bulldozer ASM code.

algo/cryptonight/cryptonight.c

@@ -44,6 +44,9 @@
 #include "options.h"
 
 
+static cn_hash_fun asm_func_map[AV_MAX][VARIANT_MAX][ASM_MAX] = {};
+
+
 void cryptonight_av1_v0(const uint8_t *input, size_t size, uint8_t *output, struct cryptonight_ctx **ctx);
 void cryptonight_av1_v1(const uint8_t *input, size_t size, uint8_t *output, struct cryptonight_ctx **ctx);
 void cryptonight_av1_v2(const uint8_t *input, size_t size, uint8_t *output, struct cryptonight_ctx **ctx);
@@ -78,6 +81,7 @@ void cryptonight_lite_av4_v1(const uint8_t *input, size_t size, uint8_t *output,
 #ifndef XMRIG_NO_ASM
 void cryptonight_single_hash_asm_intel(const uint8_t *input, size_t size, uint8_t *output, struct cryptonight_ctx **ctx);
 void cryptonight_single_hash_asm_ryzen(const uint8_t *input, size_t size, uint8_t *output, struct cryptonight_ctx **ctx);
+void cryptonight_single_hash_asm_bulldozer(const uint8_t *input, size_t size, uint8_t *output, struct cryptonight_ctx **ctx);
 void cryptonight_double_hash_asm(const uint8_t *input, size_t size, uint8_t *output, struct cryptonight_ctx **ctx);
 #endif
 
@@ -171,34 +175,20 @@ static bool self_test() {
 }
 
 
-size_t fn_index(enum Algo algorithm, enum AlgoVariant av, enum Variant variant, enum Assembly assembly)
+#ifndef XMRIG_NO_ASM
+cn_hash_fun cryptonight_hash_asm_fn(enum AlgoVariant av, enum Variant variant, enum Assembly assembly)
 {
-    const size_t index = VARIANT_MAX * 4 * algorithm + 4 * variant + av - 1;
-
-#   ifndef XMRIG_NO_ASM
     if (assembly == ASM_AUTO) {
-        assembly = cpu_info.assembly;
+        assembly = (enum Assembly) cpu_info.assembly;
     }
 
     if (assembly == ASM_NONE) {
-        return index;
+        return NULL;
     }
 
-    const size_t offset = VARIANT_MAX * 4 * 2;
+    return asm_func_map[av][variant][assembly];
-
-    if (algorithm == ALGO_CRYPTONIGHT && variant == VARIANT_2) {
-        if (av == AV_SINGLE) {
-            return offset + assembly - 2;
-        }
-
-        if (av == AV_DOUBLE) {
-            return offset + 2;
-        }
-    }
-#   endif
-
-    return index;
 }
+#endif
 
 
 cn_hash_fun cryptonight_hash_fn(enum Algo algorithm, enum AlgoVariant av, enum Variant variant)
@@ -207,10 +197,15 @@ cn_hash_fun cryptonight_hash_fn(enum Algo algorithm, enum AlgoVariant av, enum V
     assert(variant > VARIANT_AUTO && variant < VARIANT_MAX);
 
 #   ifndef XMRIG_NO_ASM
-    static const cn_hash_fun func_table[VARIANT_MAX * 4 * 2 + 3] = {
+    if (algorithm == ALGO_CRYPTONIGHT) {
-#   else
+        cn_hash_fun fun = cryptonight_hash_asm_fn(av, variant, opt_assembly);
-    static const cn_hash_fun func_table[VARIANT_MAX * 4 * 2] = {
+        if (fun) {
+            return fun;
+        }
+    }
 #   endif
+
+    static const cn_hash_fun func_table[VARIANT_MAX * 4 * 2] = {
         cryptonight_av1_v0,
         cryptonight_av2_v0,
         cryptonight_av3_v0,
@@ -263,16 +258,11 @@ cn_hash_fun cryptonight_hash_fn(enum Algo algorithm, enum AlgoVariant av, enum V
         NULL,
         NULL,
         NULL,
-#       endif
-#       ifndef XMRIG_NO_ASM
-        cryptonight_single_hash_asm_intel,
-        cryptonight_single_hash_asm_ryzen,
-        cryptonight_double_hash_asm
 #       endif
     };
 
