py/emitnative: Let Viper int-indexed code use appropriate operands.

This commit extends the generic ASM API by adding the rest of the
ASM_{LOAD,STORE}[size]_REG_REG_OFFSET macros whenever applicable.

The Viper int-indexed load/store code generator was changed to use those
API functions if they are available, falling back to backend-specific
implementations if possible and ultimately to a generic implementation.

Right now all backends except for x64 implement load16, load32, and
store32 operations (x64 only implements load16).

Signed-off-by: Alessandro Gatti <a.gatti@frob.it>

6 files changed, 46 insertions, 12 deletions

diff --git a/py/asmarm.h b/py/asmarm.h
index 42fa1ea212..2c3af41312 100644
--- a/py/asmarm.h
+++ b/py/asmarm.h
@@ -208,16 +208,18 @@ void asm_arm_bx_reg(asm_arm_t *as, uint reg_src);
 #define ASM_SUB_REG_REG(as, reg_dest, reg_src) asm_arm_sub_reg_reg_reg((as), (reg_dest), (reg_dest), (reg_src))
 #define ASM_MUL_REG_REG(as, reg_dest, reg_src) asm_arm_mul_reg_reg_reg((as), (reg_dest), (reg_dest), (reg_src))
 
-#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_arm_ldr_reg_reg((as), (reg_dest), (reg_base), 4 * (word_offset))
+#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) ASM_LOAD32_REG_REG_OFFSET((as), (reg_dest), (reg_base), (word_offset))
 #define ASM_LOAD8_REG_REG(as, reg_dest, reg_base) asm_arm_ldrb_reg_reg((as), (reg_dest), (reg_base))
 #define ASM_LOAD16_REG_REG(as, reg_dest, reg_base) asm_arm_ldrh_reg_reg((as), (reg_dest), (reg_base))
 #define ASM_LOAD16_REG_REG_OFFSET(as, reg_dest, reg_base, uint16_offset) asm_arm_ldrh_reg_reg_offset((as), (reg_dest), (reg_base), 2 * (uint16_offset))
 #define ASM_LOAD32_REG_REG(as, reg_dest, reg_base) asm_arm_ldr_reg_reg((as), (reg_dest), (reg_base), 0)
+#define ASM_LOAD32_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_arm_ldr_reg_reg((as), (reg_dest), (reg_base), 4 * (word_offset))
 
-#define ASM_STORE_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_arm_str_reg_reg((as), (reg_dest), (reg_base), 4 * (word_offset))
+#define ASM_STORE_REG_REG_OFFSET(as, reg_value, reg_base, word_offset) ASM_STORE32_REG_REG_OFFSET((as), (reg_value), (reg_base), (word_offset))
 #define ASM_STORE8_REG_REG(as, reg_value, reg_base) asm_arm_strb_reg_reg((as), (reg_value), (reg_base))
 #define ASM_STORE16_REG_REG(as, reg_value, reg_base) asm_arm_strh_reg_reg((as), (reg_value), (reg_base))
 #define ASM_STORE32_REG_REG(as, reg_value, reg_base) asm_arm_str_reg_reg((as), (reg_value), (reg_base), 0)
+#define ASM_STORE32_REG_REG_OFFSET(as, reg_value, reg_base, word_offset) asm_arm_str_reg_reg((as), (reg_value), (reg_base), 4 * (word_offset))
 
