Clean up the macros, add comment headers, add a couple of helper macros

for spilling/restoring registers.

Reorder some instructions for slightly better scheduling across rounds

2 files changed, 63 insertions, 28 deletions

diff --git a/modules/alg_ia32/asm_macr.h b/modules/alg_ia32/asm_macr.h
index 582184080..a4409a211 100644
--- a/modules/alg_ia32/asm_macr.h
+++ b/modules/alg_ia32/asm_macr.h
@@ -6,6 +6,9 @@
 #ifndef BOTAN_EXT_ASM_MACROS_H__
 #define BOTAN_EXT_ASM_MACROS_H__
 
+/*************************************************
+* General/Global Macros                          *
+*************************************************/
 #define ALIGN .p2align 4,,15
 
 #define START_LISTING(FILENAME) \
@@ -13,13 +16,22 @@
    .text;                       \
    .p2align 4,,15;
 
-#define FUNCTION(func_name)      \
-   .align   8;                   \
-   ALIGN;                        \
-   .global  func_name;           \
-   .type    func_name,@function; \
+/*************************************************
+* Function Definitions                           *
+*************************************************/
+#define START_FUNCTION(func_name) \
+   .align   8;                    \
+   ALIGN;                         \
+   .global  func_name;            \
+   .type    func_name,@function;  \
 func_name:
 
+#define END_FUNCTION(func_name) \
+   ret
+
+/*************************************************
+* Loop Control                                   *
+*************************************************/
 #define START_LOOP(LABEL) \
    ALIGN; \
    LABEL##_LOOP:
@@ -28,6 +40,9 @@ func_name:
    cmpl NUM, REG; \
    jne LABEL##_LOOP
 
+/*************************************************
+* Register Names                                 *
+*************************************************/
 #define EAX %eax
 #define EBX %ebx
 #define ECX %ecx
@@ -37,11 +52,9 @@ func_name:
 #define ESI %esi
 #define ESP %esp
 
-#define IMM(VAL) $VAL
-
-#define PUSH(REG) pushl REG
-#define POP(REG) popl REG
-
+/*************************************************
+* Memory Access Operations                       *
+*************************************************/
 #define ARRAY1(REG, NUM) NUM(REG)
 #define ARRAY4(REG, NUM) 4*NUM(REG)
 #define ARRAY4_INDIRECT(BASE, OFFSET, NUM) 4*NUM(BASE,OFFSET,4)
@@ -50,6 +63,26 @@ func_name:
 #define ASSIGN(TO, FROM) movl FROM, TO
 #define ASSIGN_BYTE(TO, FROM) movzbl FROM, TO
 
+#define PUSH(REG) pushl REG
+#define POP(REG) popl REG
+
+#define SPILL_REGS() \
+   PUSH(EBP) ; \
+   PUSH(EDI) ; \
+   PUSH(ESI) ; \
+   PUSH(EBX)
+
+#define RESTORE_REGS() \
+   POP(EBX) ;  \
+   POP(ESI) ;  \
+   POP(EDI) ;  \
+   POP(EBP)
+
+/*************************************************
+* ALU Operations                                 *
+*************************************************/
+#define IMM(VAL) $VAL
+
 #define ADD(TO, FROM) addl FROM, TO
 #define ADD_IMM(TO, NUM) addl IMM(NUM), TO
 #define SUB_IMM(TO, NUM) subl IMM(NUM), TO
diff --git a/modules/alg_ia32/sha1core.S b/modules/alg_ia32/sha1core.S
index f439a45f6..ed9d697ef 100644
--- a/modules/alg_ia32/sha1core.S
+++ b/modules/alg_ia32/sha1core.S
@@ -7,11 +7,8 @@
 
 START_LISTING(sha1core.S)
 
-FUNCTION(sha160_core)
-   PUSH(EBP)
-   PUSH(EDI)
-   PUSH(ESI)
-   PUSH(EBX)
+START_FUNCTION(sha160_core)
+   SPILL_REGS()
 
 #define PUSHED 4
 
@@ -103,19 +100,19 @@ LOOP_UNTIL(ESI, IMM(80), .EXPANSION)
    XOR(T, D)                    ; \
    ROTR_IMM(B, 2)               ; \
    ADD(E, A)                    ; \
-   ROTR_IMM(A, 5)               ; \
-   ADD3_IMM(E, T, MAGIC1)       ;
+   ADD3_IMM(E, T, MAGIC1)       ; \
+   ROTR_IMM(A, 5)               ;
    
 #define F2_OR_F4(A, B, C, D, E, T, MSG, MAGIC) \
    ROTL_IMM(A, 5)               ; \
    ADD(E, ARRAY4(EDI, MSG))     ; \
-   ASSIGN(T, B)                 ; \
-   XOR(T, D)                    ; \
+   ASSIGN(T, D)                 ; \
    XOR(T, C)                    ; \
+   XOR(T, B)                    ; \
    ROTR_IMM(B, 2)               ; \
    ADD(E, A)                    ; \
-   ROTR_IMM(A, 5)               ; \
-   ADD3_IMM(E, T, MAGIC)        ;
+   ADD3_IMM(E, T, MAGIC)        ; \
+   ROTR_IMM(A, 5)               ;
 
 #define F3(A, B, C, D, E, T, MSG) \
    ROTL_IMM(A, 5)               ; \
@@ -128,8 +125,8 @@ LOOP_UNTIL(ESI, IMM(80), .EXPANSION)
    OR(T, ARRAY4(EDI, MSG))      ; \
    ROTR_IMM(B, 2)               ; \
    ADD(E, A)                    ; \
-   ROTR_IMM(A, 5)               ; \
-   ADD3_IMM(E, T, MAGIC3)       ;
+   ADD3_IMM(E, T, MAGIC3)       ; \
+   ROTR_IMM(A, 5)               ;
 
 #define F2(A, B, C, D, E, T, MSG) \
    F2_OR_F4(A, B, C, D, E, T, MSG, MAGIC2)
@@ -144,6 +141,14 @@ LOOP_UNTIL(ESI, IMM(80), .EXPANSION)
    F(ECX, EDX, ESI, EAX, EBX, EBP, (MSG+3)) \
    F(EBX, ECX, EDX, ESI, EAX, EBP, (MSG+4))
 
+   #if 0
+   F2(EAX, EBX, ECX, EDX, ESI, EBP, 0)
+   F2(ESI, EAX, EBX, ECX, EDX, EBP, 1)
+   F2(EDX, ESI, EAX, EBX, ECX, EBP, 2)
+   F2(ECX, EDX, ESI, EAX, EBX, EBP, 3)
+   F2(EBX, ECX, EDX, ESI, EAX, EBP, 4)
+   #endif
+   
    F_BLOCK(F1, 0)
    F_BLOCK(F1, 5)
    F_BLOCK(F1, 10)
@@ -171,8 +176,5 @@ LOOP_UNTIL(ESI, IMM(80), .EXPANSION)
    ADD(ARRAY4(EBP, 3), EDX)
    ADD(ARRAY4(EBP, 4), ESI)
 
-   POP(EBX)
-   POP(ESI)
-   POP(EDI)
-   POP(EBP)
-   ret
+   RESTORE_REGS()
+END_FUNCTION(sha160_core)