 #   ifndef NDEBUG
-    const size_t index = fn_index(algorithm, av, variant, opt_assembly);
+    const size_t index = VARIANT_MAX * 4 * algorithm + 4 * variant + av - 1;
 
     cn_hash_fun func = func_table[index];
 
@@ -281,7 +271,7 @@ cn_hash_fun cryptonight_hash_fn(enum Algo algorithm, enum AlgoVariant av, enum V
 
     return func;
 #   else
-    return func_table[fn_index(algorithm, av, variant, opt_assembly)];
+    return func_table[VARIANT_MAX * 4 * algorithm + 4 * variant + av - 1];
 #   endif
 }
 
@@ -290,6 +280,16 @@ bool cryptonight_init(int av)
 {
     opt_double_hash = av == AV_DOUBLE || av == AV_DOUBLE_SOFT;
 
+#   ifndef XMRIG_NO_ASM
+    asm_func_map[AV_SINGLE][VARIANT_2][ASM_INTEL]     = cryptonight_single_hash_asm_intel;
+    asm_func_map[AV_SINGLE][VARIANT_2][ASM_RYZEN]     = cryptonight_single_hash_asm_intel;
+    asm_func_map[AV_SINGLE][VARIANT_2][ASM_BULLDOZER] = cryptonight_single_hash_asm_bulldozer;
+
+    asm_func_map[AV_DOUBLE][VARIANT_2][ASM_INTEL]     = cryptonight_double_hash_asm;
+    asm_func_map[AV_DOUBLE][VARIANT_2][ASM_RYZEN]     = cryptonight_double_hash_asm;
+    asm_func_map[AV_DOUBLE][VARIANT_2][ASM_BULLDOZER] = cryptonight_double_hash_asm;
+#   endif
+
     return self_test();
 }
 

algo/cryptonight/cryptonight_av1.c

@@ -196,6 +196,7 @@ void cryptonight_av1_v2(const uint8_t *restrict input, size_t size, uint8_t *res
 #ifndef XMRIG_NO_ASM
 extern void cnv2_mainloop_ivybridge_asm(struct cryptonight_ctx *ctx);
 extern void cnv2_mainloop_ryzen_asm(struct cryptonight_ctx *ctx);
+extern void cnv2_mainloop_bulldozer_asm(struct cryptonight_ctx *ctx);
 extern void cnv2_double_mainloop_sandybridge_asm(struct cryptonight_ctx* ctx0, struct cryptonight_ctx* ctx1);
 
 
@@ -225,6 +226,19 @@ void cryptonight_single_hash_asm_ryzen(const uint8_t *restrict input, size_t siz
 }
 
 
+void cryptonight_single_hash_asm_bulldozer(const uint8_t *restrict input, size_t size, uint8_t *restrict output, struct cryptonight_ctx **restrict ctx)
+{
+    keccak(input, size, ctx[0]->state, 200);
+    cn_explode_scratchpad((__m128i*) ctx[0]->state, (__m128i*) ctx[0]->memory);
+
+    cnv2_mainloop_bulldozer_asm(ctx[0]);
+
+    cn_implode_scratchpad((__m128i*) ctx[0]->memory, (__m128i*) ctx[0]->state);
+    keccakf((uint64_t*) ctx[0]->state, 24);
+    extra_hashes[ctx[0]->state[0] & 3](ctx[0]->state, 200, output);
+}
+
+
 void cryptonight_double_hash_asm(const uint8_t *restrict input, size_t size, uint8_t *restrict output, struct cryptonight_ctx **restrict ctx)
 {
     keccak(input,        size, ctx[0]->state, 200);

algo/cryptonight/cryptonight_r_av1.c

@@ -97,54 +97,5 @@ void cryptonight_r_av1(const uint8_t *restrict input, size_t size, uint8_t *rest
 
 
 #ifndef XMRIG_NO_ASM
-extern void cnv2_mainloop_ivybridge_asm(struct cryptonight_ctx *ctx);
-extern void cnv2_mainloop_ryzen_asm(struct cryptonight_ctx *ctx);
-extern void cnv2_double_mainloop_sandybridge_asm(struct cryptonight_ctx* ctx0, struct cryptonight_ctx* ctx1);
 