 #define ASM_LOAD8_REG_REG_REG(as, reg_dest, reg_base, reg_index) asm_arm_ldrb_reg_reg_reg((as), (reg_dest), (reg_base), (reg_index))
 #define ASM_LOAD16_REG_REG_REG(as, reg_dest, reg_base, reg_index) asm_arm_ldrh_reg_reg_reg((as), (reg_dest), (reg_base), (reg_index))
diff --git a/py/asmrv32.h b/py/asmrv32.h
index 6453b0a3d4..4f986d7bbd 100644
--- a/py/asmrv32.h
+++ b/py/asmrv32.h
@@ -732,11 +732,12 @@ void asm_rv32_emit_store_reg_reg_offset(asm_rv32_t *state, mp_uint_t source, mp_
 #define ASM_JUMP_IF_REG_NONZERO(state, rs, label, bool_test) asm_rv32_emit_jump_if_reg_nonzero(state, rs, label)
 #define ASM_JUMP_IF_REG_ZERO(state, rs, label, bool_test) asm_rv32_emit_jump_if_reg_eq(state, rs, ASM_RV32_REG_ZERO, label)
 #define ASM_JUMP_REG(state, rs) asm_rv32_opcode_cjr(state, rs)
+#define ASM_LOAD_REG_REG_OFFSET(state, rd, rs, offset) ASM_LOAD32_REG_REG_OFFSET(state, rd, rs, offset)
 #define ASM_LOAD16_REG_REG_OFFSET(state, rd, rs, offset) asm_rv32_emit_load16_reg_reg_offset(state, rd, rs, offset)
 #define ASM_LOAD16_REG_REG(state, rd, rs) asm_rv32_opcode_lhu(state, rd, rs, 0)
-#define ASM_LOAD32_REG_REG(state, rd, rs) ASM_LOAD_REG_REG_OFFSET(state, rd, rs, 0)
+#define ASM_LOAD32_REG_REG(state, rd, rs) ASM_LOAD32_REG_REG_OFFSET(state, rd, rs, 0)
+#define ASM_LOAD32_REG_REG_OFFSET(state, rd, rs, offset) asm_rv32_emit_load_reg_reg_offset(state, rd, rs, offset)
 #define ASM_LOAD8_REG_REG(state, rd, rs) asm_rv32_opcode_lbu(state, rd, rs, 0)
-#define ASM_LOAD_REG_REG_OFFSET(state, rd, rs, offset) asm_rv32_emit_load_reg_reg_offset(state, rd, rs, offset)
 #define ASM_LSL_REG_REG(state, rd, rs) asm_rv32_opcode_sll(state, rd, rd, rs)
 #define ASM_LSR_REG_REG(state, rd, rs) asm_rv32_opcode_srl(state, rd, rd, rs)
 #define ASM_MOV_LOCAL_REG(state, local, rs) asm_rv32_emit_mov_local_reg(state, local, rs)
@@ -749,10 +750,11 @@ void asm_rv32_emit_store_reg_reg_offset(asm_rv32_t *state, mp_uint_t source, mp_
 #define ASM_NEG_REG(state, rd) asm_rv32_opcode_sub(state, rd, ASM_RV32_REG_ZERO, rd)
 #define ASM_NOT_REG(state, rd) asm_rv32_opcode_xori(state, rd, rd, -1)
 #define ASM_OR_REG_REG(state, rd, rs) asm_rv32_opcode_or(state, rd, rd, rs)
+#define ASM_STORE_REG_REG_OFFSET(state, rd, rs, offset) ASM_STORE32_REG_REG_OFFSET(state, rd, rs, offset)
 #define ASM_STORE16_REG_REG(state, rs1, rs2) asm_rv32_opcode_sh(state, rs1, rs2, 0)
-#define ASM_STORE32_REG_REG(state, rs1, rs2) ASM_STORE_REG_REG_OFFSET(state, rs1, rs2, 0)
+#define ASM_STORE32_REG_REG(state, rs1, rs2) ASM_STORE32_REG_REG_OFFSET(state, rs1, rs2, 0)
+#define ASM_STORE32_REG_REG_OFFSET(state, rd, rs, offset) asm_rv32_emit_store_reg_reg_offset(state, rd, rs, offset)
 #define ASM_STORE8_REG_REG(state, rs1, rs2) asm_rv32_opcode_sb(state, rs1, rs2, 0)
-#define ASM_STORE_REG_REG_OFFSET(state, rd, rs, offset) asm_rv32_emit_store_reg_reg_offset(state, rd, rs, offset)
 #define ASM_SUB_REG_REG(state, rd, rs) asm_rv32_opcode_sub(state, rd, rd, rs)
 #define ASM_XOR_REG_REG(state, rd, rs) asm_rv32_emit_optimised_xor(state, rd, rs)
 #define ASM_CLR_REG(state, rd)
diff --git a/py/asmthumb.h b/py/asmthumb.h
index cc4213503b..570b44b96b 100644
--- a/py/asmthumb.h
+++ b/py/asmthumb.h
@@ -462,16 +462,18 @@ void asm_thumb_b_rel12(asm_thumb_t *as, int rel);
 #define ASM_SUB_REG_REG(as, reg_dest, reg_src) asm_thumb_sub_rlo_rlo_rlo((as), (reg_dest), (reg_dest), (reg_src))
 #define ASM_MUL_REG_REG(as, reg_dest, reg_src) asm_thumb_format_4((as), ASM_THUMB_FORMAT_4_MUL, (reg_dest), (reg_src))
 
