arm/hevc_idct: fix compilation on Android

Compilation error "out of range" fixed for armeabi-v7a. Compilation failed
trying to build libvlc.aar for ARM7 android on ubuntu 16.04 host. Error
messages is "Offset out of range". The reason of the error is assembler LDR
directives in function "ff_hevc_transform_luma_4x4_neon_8" need local storage
in range <1k, but no such storage provided.

Based on a patch by Ihor Bobalo <bob@eleks.com>

Suggested-by: wbs
Signed-off-by: James Almer <jamrial@gmail.com>

libavcodec/arm/hevcdsp_idct_neon.S

@@ -229,65 +229,6 @@ function ff_hevc_add_residual_32x32_10_neon, export=1
         bx              lr
 endfunc
 
-/* uses registers q2 - q9 for temp values */
-/* TODO: reorder */
-.macro tr4_luma_shift r0, r1, r2, r3, shift
-        vaddl.s16   q5, \r0, \r2    // c0 = src0 + src2
-        vaddl.s16   q2, \r2, \r3    // c1 = src2 + src3
-        vsubl.s16   q4, \r0, \r3    // c2 = src0 - src3
-        vmull.s16   q6, \r1, d0[0]  // c3 = 74 * src1
-
-        vaddl.s16   q7, \r0, \r3    // src0 + src3
-        vsubw.s16   q7, q7, \r2     // src0 - src2 + src3
-        vmul.s32    q7, q7, d0[0]   // dst2 = 74 * (src0 - src2 + src3)
-
-        vmul.s32    q8, q5, d0[1]   // 29 * c0
-        vmul.s32    q9, q2, d1[0]   // 55 * c1
-        vadd.s32    q8, q9          // 29 * c0 + 55 * c1
-        vadd.s32    q8, q6          // dst0 = 29 * c0 + 55 * c1 + c3
-
-        vmul.s32    q2, q2, d0[1]   // 29 * c1
-        vmul.s32    q9, q4, d1[0]   // 55 * c2
-        vsub.s32    q9, q2          // 55 * c2 - 29 * c1
-        vadd.s32    q9, q6          // dst1 = 55 * c2 - 29 * c1 + c3
-
-        vmul.s32    q5, q5, d1[0]   // 55 * c0
-        vmul.s32    q4, q4, d0[1]   // 29 * c2
-        vadd.s32    q5, q4          // 55 * c0 + 29 * c2
-        vsub.s32    q5, q6          // dst3 = 55 * c0 + 29 * c2 - c3
-
-        vqrshrn.s32   \r0, q8, \shift
-        vqrshrn.s32   \r1, q9, \shift
-        vqrshrn.s32   \r2, q7, \shift
-        vqrshrn.s32   \r3, q5, \shift
-.endm
-
-function ff_hevc_transform_luma_4x4_neon_8, export=1
-        vpush       {d8-d15}
-        vld1.16     {q14, q15}, [r0]  // coeffs
-        ldr         r3, =0x4a  // 74
-        vmov.32     d0[0], r3
-        ldr         r3, =0x1d  // 29
-        vmov.32     d0[1], r3
-        ldr         r3, =0x37  // 55
-        vmov.32     d1[0], r3
-
-        tr4_luma_shift d28, d29, d30, d31, #7
-
-        vtrn.16     d28, d29
-        vtrn.16     d30, d31
-        vtrn.32     q14, q15
-
-        tr4_luma_shift d28, d29, d30, d31, #12
-
-        vtrn.16     d28, d29
-        vtrn.16     d30, d31
-        vtrn.32     q14, q15
-        vst1.16     {q14, q15}, [r0]
-        vpop        {d8-d15}
-        bx lr
-endfunc
-
 .macro idct_4x4_dc bitdepth
 function ff_hevc_idct_4x4_dc_\bitdepth\()_neon, export=1
         ldrsh           r1, [r0]
@@ -1040,3 +981,63 @@ idct_32x32 8
 idct_32x32_dc 8
 idct_32x32 10
 idct_32x32_dc 10
+
+/* uses registers q2 - q9 for temp values */
+/* TODO: reorder */
+.macro tr4_luma_shift r0, r1, r2, r3, shift
+        vaddl.s16   q5, \r0, \r2    // c0 = src0 + src2
+        vaddl.s16   q2, \r2, \r3    // c1 = src2 + src3
+        vsubl.s16   q4, \r0, \r3    // c2 = src0 - src3
+        vmull.s16   q6, \r1, d0[0]  // c3 = 74 * src1
+
+        vaddl.s16   q7, \r0, \r3    // src0 + src3
+        vsubw.s16   q7, q7, \r2     // src0 - src2 + src3
+        vmul.s32    q7, q7, d0[0]   // dst2 = 74 * (src0 - src2 + src3)
+
+        vmul.s32    q8, q5, d0[1]   // 29 * c0
+        vmul.s32    q9, q2, d1[0]   // 55 * c1
+        vadd.s32    q8, q9          // 29 * c0 + 55 * c1
+        vadd.s32    q8, q6          // dst0 = 29 * c0 + 55 * c1 + c3
+
+        vmul.s32    q2, q2, d0[1]   // 29 * c1
+        vmul.s32    q9, q4, d1[0]   // 55 * c2
+        vsub.s32    q9, q2          // 55 * c2 - 29 * c1
+        vadd.s32    q9, q6          // dst1 = 55 * c2 - 29 * c1 + c3
+
+        vmul.s32    q5, q5, d1[0]   // 55 * c0
+        vmul.s32    q4, q4, d0[1]   // 29 * c2
+        vadd.s32    q5, q4          // 55 * c0 + 29 * c2
+        vsub.s32    q5, q6          // dst3 = 55 * c0 + 29 * c2 - c3
+
+        vqrshrn.s32   \r0, q8, \shift
+        vqrshrn.s32   \r1, q9, \shift
+        vqrshrn.s32   \r2, q7, \shift
+        vqrshrn.s32   \r3, q5, \shift
+.endm
+
+.ltorg
+function ff_hevc_transform_luma_4x4_neon_8, export=1
+        vpush       {d8-d15}
+        vld1.16     {q14, q15}, [r0]  // coeffs
+        ldr         r3, =0x4a  // 74
+        vmov.32     d0[0], r3
+        ldr         r3, =0x1d  // 29
+        vmov.32     d0[1], r3
+        ldr         r3, =0x37  // 55
+        vmov.32     d1[0], r3
+
+        tr4_luma_shift d28, d29, d30, d31, #7
+
+        vtrn.16     d28, d29
+        vtrn.16     d30, d31
+        vtrn.32     q14, q15
+
+        tr4_luma_shift d28, d29, d30, d31, #12
+
+        vtrn.16     d28, d29
+        vtrn.16     d30, d31
+        vtrn.32     q14, q15
+        vst1.16     {q14, q15}, [r0]
+        vpop        {d8-d15}
+        bx lr
+endfunc