-
-void cryptonight_single_hash_asm_intel(const uint8_t *restrict input, size_t size, uint8_t *restrict output, struct cryptonight_ctx **restrict ctx)
-{
-    keccak(input, size, ctx[0]->state, 200);
-    cn_explode_scratchpad((__m128i*) ctx[0]->state, (__m128i*) ctx[0]->memory);
-
-    cnv2_mainloop_ivybridge_asm(ctx[0]);
-
-    cn_implode_scratchpad((__m128i*) ctx[0]->memory, (__m128i*) ctx[0]->state);
-    keccakf((uint64_t*) ctx[0]->state, 24);
-    extra_hashes[ctx[0]->state[0] & 3](ctx[0]->state, 200, output);
-}
-
-
-void cryptonight_single_hash_asm_ryzen(const uint8_t *restrict input, size_t size, uint8_t *restrict output, struct cryptonight_ctx **restrict ctx)
-{
-    keccak(input, size, ctx[0]->state, 200);
-    cn_explode_scratchpad((__m128i*) ctx[0]->state, (__m128i*) ctx[0]->memory);
-
-    cnv2_mainloop_ryzen_asm(ctx[0]);
-
-    cn_implode_scratchpad((__m128i*) ctx[0]->memory, (__m128i*) ctx[0]->state);
-    keccakf((uint64_t*) ctx[0]->state, 24);
-    extra_hashes[ctx[0]->state[0] & 3](ctx[0]->state, 200, output);
-}
-
-
-void cryptonight_double_hash_asm(const uint8_t *restrict input, size_t size, uint8_t *restrict output, struct cryptonight_ctx **restrict ctx)
-{
-    keccak(input,        size, ctx[0]->state, 200);
-    keccak(input + size, size, ctx[1]->state, 200);
-
-    cn_explode_scratchpad((__m128i*) ctx[0]->state, (__m128i*) ctx[0]->memory);
-    cn_explode_scratchpad((__m128i*) ctx[1]->state, (__m128i*) ctx[1]->memory);
-
-    cnv2_double_mainloop_sandybridge_asm(ctx[0], ctx[1]);
-
-    cn_implode_scratchpad((__m128i*) ctx[0]->memory, (__m128i*) ctx[0]->state);
-    cn_implode_scratchpad((__m128i*) ctx[1]->memory, (__m128i*) ctx[1]->state);
-
-    keccakf((uint64_t*) ctx[0]->state, 24);
-    keccakf((uint64_t*) ctx[1]->state, 24);
-
-    extra_hashes[ctx[0]->state[0] & 3](ctx[0]->state, 200, output);
-    extra_hashes[ctx[1]->state[0] & 3](ctx[1]->state, 200, output + 32);
-}
 #endif

cmake/asm.cmake

@@ -1,28 +1,16 @@
 if (WITH_ASM AND NOT XMRIG_ARM AND CMAKE_SIZEOF_VOID_P EQUAL 8)
     set(XMRIG_ASM_LIBRARY "xmrig-asm")
 
-    if (CMAKE_C_COMPILER_ID MATCHES MSVC)
+    enable_language(ASM)
-        enable_language(ASM_MASM)
 
-        if (MSVC_TOOLSET_VERSION GREATER_EQUAL 141)
+    if (WIN32 AND CMAKE_C_COMPILER_ID MATCHES GNU)
-            set(XMRIG_ASM_FILE "crypto/asm/cnv2_main_loop.asm")
+        set(XMRIG_ASM_FILE "crypto/asm/win64/cn_main_loop.S")
-        else()
-            set(XMRIG_ASM_FILE "crypto/asm/win64/cnv2_main_loop.asm")
-        endif()
-
-        set_property(SOURCE ${XMRIG_ASM_FILE} PROPERTY ASM_MASM)
     else()
-        enable_language(ASM)
+        set(XMRIG_ASM_FILE "crypto/asm/cn_main_loop.S")
-
-        if (WIN32 AND CMAKE_C_COMPILER_ID MATCHES GNU)
-            set(XMRIG_ASM_FILE "crypto/asm/win64/cnv2_main_loop.S")
-        else()
-            set(XMRIG_ASM_FILE "crypto/asm/cnv2_main_loop.S")
-        endif()
-
-        set_property(SOURCE ${XMRIG_ASM_FILE} PROPERTY C)
     endif()
 