-#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_thumb_ldr_reg_reg_i12_optimised((as), (reg_dest), (reg_base), (word_offset))
+#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) ASM_LOAD32_REG_REG_OFFSET((as), (reg_dest), (reg_base), (word_offset))
 #define ASM_LOAD8_REG_REG(as, reg_dest, reg_base) asm_thumb_ldrb_rlo_rlo_i5((as), (reg_dest), (reg_base), 0)
 #define ASM_LOAD16_REG_REG(as, reg_dest, reg_base) asm_thumb_ldrh_rlo_rlo_i5((as), (reg_dest), (reg_base), 0)
 #define ASM_LOAD16_REG_REG_OFFSET(as, reg_dest, reg_base, uint16_offset) asm_thumb_ldrh_reg_reg_i12_optimised((as), (reg_dest), (reg_base), (uint16_offset))
 #define ASM_LOAD32_REG_REG(as, reg_dest, reg_base) asm_thumb_ldr_rlo_rlo_i5((as), (reg_dest), (reg_base), 0)
+#define ASM_LOAD32_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_thumb_ldr_reg_reg_i12_optimised((as), (reg_dest), (reg_base), (word_offset))
 
-#define ASM_STORE_REG_REG_OFFSET(as, reg_src, reg_base, word_offset) asm_thumb_str_rlo_rlo_i5((as), (reg_src), (reg_base), (word_offset))
+#define ASM_STORE_REG_REG_OFFSET(as, reg_src, reg_base, word_offset) ASM_STORE32_REG_REG_OFFSET((as), (reg_src), (reg_base), (word_offset))
 #define ASM_STORE8_REG_REG(as, reg_src, reg_base) asm_thumb_strb_rlo_rlo_i5((as), (reg_src), (reg_base), 0)
 #define ASM_STORE16_REG_REG(as, reg_src, reg_base) asm_thumb_strh_rlo_rlo_i5((as), (reg_src), (reg_base), 0)
 #define ASM_STORE32_REG_REG(as, reg_src, reg_base) asm_thumb_str_rlo_rlo_i5((as), (reg_src), (reg_base), 0)
+#define ASM_STORE32_REG_REG_OFFSET(as, reg_src, reg_base, word_offset) asm_thumb_str_rlo_rlo_i5((as), (reg_src), (reg_base), (word_offset))
 
 #define ASM_LOAD8_REG_REG_REG(as, reg_dest, reg_base, reg_index) asm_thumb_ldrb_rlo_rlo_rlo((as), (reg_dest), (reg_base), (reg_index))
 #define ASM_LOAD16_REG_REG_REG(as, reg_dest, reg_base, reg_index) \
diff --git a/py/asmx86.h b/py/asmx86.h
index af73c163b4..5011e56d0c 100644
--- a/py/asmx86.h
+++ b/py/asmx86.h
@@ -200,16 +200,18 @@ void asm_x86_call_ind(asm_x86_t *as, size_t fun_id, mp_uint_t n_args, int temp_r
 #define ASM_SUB_REG_REG(as, reg_dest, reg_src) asm_x86_sub_r32_r32((as), (reg_dest), (reg_src))
 #define ASM_MUL_REG_REG(as, reg_dest, reg_src) asm_x86_mul_r32_r32((as), (reg_dest), (reg_src))
 
-#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_x86_mov_mem32_to_r32((as), (reg_base), 4 * (word_offset), (reg_dest))
+#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) ASM_LOAD32_REG_REG_OFFSET((as), (reg_dest), (reg_base), (word_offset))
 #define ASM_LOAD8_REG_REG(as, reg_dest, reg_base) asm_x86_mov_mem8_to_r32zx((as), (reg_base), 0, (reg_dest))
 #define ASM_LOAD16_REG_REG(as, reg_dest, reg_base) asm_x86_mov_mem16_to_r32zx((as), (reg_base), 0, (reg_dest))
 #define ASM_LOAD16_REG_REG_OFFSET(as, reg_dest, reg_base, uint16_offset) asm_x86_mov_mem16_to_r32zx((as), (reg_base), 2 * (uint16_offset), (reg_dest))
 #define ASM_LOAD32_REG_REG(as, reg_dest, reg_base) asm_x86_mov_mem32_to_r32((as), (reg_base), 0, (reg_dest))
+#define ASM_LOAD32_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_x86_mov_mem32_to_r32((as), (reg_base), 4 * (word_offset), (reg_dest))
 
-#define ASM_STORE_REG_REG_OFFSET(as, reg_src, reg_base, word_offset) asm_x86_mov_r32_to_mem32((as), (reg_src), (reg_base), 4 * (word_offset))
+#define ASM_STORE_REG_REG_OFFSET(as, reg_src, reg_base, word_offset) ASM_STORE32_REG_REG_OFFSET((as), (reg_src), (reg_base), (word_offset))
 #define ASM_STORE8_REG_REG(as, reg_src, reg_base) asm_x86_mov_r8_to_mem8((as), (reg_src), (reg_base), 0)
 #define ASM_STORE16_REG_REG(as, reg_src, reg_base) asm_x86_mov_r16_to_mem16((as), (reg_src), (reg_base), 0)
 #define ASM_STORE32_REG_REG(as, reg_src, reg_base) asm_x86_mov_r32_to_mem32((as), (reg_src), (reg_base), 0)
+#define ASM_STORE32_REG_REG_OFFSET(as, reg_src, reg_base, word_offset) asm_x86_mov_r32_to_mem32((as), (reg_src), (reg_base), 4 * (word_offset))
 
 #endif // GENERIC_ASM_API
 
diff --git a/py/asmxtensa.h b/py/asmxtensa.h
index d0fd2dcc18..598111f909 100644
--- a/py/asmxtensa.h
+++ b/py/asmxtensa.h
@@ -413,7 +413,7 @@ void asm_xtensa_l32r(asm_xtensa_t *as, mp_uint_t reg, mp_uint_t label);
 #define ASM_SUB_REG_REG(as, reg_dest, reg_src) asm_xtensa_op_sub((as), (reg_dest), (reg_dest), (reg_src))
 #define ASM_MUL_REG_REG(as, reg_dest, reg_src) asm_xtensa_op_mull((as), (reg_dest), (reg_dest), (reg_src))
 