+    set_property(SOURCE ${XMRIG_ASM_FILE} PROPERTY C)
+
     add_library(${XMRIG_ASM_LIBRARY} STATIC ${XMRIG_ASM_FILE})
     set(XMRIG_ASM_SOURCES "")
     set_property(TARGET ${XMRIG_ASM_LIBRARY} PROPERTY LINKER_LANGUAGE C)

cpu.c

@@ -4,8 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- *
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -64,20 +65,20 @@ void cpu_init_common() {
 
     if (data.flags[CPU_FEATURE_AES]) {
         cpu_info.flags |= CPU_FLAG_AES;
-
-#       ifndef XMRIG_NO_ASM
-        if (data.vendor == VENDOR_AMD) {
-            cpu_info.assembly = ASM_RYZEN;
-        }
-        else if (data.vendor == VENDOR_INTEL) {
-            cpu_info.assembly = ASM_INTEL;
-        }
-#       endif
     }
 
     if (data.flags[CPU_FEATURE_BMI2]) {
         cpu_info.flags |= CPU_FLAG_BMI2;
     }
+
+#   ifndef XMRIG_NO_ASM
+    if (data.vendor == VENDOR_AMD) {
+        cpu_info.assembly = (data.ext_family >= 23) ? ASM_RYZEN : ASM_BULLDOZER;
+    }
+    else if (data.vendor == VENDOR_INTEL) {
+        cpu_info.assembly = ASM_INTEL;
+    }
+#   endif
 }
 #endif
 

cpu.h

@@ -4,8 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- *
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by

crypto/asm/cn2/cnv2_double_main_loop_sandybridge.inc

@@ -94,7 +94,7 @@
 	lea	r9, QWORD PTR [rdx+r13]
 	movdqu	xmm15, XMMWORD PTR [r9]
 
-	ALIGN 16
+	ALIGN(64)
 main_loop_double_sandybridge:
 	movdqu	xmm9, xmm15
 	mov eax, edx