-#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_xtensa_l32i_optimised((as), (reg_dest), (reg_base), (word_offset))
+#define ASM_LOAD_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) ASM_LOAD32_REG_REG_OFFSET((as), (reg_dest), (reg_base), (word_offset))
 #define ASM_LOAD8_REG_REG(as, reg_dest, reg_base) asm_xtensa_op_l8ui((as), (reg_dest), (reg_base), 0)
 #define ASM_LOAD16_REG_REG(as, reg_dest, reg_base) asm_xtensa_op_l16ui((as), (reg_dest), (reg_base), 0)
 #define ASM_LOAD16_REG_REG_OFFSET(as, reg_dest, reg_base, uint16_offset) asm_xtensa_op_l16ui((as), (reg_dest), (reg_base), (uint16_offset))
@@ -423,13 +423,14 @@ void asm_xtensa_l32r(asm_xtensa_t *as, mp_uint_t reg, mp_uint_t label);
         asm_xtensa_op_l16ui((as), (reg_dest), (reg_base), 0); \
     } while (0)
 #define ASM_LOAD32_REG_REG(as, reg_dest, reg_base) asm_xtensa_op_l32i_n((as), (reg_dest), (reg_base), 0)
+#define ASM_LOAD32_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_xtensa_l32i_optimised((as), (reg_dest), (reg_base), (word_offset))
 #define ASM_LOAD32_REG_REG_REG(as, reg_dest, reg_base, reg_index) \
     do { \
         asm_xtensa_op_addx4((as), (reg_base), (reg_index), (reg_base)); \
         asm_xtensa_op_l32i_n((as), (reg_dest), (reg_base), 0); \
     } while (0)
 