crypto/asm/cn2/cnv2_main_loop_bulldozer.inc

@@ -0,0 +1,180 @@
+	mov	QWORD PTR [rsp+16], rbx
+	mov	QWORD PTR [rsp+24], rbp
+	mov	QWORD PTR [rsp+32], rsi
+	push	rdi
+	push	r12
+	push	r13
+	push	r14
+	push	r15
+	sub	rsp, 64
+
+	stmxcsr DWORD PTR [rsp]
+	mov DWORD PTR [rsp+4], 24448
+	ldmxcsr DWORD PTR [rsp+4]
+
+	mov	rax, QWORD PTR [rcx+48]
+	mov	r9, rcx
+	xor	rax, QWORD PTR [rcx+16]
+	mov	ebp, 524288
+	mov	r8, QWORD PTR [rcx+32]
+	xor	r8, QWORD PTR [rcx]
+	mov	r11, QWORD PTR [rcx+40]
+	mov	r10, r8
+	mov	rdx, QWORD PTR [rcx+56]
+	movq	xmm3, rax
+	xor	rdx, QWORD PTR [rcx+24]
+	xor	r11, QWORD PTR [rcx+8]
+	mov	rbx, QWORD PTR [rcx+224]
+	mov	rax, QWORD PTR [r9+80]
+	xor	rax, QWORD PTR [r9+64]
+	movq	xmm0, rdx
+	mov	rcx, QWORD PTR [rcx+88]
+	xor	rcx, QWORD PTR [r9+72]
+	mov	rdi, QWORD PTR [r9+104]
+	and	r10d, 2097136
+	movaps	XMMWORD PTR [rsp+48], xmm6
+	movq	xmm4, rax
+	movaps	XMMWORD PTR [rsp+32], xmm7
+	movaps	XMMWORD PTR [rsp+16], xmm8
+	xorps	xmm8, xmm8
+	mov ax, 1023
+	shl rax, 52
+	movq xmm7, rax
+	mov	r15, QWORD PTR [r9+96]
+	punpcklqdq xmm3, xmm0
+	movq	xmm0, rcx
+	punpcklqdq xmm4, xmm0
+
+	ALIGN(64)
+cnv2_main_loop_bulldozer:
+	movdqa	xmm5, XMMWORD PTR [r10+rbx]
+	movq xmm6, r8
+	pinsrq xmm6, r11, 1
+	lea	rdx, QWORD PTR [r10+rbx]
+	lea	r9, QWORD PTR [rdi+rdi]
+	shl	rdi, 32
+
+	mov	ecx, r10d
+	mov	eax, r10d
+	xor	ecx, 16
+	xor	eax, 32
+	xor	r10d, 48
+	aesenc	xmm5, xmm6
+	movdqa	xmm2, XMMWORD PTR [rcx+rbx]
+	movdqa	xmm1, XMMWORD PTR [rax+rbx]
+	movdqa	xmm0, XMMWORD PTR [r10+rbx]
+	paddq	xmm2, xmm3
+	paddq	xmm1, xmm6
+	paddq	xmm0, xmm4
+	movdqa	XMMWORD PTR [rcx+rbx], xmm0
+	movdqa	XMMWORD PTR [rax+rbx], xmm2
+	movdqa	XMMWORD PTR [r10+rbx], xmm1
+
+	movaps	xmm1, xmm8
+	mov	rsi, r15
+	xor	rsi, rdi
+
+	mov edi, 1023
+	shl rdi, 52
+
+	movq	r14, xmm5
+	pextrq rax, xmm5, 1
+
+	movdqa	xmm0, xmm5
+	pxor	xmm0, xmm3
+	mov	r10, r14
+	and	r10d, 2097136
+	movdqa	XMMWORD PTR [rdx], xmm0
+	xor	rsi, QWORD PTR [r10+rbx]
+	lea	r12, QWORD PTR [r10+rbx]
+	mov	r13, QWORD PTR [r10+rbx+8]
+
+	add	r9d, r14d
+	or	r9d, -2147483647
+	xor	edx, edx
+	div	r9
+	mov	eax, eax
+	shl	rdx, 32
+	lea	r15, [rax+rdx]
+	lea	rax, [r14+r15]
+	shr	rax, 12
+	add	rax, rdi
+	movq	xmm0, rax
+	sqrtsd	xmm1, xmm0
+	movq	rdi, xmm1
+	test	rdi, 524287
+	je	sqrt_fixup_bulldozer
+	shr	rdi, 19
+
+sqrt_fixup_bulldozer_ret:
+	mov	rax, rsi
+	mul	r14
+	movq xmm1, rax
+	movq xmm0, rdx
+	punpcklqdq xmm0, xmm1
+
+	mov	r9d, r10d
+	mov	ecx, r10d
+	xor	r9d, 16
+	xor	ecx, 32
+	xor	r10d, 48
+	movdqa	xmm1, XMMWORD PTR [rcx+rbx]
+	xor rdx, [rcx+rbx]
+	xor rax, [rcx+rbx+8]
+	movdqa	xmm2, XMMWORD PTR [r9+rbx]
+	pxor xmm2, xmm0
+	paddq xmm4, XMMWORD PTR [r10+rbx]
+	paddq	xmm2, xmm3
+	paddq	xmm1, xmm6
+	movdqa	XMMWORD PTR [r9+rbx], xmm4
+	movdqa	XMMWORD PTR [rcx+rbx], xmm2
+	movdqa	XMMWORD PTR [r10+rbx], xmm1
+
+	movdqa	xmm4, xmm3
+	add	r8, rdx
+	add	r11, rax
+	mov	QWORD PTR [r12], r8
+	xor	r8, rsi
+	mov	QWORD PTR [r12+8], r11
+	mov	r10, r8
+	xor	r11, r13
+	and	r10d, 2097136
+	movdqa	xmm3, xmm5
+	dec	ebp
+	jne	cnv2_main_loop_bulldozer
+
+	ldmxcsr DWORD PTR [rsp]
+	movaps	xmm6, XMMWORD PTR [rsp+48]
+	lea	r11, QWORD PTR [rsp+64]
+	mov	rbx, QWORD PTR [r11+56]
+	mov	rbp, QWORD PTR [r11+64]
+	mov	rsi, QWORD PTR [r11+72]
+	movaps	xmm8, XMMWORD PTR [r11-48]
+	movaps	xmm7, XMMWORD PTR [rsp+32]
+	mov	rsp, r11
+	pop	r15
+	pop	r14
+	pop	r13
+	pop	r12
+	pop	rdi
+	jmp cnv2_main_loop_bulldozer_endp
+
+sqrt_fixup_bulldozer:
+	movq r9, xmm5
+	add r9, r15
+	dec	rdi
+	mov edx, -1022
+	shl rdx, 32
+	mov	rax, rdi
+	shr	rdi, 19
+	shr	rax, 20
+	mov	rcx, rdi
+	sub	rcx, rax
+	lea	rcx, [rcx+rdx+1]
+	add	rax, rdx
+	imul	rcx, rax
+	sub	rcx, r9
+	adc	rdi, 0
+	jmp	sqrt_fixup_bulldozer_ret
+
+cnv2_main_loop_bulldozer_endp:

crypto/asm/cn2/cnv2_main_loop_ivybridge.inc

@@ -50,7 +50,7 @@
 	punpcklqdq xmm5, xmm0
 	movdqu	 xmm6, XMMWORD PTR [r10+rbx]
 
-	ALIGN 16
+	ALIGN(64)
 main_loop_ivybridge:
 	lea	 rdx, QWORD PTR [r10+rbx]
 	mov	 ecx, r10d

crypto/asm/cn2/cnv2_main_loop_ryzen.inc

@@ -45,7 +45,7 @@
 	movq	xmm0, rcx
 	punpcklqdq xmm4, xmm0
 
-	ALIGN 16
+	ALIGN(64)
 main_loop_ryzen:
 	movdqa	xmm5, XMMWORD PTR [r10+rbx]
 	movq	xmm0, r11

crypto/asm/cn_main_loop.S

@@ -1,4 +1,8 @@
-#define ALIGN .align
+#ifdef __APPLE__
+#   define ALIGN(x) .align 6
+#else
+#   define ALIGN(x) .align 64
+#endif
 .intel_syntax noprefix
 #ifdef __APPLE__
 #   define FN_PREFIX(fn) _ ## fn
@@ -9,29 +13,42 @@
 #endif
 .global FN_PREFIX(cnv2_mainloop_ivybridge_asm)
 .global FN_PREFIX(cnv2_mainloop_ryzen_asm)
+.global FN_PREFIX(cnv2_mainloop_bulldozer_asm)
 .global FN_PREFIX(cnv2_double_mainloop_sandybridge_asm)
 
-ALIGN 16
+ALIGN(64)
 FN_PREFIX(cnv2_mainloop_ivybridge_asm):
 	sub rsp, 48
 	mov rcx, rdi
-	#include "cnv2_main_loop_ivybridge.inc"
+	#include "cn2/cnv2_main_loop_ivybridge.inc"
 	add rsp, 48
 	ret 0
+	mov eax, 3735929054
 
-ALIGN 16
+ALIGN(64)
 FN_PREFIX(cnv2_mainloop_ryzen_asm):
 	sub rsp, 48
 	mov rcx, rdi
-	#include "cnv2_main_loop_ryzen.inc"
+	#include "cn2/cnv2_main_loop_ryzen.inc"
 	add rsp, 48
 	ret 0
+	mov eax, 3735929054
 
-ALIGN 16
+ALIGN(64)
+FN_PREFIX(cnv2_mainloop_bulldozer_asm):
+	sub rsp, 48
+	mov rcx, rdi
+	#include "cn2/cnv2_main_loop_bulldozer.inc"
+	add rsp, 48
+	ret 0
+	mov eax, 3735929054
+
+ALIGN(64)
 FN_PREFIX(cnv2_double_mainloop_sandybridge_asm):
 	sub rsp, 48
 	mov rcx, rdi
 	mov rdx, rsi
-	#include "cnv2_double_main_loop_sandybridge.inc"
+	#include "cn2/cnv2_double_main_loop_sandybridge.inc"
 	add rsp, 48
 	ret 0
+	mov eax, 3735929054

crypto/asm/cn_main_loop.asm

@@ -1,24 +1,35 @@
 _TEXT_CNV2_MAINLOOP SEGMENT PAGE READ EXECUTE
 PUBLIC cnv2_mainloop_ivybridge_asm
 PUBLIC cnv2_mainloop_ryzen_asm
+PUBLIC cnv2_mainloop_bulldozer_asm
 PUBLIC cnv2_double_mainloop_sandybridge_asm
 
-ALIGN 64
+ALIGN(64)
 cnv2_mainloop_ivybridge_asm PROC
-	INCLUDE cnv2_main_loop_ivybridge.inc
+	INCLUDE cn2/cnv2_main_loop_ivybridge.inc
 	ret 0
+	mov eax, 3735929054
 cnv2_mainloop_ivybridge_asm ENDP
 
-ALIGN 64
+ALIGN(64)
 cnv2_mainloop_ryzen_asm PROC
-	INCLUDE cnv2_main_loop_ryzen.inc
+	INCLUDE cn2/cnv2_main_loop_ryzen.inc
 	ret 0
+	mov eax, 3735929054
 cnv2_mainloop_ryzen_asm ENDP
 
-ALIGN 64
+ALIGN(64)
-cnv2_double_mainloop_sandybridge_asm PROC
+cnv2_mainloop_bulldozer_asm PROC
-	INCLUDE cnv2_double_main_loop_sandybridge.inc
+	INCLUDE cn2/cnv2_main_loop_bulldozer.inc
 	ret 0
+	mov eax, 3735929054
+cnv2_mainloop_bulldozer_asm ENDP
+
+ALIGN(64)
+cnv2_double_mainloop_sandybridge_asm PROC
+	INCLUDE cn2/cnv2_double_main_loop_sandybridge.inc
+	ret 0
+	mov eax, 3735929054
 cnv2_double_mainloop_sandybridge_asm ENDP
 
 _TEXT_CNV2_MAINLOOP ENDS

crypto/asm/win64/cn_main_loop.S

@@ -0,0 +1,31 @@
+#define ALIGN(x) .align 64
+.intel_syntax noprefix
+.section .text
+.global cnv2_mainloop_ivybridge_asm
+.global cnv2_mainloop_ryzen_asm
+.global cnv2_mainloop_bulldozer_asm
+.global cnv2_double_mainloop_sandybridge_asm
+
+ALIGN(64)
+cnv2_mainloop_ivybridge_asm:
+	#include "../cn2/cnv2_main_loop_ivybridge.inc"
+	ret 0
+	mov eax, 3735929054
+
+ALIGN(64)
+cnv2_mainloop_ryzen_asm:
+	#include "../cn2/cnv2_main_loop_ryzen.inc"
+	ret 0
+	mov eax, 3735929054
+
+ALIGN(64)
+cnv2_mainloop_bulldozer_asm:
+	#include "../cn2/cnv2_main_loop_bulldozer.inc"
+	ret 0
+	mov eax, 3735929054
+
+ALIGN(64)
+cnv2_double_mainloop_sandybridge_asm:
+	#include "../cn2/cnv2_double_main_loop_sandybridge.inc"
+	ret 0
+	mov eax, 3735929054

crypto/asm/win64/cnv2_main_loop.S

@@ -1,21 +0,0 @@
-#define ALIGN .align
-.intel_syntax noprefix
-.section .text
-.global cnv2_mainloop_ivybridge_asm
-.global cnv2_mainloop_ryzen_asm
-.global cnv2_double_mainloop_sandybridge_asm
-
-ALIGN 16
-cnv2_mainloop_ivybridge_asm:
-	#include "../cnv2_main_loop_ivybridge.inc"
-	ret 0
-
-ALIGN 16
-cnv2_mainloop_ryzen_asm:
-	#include "../cnv2_main_loop_ryzen.inc"
-	ret 0
-
-ALIGN 16
-cnv2_double_mainloop_sandybridge_asm:
-	#include "../cnv2_double_main_loop_sandybridge.inc"
-	ret 0

options.c

@@ -127,7 +127,7 @@ static struct option const options[] = {
     { "cpu-affinity",  1, NULL, 1020 },
     { "donate-level",  1, NULL, 1003 },
     { "help",          0, NULL, 'h'  },
-    { "keepalive",     0, NULL ,'k'  },
+    { "keepalive",     0, NULL, 'k'  },
     { "max-cpu-usage", 1, NULL, 1004 },
     { "nicehash",      0, NULL, 1006 },
     { "no-color",      0, NULL, 1002 },
@@ -167,7 +167,8 @@ static const char *asm_names[] = {
     "none",
     "auto",
     "intel",
-    "ryzen"
+    "ryzen",
+    "bulldozer"
 };
 
 

options.h

@@ -65,6 +65,7 @@ enum Assembly {
     ASM_AUTO,
     ASM_INTEL,
     ASM_RYZEN,
+    ASM_BULLDOZER,
     ASM_MAX
 };