-#define ASM_STORE_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_xtensa_s32i_optimised((as), (reg_dest), (reg_base), (word_offset))
+#define ASM_STORE_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) ASM_STORE32_REG_REG_OFFSET((as), (reg_dest), (reg_base), (word_offset))
 #define ASM_STORE8_REG_REG(as, reg_src, reg_base) asm_xtensa_op_s8i((as), (reg_src), (reg_base), 0)
 #define ASM_STORE16_REG_REG(as, reg_src, reg_base) asm_xtensa_op_s16i((as), (reg_src), (reg_base), 0)
 #define ASM_STORE16_REG_REG_REG(as, reg_val, reg_base, reg_index) \
@@ -438,6 +439,7 @@ void asm_xtensa_l32r(asm_xtensa_t *as, mp_uint_t reg, mp_uint_t label);
         asm_xtensa_op_s16i((as), (reg_val), (reg_base), 0); \
     } while (0)
 #define ASM_STORE32_REG_REG(as, reg_src, reg_base) asm_xtensa_op_s32i_n((as), (reg_src), (reg_base), 0)
+#define ASM_STORE32_REG_REG_OFFSET(as, reg_dest, reg_base, word_offset) asm_xtensa_s32i_optimised((as), (reg_dest), (reg_base), (word_offset))
 #define ASM_STORE32_REG_REG_REG(as, reg_val, reg_base, reg_index) \
     do { \
         asm_xtensa_op_addx4((as), (reg_base), (reg_index), (reg_base)); \
diff --git a/py/emitnative.c b/py/emitnative.c
index 4b470f3c93..577bcddfbf 100644
--- a/py/emitnative.c
+++ b/py/emitnative.c
@@ -1537,6 +1537,9 @@ static void emit_native_load_subscr(emit_t *emit) {
             switch (vtype_base) {
                 case VTYPE_PTR8: {
                     // pointer to 8-bit memory
+                    #ifdef ASM_LOAD8_REG_REG_OFFSET
+                    ASM_LOAD8_REG_REG_OFFSET(emit->as, REG_RET, reg_base, index_value);
+                    #else
                     #if N_THUMB
                     if (index_value >= 0 && index_value < 32) {
                         asm_thumb_ldrb_rlo_rlo_i5(emit->as, REG_RET, reg_base, index_value);
@@ -1561,10 +1564,14 @@ static void emit_native_load_subscr(emit_t *emit) {
                         reg_base = reg_index;
                     }
                     ASM_LOAD8_REG_REG(emit->as, REG_RET, reg_base); // load from (base+index)
+                    #endif
                     break;
                 }
                 case VTYPE_PTR16: {
                     // pointer to 16-bit memory
+                    #ifdef ASM_LOAD16_REG_REG_OFFSET
+                    ASM_LOAD16_REG_REG_OFFSET(emit->as, REG_RET, reg_base, index_value);
+                    #else
                     #if N_THUMB
                     if (index_value >= 0 && index_value < 32) {
                         asm_thumb_ldrh_rlo_rlo_i5(emit->as, REG_RET, reg_base, index_value);
@@ -1589,10 +1596,14 @@ static void emit_native_load_subscr(emit_t *emit) {
                         reg_base = reg_index;
                     }
                     ASM_LOAD16_REG_REG(emit->as, REG_RET, reg_base); // load from (base+2*index)
+                    #endif
                     break;
                 }
                 case VTYPE_PTR32: {
                     // pointer to 32-bit memory
+                    #ifdef ASM_LOAD32_REG_REG_OFFSET
+                    ASM_LOAD32_REG_REG_OFFSET(emit->as, REG_RET, reg_base, index_value);
+                    #else
                     #if N_THUMB
                     if (index_value >= 0 && index_value < 32) {
                         asm_thumb_ldr_rlo_rlo_i5(emit->as, REG_RET, reg_base, index_value);
@@ -1617,6 +1628,7 @@ static void emit_native_load_subscr(emit_t *emit) {
                         reg_base = reg_index;
                     }
                     ASM_LOAD32_REG_REG(emit->as, REG_RET, reg_base); // load from (base+4*index)
+                    #endif
                     break;
                 }
                 default:
@@ -1809,6 +1821,9 @@ static void emit_native_store_subscr(emit_t *emit) {
             switch (vtype_base) {
                 case VTYPE_PTR8: {
                     // pointer to 8-bit memory
+                    #ifdef ASM_STORE8_REG_REG_OFFSET
+                    ASM_STORE8_REG_REG_OFFSET(emit->as, reg_value, reg_base, index_value);
+                    #else
                     // TODO optimise to use thumb strb r1, [r2, r3]
                     #if N_THUMB
                     if (index_value >= 0 && index_value < 32) {
@@ -1837,10 +1852,14 @@ static void emit_native_store_subscr(emit_t *emit) {
                         reg_base = reg_index;
                     }
                     ASM_STORE8_REG_REG(emit->as, reg_value, reg_base); // store value to (base+index)
+                    #endif
                     break;
                 }
                 case VTYPE_PTR16: {
                     // pointer to 16-bit memory
+                    #ifdef ASM_STORE16_REG_REG_OFFSET
+                    ASM_STORE16_REG_REG_OFFSET(emit->as, reg_value, reg_base, index_value);
+                    #else
                     #if N_THUMB
                     if (index_value >= 0 && index_value < 32) {
                         asm_thumb_strh_rlo_rlo_i5(emit->as, reg_value, reg_base, index_value);
@@ -1864,10 +1883,14 @@ static void emit_native_store_subscr(emit_t *emit) {
                         reg_base = reg_index;
                     }
                     ASM_STORE16_REG_REG(emit->as, reg_value, reg_base); // store value to (base+2*index)
+                    #endif
                     break;
                 }
                 case VTYPE_PTR32: {
                     // pointer to 32-bit memory
+                    #ifdef ASM_STORE32_REG_REG_OFFSET
+                    ASM_STORE32_REG_REG_OFFSET(emit->as, reg_value, reg_base, index_value);
+                    #else
                     #if N_THUMB
                     if (index_value >= 0 && index_value < 32) {
                         asm_thumb_str_rlo_rlo_i5(emit->as, reg_value, reg_base, index_value);
@@ -1896,6 +1919,7 @@ static void emit_native_store_subscr(emit_t *emit) {
                         reg_base = reg_index;
                     }
                     ASM_STORE32_REG_REG(emit->as, reg_value, reg_base); // store value to (base+4*index)
+                    #endif
                     break;
                 }
                 default: